Eversolo AMP F2 Amplifier Review

[shuppatsu]

Ironically, @Davey's original ask for graphs showing the FR with reactive loads seemingly supports Amir's contention, whereas the static loads seemingly support Davey's. Sokel's post showing the Wiim Amp's FR with two simulated reactive loads shows that load dependency can show wide boosts in more sensitive frequency ranges. I'm personally unsure why that's the case, or at least it doesn't seem to track entirely with the impedance magnitude of the simulated loads. I'm going to assume that it has to do with phase angle without really understanding what's going on.

At any rate, for ignorant people like me, the static load graphs imply that this really only affects the high frequencies, say 15KHz and up. In which case it's not a big deal at all, especially for a person of my um...experience. But if I'm reading Sokel's graph correctly, it's a (slightly) bigger problem.

 

[JSmith]

Instead of all of us discussing whether the load dependency is audible... maybe Eversolo should go back to the drawing board and release an amp using PFFB that doesn't have this issue?


JSmith

 

[shuppatsu]

Instead of all of us discussing whether the load dependency is audible... maybe Eversolo should go back to the drawing board and release an amp using PFFB that doesn't have this issue?


JSmith

I like discussing whether x or y is audible. I really only have need for one DAC, one amp, one set of speakers, etc. The only reason why I’m interested in this is for its own sake.

Also, we can’t control what Eversolo or Starke does, at least not directly. Point is, this and many other Class D amps don’t have PFFB. It looks like that’s changing, but they’ll be around for a while.

Do we have any industry insiders who have a good idea how much it costs to implement PFFB, or is it all speculation? The engineering should be straightforward since TI laid it all out, right? Is PFFB simply a feature used to differentiate different price points for marketing purposes, or is one significantly more expensive to make?

 

[JSmith]

is one significantly more expensive to make

Not really, however it requires some expertise and experience... so unsure if they just cheaped out, or whether the expertise is missing.

I'd doubt they would be believers in negative feedback is, well, negative. But it's surely created some negative feedback in this thread.

People don't want to be wondering if this will be audible and will likely go for another brand that doesn't show this in testing. Those that proceed have been warned that it could become audible in some circumstances.


JSmith

 

[Gremlins]

Eversolo should stick to streamers, thanks Amir.
Keith

I disagree
They should improve their amp... That's the way to go

 

[ICIETDIYEUR]

People hi

Here is the diagram recommended by TI for the PFFB in its dedicated explanatory note ->

In total there are 8 resistors and 8 capacitors, which represents a lower cost knowing that in addition no development costs are to be expected since TI did it for us.

Even the smallest companies that offer this type of amplifier at very low prices use this principle (PFFB) since it is known to everyone and has proven itself in terms of operation.

The PFFB has been around for years and the first to use it were TOPPING with the PA5 as well as SYLPH AUDIO and 3eAudio with their PCBs offered for integration into a DIY device

So I really don't understand why EVERSOLO 'failed' at this point with this amplifier when they know how to make excellent DACs which are much more technically complex in terms of development and production...

Here is what I got myself (and I am not EVERSOLO) on a modified O-NOORUS D1 ->

 

[amirm]

I'm personally unsure why that's the case

That's because those complex loads have wider impedance ranges than my 4 to 8 ohm. They also have a lot of variation which means you can't interpolate from them to other load without doing fair bit of math. In contrast, my measurements let you compute the output impedance of the amp and with it, we can a) decide no interest in the amp due to this or b) what the impact might be on a speaker you own.

 

[Sokel]

Ironically, @Davey's original ask for graphs showing the FR with reactive loads seemingly supports Amir's contention, whereas the static loads seemingly support Davey's. Sokel's post showing the Wiim Amp's FR with two simulated reactive loads shows that load dependency can show wide boosts in more sensitive frequency ranges. I'm personally unsure why that's the case, or at least it doesn't seem to track entirely with the impedance magnitude of the simulated loads. I'm going to assume that it has to do with phase angle without really understanding what's going on.

At any rate, for ignorant people like me, the static load graphs imply that this really only affects the high frequencies, say 15KHz and up. In which case it's not a big deal at all, especially for a person of my um...experience. But if I'm reading Sokel's graph correctly, it's a (slightly) bigger problem.

The problem with load dependency is that its effect is random depending on each and one speaker.
The two simulated loads there are just that, specific to them, at any other speaker could be different, more or less severe.

Down the thread I posted about it is all the math and many more examples, some with popular speakers too.

That's why load dependency is a mess, it's random and fairly unpredictable for the average user.

Link: the thread

 

[Purité Audio]

I disagree
They should improve their amp... That's the way to go

Their latest AMP F-10 was also nothing to write home about.
Keith

 

[Davey]

Neither Eversolo or Starke Sound did the electronic design here.
These amplifier and PS modules were purchased from a supplier and incorporated into the final product we see here.
('Not unlike what many companies have done with Hypex OEM modules.)

In this case, modules that were probably quite dated and sourced for cheap.

I hope that's clear to everyone.

 

[Ropeburn]

Neither Eversolo or Starke Sound did the electronic design here.
These amplifier and PS modules were purchased from a supplier and incorporated into the final product we see here.
'Not unlike what many companies have done with Hypex OEM modules.

In this case, some modules that were probably quite dated and sourced for cheap.

I hope that's clear to everyone.

That's exactly the problem I have with it.

There's nothing wrong with this simple approach in general, it's a simple poweramp that needs to perform and be put together properly, and that's pretty much it - for an appropriate price. This is a 200-300 moneys product at best, and even there the competition is fierce. Repeating myself, for less money than the asked 750 you get performance worlds above this, and in quite nice alu cases too.

