Fosi Audio ZA3 Amplifier Review

[Bruce Morgen]

Erin's corner has a very good review of the amp. It may explain why some users of digital amps have reported experiencing different results to what you would expect after just seeing one set of measurements/review.

An interesting claim regarding op-amp "rolling" from a comment (by a former ASR member) left on Erin's U-Toob review:

"On the op-amp thing... (edit: In relation to these mini-amps) Get a copy of the NE5532 spec sheets and look at the various curves and ratings. Two things become readily apparent ... the chip has uneven gain below about 1 volt of input (which is where we use them!) and they don't even specify differential input specs because the two inputs are diode clamped together giving a maximum of 1.2 volts across the two inputs. (On the TI data sheet take a look at graphs 5, 6, and 7) Now you have your difference... a cheap as chips chip with uneven gain and very uneven input characteristics vs. any other chip that does not suffer these limitations. The real difference in op-amps with these small amps is NOT putting in an expensive chip ... it is getting rid of the stock chips. I commonly pull out all the NE5532 chips supplied and put in either RC4558 or TLO72 ... under $1.00 each in lots of 10. Yes that first step does make a noticeable difference ... beyond that, not so much."

Perhaps one of the multitudes who, unlike me, have the EE smarts to properly interpret said "spec sheets" would like to comment on that comment. Is the NE5532 actually a poor (or at least suboptimal) component choice for "chip amp" input stages?

 

[ROOSKIE]

Annoying to see this guy advocating for a high pass filter. He really came off ignorant regarding that.

1st how would doing the high pass in the digital domain work conveniently with the Fosi amp.
Many don't want to lose the volume control. Plus if using something like a miniDSP pre-Fosi to generate the high and low pass then the Fosi doesn't even need the sub out.
So if using the sub out here that means you don't have other options & need the sub out and so it ought to have a high pass. Having one to free the mains from bass duty would be far superior in almost all cases vs fretting about it not happening in the digital domain.
It is incredibly easy to do it well enough in the analog domain. Especially if just using a fixed LR high pass @80hrz and sending a full range out the sub output, then using the subwoofer to adjust the subs own low pass.

2nd what do mean by 'this guy'? You don't know who Erin is. I don't agree with all the things he says, yet I am pretty sure he is not ignorant.

@312elements "If there really is a ton of demand for a cheap analog high pass filter then they should bring a new product to market instead of crippling an otherwise decent amplifier."
The high pass could (&should since it would be so easy )be made defeatable. Just a switch to engage/disengage.

 

[312elements]

The subwoofer low pass is so poorly implemented that I have zero confidence in their ability to get the high pass correct and even if they somehow did manage to get it right it’s going to add substantial cost to an amp that’s already getting too expensive for the space in the market that it occupies.

 

[hmt]

The Wiim amp is doing it correctly.

 

[312elements]

The Wiim amp is doing it correctly.

I haven't figured out how the Wiim amp is doing it but given that it's app controlled as opposed to some kind of a dial, I assume it's somehow being done in conjunction with the dac. If someone knows for sure, I'd love to understand that better.

 

[somebodyelse]

An interesting claim regarding op-amp "rolling" from a comment (by a former ASR member) left on Erin's U-Toob review:

"On the op-amp thing... (edit: In relation to these mini-amps) Get a copy of the NE5532 spec sheets and look at the various curves and ratings. Two things become readily apparent ... the chip has uneven gain below about 1 volt of input (which is where we use them!) and they don't even specify differential input specs because the two inputs are diode clamped together giving a maximum of 1.2 volts across the two inputs. (On the TI data sheet take a look at graphs 5, 6, and 7) Now you have your difference... a cheap as chips chip with uneven gain and very uneven input characteristics vs. any other chip that does not suffer these limitations. The real difference in op-amps with these small amps is NOT putting in an expensive chip ... it is getting rid of the stock chips. I commonly pull out all the NE5532 chips supplied and put in either RC4558 or TLO72 ... under $1.00 each in lots of 10. Yes that first step does make a noticeable difference ... beyond that, not so much."

Perhaps one of the multitudes who, unlike me, have the EE smarts to properly interpret said "spec sheets" would like to comment on that comment. Is the NE5532 actually a poor (or at least suboptimal) component choice for "chip amp" input stages?

I don't think the commenter understands opamps or the datasheet. The nonlinearity below ~1v is because that is for a single supply application (0V and 15V rails) and the 5532 is not a rail-to-rail opamp (and nor is the TL072 they suggest as areplacement). Contrary to the commenter's assertion, this is NOT where they will be used. Instead it will be used around the mid point between the rails. This can be seen from the schematic of TI's evaluation module where they use the 5532 to achieve datasheet performance from the TPA3255.

 

[amix]

Are you telling me if I get a good deal on a non-functioning KBA75 I cant send it to you to get repaired?

