TPA3255 And Postfilter Feedback (PFFB)
For some time there seems to be some hype about postfilter-feedback in this forum. When I started evaluating TPA3255 I did my own investigations on that issue ending in an approach different from TI. Some aspects not covered by TI app note SLA788 reveal additional insights and may change your point of view.
Adding PFFB to the internal prefilter feedback reduces nonlinearity of output filter and overall linearity as well. This is achieved by reducing the gain of the power amp and adding external gain for compensation. With a fixed gain of approximately 21.5dB there not very much you can reduce furthermore. According to Tis appnote SLAA788 a gain reduction of 5.5dB was achieved giving the same amount of negative feedback. To take advantage of less than 6dB lower THD figures the driving op-amp must deliver 6dB more output level – at a very low distortion. It should be mentioned that THD and noise measurements done by Tis labs were performed by feeding the EVM directly with the symmetrical output of the AP signal generator. No recommendation is given for a real-world audio-frontend which certainly would further degrade the numbers. So what can we expect? Some 4-5dB lower THD- and noise figures or less at the expense of an exceptional op-amp in the audio frontend. Speaking frankly, for me this is not worth the effort. THD and noise are fine even without PFFB. But there is more to say about PFFB.
Besides the sparse improvements of linearity the PFFB improves frequency response above the audioband by damping the output LC-filter resonant tank. The resonant peak in the area of 50kHz can be considerable specially with the unloaded output. LC output filters are calculated for critical damping with 4 or 8 ohms speaker impedance. It should be pointed out that damping of a 50kHz resonant tank requires a resistor with a real impedance of 4 or 8 ohms at 50kHz. There is very little chance to find a real-world loudspeaker with 4 or 8 Ohms real impedance upto 50kHz. Accordingly square wave response is measured in the testlabs with low inductance dummy loads yielding impressive oscilloscope plots. Certainly it is a fact that the resonant peak is considerably damped by PFFB as implemented by TI. But being a wicked guy I assume a real speaker with its self-inductance provides an open circuit at 50kHz. So testing the unloaded output filter resonance with square waves is the real way to go.
Anyway the relevance of these considerations is debatable. With band-limited signal sources like digital audio with sample rates of 44kHz you cannot excite that resonant tank – and you will never encounter any related problem at all. As there is no content above 22kHz in the signal, so why care about a peak of 50kHz?! It is really simple like this. Test you amp with the output of your soundcard with 44kHz sample rate and you will observe no strange behaviour at all. Do we have to consider frequency response of our audio equipment above the audio band? This has been debated for decades, and my opinion is firm: There is no audible relevant need for such improvements. But for me it is fun to expand the limits of amps beyond personal perception, nothing more nothing less.
A close to perfect square response always intrigued me. It gives a feeling of stability and wide bandwidth. While testing PFFB around TPA3255 it turned out that square response can be improved considerably by applying a basic PFFB circuit consisting of a simple high-pass filter. This has little effect in the audio band but dampens perfectly the LC resonant peak even without load. By fine tuning a maximal flat frequency response upto 20kHz and above can be achieved. As the gain of 21.5dB is not reduced there is no need for any op-amp extravaganca, any (non-fake) NE5532 will do the job as audio front delivering the symmetrical drive signal.
I append measured 1kHz square response with and without load (10R dummy).
Attachments
Last edited:
Yo Bro ), this is why I was perplex.
I also wonder why 3E Audio does not enable PFFB on their new line....
I think it is again a question of price ...
Many thanks for this clear and great explanation.
I will let some expert to answer with relevance to this detailed analysis.
I know that DrMordor advocates the use of PFFB and I imagine that if he speaks with devotion, he should be able to respond with an expertise worthy of the name)
Thanks again )
I am sorry to say but hearing experience varies from person to person and does not deliver revisable data. So I do keep away from any discussions based on non measured data.
You may call me an objectivist hardliner
Just my 2 c
Last edited:
Also why do they use Bi-Polar Caps in the signal path ?
Why the choice of Op-Amps when newer versions such as OPA1656 would appear to be preferable.
It appears a little disappointing from 3e-Audio - I expected something better.
I am sorry to say but hearing experience varies from person to person and does not deliver revisable data. So I do keep away from any discussions based on non measured data.
You may call me an objectivist hardliner
Just my 2 c
I also evaluated PFFB and AP and listening tests agree. PFFB makes both better.
You say that loss of gain in the amp has to be compensated in opamp stage , but as you know we have a maximum input voltage on the analog inputs...there simply is no way we can increase by 5db a 4Vrms signal without going over max input voltage
The increased drive voltage is not directly applied to the input terminal, but to the PFFB-network. Theoretically, you could increase drive level without limits - if your AFE delivers
I am sorry to say but hearing experience varies from person to person and does not deliver revisable data. So I do keep away from any discussions based on non measured data.
You may call me an objectivist hardliner
Just my 2 c
from a scientific point of view you are right... no complaints)
But sometimes our bodies, our ears; feel things that go beyond purely rational and metric things.
Do we provoke our buying pulse on a crush or on readings of measurements concerning our amplifiers which are so dear to us ? ...
Rather I think we are looking for the best product with a nice price performance ratio) This is what we see with the affordables TPA32XX.
By the way, this is why I asked Audio Science Review to measure the performance of the AIYIMA TPA3251. It's always interesting to know what it is part of subjectivity in the appreciation of a product that seems to please everyone, that said : I wish Rhing member could give an update about this ?
Last edited:
I agree about your 3E Audio feedback.
Many of us appreciate the proven 3E Audio boards but I wonder however; why in their strategy, they opted for a TPA3250 chip instead of the TPA3251 for example ... by repeating myself, we would have appreciated an evolution with the implementation of PFFB also and as you said the best affordable OPA amps.
I imagine the guys from 3E audio will read us )
Packaging; the 3250 does not need a 'heatsink' in that it is a pad down chip, using the PCB to dissipate heat.
I put a Khadas Tone Board DAC inside my amp. It has lots of favourable reviews and Audio Science Review rates it highly. I used a usb c panel mount extension and the rca plugs with screw terminals to connect it. It was about £100. It also comes with 2mm standoffs. It also has a coax input which I have not used but plan to in the future. I will make the connection for the coax with a female to female rca socket and a 20cm rca lead.
Compared to the DAC in my DX160 it has a more analog sound without being less precise. I’m sure there are better DACs out there but for the money nothing comes close to the KTB.
I am quite aware that we all are prone to subjective influenced perceptions when judging things and choosing stuff to buy. Nothing wrong with that, we are only humans. I just do not take such experience for proven evidence. In my younger years I did some aural examinations on my own focusing on well perceivable noise and THD. But as my actual hearing capabilities are in the ballpark of 8bit resolution these things get out of focus...
Last edited:
The increased drive voltage is not directly applied to the input terminal, but to the PFFB-network. Theoretically, you could increase drive level without limits - if your AFE delivers
I am unable to see it. PFFB is connected to input of TPA325X (and its limitations)
My main source has been a Mac Mini and the DAC used in this model is not bad and I have been happy to date. Apple seem to do a good job with their DAC implementations.
I recently started to use an Apple Airport Express (latest model) to feed directly into the amplifier (RCA Inputs). The DAC in this model also has quite a good reputation and it makes it really easy to stream music from my various Apple Devices - including the Mac Mini to the amplifier and speakers which have now been moved to the other side of the room.
I don't see many comments on here about using Apple devices as a source.