Handling a complex load by power amplifiers

[Sokel]

I think we have to draw a Price vs Tolerance chart (PriTo™),it will be surely a steep slope one but hopefully will eliminate most of the debate.

Or stop having double standards regardless the price as any technical nerd would do.

Either would work.

 

[NTK]

My reply was harmonised it seems. Aparently posts referencing the TI EVM design are not welcome.

Put simply, the TI EVM does not show this behaviour. What is more, the EVM contains additional non-populated positions for parts that can be added, if changing (for example) the filter inductors lead to problems. Their purpose is explained in the EVM Documentation.

Bottom line, copies of the reference designing do not have peaking filters. So don't blame TI. Blame the designer.

Thor

The behavior of the Aiyima A07 as shown in posts #25, #30 and #35 can be easily explained by the response of the output LC low pass filter, which is basic stuff.

Here is the example given in TI's application guide. In the TI example, L = 10 μH and C = 2 × 0.68 μF. In the app guide, the LC filter is designed for a nominal load of 4 Ω. Replacing the load with the circuit as shown in post #23, you get a response that matched what where shown in the posts mentioned above. The is no mystery. Performance is just as predicted using TI's design guidelines.

Impedance plot of the simulated load that matched what were shown in post #23.

Calculated response using the simple LC low pass circuit as shown in TI's app guide (Fig 10).

 

[retroflex]

If it afflicts >90% of all amplifiers/speaker combos, but seemingly nobody notices, is that the dictionary definition of a non-problem?

I think this is a very relevant discussion on a forum where people will (implicitly or explicitly) recommend a -120 dB THD+N @ 1KHz FS DAC over a -110 dB one, and will laugh a -80 dB DAC out of the room. We are not just concerned with audible perfection, we are also concerned with measurable perfection. And TBH, I think it would be good for a lot of people here, myself very much included, to gain a deeper understanding of how the numbers on the AP dashboard actually relate to the sound waves that hit our ears. It seems to me that the investigations in this thread are a relevant part of that puzzle.

We can also develop better tests to detects ghosts and spirits!

When audiophiles claim to hear unmeasurable problems, we disregard it, because our measuring equipment has proven to be way more sensitive than human hearing. When PMA finds measurable problems when an amplifier drives a capacitive load, we disregard it, because that's just not how amplifiers are used in practice. When PMA finds measurable problems when an amplifier drives a realistic speaker load, we disregard it, because.......ghosts?........

 

[IPunchCholla]

Recently a Buckeye amp using Purfi modules had a slight increase in distortion as it approached peak power and was not recommended despite SINAD over 100 because it did not match the reference design. It was then "fixed" by replacing the binding posts of all things and SINAD increased by 3 points and it was recommended and applauded which was well deserved. In this case an amp potentially becomes unstable because it did not properly copy the reference design and it is supposed to get a pass. It does not require heroic or expensive engineering to copy the reference design.

I think people here are thrown off because as bad as this oscillation looks it may not be all that audible in many cases. I built a DIY amp and used it for a couple of years and was very happy with it. I then got an Oscilloscope and noticed over a certain output it would being to oscillate (at HF I could not hear) I then fixed it (it was easy, changing a cap value in the feedback loop) and I can't say it was a big improvement for my use case. None the less not fixing it and keeping a defective amp the way it was because it was not noticeable to me would not make sense and I am surprised how many people seem to think potential instability is OK when it could be easily fixed.

Personally, I thought that rating was odd, given the performance. And I really admire how Buckeye handled the issue. But I would also like to note that I could buy 12 of the AO7 for the cost of the Buckeye. I am sure it is an easy fix. But I would also be surprised if there weren’t reasons at the boundary of economics and engineering that led to the decisions to do them the way they did.

I personally don’t think any of this is black and white, nor do I see black and white statements from very many people on this site. There are a few on various sides of the question, but most of the discussion is in the gray area. I really appreciate PMAs doing these tests. I learn a lot every time. Both from the test and from the discussions.

I’m in the very beginning of what will be a very long process of casing an amp. Not even building one just building a control board and hooking up a PSU and pre-made boards. These discussions help enormously as I climb the learning curve. But, one thing that doesn’t help is people doing ad hominem attacks, insulting others for not accepting their viewpoint 100%, for not being willing accept the possibility of maybe being a bit wrong or hypocritical. We are all wrong and hypocritical. Just like we are all biased during sighted listening.