 

[shuppatsu]

That's because those complex loads have wider impedance ranges than my 4 to 8 ohm.

This explanation occurred to me but it doesn't fully explain the phenomenon, which is why I wrote, "or at least it doesn't seem to track entirely with the impedance magnitude of the simulated loads." What I meant by that is that the FR deviation did not track directly with the impedance magnitude of Erin's load.

Down the thread I posted about it is all the math and many more examples, some with popular speakers too.

I had looked at and appreciated the thread, though admittedly I did not grok the math. But I will note that @NTK's formula (post #96 of that thread) incorporates the phase angle of the speaker/load. So I still think that's the missing piece, or at least a missing piece.

They also have a lot of variation which means you can't interpolate from them to other load without doing fair bit of math.

The problem with load dependency is that its effect is random depending on each and one speaker.
The two simulated loads there are just that, specific to them, at any other speaker could be different, more or less severe.

I agree, and did not intend to imply otherwise. Nevertheless, the simulated loads show that broad swathes of sensitive frequencies can be influenced by load dependency, and this is not intuitive from the static graphs. That's the point I was trying to make, that @Davey's belief that the load dependency was a non-issue may have been influenced by the static graphs.

In contrast, my measurements let you compute the output impedance of the amp and with it, we can a) decide no interest in the amp due to this or b) what the impact might be on a speaker you own.

How would one go about computing the output impedance of the amp from these measurements? To my knowledge the calculations done in @Sokel's thread did not use these measurements. Rather, @NTK had a hunch that the Wiim Amp used TI's reference LC filter design, applied TI's formula to the simulated load's impedance measurements, and confirmed that Erin's results matched the formula. Thus, strongly implying that @NTK's hunch was correct. From there, impedance measurements and TI's formula were used to calculate the FR deviation for other speakers.

 

[ICIETDIYEUR]

At the price at which this model of amplifier is offered, they could have at least used a PCB from SYLPH AUDIO or 3eAUDIO: these models are really very good.

 

[shuppatsu]

Neither Eversolo or Starke Sound did the electronic design here.
These amplifier and PS modules were purchased from a supplier and incorporated into the final product we see here.
('Not unlike what many companies have done with Hypex OEM modules.)

In this case, modules that were probably quite dated and sourced for cheap.

I hope that's clear to everyone.

I'm not saying Starke designed the modules, because I don't know, but their switching frequency is 600KHz and a cursory look around the Internet suggests that's pretty nonstandard. It's the same AD4 module used in the Starke class D amps since at least 2000, so in that sense they are indeed dated. If I were to guess I'd say Starke had too many of these modules floating around, and rather than encheapening their brand reputation by pricing them to sell, they found a collab partner who could take them off their hands.

 

[Davey]

That's exactly the problem I have with it.

There's nothing wrong with this simple approach in general, it's a simple power amp that needs to perform and be put together properly, and that's pretty much it - for an appropriate price. This is a 200-300 moneys product at best, and even there the competition is fierce. Repeating myself, for less money than the asked 750 you get performance worlds above this, and in quite nice alu cases too.

In general, this simple approach will not cause any problems.
Below is one of Amir's favorite speakers. This speaker attached to this amplifier will not excite the "load dependency" issue.

This could be said for many conventional speakers tested here (and elsewhere.) Most dome tweeters are nominal 4 ohm loads with very little inductance and, (depending upon how they're padded in the crossover design,) will show a nominal system load near 4 ohms at 20khz and beyond.

 

[ICIETDIYEUR]

Hi.

I think it's time for 'some people' to tell the TI engineers that they wasted time designing the PFFB circuit (LOL)

 

[Davey]

Hi.

I think it's time for 'some people' to tell the TI engineers that they wasted time designing the PFFB circuit (LOL)

Not at all. It improves/flattens the response for varying reactive loads.
It not being used in this particular amplifier doesn't render the unit unusable.

There's a lot of gray area in audio interfacing where application-specific products can fit right in.
In fact, if the LC output filter were optimized for an 8 ohm load, the ascilab speaker would have an over-damped response on the top end.
Not everything is black and white.

 

[shuppatsu]

In general, this simple approach will not cause any problems.
Below is one of Amir's favorite speakers. This speaker attached to this amplifier will not excite the "load dependency" issue.

This excited my curiosity so I figured out WebPlotDigitizer and VituixCAD, not enough to be proficient, but enough to recreate and use the work that @ctrl and @NTK did re: the Wiim Amp. So I plugged in the impedance and phase measurements of the AsciLab and got a ~+1dB bump from 7-12KHz. Now of course I could have done something wrong. Also, this is using the reference TI formula for calculating the frequency response, which happened to map exactly to the Wiim Amp. This amp's load dependency is very similar but not identical to the Wiim Amp, so this is more of a demonstration that an issue exists than anything definitive regarding this amp.

EDIT: My PSB Platinum M2s, OTOH, have absolutely no problems. Glad I bought the Wiim Apm at $240 rather than the Pro at $380.

 

[Davey]

@shuppatsu Where did you get the component values for the "T"-filter?
Me thinks this is not an accurate simulation of the deviation you'd get with this amp/speaker combination.

 

[shuppatsu]

@shuppatsu Where did you get the component values for the "T"-filter?
Me thinks this is not an accurate simulation of the deviation you'd get with this amp/speaker combination.

I didn't. I used other people's work. And they didn't either. See this post for how they went about it.

Like I said, this is for Wiim Amp, not this particular amp. But from the static load measurements they behave nearly identically. They both likely follow the TI reference design, and hence the function formula provided by TI should be usable.

This:

versus this: