diyAudio

diyAudio (http://www.diyaudio.com/forums/)
-   Class D (http://www.diyaudio.com/forums/class-d/)
-   -   Rebuilt my old Tripath TA0104-based ampifier (http://www.diyaudio.com/forums/class-d/123221-rebuilt-my-old-tripath-ta0104-based-ampifier.html)

jahonen 17th May 2008 12:00 PM

Rebuilt my old Tripath TA0104-based ampifier
 
Or, actually my friend rebuilt his (I haven't got all the new parts yet).

I'd like to share my story with Tripath amplifier in a hope it helps someone else. Of course the ideas can be applied to other designs also.

Well, the problems I experienced were that there was quite high levels of radiated EMI from the signal cables (caused interference on the TV reception), and my friend noticed that conducted EMI level was high enough to cause noise on the audio also (probably due to that it drives low frequency bipolar circuits in non-linear operation, and RF rectifies there), when the amplifier was connected to the signal source. Also, there were some kind of strange instability when amplifier was driven without load. I also integrated soft-start and protection features on the same PCB, since using 0603-sized passive components gave me plenty of extra room to play with, as the previous one was done with through-hole components. Board dimensions were duplicated from the earlier design, so that replacing would be easy.

What I then did to improve the situation, I redesigned the board using 4-layer construction, using one layer as contiguous ground plane, as required by RF-design principles. Switching node is now also buried inside the PCB, against the ground plane. Top and bottom sides are also filled with ground, only auxliary 15 V regulator area is left unfilled. The board is naturally partitioned so that the power stages are on the one side of the board and low-level circuits are on the other side. Here is a picture from my board, half assembled. SMA connectors are for the audio inputs. I like them, they are mechanically very solid.

To compare the designs, the ground EMI levels were measured from both of the boards. My friend works in a RF-design house and he measured the ground noise levels from both boards using Agilent MXA spectrum analyzer. Result was that the 4-layer design with contiguous ground plane throughout the PCB produces about 5-30 dB less ground EMI than 2-layer design with separate grounds (connected through Tripath module as suggested by Tripath datasheet), as can be seen from the ground noise measurement plot. On the new board both grounds are connected directly to the common ground plane. The only exception is that the noise level around 200 MHz rised somewhat, but levels are still quite low. The 950 MHz spike is from the local cellular base station, it is not from the amplifier. Note that measurement point was where the low-level inputs connect to the board.

Listening test also verified that audio noise problem also disappeared. The amplifier is now very quiet, and no noise level degradation is observed when the signal source is connected.

Regards,
Janne

theAnonymous1 17th May 2008 04:13 PM

Very nice! What is all the other active circuitry on the board?

jahonen 17th May 2008 04:39 PM

Quote:

Originally posted by theAnonymous1
Very nice! What is all the other active circuitry on the board?
Thanks!

There is a buffer for the input signals with possibility for little more gain and/or phase inversion for bridged application.

In addition to that, a "mains voltage watchdog", which quickly shuts down the amplifier in case the mains voltage disappears, for avoiding the nasty uncontrolled output transients. The same circuitry also supervises output DC-offset (3 volts allowed before the protection shuts down the amplifier). Same logic also provides controlled time delays during startup. Controlled soft-start is necessary to avoid large uncontrolled current transients during rail capacitor charging.

I also added a 74HC4538 retriggerable monostable for lengthening the overload indicator for about 1 second, since the OVERLOADB-output is only active when the waveform actually clips. That makes it easier to observe.

Regards,
Janne

theAnonymous1 17th May 2008 04:48 PM

Quote:

Originally posted by jahonen
Thanks!

There is a buffer for the input signals with possibility for little more gain and/or phase inversion for bridged application.

In addition to that, a "mains voltage watchdog", which quickly shuts down the amplifier in case the mains voltage disappears, for avoiding the nasty uncontrolled output transients. The same circuitry also supervises output DC-offset (3 volts allowed before the protection shuts down the amplifier). Same logic also provides controlled time delays during startup. Controlled soft-start is necessary to avoid large uncontrolled current transients during rail capacitor charging.

I also added a 74HC4538 retriggerable monostable for lengthening the overload indicator for about 1 second, since the OVERLOADB-output is only active when the waveform actually clips. That makes it easier to observe.

Regards,
Janne

:)

I'm half tempted to ask you to share as I have two extra TA0104A. I have way too many Tripath amps at the moment and couldn't justify a new project though.

jahonen 17th May 2008 07:57 PM

Quote:

Originally posted by theAnonymous1


:)

I'm half tempted to ask you to share as I have two extra TA0104A. I have way too many Tripath amps at the moment and couldn't justify a new project though.

Well, I have 2 extra PCBs still around if someone is interested :) I can share the gerbers and design files (drawn with PADS), no problem. I have no commercial interest for making a PCB for now long time ago obsolete amplifier module, especially now as Tripath has gone away.

Regards,
Janne

theAnonymous1 17th May 2008 08:40 PM

Quote:

Originally posted by jahonen
Well, I have 2 extra PCBs still around if someone is interested :) I can share the gerbers and design files (drawn with PADS), no problem. I have no commercial interest for making a PCB for now long time ago obsolete amplifier module, especially now as Tripath has gone away.

Regards,
Janne

Hmmmm, very very tempting. What would you want for a PCB? Could you supply a parts list as well?

jahonen 18th May 2008 09:07 AM

Quote:

Originally posted by theAnonymous1


Hmmmm, very very tempting. What would you want for a PCB? Could you supply a parts list as well?

I'd say that 70/piece + postage would be nice.

Of course, here are the documents:
Schematics (PDF)
BOM (PDF)
Zipped gerber files
Gerber checkplots (PDF)
Top Locator/assembly (PDF)

Regards,
Janne

speaker 1st June 2008 11:42 AM

Janne:

How serious are you about building up a PCB and how many do you need to do in order for it to be feasible?

Thomas

jahonen 1st June 2008 04:26 PM

Quote:

Originally posted by speaker
Janne:

How serious are you about building up a PCB and how many do you need to do in order for it to be feasible?

Thomas

Hi Thomas,

I already have the PCBs. I originally just wanted to share the experiences of this particular design, but if anybody wants to buy the extra two PCBs, I can sell those. Or, alternatively, if someone has extra TA0104's to sell, I'm interested.

Regards,
Janne

jahonen 8th June 2008 06:40 AM

Hi everyone, small update on the rebuild project.

Yesterday I finally assembled mine. My friend assembled his on the same PCB earlier. Here is a picture from the assembled PCB, the FETs are not yet in place. They are assembled together with the heat sink into the enclosure.

I was somewhat sceptical if the Tripath module would work at all, since the first 2-side PCB-design started to oscillate and the output snubbers smoked :hot:. I hadn't used the module ever since. My friend then said that he would be interested on the PCB too and output FETs and the module driver outputs seemed (by measuring the resistance between source and gate) okay, so I decided to work on the new multilayer PCB.

I first tested the protection circuit, and it seems to react quite fast on removal of the mains AC, it takes just under 30 milliseconds to mute the amplifier. CH1 = tripath mute, CH2 = incoming AC, CH3 = timing capacitor voltage. Only trouble I had was that I accidentally had placed Y5V capacitor into place where capacitance invariance against the bias voltage is a necessity. That was pretty hard to find :) I had thought they were X7R, but I had accidentally ordered Y5V ones :mad:

Desipte of the doubts I had, when I assembled and fired up this one, everything seemed to work just fine. Switch output from the fets seemed quite clean. This is the first time I noticed that the switching times seem to be time-quantized. Previously I had only looked the switch output with low-bandwidth analog oscilloscope. This one is measured with a 300 MHz scope, using very short ground wire from the probe ground barrel.

The previous design was somewhat unstable when driven without load. This time, no unstability of any kind.

Overall, I'm personally very satisfied on the project. :cool:

Regards,
Janne


All times are GMT. The time now is 07:04 AM.


vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2014 DragonByte Technologies Ltd.
Copyright 1999-2014 diyAudio


Content Relevant URLs by vBSEO 3.3.2