Leave poor @Doodski alone and contact this guy, instead. ;-) (Make sure to read the long text)

 

[nvidia_7]

@amirm can you confirm that when you tested this amp, the gain knob was at max? I just ran some tests on two different ZA3's (my broken one with the broken right RCA in, and the replacement one) and with 1v in at 1Khz test tone, I get 29.32dB in stereo setting and 31.26dB in mono setting. I ran and reran the test multiple times, same results on both amps. I have the 48v power supply.

This was with a SE connection.

 

[EddNog]

Rolled! Burson Vivid V7 Pro on right channel, Sparkos SS3602 on the XLR input. Pair of these in monobloc.

-Ed

 

[solderdude]

@amirm can you confirm that when you tested this amp, the gain knob was at max? I just ran some tests on two different ZA3's (my broken one with the broken right RCA in, and the replacement one) and with 1v in at 1Khz test tone, I get 29.32dB in stereo setting and 31.26dB in mono setting. I ran and reran the test multiple times, same results on both amps. I have the 48v power supply.

This was with a SE connection.

Looks like Amir measured 29.17dB gain (0.222V in 5W/8ohm out)
Amir stated volume control at max.

The power supply voltage does not determine the gain, just the max output voltage.
Amir stated that it takes around 1.2V to reach clipping levels which is about 31V/8ohm which makes sense as 48V (when there would be a Rail-to-Rail output) would be 34V.

 

[howard416]

Anyone know when the latest batch is shipping?

Also, has anyone recreated Erin’s issue where the amp doesn’t work with 4 ohm stereo loads?

 

[regan]

Doesnt look so good. It almost seems that it was optimised to look good on the ASR measurument protocol.

Yeah, cant blame them though

 

[amix]

Yeah, cant blame them though

You should!

 

[mike70]

Anyone know when the latest batch is shipping?

Also, has anyone recreated Erin’s issue where the amp doesn’t work with 4 ohm stereo loads?

If Fosi Audio doesn't say anything about it ... silence talks. A great difference between the measurements I can see is that Erin bought the unit, Amir takes a unit sended by Fosi Audio.

 

[nvidia_7]

Looks like Amir measured 29.17dB gain (0.222V in 5W/8ohm out)
Amir stated volume control at max.

The power supply voltage does not determine the gain, just the max output voltage.
Amir stated that it takes around 1.2V to reach clipping levels which is about 31V/8ohm which makes sense as 48V (when there would be a Rail-to-Rail output) would be 34V.

I must be completely missing something with amirs testing here. When you do the math x=20*log((6.32455532)/(.222)) you get the 29dB you stated, but the visual data in the review (green bars) shows a gain of 26dB.

 

[EddNog]

If Fosi Audio doesn't say anything about it ... silence talks. A great difference between the measurements I can see is that Erin bought the unit, Amir takes a unit sended by Fosi Audio.

Are you suggesting something about Amir’s objectivity in the face of continuing to receive samples?

-Ed

 

[mike70]

Are you suggesting something about Amir’s objectivity in the face of continuing to receive samples?

-Ed

No. Some people suggest that Fosi made the amp to pass ASR tests and if the company sends the unit to you, that's more possible than a random unit from the market.

I don't say that's the true, because I cannot know it (of course) but we have 2 different objective (not subjective) "visions" or results ... maybe is only a difference in the method ... maybe is a difference in the product. I really like to know what happened. That's all.

 

[solderdude]

I must be completely missing something with amirs testing here. When you do the math x=20*log((6.32455532)/(.222)) you get the 29dB you stated, but the visual data in the review (green bars) shows a gain of 26dB.

Ah yes... it looks like Amir uses a 4ohm load (thus explaining the 3dB difference).
My calc. assumed 8ohm load and you probably also used an 8ohm load.

 

[hmt]

No. Some people suggest that Fosi made the amp to pass ASR tests and if the company sends the unit to you, that's more possible than a random unit from the market.

I don't say that's the true, because I cannot know it (of course) but we have 2 different objective (not subjective) "visions" or results ... maybe is only a difference in the method ... maybe is a difference in the product. I really like to know what happened. That's all.

I dont think amir got a "review sample". The measurements that he and erin did look similar. It is just that ernin did some additional measurements where the Fosi shows some weaknesses. And btw that goes as well for the Wiim amp. So yes, when I say I suggest they optimzied it for the ASR protocoll I mean that the whole product is designed like that, not that they shipped a golden sample. Point is that Class D amps often dont behave predictable. With a class AB amp just a 1KHz SINAD measurement is enough to determine max output. With class D amps that is the also the case with the really good ones (Hypex etc) but obviously not with some cheap chip amps.

 

[mike70]

but obviously not with some cheap chip amps

Not as obvious for me ... and surely for many others.
As consumers taking decisions based on objective data ... we need the "real photography", in other way we're acting equally as subjective consumers, guided by "warm sound, sweet highs and thunderous bass" literature