No Specific Topic

Low Power Mini ITX System II (17.5W!): Intel DH67CF and Core i5-2500T -

2011-12-18T14:35:00.001-08:00

After my reasonably unsuccessful attempt to build a PC with a very low idle power consumption I decided to replace the motherboard and CPU. I opted for an Intel DH67CF motherboard and a Core i5-2500T and achieved much better results! The CPU is a Sandy Bridge model with 4 cores and a TDP of just 45W.

With this combo powered by a 150W PicoPSU and using default BIOS settings I was able to achieve an idle power consumption of just 17.5W!This was with an OCZ vertex 3 SSD and a USB keyboard and mouse connected.

Idle power consumption was 19.6W once I connected a second 500GB non-SSD 2.5" hard drive and USB wifi adapter, still really low. Peak power when running Prime95 was 62W. I also tried running the system with a standard power supply, an Enermax Liberty ELT500AWT 500W, which idled at 24.8W.

How to remove the Lock Screen PIN on an Android Phone

2011-08-26T14:13:00.000-07:00

Recently I got in to a situation where I set a lock screen PIN but the once the phone was locked, the PIN did not unlock the phone. This was due to a bug in DarkyROM 10.2 on my Galaxy S. To make matters worse, there was no option to unlock the phone with my Google credentials. Great.

So what to do? Well, there is a really handy app in the market called Screen Lock Bypass. If you install this from the market it will install on your phone automatically as long as your phone has connectivity to the internet. Once you have this installed, I think you need to install another app, any app, not sure why. Then when you wake your phone, you will not have to enter your PIN. This worked fine for me but it is a hack as pressing the home button will make your phone lock. It also means you are unable to change the lock screen settings as you need to enter the original PIN. But it will enable you to backup your phone before restoring factory settings which is a life saver.

There is another option though! You need a rooted phone and you need to be able to get an adb shell on your phone. I'm not going in to details of using adb, if you don't know what I'm talking about then you probably shouldn't be doing it. Anyway, here's how...

You need to run some commands using sqlite3 on a DB file, called settings.db, which can be in one of two locations. To do this you can either put the sqlite3 binary on your phone (I think busybox includes it) and run the commands on your phone, or you can 'adb pull' the settings.db file, run the sqlite3 commands on your computer, then 'adb push' the file back to your phone.

Option 1
Use this option if this file exists on your phone:
/dbdata/databases/com.android.providers.settings/settings.db
You need to run these sqlite3 commands:

sqlite3 <path to settings.db>
SQLite version 3.5.0
Enter ".help" for instructions
sqlite> update secure set value=0 where name='lockscreen.password_salt';
sqlite> update secure set value=0 where name='lockscreen.password_type';
sqlite> update secure set value=0 where name='lockscreen.lockoutattemptdeadline';
sqlite> update secure set value=0 where name='lock_pattern_visible_pattern';

Option 2
Use this option if this file exists on your phone:
/data/data/com.android.providers.settings/databases/settings.db
You need to run this sqlite3 command:

sqlite3 <path to settings.db>
SQLite version 3.5.0
Enter ".help" for instructions
sqlite> update system set value=0 where name='lockscreen.lockexchange.enable';

Update: Just use CyanogenMod 7.1, don't use Darky's Rom.

After spending a few months using Darky ROM 10.x I decided to switch to CyanogenMod as they added support for the Galaxy S in version 7.1. CyanogenMod is much more stable and most importantly doesn't suffer from really annoying bugs like the Darky ROM does, such as the first touch not being recognised and serious browser instability issues. I tried Darky's ROM 10.1, 10.2 and even 10.2.2.

Building an Arduino based Preamp - Relay based Attenuator and Input Selector

2011-03-19T09:51:00.000-07:00

Inspired by the great posts on HiFiDUINO and work by linuxworks I decided I wanted to create a relay based attenuator and input selector capable of being controlled from an Arduino. I looked at some other peoples designs, such as the Joshua Tree, Delta 1 and the RelaiXed and decided to try one that is similar to the Joshua Tree (as I've had success with that in the past) and one that is like the AMB Delta 1. The difference between these two designs is that the Joshua Tree's input impedance varies between 2.2K and 10K and has a fixed output impedance of 750R, where the Delta 1 is just like a Potentiometer with a constant input impedance and variable output impedance.

Here are a couple of calculators kindly supplied by AMB and Jos van Eijndhoven:
jos.vaneijndhoven.net/switchr/switchdesign.html
www.amb.org/audio/delta1/r2r.cgi

For controlling the relays from an Arduino I went with I2C and chose the MPC23008 as its easy to find, cheap and there were plenty of examples in the forums.

I made a schematic:

Designed a few different PCBs using SMD resistors, here is one:

Had them fabricated by Seeed Studio:

Assembled them:

I tried both 1206 and MiniMelf (MMA) 0204 package resistors:

Here is my code for controlling the attenuator with a rotary encoder and the input selector with a push button:

#include "Wire.h"

// Set up the Variables
int state, prevstate, attenuation_value, level = 0;
int input = 1;
int nextEncoderState[4] = { 2, 0, 3, 1 };
int prevEncoderState[4] = { 1, 3, 0, 2 };
int buttonState = 0;
const int buttonPin = 12;

// Define the I2C addresses for the Attenuator and Selector. Use I2CScanner to discover your addresses if you don't know them.
#define ATTAddr 32
#define SELAddr 39

void setup()
{
  // initialise the wire library and MCP23008s
  Wire.begin();
  Wire.beginTransmission(ATTAddr);
  Wire.send(0);
  Wire.send(0);
  Wire.endTransmission();
  Wire.begin();
  Wire.beginTransmission(SELAddr);
  Wire.send(0);
  Wire.send(0);
  Wire.endTransmission();
  
  // Set up for the encoder
  pinMode(5, INPUT);
  pinMode(6, OUTPUT);
  pinMode(7, INPUT);
  digitalWrite(5, HIGH);
  digitalWrite(6, LOW);
  digitalWrite(7, HIGH);
  
  // Set up for the button
  pinMode(buttonPin, INPUT);
  pinMode(11, OUTPUT);
  digitalWrite(11, HIGH);
  
  // Start the serial console
  Serial.begin(9600);
}

// Function to increase the volume
int VolumeUp()
{
  if (level < 127) {
  level = level + 1;
  }
}

// Function to decrease the volume
int VolumeDown()
{
  if (level > 0) {
  level = level - 1;
  }
}

// Function to change to the next input
int NextInput()
{
  input = input + 1;
  delay(500);
  if (input > 5) {
    input = 1;
  }
  if (input == 1) {
    Wire.beginTransmission(SELAddr);      // start talking to the device
    Wire.send(9);                         // select the GPIO register
    Wire.send(B00100000);                 // set register value
    Wire.endTransmission();               // stop talking to the device
  }
  if (input == 2) {
    Wire.beginTransmission(SELAddr);      // start talking to the device
    Wire.send(9);                         // select the GPIO register
    Wire.send(B00010000);                 // set register value
    Wire.endTransmission();               // stop talking to the device
  }
  if (input == 3) {
    Wire.beginTransmission(SELAddr);      // start talking to the device
    Wire.send(9);                         // select the GPIO register
    Wire.send(B00001000);                 // set register value
    Wire.endTransmission();               // stop talking to the device
  }
  if (input == 4) {
    Wire.beginTransmission(SELAddr);      // start talking to the device
    Wire.send(9);                         // select the GPIO register
    Wire.send(B00000100);                 // set register value
    Wire.endTransmission();               // stop talking to the device
  }
  if (input == 5) {
    Wire.beginTransmission(SELAddr);      // start talking to the device
    Wire.send(9);                         // select the GPIO register
    Wire.send(B00000010);                 // set register value
    Wire.endTransmission();               // stop talking to the device
  }
}

// Function to change the volume level variable to the actual value that gets sent to the Attenuator
int nibbleShift(int num) {
  int var = 0;    
  int i, x, y, p;
  int s = 8;    // number of bits in 'num'. (This case a 4bit nibble)
  for (i = 0; i < (s / 2); i++) {
    // extract bit on the left, from MSB
    p = s - i - 1;
    x = num & (1 << p);
    x = x >> p;  
    // extract bit on the right, from LSB
    y = num & (1 << i);
    y = y >> i;
    var = var | (x << i);       // apply x
    var = var | (y << p);       // apply y
  }
  return var;
}

void loop()
{
  // This read the rotary encoder and changes the volume accordingly
  state = (digitalRead(7) << 1) | digitalRead(5);
  if (state != prevstate) {
    if (state == nextEncoderState[prevstate]) {
      VolumeUp();
    } else if (state == prevEncoderState[prevstate]) {
      VolumeDown();
    }
    // The level variable is just a number between 0 - 127 to represent the volume level. 0 being lowest and 127 being the highest.
    attenuation_value = nibbleShift(level);
    // The deactivates the mute relay
    bitWrite(attenuation_value, 0, 1);
    // Set the attenuator to the required value
    Wire.beginTransmission(ATTAddr);     // start talking to the device
    Wire.send(9);                        // select the GPIO register
    Wire.send(attenuation_value);        // set registers
    Wire.endTransmission();              // stop talking to the device
    // Print the volume level to the serial console
    Serial.print( "Volume " );
    Serial.println(level, DEC);
    prevstate = state;
  }

  // Read the button pin and change input if its pressed
  buttonState = digitalRead(12);
  if (buttonState == HIGH) {
    NextInput();
    // Print the input to the serial console
    Serial.print( "Input " );
    Serial.println( input );
   }
}

And how to connect it up:

And they work really well. No hum, grounding problems, clicks or anything. Really smooth volume adjustment. Result! I'll build these in to a preamp soon..

