Attachments
Let me explain that I built PBTL modules - i.e. mono blocks with 1 differential input and parallell bridge tied outputs requiring 2 caps in total.
The modification has been verified here with TPA3118, TPA3251 and TPA3255 and should work similarly with most other chips.
The modification has been verified here with TPA3118, TPA3251 and TPA3255 and should work similarly with most other chips.
Near the op amp, otherwise it will be an antenna. The connection to the speaker has the (low) resistance of the driver and is less problematic.
I tried implementing it on my TPA3255 amp. I took two 33pF C0G caps (66pF total) rated for 2500v and paralled them at the output pins. Then used tiny wire wrap silver wire to connect to the chip resistor closest to the TPA3255 input pins. This was done for 4 of the outputs and amp is running in BTL stereo mode. Will look at Oscope trace and add another 33pF if needed.
For obvious reason I would have placed the caps at the opposite side of the wires.
Actually, I am not sure why except maybe the capacitor is working with input series resistor as a snubber and we are minimizing length of wire without a snubber. The inconvenience of not having enough room for the capacitors was one main issue for me.
Looks like some misunderstanding. The 100pF caps should be placed direktly on the sensitive input side thus decoupling the antenna of several cm wire to the output filter.
You could install the capacitor with a short end of wire (to the input) and use some shrink tube to protect it. So you dont have to place them at the bord, making installation much simpler.
I only ask my self why TI has not mentioned this way to improve the output signal.
I only ask my self why TI has not mentioned this way to improve the output signal.
Great long thread here. Thanks for sharing 
Just started my own design of a TPA3255 implementation.
Intention is to do a prototype/POC of just the amp section, before integrating a ADAU1701 to forma 2 channel plate amp. Think I can fit it all on a 10x10cm 4 layer to stay cheap at JLCPCB.
As for Post Filet FB there is the suggested design by TI here:
https://www.ti.com/lit/an/slaa788a/slaa788a.pdf?ts=1735744921637
You are probably all aware, and it is probably already referred here in the thread somewhere.
Just started my own design of a TPA3255 implementation.
Intention is to do a prototype/POC of just the amp section, before integrating a ADAU1701 to forma 2 channel plate amp. Think I can fit it all on a 10x10cm 4 layer to stay cheap at JLCPCB.
As for Post Filet FB there is the suggested design by TI here:
https://www.ti.com/lit/an/slaa788a/slaa788a.pdf?ts=1735744921637
You are probably all aware, and it is probably already referred here in the thread somewhere.
Certainly I studied these files. But a ref design is primarily intended to verify manufacturers data.
So it should be considered the STARTING point - and not the endpoint - of any development.
In those days one of the TI developer team was writing on this forum and I contacted him.
A danish guy, sadly I lost his name.
We discussed ways for post filter feedback, and although he liked my approach this was too late as the development project came to its end("hire and fire...")
In other words, there may be room for improvement
Last edited:
Hi bucks bunny
Fully agree 🙂
(the post was also mostly intended for more readers to take a look at the app note 😉 )
A single cap will "only" give further FB at higher freq.
The approach in the app note gives overall 5 dB FB.
A drawback is the lower overall gain, but that's of course just the backside of FB 🙂
So you'll have to add more gain in a pre-amp stage to make it work with normal pre in level (0.775V RMS).
An approach could be to include an opamp in the overall PFFB loop. Alternatively just a good opamp stage.
Overall I fully doubt anyone can really hear any difference with a 5 dB decrease of THD.
As you have also shown the most important improvement is the reduction of the peak in the freq response at higher freq caused by the output filter.
But if you want performance like shown by 3e Audio, you'll need the extra FB 😉
Fully agree 🙂
(the post was also mostly intended for more readers to take a look at the app note 😉 )
A single cap will "only" give further FB at higher freq.
The approach in the app note gives overall 5 dB FB.
A drawback is the lower overall gain, but that's of course just the backside of FB 🙂
So you'll have to add more gain in a pre-amp stage to make it work with normal pre in level (0.775V RMS).
An approach could be to include an opamp in the overall PFFB loop. Alternatively just a good opamp stage.
Overall I fully doubt anyone can really hear any difference with a 5 dB decrease of THD.
As you have also shown the most important improvement is the reduction of the peak in the freq response at higher freq caused by the output filter.
But if you want performance like shown by 3e Audio, you'll need the extra FB 😉
I think I can do this by series butt joints on two 220pF 0805 caps to get 110pF equivalence and it will be small enough to solder to the input side of the amp.
The reduction in noise would be very important for a nearfield application
You made a point. There is no solution that fits all demands - so this is a matter of personal priority.
The TPA3255 without PFFB and a good PSU had very low intrinsic noise. One of the quietest amps I have measured. I can press my ear to the speaker cone and cannot tell the amp is on. PFFB reduces the THD, but the background noise is already low.
This is the measured noise floor on my amp without PFFB using a MicroAudio SMPS650-SO PSU.
This is the measured noise floor on my amp without PFFB using a MicroAudio SMPS650-SO PSU.