Fosi Audio V3 Amplifier Review

[somebodyelse]

There is never a reason why op-amp rolling will make any difference to the measured response of a unit.

I'd go with rarely, not never. For an example see https://www.audiosciencereview.com/...-audio-sy-dap2002-review-dsp-amplifier.20114/ - IIRC traced to some caps that hadn't been populated.

 

[Spkrdctr]

Can one really hear a noticeable difference of + or - 1db at 20khz? I'm presuming that most people spending $100 dollars on an amplifier are not hooking them up to expensive speakers with ruler flat frequency responses.

Nope. It is insignificant when playing anything besides a test tone. Not a worry for any real world use. Buy it and enjoy it!

 

[Colonel7]

That's not what happens with real speakers. This is what happens with real speakers, it will vary with load but this is what I got from an aiyima a07. I tested three a07's and got different rising responses towards the top end with each one but similar behavior of things starting to rise around 5khz (just the tweeter here and it's passive network, c note speaker).

While I'm not a big fan of the youtuber Andrew Robinson, he put out a video documenting the same problem, but he wasn't sure what the cause was, just that he was getting totally different HF responses with different amplifiers.

View attachment 293761

Same tweeter with amp that doesn't have load dependency (had woofer connected for this one, can ignore below 2k,). This seems like a good opportunity to market "non-load dependant" amplifiers from the budget amp makers. I've had people tell me it's just these speakers that an outlier, nope found plenty of commercial offerings with very similar impedence charts, like kef q150. Most people are pairing these speakers with cheap amps, I wonder how many complaints about brightness are because of this issue. I will continue to rate any amplifier with this issue as poor as I don't consider it acceptable and it is a design flaw in my eyes.

View attachment 293762

Appreciate the input but this is not an A07 and your experience with an odd phenomenon (15db rise - yikes). Like some others, I've not had any issues testing raw driver responses and 3116 chip amps with various mics and cheap class D compared to a Yamaha A/B. I wouldn't hesitate to get this Fosi or the BT20a pro

 

[Sokel]

That's not what happens with real speakers. This is what happens with real speakers, it will vary with load but this is what I got from an aiyima a07. I tested three a07's and got different rising responses towards the top end with each one but similar behavior of things starting to rise around 5khz (just the tweeter here and it's passive network, c note speaker).

While I'm not a big fan of the youtuber Andrew Robinson, he put out a video documenting the same problem, but he wasn't sure what the cause was, just that he was getting totally different HF responses with different amplifiers.

View attachment 293761

This chart is the horror.I expect anyone listening to that getting a strong headache after a couple of songs.
Yes.that's as poor as it gets.

 

[shuppatsu]

https://www.ti.com/lit/an/slaa788a/slaa788a.pdf

[TI's reference PFFB implementation] Does not solve the load-dependency issue, at least in TI's implementation. The other performance benefits are massive, esp when handling complex loads.

I don't know very much about amps. Under my probably incorrect mental model of how all this works, the biggest "problem" with this amp is the load dependency issue. Per the linked reference, the other benefits of PFFB relate to noise and THD, both of which are likely "good enough" without the PFFB?

Is the load dependency just a consequence of this TI chip, or can it be rectified with a different PFFB implementation or some other circuit?

Back to my faulty mental model. I have trouble understanding how big of a problem load dependency is. My speakers are 4ohm nominal, but is that the value I use to interpret the effect of the load dependency? I see that some (but far from all) of Amir's reviews have a frequency sweep, which usually shows much higher impedance at various frequencies. Are these the relevant values? If so, is it useful to only show 4 and 8ohm loads when speakers can have much higher impedances, including at much lower (i.e., audible to 40-year olds) frequencies?

 

[badspeakerdesigner]

Appreciate the input but this is not an A07 and your experience with an odd phenomenon (15db rise - yikes). Like some others, I've not had any issues testing raw driver responses and 3116 chip amps with various mics and cheap class D compared to a Yamaha A/B. I wouldn't hesitate to get this Fosi or the BT20a pro

I have experienced this is several other Class D amplifiers that are low cost, which includes some models from fosi and smsl. The a07 was simply the worst offender.