Controlling a PGA2310 with an Arduino

2011-01-16T12:54:00.000-08:00

Here is a quick example of how to control a PGA2310 from an arduino. This will just loop through volume levels between 0 - 127.

Here is how to connect them:

Zero Crossing Detection is enabled (ZCEN to 5V)

Mute is disabled (Mute pin to 5V)

Note that I have left out the audio and analog power connections on the right side of the IC for simplicity. Check the datasheet for complete pinout.

Here is the code:

#include "SPI.h"

const int PGA_CS_PIN = 11;
const int PGA_SCLK_PIN = 13;
const int PGA_SDI_PIN = 12;

int timedelay = 100;
int volumeLevel = 1;

void setup() {
pinMode(PGA_CS_PIN,OUTPUT);
pinMode(PGA_SCLK_PIN,OUTPUT);
pinMode(PGA_SDI_PIN,OUTPUT);
digitalWrite(PGA_CS_PIN,HIGH);
digitalWrite(PGA_SCLK_PIN,HIGH);
digitalWrite(PGA_SDI_PIN,HIGH);
}

void PGA_set_volume(byte value)
{
byte shifted_val = (value << 1);
digitalWrite(PGA_CS_PIN, LOW);     // assert CS
SPI_write(shifted_val);            // right value (0..255)
SPI_write(shifted_val);            // right value (0..255)
digitalWrite(PGA_CS_PIN, HIGH);    // deassert CS

}

static inline void SPI_write(byte out_spi_byte)
{
byte i;
// loop thru each of the 8-bits in the byte
for (i=0; i < 8; i++) {
// strobe clock
digitalWrite(PGA_SCLK_PIN, LOW);
// send the bit (we look at the high order bit and 'print' that to the remtoe device)
if (0x80 & out_spi_byte)    // MSB is set
digitalWrite(PGA_SDI_PIN, HIGH);
else
digitalWrite(PGA_SDI_PIN, LOW);
// unstrobe the clock
digitalWrite(PGA_SCLK_PIN, HIGH);
out_spi_byte <<= 1;   // left-shift the byte by 1 bit
}
}

void loop()
{
// Reset the volumeLevel if it goes over 128. I think it can actually go to 255.
if (volumeLevel > 127) {
volumeLevel = 1;
}
// Set the volume
PGA_set_volume(volumeLevel);
delay(timedelay);
volumeLevel = volumeLevel + 1;
}

Credit goes to linux-works on the arduino forums for the code, I just tweaked it and made a

diagram.

Using an encoder
And here is an example using a rotary encoder to control the volume level between 0 and 127:

#include "SPI.h"

const int PGA_CS_PIN = 11;
const int PGA_SCLK_PIN = 13;
const int PGA_SDI_PIN = 12;
int state, prevstate, volumeLevel, level = 0;
int nextEncoderState[4] = { 2, 0, 3, 1 };
int prevEncoderState[4] = { 1, 3, 0, 2 };

void setup() {
pinMode(PGA_CS_PIN,OUTPUT);
pinMode(PGA_SCLK_PIN,OUTPUT);
pinMode(PGA_SDI_PIN,OUTPUT);
digitalWrite(PGA_CS_PIN,HIGH);
digitalWrite(PGA_SCLK_PIN,HIGH);
digitalWrite(PGA_SDI_PIN,HIGH);
pinMode(5, INPUT);
pinMode(6, OUTPUT);
pinMode(7, INPUT);
digitalWrite(5, HIGH);
digitalWrite(6, LOW);
digitalWrite(7, HIGH);
Serial.begin(9600);
}

void PGA_set_volume(byte value)
{
byte shifted_val = (value << 1);
digitalWrite(PGA_CS_PIN, LOW);     // assert CS
SPI_write(shifted_val);            // right value (0..255)
SPI_write(shifted_val);            // right value (0..255)
digitalWrite(PGA_CS_PIN, HIGH);    // deassert CS
}

static inline void SPI_write(byte out_spi_byte)
{
byte i;
// loop thru each of the 8-bits in the byte
for (i=0; i < 8; i++) {
// strobe clock
digitalWrite(PGA_SCLK_PIN, LOW);
// send the bit (we look at the high order bit and 'print' that to the remtoe device)
if (0x80 & out_spi_byte)    // MSB is set
digitalWrite(PGA_SDI_PIN, HIGH);
else
digitalWrite(PGA_SDI_PIN, LOW);
// unstrobe the clock
digitalWrite(PGA_SCLK_PIN, HIGH);
out_spi_byte <<= 1;   // left-shift the byte by 1 bit
}
}

void loop()
{
state = (digitalRead(7) << 1) | digitalRead(5);
if (state != prevstate) {
if (state == nextEncoderState[prevstate]) {
if (volumeLevel < 127) {
volumeLevel = volumeLevel + 1;
PGA_set_volume(volumeLevel);
Serial.println(volumeLevel, DEC);
}
} else if (state == prevEncoderState[prevstate]) {
if (volumeLevel > 0) {
volumeLevel = volumeLevel - 1;
PGA_set_volume(volumeLevel);
Serial.println(volumeLevel, DEC);
}
}
prevstate = state;
}
}

And here is how to connect it all:

Here is a PGA based module I made specifically for use with an arduino:

Chrome Issue: One tab won't scroll while another is loading. Resolved!

2010-12-17T09:24:00.001-08:00

I had this super annoying problem with chrome: When I opened links from one page into new background tabs (ctrl + click) the page I was on would stop scrolling. Soooo annoying!

This would happen on pages that had a lot of rich content like videos etc. My PC had 4x cores and 4GB of RAM so I knew it wasn't an issue of resources. It was caused by the way chrome groups tabs under one process. You can see this by watching the number of chrome processes with a task manager and then ctrl-clicking links from Google search results. You will see that all tabs opened from the search results will all use the same process as the search results page, and no increase in the number of chrome processes running. You can also see this in the chrome task manager (shift + escape in chrome), it will show many tabs grouped under one process.

The solution? Start chrome with the "--process-per-tab" switch. Go to the properties of your shortcut and add it to the end of the target:

This solved the issue for me. I can still see that sometimes tabs are grouped to one process. I'm not quite sure what logic chrome uses to group tabs into single processes but I haven't seen the scrolling problem since. Result.

Low Power Mini ITX System: Xeon L3406 and H55N-USB3

2010-10-06T02:29:00.000-07:00

If Toms Hardware can build a 25W PC using a 220W PSU and 73W TDP CPU I figured could do better. I decided on the Gigabyte H55N-USB3 due to favourable reviews and price but there are no Intel Core i3/i5/i7 CPUs with a TDP less than 73w! Looking through the Intel Ark I did find another CPU for an LGA1156 socket though, the Xeon L3406 with a TDP of 30W!

I wasn't sure if this CPU would work in my Gigabyte H55 based motherboard, it wasn't on the CPU compatibility list. I googled plenty but couldn't find anyone using this CPU in my desktop motherboard. I did find this though:

http://bbs.kakaku.com/bbs/K0000079526/SortID=11177770/ (needs translation from Japanese), so it looked liked the Xeon worked in the bigger Gigabyte GA-H55-UD3H so figured it would be OK.

For a power supply I used a 150W PicoPSU which is over 96% efficient.

So in summary:

Gigabyte H55N-USB3

Intel Xeon L3406

Thermalright AXP-140 with a Fractal Design 140mm Fan

XFX HD 5450 512MB Graphics Card

4GB RAM

30GB SSD + 500GB HDD

150W PicoPSU

Lian Li PC-Q07B mini-ITX Case

Intel Core i5 650 temperature idle/load: 37/57

Intel Xeon L3406 temperature idle/load: 36/48

Clearance around the AXP-140 Cooler:

Had to use the AXP-140 mounting plate in reverse due to some components around the mounting holes:

The H55N-USB3 posting with the Xeon L3406:

In the Lian Li PC-Q07B mini-ITX Case:

UPDATE 20/02/2011

I got a discounted power meter for £7. The specs state the accuracy is +/- 0.5%. Here are the measurements:

Off: 4 Watts

Idle: 44 Watts

Load (Prime95): 65 Watts

So, not exactly what Tom's Hardware measured but the same as one of the comments on this post (using a i3-530). Any idea how Tom's Hardware get 26W with their system? Surely a 30W Xeon L3406 doesn't idle that much higher than a 87W i5-661!

UPDATE 26/02/2011

After tweaking some BIOS settings (enabling C1E and EIST) I was unable to get it any lower. I switched the Xeon L3406 for a 73W Core i5-650 and this is what I got:

Off: 4 Watts

Idle: 38 Watts

Load (Prime95): 90 Watts

So perhaps the power management functions for an unsupported CPU are just not quite right. Also I noted that the keyboard, mouse and USB wifi consumed about 4W.