 

[amirm]

Bruno raises an interesting point - quite a lot of HiRez Flacs are actually just DSD ports which included the ultrasonic noise. They shouldn't be like this, but they are. I've got about three of them. These sorts of files played into difficult speaker loads could viably cause some of the instability discussed here.

Vast majority of high-res files are PCM encoded, not DSD. DSD is only popular from small boutique labels. That aside, the amplitude of noise you are showing in your graph is hugely lower than bass. Something like -50 dB so it is not remotely going to trigger anything here. But even if it did, nothing bad will happen. There will be a tiny ultrasonic resonance that will stick out already present DSD noise.

 

[amirm]

As I said above, an Amplifier with such measured behaviour is in my view broken by design.

Well, your view is wrong as evidenced by my testing and experience of literally thousands of users. Inventing a test to find a problem people are not having makes zero sense. The reason people are not having is because the test is not simulating any kind of real situation.

Keep in mind that this amplifier beats over half of all amplifiers I have ever tested regardless of price! The losing amps have more noise and distortion with any load, not just a contrived one.

The amplifier got my standard recommendation:

It was not put on a pedestal as state of the art of anything like that. It was a great find and continues to be so today.

 

[amirm]

Personally, I thought that rating was odd, given the performance.

??? Here is the comparison of first vs second sample:

You would have wished to leave this much performance left on the table and the cause not found??? The second measurement shows an amplifier that basically has no distortion and is completely noise limited. Whereas the first shows large amount of distortion taking over noise at just 10 watts.

 

[amirm]

Fun part, as a variability in sound quality. CD recordings may sound great, but only with DAC's using brickwall filters. Playing MQA for example may sound aggressive, cymbals may sound harsh.

You don't need brickwall filters. Even crappy DACs have 60 or more dBs worth of attenuation above audible band. Quick random example:

That is nearly 80 dB reduction of the spectrum that is already there. Of course anyone who is purchasing products based on my DAC recommendations would get one that is superbly better than that, leaving no spectrum to trigger the resonance. As for those with lazy filters, their response in audible band would be compromised anyway so this is least of your problems.

 

[IPunchCholla]

You would have wished to leave this much performance left on the table and the cause not found???

I certainly did not mean to imply that. But again, it doesn’t have to be black and white. There was certainly performance left on the table as the first review showed. But even with that performance left off it is a well performing amp. It’s not my call at all and I am not trying to criticize. I just thought it was interesting call that I might have done differently. I might have recommended with the caveat that there was clearly an issue, though likely non-audible, but pointing out that issue as strongly as you did.

 

[Thorsten Loesch]

The behavior of the Aiyima A07 as shown in posts #25, #30 and #35 can be easily explained by the response of the output LC low pass filter, which is basic stuff.

Here is the example given in TI's application guide. In the TI example, L = 10 μH and C = 2 × 0.68 μF. In the app guide, the LC filter is designed for a nominal load of 4 Ω. Replacing the load with the circuit as shown in post #23, you get a response that matched what where shown in the posts mentioned above. The is no mystery. Performance is just as predicted using TI's design guidelines.

View attachment 280513

Impedance plot of the simulated load that matched what were shown in post #23.

View attachment 280514

Calculated response using the simple LC low pass circuit as shown in TI's app guide (Fig 10).

View attachment 280516

Note the parts in the red ellipses. Fit them and the resonance peak goes away. Specifically the 3R3 + 1uF.

Alternatively include a note that the Amplifier should not be used with speakers, unless they are purely resistive in nature in the HF fegion and the amplifier ideally should not be used with speakers but only dummy load resistors.

Thor

 

[Joe Smith]

Thanks to Amir for staying with this thread! The AO7 is a good amp for many use cases that will not cause significant problems, in my view. I fit equipment to specific use cases. Is it a be all and end all? No. Is it perfectly good for amp duty with normal speakers in medium sized rooms not requiring gobs of power? Yes.

 

[sarumbear]

Alternatively include a note that the Amplifier should not be used with speakers, unless they are purely resistive in nature in the HF fegion and the amplifier ideally should not be used with speakers but only dummy load resistors.

Amazon.com and Amazon.co.uk between them have 1200 reviews. 90% of them are 5 or 4 stars.

Do you suppose that thousand owners with issues would give rave reviews? Had they all been using dummy load resistors? How come their amplifiers not failed as per @pma’s prophecy?