I don't believe raw driver response is relevant to the issue, you need a passive crossover network before the driver as well. Here is the tweeter adding components 1 by 1, you can see the point at which the load starts to present problems. Many of the graphs overlap here but if you're familiar with measuring raw drivers you should be able deduce what is what, the highest spl obviously being the raw driver, followed by a resistor, and so on (in context of cnote xover). This data was actually taken with an SMSL SA36, so it's not just an a07 problem.

Be nice if someone with better tools and measurement environment could delve further into this problem, and hopefully this is the next thing amp makers focus on to fix.

 

[sarumbear]

I'd go with rarely, not never. For an example see https://www.audiosciencereview.com/...-audio-sy-dap2002-review-dsp-amplifier.20114/ - IIRC traced to some caps that hadn't been populated.

Which means it was not caused by the opamp but by faulty build/design. Thank you for the reference though.

 

[Colonel7]

I have experienced this is several other Class D amplifiers that are low cost, which includes some models from fosi and smsl. The a07 was simply the worst offender.

I don't believe raw driver response is relevant to the issue, you need a passive crossover network before the driver as well. Here is the tweeter adding components 1 by 1, you can see the point at which the load starts to present problems. Many of the graphs overlap here but if you're familiar with measuring raw drivers you should be able deduce what is what, the highest spl obviously being the raw driver, followed by a resistor, and so on (in context of cnote xover). This data was actually taken with an SMSL SA36, so it's not just an a07 problem.

Be nice if someone with better tools and measurement environment could delve further into this problem, and hopefully this is the next thing amp makers focus on to fix.

View attachment 293822

I haven't had issues after the crossover in place. I think we'll have to agree to disagree

 

[ocinn]

Per the linked reference, the other benefits of PFFB relate to noise and THD, both of which are likely "good enough" without the PFFB?

Not really, this amp can't clear 16bit @ 5w. Def room for improvement.

Considering how cheap and easy it is, and the 5 to 9db benefit across the whole range of measurements, PFFB is a no-brainer, and I'm not sure why Fosi didn't implement it.

Not implementing the manufacturer's feedback circuit which has already been optimized is much more critique-able than not spending R&D money on a custom solution to fix load dependency (which maybe 0.01% of the people buying this amp could even dream of hearing) on a <$100 chip amp.

 

[badspeakerdesigner]

This chart is the horror.I expect anyone listening to that getting a strong headache after a couple of songs.
Yes.that's as poor as it gets.

They were absolutely horrible to listen to. Instant tear your head off brightness.

I haven't had issues after the crossover in place. I think we'll have to agree to disagree

No, you're just not understanding the problem. CERTAIN loads cause the issue, not all loads. Your xover likely simply doesn't cause the amp to misbehave. I have other speakers with much flatter loads that don't exhibit the problem.

Like I said this really needs to be tested in depth by someone like Amir or preferably Erin.

 

[shuppatsu]

Not really, this amp can't clear 16bit @ 5w. Def room for improvement.

By "good enough" I did not mean to imply that there was no room for improvement. But 91dB SNR@5W is plenty good for me and I suspect many other people.

Considering how cheap and easy it is, and the 5 to 9db benefit across the whole range of measurements, PFFB is a no-brainer, and I'm not sure why Fosi didn't implement it.

How cheap and easy is it? I know nothing of amp design and construction. From the TI reference:

The following components are required for PFFB implementation:
• R_fb – Feedback resistor
• C_fb_in – Capacitor on input side of the feedback network
• C_fb_out – Capacitor on output side of the feedback network
• R_fb_gnd – Resistor between C_fb_in and C_fb_out to GND in the feedback network
• R_in – Input summing junction resistor
• C_z – Zobel network capacitor
• R_z – Zobel network resistor
• C_op – Op-Amp feedback capacitor
• R_op-fb – Op-Amp feedback resistor

Seems like a lot, but maybe these components are very small and there are integrated solutions?

custom solution to fix load dependency (which maybe 0.01% of the people buying this amp could even dream of hearing)

They were absolutely horrible to listen to. Instant tear your head off brightness.