After lowering some voltage settings in the BIOS and disconnecting USB wifi these are the results:

Idle: 34 Watts

Load (Prime95): 86 Watts

How to install VMWare Tools on RHEL/Centos in one command

2010-07-20T08:38:00.000-07:00

[root@host ~]# rpm --import http://packages.vmware.com/tools/VMWARE-PACKAGING-GPG-KEY.pub && echo -e "[vmware-tools]\nname=VMware Tools for Red Hat Enterprise Linux \$releasever – \$basearch\nbaseurl=http://packages.vmware.com/tools/esx/4.0u1/rhel5/x86_64\nenabled=1\ngpgcheck=1" > /etc/yum.repos.d/vmware-tools.repo && yum install -y vmware-tools-nox && /etc/init.d/vmware-tools start

Works for me in RHEL5 and CentOS 5. Easy!

How to dump a Postgresql DB over SSH with compression and other handy commands!

2010-07-20T08:34:00.000-07:00

If you want to do it without password prompts you will need to set up ssh keys and .pgpass for the associated users. If you want to dump all databases on a host use pg_dumpall and if you want all target databases to be dropped then use the "--clean" switch. Easy!

Dumping

To dump a DB to a text file:

[root@host ~]# pg_dump --clean --username=postgres --host=database-server.domain.com ExampleDbName > /path/to/text-file-dump.sql

To dump all DBs on a host to a text file:

[root@host ~]# pg_dumpall --clean --username=postgres --host=database-server.domain.com > /path/to/text-file-dump.sql

To dump a DB to a compressed text file:

[root@host ~]# pg_dump --clean --username=postgres --host=database-server.domain.com ExampleDbName | gzip > /path/to/compressed-file-dump.gz

To dump a DB to a compressed text file on a remote host over SSH:

[root@host ~]# pg_dump --clean --username=postgres --host=database-server.domain.com ExampleDbName | gzip | ssh remote-host.domain.com cat ">" /remote/path/to/compressed-file-dump.gz

To load a dump into a DB from a text file:

[root@host ~]# psql --username=postgres --host=database-server.domain.com ExampleDbName --file /path/to/text-file-dump.sql

[root@host ~]# cat /path/to/text-file-dump.sql | psql --username=postgres --host=database-server.domain.com ExampleDbName

To load a dump in to a DB from a compressed text file:

[root@host ~]# gunzip -c /path/to/compressed-file-dump.gz | psql --username=postgres --host=database-server.domain.com ExampleDbName

Combos!

To dump a DB and load it into a remote host:

[root@host ~]# pg_dump --clean --username=postgres --host=database-server.domain.com ExampleDbName | psql --username=postgres --host=another-database-server.domain.com ExampleDbName

To dump a DB and load it in to a remote host over SSH with gzip compression:

[root@host ~]# pg_dump --clean --username=postgres --host=database-server.domain.com ExampleDbName | gzip | ssh remote-host.domain.com gunzip -c "|" psql --username=postgres --host=another-database-server.domain.com ExampleDbName

To dump all DBs on a host, load them into a remote host over SSH with gzip compression:

[root@host ~]# pg_dumpall --clean --username=postgres --host=database-server.domain.com | gzip | ssh remote-host.domain.com gunzip -c "|" psql --username=postgres --host=another-database-server.domain.com

How to get the WDTV working with a RT2870 wifi adapter and WPA2

2010-01-10T07:48:00.000-08:00

I wanted to add wireless to my WD TV so I didn't have to copy video files to a USB flash drive every time. I read all the instructions, flashed my WD TV with the WDLXTV firmware and tried it with my Linksys WUSB54GC but couldn't get it to work. This adapter was huge anyway:

So I decided to find a compatable USB wifi adapter that was really small. I got the Edimax EW-7711UTn which has a Vendor ID of 0x7392 and a product ID of 0x7711. According to info attached to this forum post, this device will run with the RT3070 driver.

Here is how to get it working with WPA2, DHCP and samba so you can mount share from your windows desktop and play videos from there. First you need to flash your WD TV with the WDLXTV firmware.

Get these three files and extract them to the root of a USB drive:

log-saver.app.bin.zip: http://sites.google.com/site/nospecifictop/log-saver.app.bin.zip

samba.app.bin.zip (version 0.2): http://sites.google.com/site/nospecifictop/samba.app.bin.zip

wireless-mod-rt.app.bin.zip (verison 1.2): http://sites.google.com/site/nospecifictop/wireless-mod-rt.app.bin.zip

(Its so annoying how the people who produce these apps put them on mediafire and rapid share!)

Create the following text files in the root of your flash drive.

net.config:

HOST=wdtv

#one of rt61.ko, rt73.ko, rt2400.ko, rt2500.ko, rt2570.ko, rt2870sta.ko, rt3070sta.ko

WLAN_LOAD=rt3070sta.ko

#one of rt61sta.dat, rt73sta.dat, RT2400.dat, RT2500STA.dat, RT2570STA.dat, RT2870STA.dat (used by rt3070 as well)

WLAN_CFG=RT2870STA.dat

WLAN_UP=yes

DHCP=yes

net.mounts:

xmount //192.168.1.100/Videos Videos cifs

RT2870STA.dat:

Default

CountryCode=GB

SSID=your-ssid-name

NetworkType=Infra

WirelessMode=0

Channel=your-ap-channel

AuthMode=WPA2PSK

EncrypType=AES

WPAPSK=your-password

Some of these options may need customising to your environment, specifically the dat file.

Plug your USB wifi adapter and flash drive in to your WD TV and turn it on. If all has worked well you should see an IP address on the boot up screen. Wait for the home screen to load up. Now if everything didn't work as planned, power off the WD TV and plug the flash drive in to a PC and open the messages.txt log file. This will give you clues about the problem. When everything works it should look like this near the end:

Jan 1 00:00:35 WDTV-6e0d5 user.notice root: wireless-mod-rt: Starting..

Jan 1 00:00:35 WDTV-6e0d5 user.notice root: wireless-mod-rt: Creating read+write /etc under /tmp/.root

Jan 1 00:00:36 WDTV-6e0d5 user.notice root: wireless-mod-rt: Installing new device configs to /etc/mdev.conf

Jan 1 00:00:36 WDTV-6e0d5 user.warn kernel: wireless_ext: No versions for exported symbols. Tainting kernel.

Jan 1 00:00:36 WDTV-6e0d5 user.info kernel: Loading wireless extentions module...

Jan 1 00:00:36 WDTV-6e0d5 user.notice root: wireless-mod-rt: WLAN_LOAD=rt3070sta.ko, config=/tmp/mnt/18A2-EDDB/RT2870STA.dat

Jan 1 00:00:36 WDTV-6e0d5 user.notice root: wireless-mod-rt: Copying /tmp/mnt/18A2-EDDB/RT2870STA.dat to /etc

Jan 1 00:00:36 WDTV-6e0d5 user.notice root: wireless-mod-rt: Install modules /apps/wireless-mod-rt/modules/rt3070sta.ko

Jan 1 00:00:37 WDTV-6e0d5 user.warn kernel: rtusb init --->

Jan 1 00:00:37 WDTV-6e0d5 user.warn kernel:

Jan 1 00:00:37 WDTV-6e0d5 user.warn kernel: === pAd = c9206000, size = 462696 ===

Jan 1 00:00:37 WDTV-6e0d5 user.warn kernel:

Jan 1 00:00:37 WDTV-6e0d5 user.warn kernel: <-- RTMPAllocAdapterBlock, Status=0 Jan 1 00:00:37 WDTV-6e0d5 user.notice root: wireless-mod-rt: hotplug action=add Jan 1 00:00:37 WDTV-6e0d5 user.info kernel: usbcore: registered new driver rt2870 Jan 1 00:00:37 WDTV-6e0d5 user.notice root: wireless-mod-rt: Bringing up wireless device 'ra0'... Jan 1 00:00:38 WDTV-6e0d5 user.warn kernel: <-- RTMPAllocTxRxRingMemory, Status=0 Jan 1 00:00:38 WDTV-6e0d5 user.warn kernel: -->RTUSBVenderReset

Jan 1 00:00:38 WDTV-6e0d5 user.warn kernel: <--RTUSBVenderReset

Jan 1 00:00:43 WDTV-6e0d5 daemon.info utelnetd[1322]: utelnetd (port: 23, ifname: any, login: /bin/login) startup succeeded

Jan 1 00:00:47 WDTV-6e0d5 user.warn kernel: 1. Phy Mode = 0

Jan 1 00:00:47 WDTV-6e0d5 user.warn kernel: 2. Phy Mode = 0

Jan 1 00:00:47 WDTV-6e0d5 user.warn kernel: NVM is Efuse and its size =2d[2d0-2fc]

Jan 1 00:00:47 WDTV-6e0d5 user.warn kernel: 3. Phy Mode = 0

Jan 1 00:00:47 WDTV-6e0d5 user.warn kernel: MCS Set = 00 00 00 00 00

Jan 1 00:00:47 WDTV-6e0d5 user.warn kernel: <==== rt28xx_init, Status=0

Jan 1 00:00:47 WDTV-6e0d5 user.warn kernel: 0x1300 = 00073200

Jan 1 00:00:47 WDTV-6e0d5 user.notice root: wireless-mod-rt: Waiting for AP association...