Get a grip, please.

 

[restorer-john]

But you seem to argue that your parameters should be everyone’s.

I don't argue at all. I merely point out, in a nice way, that the bar for high fidelity is considerably higher that many of these cheap amplifiers are remotely capable of clearing. It's nearly a dB down at 20kHz, it shuts down when driven into 4R before clipping, its S/N at full power is poor, due to excessive residual noise and it doesn't come anywhere near its ridiculous advertised power specifications.

They are cute little application note designs in an extruded aluminium sleeve. That's the truth. There's zero 'engineering' in most of them (Topping excluded) and I'll bet they don't have an EE anywhere on staff, letalone an actual analogue designer specialising in audio amplification.

 

[tmtomh]

Yes, let's just stop making measurements of gear that show problems, when we can instead focus on only the good parts.

Let's be like the Curate and if we get a bad egg, let's focus on the parts that are excellent. Unless of course it is an expensive audiophile egg...

Thor

No, it’s the converse - let’s stop using measurements specifically designed to exploit known characteristics of a particular amp design/topology and then calling those characteristics a flaw or broken design.

“This egg tastes terrible!”

“That’s because you didn’t remove the shell. Eggs are meant to be eaten without the shell.”

“You can’t count on the shell being removed 100% of the time. An egg that tastes bad with the shell on is not a robust egg - in fact it’s a broken egg design, and that’s a fact. It’s just that some people get emotional when such facts are presented, even though this is an engineering forum and we are seeking TRUTH.”

 

[restorer-john]

Amazon.com and Amazon.co.uk between them have 1200 reviews. 90% of them are 5 or 4 stars.

Quoting Amazon review star numbers as gospel is nearly as bad as using wikipedia as a reference...

 

[dlaloum]

Uncontrolled amplifier oscillation, even if not readily apparent when listening, is always sub optimal. It happens more than one would think with any class of amplifier including tube amps. Stability should always be part of a good amp design. While these tests are approaching being edge cases they are none the less interesting and I have learned a lot from them. It seems like some people get all "tribal" on certain amp topologies or brands and don't want anything discussed that shows any "weakness" even if it is just an academic exercise showing issues that could be improved with better engineering. The SINAD chase, even though it has no audible effects past a certain point, is cheered on as show casing good engineering, so why should less than optimal stability be overlooked?

I'm a big fan of the classic Quad current dumping amps... "unconditionally stable into any load"...

 

[IPunchCholla]

I don't argue at all. I merely point out, in a nice way, that the bar for high fidelity is considerably higher that many of these cheap amplifiers are remotely capable of clearing. It's nearly a dB down at 20kHz, it shuts down when driven into 4R before clipping, its S/N at full power is poor, due to excessive residual noise and it doesn't come anywhere near its ridiculous advertised power specifications.

They are cute little application note designs in an extruded aluminium sleeve. That's the truth. There's zero 'engineering' in most of them (Topping excluded) and I'll bet they don't have an EE anywhere on staff, letalone an actual analogue designer specialising in audio amplification.

Fair enough. And if you said it in the same way you did in your reply to me, I wouldn’t have taken issue. That being said 1 dB down at 20 kHz? Inaudible to almost anyone over 18. Shutting down isn’t cool, but all you are missing is the clipping end of the power range. S/N at full power is 98 dB/16 bits? Isn’t that also inaudible. The power rating is pure fantasy, yep. But it honestly seems like a fine amp for purpose at levels of fidelity that are probably transparent to almost everyone? It could absolutely be better. But could it be better for $80?

 

[BDWoody]

My reply was harmonised it seems. Aparently posts referencing the TI EVM design are not welcome.

No, it was because of the backhanded racism.

That's what isn't welcome.

Not surprised that you tried to misrepresent it though.

 

[Deleted member 48726]

I think we have to draw a Price vs Tolerance chart (PriTo™),it will be surely a steep slope one but hopefully will eliminate most of the debate.

Or stop having double standards regardless the price as any technical nerd would do.

Either would work.

Yeah, or stop praising bad stuff because it's cheap.
Either the reviews should be 100 % objective and the price just stated at the time of review. And the reviewer should not be recommending. -How could he anyway, as he does not know either budget or usecase for the individual??

 

[312elements]

If one were correcting the in-room response with DSP would it render the “problem” solved from a listening experience standpoint?