I'm trying to reconcile these two strains in the thread. I understand that @badspeakerdesigner doesn't have direct experience with this particular implementation, but he argues that it's a shared problem among these basic TPA3255 designs. Is it your position that @badspeakerdesigner is wrong, or that the specific crossover problem that he had puts him in the "maybe 0.01%" category?

 

[maudio]

Appreciate the input but this is not an A07 and your experience with an odd phenomenon (15db rise - yikes). Like some others, I've not had any issues testing raw driver responses and 3116 chip amps with various mics and cheap class D compared to a Yamaha A/B. I wouldn't hesitate to get this Fosi or the BT20a pro

Maybte it's badspeakerdesign ...15 dB+!? It must be badspeakerdesign!

Anyway, it wasn't Fosi V3, so..irrelevant speculations.

 

[Saponetto]

Damn... Another fine candidate for our office!
Just kidding, have to recognize the significative jump Fosi did on positioning their products in the market.
Kudos to their marketing staff.

 

[mccririck]

This chip is from 2016. I have no idea how much progress has been made after that. All these cheap amps seem to use quite "old" tech. It would be interesting to know if there is a cheap option available but not utilized because buyers don't care or some other non-technical reason.

There may be a newer, better chip which is currently much more expensive?

 

[ocinn]

I'm trying to reconcile these two strains in the thread. I understand that @badspeakerdesigner doesn't have direct experience with this particular implementation, but he argues that it's a shared problem among these basic TPA3255 designs. Is it your position that @badspeakerdesigner is wrong, or that the specific crossover problem that he had puts him in the "maybe 0.01%" category?

They are right in theory but it's a huge generalization to say its an "all amps with minor load dependency" issue.

Go on Stereophile and look at Class D amps with load dependency, and their FR charts with simulated loudspeaker loads.

Mytek HiFi Brooklyn AMP+ power amplifier Measurements

Sidebar 3: Measurements

www.stereophile.com

NAD Masters Series M22 power amplifier Measurements

Sidebar 3: Measurements I performed a full set of measurements on the Masters Series M22, using my Audio Precision SYS2722 system (see the January 2008 "As We See It"). As the M22 has a class-D output stage, for almost all measurements I used an Audio Precision AUX-0025 passive low-pass filter...

www.stereophile.com

Mark Levinson No.53 Reference monoblock power amplifier Measurements

Sidebar 3: Measurements

www.stereophile.com

none of these exhibit the "horrible" behavior that was alleged. They are comparable to the fixed resistor responses and therefore easy to eq.

 

[amirm]

What about crosstalk? I believe that because of the use of 1 TPA3255 chip, measuring the channel crosstalk is essential.

Oops. I had tested it but forgot to post. It is now in the review:

 

[CALCULATOR]

Save a couple hundred more and a Topping PA7 can be yours. But it runs super frickin hot... I love mine. I'm so impressed with it.

 

[amirm]

The hf response goes a bit awry into 8 ohm loads. Lower in frequency, is the output basically the same level?

Correct. The output impedance of the amplifier is variable with frequency. At low frequencies it is essentially zero so doesn't interact in any meaningful way with the speaker. As frequencies rise, this gradually gets higher and higher until it becomes a significant ratio of speaker's impedance. There, you get a simple "voltage divider" causing the response to change.

 

[amirm]

Something else which I don't *think* is covered in these reviews - what about max current output?

This is like asking what is city's max pressure feeding your house. It could be 10 times or 1000 times your house inlet but doesn't matter. As long as it is as much as your inlet, you are good to go. Here, I measure at 4 ohm which stresses current delivery and happens to be close to the lowest impedance of vast majority of amplifiers.

 

[amirm]

I find the power cube and power at various loads to be more important than the new efficiency plot. Impressive little amp and a great review, thanks Amir.

My pleasure. On Powercube, it doesn't like amps like this with high floating ground voltage on one channel. The cause is unknown. Because of that, and in the interest of getting some sleep , I chose to not try to measure it.