Jan 1 00:00:47 WDTV-6e0d5 user.notice root: wireless-mod-rt: AP not associated ... sleeping for 1 sec

Jan 1 00:00:48 WDTV-6e0d5 user.notice root: wireless-mod-rt: AP not associated ... sleeping for 1 sec

Jan 1 00:00:49 WDTV-6e0d5 user.notice root: wireless-mod-rt: AP not associated ... sleeping for 1 sec

Jan 1 00:00:50 WDTV-6e0d5 user.debug kernel: ir: Enable NEC decoder (0x00000000)

Jan 1 00:00:50 WDTV-6e0d5 user.debug kernel: ir: Enable RC5 decoder (0x00000000)

Jan 1 00:00:50 WDTV-6e0d5 user.notice root: wireless-mod-rt: AP not associated ... sleeping for 1 sec

Jan 1 00:00:52 WDTV-6e0d5 user.notice root: wireless-mod-rt: AP associated ... after 4 seconds

Jan 1 00:00:52 WDTV-6e0d5 user.notice root: wireless-mod-rt: ra0 RT2870 Wireless ESSID:"OMGWTF" Nickname:"RT2870STA" Mode:Managed Frequency=2.452 GHz Access Point: 00:18:39:1E:8A:AD Bit Rate=54 Mb/s RTS thr:off Fragment thr:of

Jan 1 00:00:52 WDTV-6e0d5 user.notice root: net.agent: Called for ra0

Jan 1 00:00:52 WDTV-6e0d5 user.warn kernel: ERROR!!! GetSuiteFromRSNIE : it can't get any more information beyond AKM

Jan 1 00:00:52 WDTV-6e0d5 user.notice root: net.agent: No IP provided, doing DHCP

Jan 1 00:00:53 WDTV-6e0d5 user.notice net.agent: net.agent: Using to set system time

Jan 1 00:00:54 WDTV-6e0d5 user.notice root: Device ra0 configged as 192.168.1.103/192.168.1.1/255.255.255.0

Jan 1 00:00:55 WDTV-6e0d5 user.notice net.agent: net.mounts: loading /tmp/net.mounts

Jan 1 00:00:56 WDTV-6e0d5 user.notice xmount: xmounted //192.168.1.100/Videos

Jan 10 16:53:06 WDTV-6e0d5 user.notice mount.watch: Unmounting official device /tmp/media/usb/USB256MB

Jan 10 16:55:12 WDTV-6e0d5 user.warn kernel: ERROR!!! RTMPCancelTimer failed, Timer hasn't been initialize!

Jan 10 16:56:24 WDTV-6e0d5 user.warn kernel: program dmaosd is using a deprecated SCSI ioctl, please convert it to SG_IO

Jan 10 16:56:26 WDTV-6e0d5 user.warn kernel: program dmaosd is using a deprecated SCSI ioctl, please convert it to SG_IO

Jan 10 16:56:33 WDTV-6e0d5 user.warn kernel: program dmaosd is using a deprecated SCSI ioctl, please convert it to SG_IO

Jan 10 16:56:35 WDTV-6e0d5 user.warn kernel: program dmaosd is using a deprecated SCSI ioctl, please convert it to SG_IO

So how well does the Edimax EW-7711UTn wifi adapter perform with the WD TV? Well its not fast enough for 1080p or 720p but it works fine with 1 or 2 CD sized DVD rips which is fine by me. I tested the speed and to copy a 1400MB file from a windows share to the flash drive took 44 minutes. That means the speed is about 4.4Mbit/Sec. Pretty slow but bear in mind this is over CIFS (windows/samba) which is not a fast file transfer protocol and I am using WPA2 encryption. I bet it would be considerably faster with NFS over a non or weakly encrypted connection.

By the way, you need to disable the media library feature otherwise the WD TV just sits there scanning the windows share for ages.

I hope this info helps someone else wanting to do that same thing!

Nixon Player Movement

2009-12-31T03:14:00.000-08:00

In case you've ever wondered what kind of movement is inside your Nixon watch, well here ya go:

The movement is made my Miyoto who are owned by Citizen. Not particularly flash but not a cheap chinese made movement!

Linux Patch Management

2009-02-08T18:09:00.001-08:00

Introduction
While contracting for a large organisation I was tasked with finding a product for managing software and patching for 100+ Linux servers. Finding a provisioning solution was part of this project too.

The solution needed to support RHEL 5 and SLES 10 both 32 and 64bit. Once again I would have much rather use the standard Linux software management tools such as yum and mrepo but they had requirements and policies dictated from above that demanded a commercially supported product with a very specific feature set. Unfortunately I did not have time to install and test each option very thoroughly, just a quick play to get a feel for the application and it's possible limitations.

After some amount of research by the architect the options were PatchLink Update Server, Novell ZENworks Linux Management and BigFix Enterprise Suite. Other options were considered but ruled out for the following reasons:
KACE - KBOX only supports RHEL and they have no plan to support any other distros.
Redhat Satellite/Spacewalk - Only supports RH based distributions.
Shavlink - Used to support Linux agents but does not anymore.
Altiris - Bought by Symantec. Altiris CMS only supports windows but after speaking with a Symantec Product Manager via the Patch Management mailing list I found out the next version will (in beta) will support RHEL5/SLES10.
LANDesk Patch Manager - Most features are not supported on Linux hosts.

Results
Novell ZENWorks Linux Management (ZLM)
URL: www.novell.com/products/zenworks/linuxmanagement/
As mentioned in my Linux Provisioning post, ZLM works on the concept of "bundles" that are a collection of files or rpm packages that can be pushed out to servers or groups of servers. This process is used for patching and software management. It has built in capabilities to mirror the distribution repositories which you get access to when you buy a subscription for RHEL or SLES.
It comes with instructions on how to extract the contents of your distro CD and use it as a "catalog" for your ZLM managed hosts. This basically makes the ZLM server a repository for all the base packages which can be used to resolve dependencies when you push out software. For example, suppose you make a bundle containing an emacs rpm but it depends on the emacs-info rpm. If emacs-info is in the catalog it will be automatically installed to resolve the dependency. This system is pretty much identical to traditional Linux package management systems that use repositories.
After playing with ZENworks for a while, pushing out a bunch of patches and random applications to a couple of hosts I thought it's function was good and flexible but a bit basic. If you want to push a patch out to 100 servers you have to specifically add that rpm to a bundle then assign it to a server group or single server. That is fine but suppose you just want to blanket update all packages in one go? Could be very fiddly. One option is to create a un-hidden catalog which is a mirror of the online updates repo of RHEL/SLES and then do a "# rug update" on each server but this is no different from just using mrepo and using yum. One good feature is you are able to include pre/post scripts in bundles which would give anyone with shell scripting skills a lot of flexibility.
ZLM also has provisioning system where you can create an image of one server and apply it to another which could be good if you wanted to clone a system to many more hosts with identical hardware.
I get the feeling this product would be good for people who are not that familiar with Linux, maybe Windows people who want a GUI.
Screenshots of ZLM:

BigFix Enterprise Suite (BES)
URL: www.bigfix.com
This product supports Windows and Linux so I was a bit dubious as to how it would accommodate the two platforms due to their software management process being so different. BES revolves around a concept of "fixlets" which can be applied to hosts or groups of hosts. A fixlet is obtained from BigFix and addresses a specific software update or a patch. It contains the actual patch (RPMs that are downloaded to the client) and a script that checks for dependencies and it also details the process for installing the patch. In the UI a fixlet clearly shows you all this information and also links to the relevant web page of the vendor supplying the patch, such as an RHN errata advisory.
Bigfix have developed their own scripting language to check for dependencies and to check that the fixlet is being applied to the right platform. This language is very easy to understand and write so you can easily create your own custom fixlets to push out your own application or files. I can see how this is a powerful solution in a multi-platform environment but we were all Linux and this system basically doesn't utilize the dependency management systems already in all versions of Linux. I wasn't happy about this as it would mean manually scripting the installation of packages and their dependencies when pushing out custom packages. It will not mean that an RPM will be forcefully installed when dependencies aren't met but it could see it leading to troubles later on, for example: What if you want to push out a package to 100 servers but not all servers have the same packages installed? Some servers might have all the dependencies installed but some might not. Your fixlet will have to accommodate all possible options which involves a lot of scripting. Sure if all servers have the same configuration and software installed you can write one fixlet and it will work on them all but what if people have custom or odd versions of packages installed?
Apart from this concern, the BES UI is very good and fast to navigate. It has good host grouping functionality and a lot of options to support fixlets for a vast amount of software, not just OS patches.
BigFix Enterprise Suite Screenshots:

Patchlink Update Server (PLUS)
URL: www.lumension.com/patch-management.jsp
Just like BES, PLUS supports windows and Linux platforms, and also revolves around the same concept:
Software update packages are downloaded from Patchlink via the internet and can be applied to hosts or groups of hosts. These packages, called "Vulnerabilities" in PLUS, contain information like severity and dependencies etc. Also just like BES it works very well out of the box when applying standard updates but I was more interested in flexibility and how we could use the product to deploy our own packages, software, scripts or configuration files. This is were PLUS falls short in Linux. If you want to deploy a single RPM you need to upload the RPM and then include a post script that actually installs it, ie "# rpm -ivh xyz-1.23.rpm". This is no good for managing 100+ servers. You would have to write script for every single rpm file you wanted to deploy = Ballache.
The PLUS UI is via a web browser and it's OK but not great.
Patchlink Update Server (PLUS) Screenshots:

Conclusion
Both Patchlink Update Server and BigFix Enterprise Suite were very capable enterprise level products with excellent features and reporting capabilities but I think few of these features lent well to the Linux platform. They don't follow the conventional repository approach for Linux package management like ZLM does which makes me worry about having to manually sort out dependency problems down the track. They kind of felt like Linux support has been added after the product was designed for Windows.
I chose ZENworks Linux Management because it acts like a normal Linux package repository allowing you to push out packages to many host and have them resolve dependencies automatically using the repository. This process is much more comfortable for a Linux admin to use as opposed to BES and PLUS which require you to basically work out and cater for all dependencies manually when rolling out software. ZLM also has options for deploying scripts or configuration files and applying policies to desktop applications like Evolution. ZLM mirrors the repositories of Redhat and Novell so when they release a patch you get it straight away but when using BES and PLUS you have to wait for them to produce a "fixlet" or "vulnerability" for each updated RPM before applying it.
If we were deploying only Redhat hosts I would have much rathered use Redhat Satellite or Spacewalk but we needed to support SLES as well. Satellite/Spacewalk is much better than ZLM with many more features and options but ZLM is still very functional.

Linux Provisioning

2009-02-01T19:29:00.000-08:00

Introduction
While contracting for a large organisation I was tasked with finding a product for the mass provisioning of Linux servers. That is, automated installations of Linux to many hosts over the network with minimal administration and configuration.

They were deploying RHEL 5 and SLES 10. While most Linux administrators would rather use open source tools included on all Linux distributions such as netboot, tftp, apache and dhcpd to achieve this, they had requirements and policies from above that demanded a commercially supported product.

After some amount of research by the architect the options were only Linmin and Novell ZENworks. I also planned on testing Redhat Spacewalk/Satellite and Cobbler just for fun but these were not officially considered as Spacewalk and Satellite do not (fully) support provisioning of SUSE Linux and Cobbler is not a commercial product.

Results

Linmin
This product is basically a PXE, dhcp and tftp server with a basic web interface and process for importing distributions so they can be deployed. It can also deploy Windows and create or apply disk images over the network but we were not interested in this functionality. It feels like underneath it's just the regular open source deployment tools (like netboot, tftp and dhcpd) that have been preconfigured but I didn't really have time to investigate this. Apparently it's partly based on webmin.

It is very easy to install and setup as it is just one install file which is a shell script and binary combined. It asks you a bunch of questions during the install process and the documentation is adequate. Importing distributions is easy for all the main ones, you just run a perl script, select your distribution, insert your CD/DVD and it copies all the contents. You can then use the web interface to setup a profile and you're good to go. You can have this installed and ready to go in less than an hour which is good.

Overall though I found the web UI clunky, crude and unfriendly to use. Yes this product does save you time when compared to using the open source tools but only by an of hour or so if you are savvy. The open source tools are more flexible and the web UI of Cobbler is much better. It's very cheap to buy but for these reasons I still felt no inclination to choose Linmin over Cobbler.

Linmin screenshots:

Novell ZENWorks Linux Management (ZLM)
ZLM is a big expensive product and you would not buy it just for provisioning but we were looking for a patch management solution too so it was attractive.

ZLM is a total ball-ache to install unless your OS (SLES or RHEL) installation is "default", ie you don't change any of the package selections and just install all the default package groups which is stupid because what admin doesn't do this? The install script doesn't sort out any dependencies for you and there are lots. If something is missing it dies with unhelpful error messages like this:

Traceback (most recent call last):
File "/root/iso-mount/zlm-server/data/zlminstall.py", line 564, in ?
main()
File "/root/iso-mount/zlm-server/data/zlminstall.py", line 437, in main
downloadID, transactID = zmd.transact(packages, None)
File "/root/iso-mount/zlm-server/data/ZLMZMDLib.py", line 413, in transact
return(downloadID, transactID)
UnboundLocalError: local variable 'downloadID' referenced before assignment

But, if you run the installer on a default installation is works fine.

ZLM works on the concept of "bundles" that are a collection of files or rpm packages that can be pushed out to servers or groups of servers. This process is used for patching and software management. You can also include pre/post scripts in bundles so it is quite flexible if you are good at scripting. It has built in capabilities to mirror the distribution repositories which you get access to when you buy a subscription for RHEL or SLES.

For provisioning it just behaves like a normal PXE server and boots the kernels with the path to the installation source and the installation script (Yast for SUSE and kickstart for RHEL) so it's pretty basic. It does have one interesting feature that is policy based provisioning. For example you can specify portions of MAC addresses, BIOS serial number, BIOS asset tag, RAM or HDD size or any combination of these to dictate which installation profile to use. You have to manually copy all the content from the distribution CDs to a server that is available to the provisioned hosts. It is not done automatically like Cobbler and Linmin but this is a minor step so not really a problem.

The web UI is OK, not pretty but very functional:

Redhat Satellite and Spacewalk
Spacewalk/Satellite was never considered as it does not officially support SLES but I thought I would include it anyway just for comparison. Spacewalk is the upstream open source version of Satellite so I was very interested to see it in action as I have used Satellite extensively before. RH Satellite you have to pay for but Spacewalk you do not.

It is pretty much identical to RH Satellite in terms of the basic layout and functionality as you would expect. I noticed quite a few more options available for bare-metal kickstarting compared to the version of Satellite I had used previously.

Because the under lying provisioning system of Spacewalk is based on cobbler you can also deploy SUSE and Debian based distributions. Spacewalk/Satellite has a really good system for editing the kickstart scripts that are used for an automated installation. The different parts of the script are broken down in the web UI so you can change them individually. Sure you can just edit text files if you wish but this system is much easier. But this ability is lost when provisioning any distribution that is not Redhat/kickstart based (ie. SLES, debian etc).

After having a quick look around the UI it looks like it gives you all the options of cobbler and more. I did not have time to test this though as it takes a long time to set Spacewalk up, and then more time again to import the distribution packages and tree structure etc.

While you can provision SLES and RHEL with spacewalk, I didn't consider using it as its a big product to use just for that tiny functionality even if it does have the best UI out of all the products. I would love the chance to use Spacewalk in the future but it won't be with this employer.

Spacewalk Screenshots:

Cobbler
After reading through the Cobbler website I thought it sounded good but was weary of bugs, lack of documentation and thorough testing which can be very common with small open source applications that are newish. A lot of the time you follow the documentation exactly, at some point it fails and the only option you have left is google and the applications mailing list. But after installing it and testing it out I was pleasantly surprised!

Once installed you set a few settings in the config file, import some distributions, add some profiles and you're good to go. I managed to get it all working from following the documentation online in about half hour with little fuss. It is quite a complete solution as it manages the configuration of dhcpd and DNS for you. It can also register each device in to the Cobbler system after the OS installation so you can see a list in the web UI of your provisioned systems. It has many, many features and looks really flexible. I like it a lot.

Conclusion
We chose to use Novell ZENworks Linux Management because we would also be using it for software management and because its a commercially supported product. I would have preferred Cobbler because its a lot more flexible. But ZLM is fine, it offers the basic OS install via PXE boot menu which is adequate.

I would recommend each of these products in this way:
Linmin - If you are not so confident manually setting up a Linux provisioning server or Cobbler, want Windows support and disk imaging functionality.
Cobbler - If you like tweakability, many options and open source software.
Spacewalk - If you want a free alternative to Satellite with all it's software management and monitoring etc and need to support many RH based hosts.
ZLM - If you can get it cheaply or already have it somewhere in your infrastructure otherwise use Cobbler. Novell offer academic licensing which brings the cost down a lot. Otherwise, ZLM is very expensive.

DIY Preamp

2006-01-01T01:54:00.000-08:00

This preamp is basically made up of a few separate parts: Case, input selector, attenuator (volume control), line stage, phono stage and soft power circuit.

Input Selector and Attenuator
The input selector uses relays and a rotary switch to control them. Its called Darwin and you can get one from Twisted Pear Audio.

The attenuator also uses relays but it is digitally controlled. It uses a pot to set the level but the audio signal does not actually go through the pot. The value of the pot is read by the control section which adjusts the level accordingly. Its a very neat way of controlling a relay based attenuator. Its called a Joshua Tree Attenuator and is also available from Twisted Pear Audio.

Phono Stage
The phono stage is based on the VSPS by RJM. I designed a compact PCB lay out and included an on board regulated PSU using LM317/LM337 regulators.

Schematic:

Layout in Eagle PCB:

Finished:

Linestage
I wanted to make a linestage for it as definitely needs more gain but I haven't had time!

Soft-Power Circuit
This enables you to use a momentary switch to control a relay which turns on the main power. This is a much more professional way of controlling power IMHO as opposed to using a 220/110V power switch. Most commercial audio equipment uses a similar system. There is a much wider choice of switches which are low voltage and momentary too. I chose to use the Bulgin MP0037 vandal resistant switch.

I copied this design from a thread on DIYAudio.com and made the layout to suit the transformer I could get from Jaycar. It uses a 4013 flip flop IC to control a power relay. I used this method for making the PCB.

Layout in Eagle PCB:

Finished:

The Case
It took way more work to finish the case than any other part of this preamp but I'm very happy with how it turned out.

I bought the aluminium from a metal dealer and got them to cut it for me:

I used a wood router and bit to round the end edges of the front panel, it worked very well:

I used 12mm square aluminium for the corners, drill and tapped all the threads my self:

An engineer cut the large holes in the front panel for the volume and input control knobs:

The knobs and RCA sockets are from THLAudio in Taiwan. I mounted the volume pot and input selector switch on a lexan panel behind the front panel to create a ring of light around each knob:

I ended up making two cases, one for a preamp and one for a matching power amp:

I got the metal polished!! OMG it came up so shiny! Here is an HDR photo:

2-Way Speakers

2005-10-01T02:51:00.000-07:00

I made these using these 2 drivers:

Jaycar 6.5inch Carbon Fibre Cone Woofer, Catalog number CS2240:

"This quality 165mm driver has a carbon fibre woven cone, a huge power handling of 80 watts RMS and an extended smooth frequency response right up to 5,500Hz. Excellent for a quality two way system or as a mid bass woofer in a high power high quality car audio application. Driver includes Kapton voice coil and barium ferrite magnet."

- Nom impedance: 4ohms
- Power handling: 80watts rms
- Freq range: 43Hz - 5500Hz

- Sensitivity: 89dB 1watt 1metre
- Voice Coil Resistance (Re): 3.79ohm
- Resonant frequency (fs): 43Hz
- Mechanical Q factor (Qms): 9.758
- Electrical Q factor (Qes):0.471
- Total Q factor (Qts): 0.449 ]
- Equivalent Volume (Vas): 15.1Lt
- Cone Area (square meter): 0.0133
- Mounting Depth: 75mm

Vifa D26TF-05-06:

These are the Specs from Vifa, I had to email Vifa to get the datasheet:

* Fabric Diaphram
* "Butterfly" VC Assembly
* Magnetic Fluid
* Prepared for customised front
* Uncoloured sound

Nominal Impedance 6 Ohm
Nominal Power (IEC 268-5) 90W
Frequency range 3-30kHz
Sensitivity 88dB
Effective Diaphram 7.1 cm2
Voice Coil Resistance 4.6 Ohm
Operating power 6.4W
Voice coil Diameter 26mm
Voice coil Height 1.6mm
Air gap height 2mm
Free air resonance 1500 Hz
Moving Mass 0.27g
Force factor 2.3 Txm
Magnet weight 104g

Some test boxes I made to decide if I wanted to go sealed or vented:

I tuned it to 44Hz in a 21L box with a 4.5kHz crossover point at 12dB. The response plotted in WinISD:

The Crossovers:

Building and finishing the boxes:

I used Tee-nuts to fix the drivers to the front panel, they are much better than a plain screw:

This shows the port I made with acrylic tube:

The inside with stuffing and horizontal brace:

Spiked!:

Finished but no paint:

Multi-Hybrid Headphone Amp

2005-08-01T02:50:00.000-07:00

This is a Tube Mosfet hybrid amp designed by Sijosae. It uses the tube as an input and MOSFETs for the output. I used a Electro-Harmonix 6922EH tube bought from Ebay and IRF610's but you can use a number of different tubes and MOSFETs. I am very happy with the sound. Its dead quiet and is the main headphone amp I use now. You can see a schematic and measurements etc here on Sijosae's page: www.headphoneamp.co.kr/bbs/zboard.php?id=diy_sijosae&no=217

Detailed discussion here in the Headwize forums:
www.headwize.com/ubb/showpage.php?fnum=3&tid=5614
Pics....

Here is a picture of my PCB design:

I mounted a few parts on the bottom of the PCB:

The tube is lit up using a UV LED:

You cant using a switching power supply for any power supply that produces noise so I used a LM317 regulated 24V power supply using the same schematic as the TREAD but I made my own PCB to fit into this case:

Here is is next to my PIMETA, they use the same Hammond case:

LM3875 Gainclone

2005-07-01T02:52:00.000-07:00

This is the build process of my Gainclone amplifier.
What is a Gainclone?
A GainClone is a clone of a product called the "GainCard", made by 47-Labs. This "GainCard" cost USD$4000 for the Amplifier and USD$1800 for the power supply. They have VERY few components and are based on the National Semiconductor LM3875 IC. If you want to know more about the Gainclone just google as they are VERY popular and there is heaps of info on the internet about them.
I bought a kit from BrianGT, you can see it here.

The PCBs and components are very simple and quick to make, only took about 20 mins to assemble both amps and rectifier board:

Testing the assembled amps on a desk:

DC offset was about 80mV on one channel and about 40mV on the other. I used the optional Ci capacitor in the national datasheet for the IC which reduced it to between 0-4mV:
This is the capacitor I chose, its an Elna Starget (expensive):

The case was MUCH more time consuming and difficult to make though. I bought all the aluminium from a scrap metal yard including the heatsink.

I got my aluminium panels cut at a sheet metal shop as I cant make straight cuts with a hack saw:

And after powder coating the panels and dying the wood with spirit dye:

Conclusion
I haven't heard sound this good before. I had no idea a better amp would make a such a big difference to my system. I wont go on, it just sounds amazing and its so cheap to build!
Here are some very helpful and interesting links regarding Gainclones:
http://www.mhennessy.f9.co.uk/gainclone/
http://www.decdun.fsnet.co.uk/gainclone.html
http://www.vikash.info/audio/gainclone02/
http://www.diyaudio.com/forums/
And you can buy a kit like mine from here:
http://www.chipamp.com/

How to make PCBs with a laser printer

2005-06-01T02:56:00.000-07:00

First thing to do is make a schematic in EAGLE Layout Editor. Once you have done this you can then design the layout using Eagle too. Making schematic first helps because Eagle will show you clearly if you make any mistakes with the PCB.

Here is an example I will use, it is the prototype line stage used in my preamp:

Print it out onto some semi glossy photo paper. Be sure to select "solid black" in the eagle printing preferences and also un-select silk screen layers so that you only see pads and traces:

Then quickly take this paper, cut out your design and place it face down onto some blank PCB board. Then use an iron and press the design on to the board. Use A LOT of pressure and ensure all parts of the design make good contact with the blank PCB. Keep it pressed for at least 5 mins. I have found that having steam on/off makes no difference.

Then take your blank PCB with the design stuck on to it and soak it in soapy water for about 10 mins to loosen the paper:

Then CAREFULLY remove the paper. Be sure not to remove any of the black ink traces because these are what stops the acid in the etching process. I use a kind of rubbing motion to remove the paper. Keep it wet all the time so that it separates easily from the copper board.

Done and ready for etching:

Get some etchant and just follow the instructions for the dilution level. I just use a normal ceremic or glass bowel. Ghetto style:

Add boiling water from a jug:

And slowly swirl or rock the bowl around. The liquid must be always moving over the copper board to get it to etch properly.

Done!

Now give it a rinse in cold water and scrub the black ink off the PCB with a scratchy dish cleaner thing or steel wool.
Voila, you have a custom designed PCB:

Then all you need to do is drill out the holes and start mounting the components:

It will take maybe a couple of tries to get the hang of it but once you've got it you can produce perfect PCBs every time.
Tips for making PCBs:
www.expresspcb.com/ExpressPCBHtm/Tips.htm
Another guide like mine:
http://homepage.tinet.ie/~ei9gq/pcb.html
Some points to note:

You can't do it with Inkjet printers, the ink doesn't stick, don't bother trying.
OKI LED and regular laser printers work well.
I have only used matt photo paper and have always had perfect results, although glossy would probably work too. Otherwise you will have to experiment.
Once the design is stuck to the PCB put a blank piece of paper over the design and then rub and twist the iron while pressing the print on the PCB. This gives better coverage and makes sure it sticks on all parts of the PCB.

Here is some pics of another design I made (the soft-power circuit in my preamp):

LM3886 Gainclones

2005-04-01T01:53:00.000-08:00

This is my second Gainclone amp, after my first. This one is also based on a brianGT kit but it uses the LM3886 IC which is more suited to 4 ohm loads, unlike the LM3875. It will output 70 Watts into 4 ohms. This version also uses a very high capacitance PSU and a snubber.
You can see great discussion about the snubber idea here.
Schematic of the snubber:

I decided to make this one Dual mono:

I made it entirely out of recycled metal bought from scrap metal yards, you can see some of the metal has deep scratches still. I don't mind, I think it looks kind of cool:

Here you can buy the same kit:
www.chipamp.com/lm3886.shtml

LM4780 BPA-200 Amplifier

2005-03-01T01:53:00.000-08:00

This is design is very similar to the National Semiconductor BPA-200 (Bridge/Parallel Amplifier) which uses 4x LM3886 per channel and an input buffer:

My one will use 2x BrianGT LM4780 kits and a balanced line driver to bridge them, like this:

The total effect is (2x LM3886's paralleled) x2 Bridged and should give approx 225 watts into 8 ohm and 335 watts into 4 ohm when used with a sufficient power supply.

The line driver part is based on the DRV134 and a PCB from digi01, you can see the details here.

In the end mine was dual mono in a recycled amp case using 2x 300VA (18VACx2) toroidal transformers, 54,400uF of capacitance per channel with a snubber. The DRV134s add 6dB of gain so I set the LM4780s to about 16x gain to compensate (they are set to 33 normally). I get no DC offset, no hum, no crackle or thump on turn on or turn off, silent as.

Pics.....

Stereo subwoofers

2004-11-01T01:52:00.000-08:00

These are my stereo subs. They use 2x 10" drivers from Jaycar.

10inch Re/Sponse Titanium Subwoofer

Catalog number CS2274

"You often see car sound enthusiasts using banks of 10" subs in their speaker designs. These drivers, with their Titanium finish polypropylene cones, look great installed as single drivers or in banks to produce thumping bass! They also feature black Titanium frames as well as heavy duty gold plated binding posts to accept heavy speaker cable. Great performance that's difficult to beat for the price, this driver clearly represents excellent value for money"

SPECIFICATIONS
Nominal impedance:4 Ohms
Power Handling: 125 Watts rms
Sensitivity - SPL - 1Watt @ 1 metre: 88.17
Voice Coil Resistance (Re): 3.5 Ohms
Resonant Frequency (Fs): 33.298Hz
Mechanical Q Factor (Qms): 8.193
Electrical Q Factor (Qes): 0.534
Total Q Factor (Qts): 0.502
Equivalent Volume (Vas): 29.55 Litres
Cone Area: 0.0346 Square metres.
Voice Coil Diameter: 50mm
Xmax: 9mm
Speaker dimensions: A 255, B 245, C 233

I am putting them into 55 litre ported enclosers tuned to approx 25 Hz which should give an F3 (-3db) of approx 27 Hz (loooww!).

Finished (ish):

My Audio System

2004-09-01T02:55:00.000-07:00

My audio system consists of:
DIY Preamp
DIY Dual mono Gainclones
Sony ST-JX661 Tuner
Technics SL-1600MKII Turntable:

Pioneer DV-383 DivX DVD Player
Arcam Alpha One CD Player with opamps replaced with OPA2132:

The Arcam Alpha One uses a Burr Brown PCM1710 DAC:

Chu Moy headphone Amp

2004-07-01T02:49:00.000-07:00

This is a Chu Moy headphone amp I made. I got the PCB design from Guzzler on Head-Fi and the original design is very popular and is here.
I changed Guzzler's layout so I could use a Taiwan Alpha pot on the PCB. These pots a high quality (some say better than ALPS blue velvet) and are available from Jaycar.
I also made a DIY Brown dog adapter so I could use two OPA627 single opamps in a dual opamp socket.
I was try it out on paper to make sure all the parts fit:

PCBs after etching:

PIMETA Headphone Amp

2003-11-01T01:48:00.000-08:00

There is not much to say about this as I didn't design it. Go here for schematics, design notes, suggested parts etc:

www.tangentsoft.net/audio/pimeta/

The BUF634U buffers are surface mount so are on the bottom of the PCB. I used a Hammond case and Solarfilm from a model shop to cover it which gives it the shinny look.
Sourcing the parts cheaply was tricky though, I used these suppliers: Tangent, RS Componets, DSE, Jaycar and South Island Components. Total cost was about $225 NZD.

Pics....

DIY Audio Rack

2003-08-01T02:56:00.000-07:00

There is not much to say.....its a pretty simple design using threaded rods, a bunch of nuts and some flat pieces of wood for shelves. It cost me about NZ$60 in parts and I'm stoked with how it turned out:

Fibre Channel at home

2002-01-01T01:58:00.000-08:00

This shows you how to get a fibre channel system running on a PC without having to pay for a large FC enclosure array. What you need:
Commen sense - some knowledge of what Fibre Channel is and what its for. Google if you dont know already.
Fibre Channel Host Bus Adapter - Obviously you need this, it connects your PC to the FC drives or any other fibre channel devices. There are alot to choose from but if you are reading this you probably want a cheap one, Like the QLA2100 or any other 1 GB/sec HBA. They either have copper interfaces or Fibre Optic. Copper is easier for this project or if you have a fibre optic one you have to use a Media Interface Adapter (MIA) to convert from optical signals to electrical (this is what I did).

I used a Emulex LP8000 with a Stratos Lightwave MDB9-8-1 MIA.

Pic of MIA:

Hard Drive - Any FC-AL drive will do. Seagate 9 gig and 18 gig ones are very common and pretty cheap. They have 40 pin SCA2 connectors on the back of them
cables - you can just use UTP CAT5 cables for soldering etc but use STP CAT5e for any length.
T-card - This is an adapter that goes from DB9 > SCA2 (a 40 pin connector on the back of FC-AL hard drives). You dont have to use DB9, you could use any connector you with eg. RJ45. You can buy these from places in the states:
http://sierra-cables.com/
http://www.sca40.com/
http://www.cs-electronics.com/
http://www.cinonic.com/
but they are expensive for what they are and were not feasable for me to buy so this guide shows you how to make them.You will need a SCA-2 connector:

Manufacturer	Part number
AMP	787317-1
Berg	71781
Methode	512-220-91-101N
Molex	717431040

This is probably the hardest part to get (well it'd be easy if you lived in the states). You can also get some of the above companies to send you free samples if you pretend you are a company looking to buy a large amount but want to try a few first.

I had to get mine off a dead Dell Power Vault 650F backplane with a blow tourch:

You will also need:
- Blank copper circuit board
- Etchant and etch resistant pen
- Other connector (you can choose, I used DB9)
You can get all of these from an electronics store (eg. DSE)
DB9 Pinout:

Pin	Signal	Direction
1	TX+	Output
2	VCC	VCC
3	Fault -	Input
4	Key	NC
5	RX+	Input
6	TX-	Output
7	ODIS +	Output
8	GND	GND
9	RX-	Input

How to make the T card
Use a laser printer to print this document (as A4 paper size) on the glossy side of the photo quality paper.
I found that its easier to remove the soldering points and corresponding pins from the FC-SCA connector that aren't used. This is a diagram of my design, the black dots are the points where the FC-SCA connector goes:

Click here if you want the Raw BMP file (you will need to resize it before printing) Clean the copper clad board with steel wool or very fine wet sandpaper. Dry the board thoroughly. Make sure that the board is clean and free from fingerprints. Place the photo paper face down on the copper clad board. Use masking tape to hold the paper in position. Don't use vinyl tape. Place the board on a flat surface. You will be using a very hot iron, so don't use the dining room table. I use the back of an old telephone directory. Use a hot clothes iron to transfer the track pattern from the paper to the copper board. Don't be afraid to use lots of heat and pressure.

Allow the board to cool. Don't be tempted to lift the paper. Put the board in a container full of warm soapy water.

After about twenty minutes the paper will begin to dissolve and disintegrate. VERY Carefully remove the paper from the copper board. Rinse under a cold tap to remove paper residue.

You may need to touch up any broken tracks with an etch resist pen. I use a fine Staedtler laundry marker. PRACTICE MAKES PERFECT for this whole process. Heres an example of one that didn't work:

Etch the board in a Ferric Chloride etching solution. You can buy the etchant in liquid form or as anhydrous Ferric Chloride powder. Follow the instructions. NEVER add water to dry Ferric Chloride. Don't get any on your clothes. The guy at Jaycar told me that I needed an air pump, a etching tank and some other stuff but I just bought the etching powder, put some in a jar, added boiling water, rocked that jar back and forth for 5-10 mins until all the copper was gone and it worked a treat:

(Mine looks slightly different because I didn't have the right power plug and had to reverse the orientation of the 5V and 12V points) I think maybe you need the tank and everything if you are making big PCBs but Ghetto Style is always better IMHO! It may need a couple of cuts with a craft knife afterwards just to separate a couple of the close tracks but BE CAREFULL when doing this:

After etching, rinse the board under a cold tap. Remove the etch resist (black ink from printer) with some steel wool. Dry the board. Use a 0.8 or 1.0 mm drill to make the holes for component leads. I found 0.8 mm was alot better than 1.0 mm.

Now obviously you need some soldering skillz and its pretty obvious where to put the connectors etc. Heres how my final attempt turned out:

You can use the supplied circuit diagram or you can easily design and make to suite your needs from the following pin outs.

SCA2 connector:
Pin	Description	Connected To	Notes
1	-EN bypass port A	Ground
2	+12 Volts	+12 Volts	Drive Power
3	+12 Volts	+12 Volts	Drive Power
4	+12 Volts	+12 Volts	Drive Power
5*	-Parallel ESI	NC	Optional ESI, not implemented
6	-Drive Present (GND)	Ground	Ground drive detect
7	Active LED out	NC	HDD activity (flashes). Connect a 2.2k resister in series with an LED to +5v
8	Power Control	NC	Optional, not implemented
9	Start_1	+5 Volts
10	Start_2	Ground
11	-EN bypass port B	NC
12*	SEL_6/-P_ESI_6	Ground	Used to set ID, see notes below about IDs.
13*	SEL_5/-P_ESI_5	Ground	Used to set ID, see notes below about IDs.
14*	SEL_4/-P_ESI_4	Ground	Used to set ID, see notes below about IDs.
15*	SEL_3/-P_ESI_5	Ground	Used to set ID, see notes below about IDs.
16	Fault LED out	NC	Fault LED. Connect a 2.2k resister in series with an LED to +5v
17	DEV_CTRL_CODE_2	NC
18	DEV_CTRL_CODE_1	NC
19	+5 Volts	+5 Volts	Drive Power
20	+5 Volts	+5 Volts	Drive Power
21	+12 Volts Charge	+12 Volts	Drive Power
22	12V Ground	Ground
23	12V Ground	Ground
24	+Port A_in	DB9 pin 5	Drive 1, Loop 1, positive input
25	-Port A_in	DB9 Pin 9	Drive 1, Loop 1, negative input
26	12V Ground	Ground
27	+Port B_in	NC	Drive 1, Loop 2, positive input
28	-Port B_in	NC	Drive 1, Loop 2, negative input
29	12V Ground	Ground
30	+Port A_out	DB9 pin 1	Drive 1, Loop 1, positive output
31	-Port A_out	DB9 pin 6	Drive 1, Loop 1, negative output
32	5V/3.3V Ground	Ground
33	+Port B_out	NC	Drive 1, Loop 2, positive output
34	-Port B_out	NC	Drive 1, Loop 2, negative output
35	5V/3.3V Ground	Ground
36*	SEL_2/-P_ESI_2	+5v	Used to set ID, see notes below about IDs.
37*	SEL_1/-P_ESI_1	Ground	Used to set ID, see notes below about IDs.
38*	SEL_0/-P_ESI_0	+5v	Used to set ID, see notes below about IDs.
39	DEV_CTRL_CODE_0	NC
40	+5 Volts Charge	+5 Volts	Drive Power

You can loop as many as 126 FC devices together really easily. Basically each device has 2 loops each having an IN and an OUT port made of two connectors, a positive and a negative:

Each device also has to have an unique ID (just like with SCSI) which is set by having jumper like this:

Pin	12	13	14	15	36	37	38
ID	64	32	16	8	4	2	1

To select an ID connect the pin or pins to 5V and the rest to ground. eg. say you want an ID of 42, you would connect pins 13, 15 and 37 to 5V and all the others to ground (32+8+2=42).
So you could make a T-card with any number of connector on it all in a loop and this would work. But because its a loop if one device fails or if one slot is not filled then the loop is broken and nothing will go.....I went from knowing nothing about FC to knowing everything you see here (and more) purely from searching the internet. You can just follow my instructions and make a T card or if you want to know more here are the resources I used: A crude T card design: www.overclockers.com/tips681/

A project like mine: www.phfactor.net/fc/intro.html
Another project like mine but with more usefull information and a pcb design too: http://techmav.ronin-research.org/project1/
Another project like this: www.rockhounding.net/projects/fc-al/

Different Parts: www.ebay.com and www.trademe.co.nz
How to make PCBs: http://homepage.tinet.ie/~ei9gq/pcb.html
Tips for making PCBs: www.expresspcb.com/ExpressPCBHtm/Tips.htm
And most of all a big thanks to Kelly Park who gave me the original diagram for the T-card. I was about to give up when he helped me out :)

DIY Fanless ATX power supply

2001-07-01T02:57:00.000-07:00

There are a few different ways of doing this but basically you have to make the heatsinks NOT live and you have to provide better cooling for the transistors.

The heatsinks used in PSUs are generally soldered to the PCB via little metal tabs and current usually flows through them, ie they are live with voltage and act as part of the circuit. What you have to do is cut the tabs which takes the heatsinks out of the circuit and them use wire to replace them in part of the circuit.

It might look like this when done (the wires are where the metal tabs were):

The second part can be done various ways eg. make heatsinks that replace the existing ones and attach them onto a bigger heatsink (that's what I did below) or extend the transistors with wires and attach them directly to the big heatsink. You can extend them like this:

Here are 3 other articles that have used the extended the transistors method:
www.silentpcreview.com
www.zerofanzone.co.uk
www.digital-explosion.co.uk
I tried this method on three PSUs all with no luck. The PSU would whine or hiss after adding the wires. I tried making the wires really short and I checked all my connections repeatedly. I asked a PSU repair guy why and he said that it was the extra resistance the wires made in the circuit that caused the noise. He also said that it may seem like a really tiny difference but it was because the operate at a very high frequency. I wanted a REALLY silent PSU so I was particularly fussy about the noise it produced so if you did the same you might not notice or care. Maybe I just had bad luck and picked the wrong PSUs, I don't know but it worked like this anyway...

I was lucky and got a 150 watt Micro ATX PSU in which the heatsinks were grounded:

I had to remove a capacitor to unscrew the transistors from one the heatsinks:

The other I had to completely remove transistors and all because there was no way to get a screw driver in there to remove the screws on the transistors:

I then made similar heatsinks out of plain 5mm aluminium. I did this because the original ones were going to be very difficult to bolt to a big heatsink.

And then I put the PSU all back together:

Then I cut the space where the CDROM would have gone in my micro ATX case to make room for the big heatsink:

Then I bolted it onto the chassis:

Then I made 90° aluminium parts that would bolt onto the big heatsink and the little ones on the PSU:

And attached them to the PSU like this:

And I drilled some holes for passive ventilation:

And I made a plate for the back to hold the IEC power socket and switch:

I use a socket 7 Pentium 266 (Tullimook core) at 100x3 Mhz at 1.9V, this CPU was from laptop and runs very cool. It was for a Linux firewall.

Making a PC Silent

2001-04-01T03:43:00.000-07:00

I have a water-cooled PC because its quiet (I hate computer noise) and it keeps the CPU cool so I can overclock it more. I made these blocks with a drill press and hand tools - no mill.

"I water cool my computer by placing the mother board and components in my bathtub. I fill the bath tub with ice, salt, and water. My hard drive is currently in my toilet. I suggest rinsing the processor off, and then drying it in the microwave, before you put it on the mother board. For added effect, I turn my hair dryer on, and drop it in the tub full of water and computer. The extra voltage helps speed the processor up. The keyboard is grouted to the shower head, and the monitor hangs off the towel rack. Currently I have a 5mhz 8086 clocked to 200 Terahertz. Asside from the occasional underwater fire, it doesn't seem to mind. I do recommend that you ground yourself to the hot end of a High Voltage power line, before you get in the tub. Also note that welding goggels may be necessary, as the processor is near the Nuclear Fusion temprature, and may emit up to 10megawatts of pure UltraViolet radiation. You may experience the occasional rip in the time space near the processor, if this keeps occuring, lower the clock speed by 50 Terahertz. On rare occasions, black holes may develope near the power supply. UNDER NO CIRCUMSTANCE should you pass the event horrizon of the black hole. Do not turn the power off , as this may collapse the universe into your bathroom. There are no known workarounds for the Black Hole problem." - Unknown quote

My current system:
Abit NF7-S rev 2.0 motherboard
CPU XP1700 (1460 Mhz), T-Bred core, DLT3C JIUHB Core @ 2300 Mhz
Radeon 9600 Pro
Cooling: Homemade CPU and North Bridge block, SpringFlex Tubing, modified Eheim 1048 and a heatercore from a Toyota Camry ('94 I think)

AMD Athlon XP1700 @ 2300Mhz:

I soldered new barbs on to the heater core:

And put new barbs on to the pump:

An old waterblock I made once:

Update

After years if watercooling I have decided to move to something more practical. This wasn't really possible before but in the last couple of years cooling companies have realised there is a big market for silent solutions so there are quite a few good products for people concerned about noise now. The product I chose was the Thermaltake Sonic Tower. This thing really is a big-ass mutha-fucka. It weighs close to 700g and is about 150mm tall. It uses the AMD mounting holes which fortunately I have on my old motherboard but I would be hesitant if using clipped to the socket. It had mixed reviews but I was pretty sure I could make it work and its so cheap too.

Anyway, pics:

The power supply is an Enermax 430W one that I modified with a 120mm fan that is near silent. I cant hear it when its under my desk at all. I have an 80mm fan at the rear, an Vantec stealth fan running at 7V, which is VERY quit too. So how well does the Sonic Tower cool my CPU?

I am using the same CPU as above running at 2000Mhz @ 1.45V. Average idle temp is 40C. When I'm using my PC for email and browsing it sits on about 44 and when playing a game etc it gets to about 48. I am very happy with these temps considering it has no cooling actually on the CPU heatsink and the two fans I have are VERY quiet. This is much quieter than the watercooling setup I had.

I also suspended my hard drive in rubber bands which cuts seeking noise down massively:

So now I am very happy with the noise (or lack of) that my PC produces and I am moving on with other projects!