Complex impedance load amp FR influence - WiiM Amp review by Erin

[Yorkshire Mouth]

The 1 dB change is quite audible. Specially in the mid-high area. I've fine tuned my Dirac Live curves with less than 1 dB variance over that area and by quickly changing between profiles on the miniDSP it's obvious.

So, built in PEQ is your friend. It might not make FR ruler flat, but it’ll take it way below 1 dB.

 

[dominikz]

The 1 dB change is quite audible. Specially in the mid-high area. I've fine tuned my Dirac Live curves with less than 1 dB variance over that area and by quickly changing between profiles on the miniDSP it's obvious.

It is quite clear these kinds of deviations are generally audible in A-B tests.

IMO the question is more if these are a practical concern given the typical FR variability of loudspeakers, and especially when having easy access to PEQ.

If a device satisfies all other functional requirements some load dependency is IMO not an unforgivable sin - though I'd definitely much prefer if it was load independent!
I do find it strange that so many implementations suffer from this.

 

[antcollinet]

Was pleasantly surprised today when I read the review of the WiiM Amp on Erin's website. Erin will be using two complex impedance profiles in all of his amplifier reviews in the future to measure their impact on the frequency response of the tested amplifiers.

My only criticism is that the impedance profiles, with two peaks around 70Hz and 1200Hz at 20 ohms impedance, seem a bit low:

View attachment 350340

To cover around 90% of speaker impedance, impedance profiles with peaks of around 60 ohms would actually be necessary - for example speaker with 67 Ohm peak at 62 Hz:

View attachment 350343

This would lead to significantly stronger effects on the frequency response of certain amplifiers than what is observed with only 20 ohms.

However, even with the impedance profiles used with 20 ohm peaks, the benefits of such measurements are evident. In the tested WiiM Amp. A 0.6dB SPL increase is observed in the presence range of 2-4kHz, which can result in an aggressive perceived sound:

View attachment 350341

When using speakers that have an impedance peak over 10 ohms in the frequency range of 1-4kHz (caused by the crossover), it would be advisable to perform impedance equalization (RLC circuit parallel to the speaker terminal) in this frequency range for roughly $20 when using such an amplifier to avoid this kind of FR change.

Such impedance equalization can be easily performed individually for each speaker model for which impedance measurements are available, using tools like VCAD. Which in turn leads to amplifier frequency responses as measured with the typical constant loads of 4 or 8 ohms. So the speaker can typically be used with the amplifier without significant loss of sound quality.

Have you calculated the difference in level between a 20 ohm peak and a 60 ohm peak with a typical amp output impedance? You'll find its not as significant as you think it it.

 

[ctrl]

Have you calculated the difference in level between a 20 ohm peak and a 60 ohm peak with a typical amp output impedance? You'll find its not as significant as you think it it.

I'm not sure how you intend to calculate the difference in frequency response deviation of the load-dependent amplifier with a load of, for example, 20 ohms versus 60 ohms if you don't know the load-dependent behavior of the amplifier.
Wouldn't you need to simulate the entire amplifier circuit to make a prediction about the frequency response with a specific load profile?

But I have limited knowledge of electrical circuits; if you have the answer, please share it.

 

[Sokel]

So, built in PEQ is your friend. It might not make FR ruler flat, but it’ll take it way below 1 dB.

You have to have a speaker that's EQable up there (not all are) AND the combined (specific amp and speaker) anechoic data to correct up there.
That will also defeat the purpose of the (limited) PEQ (which is mostly to correct lows) .

There are a lot of reasons that all these years a amp with dependencies is considered bad.

 

[Rick Sykora]

As more folks can afford audio analyzers, am hoping the vendors focus more on better performance. Like that @VintageFlanker used an actual speaker in his measurements of this amp. Might be hard to do at high power. For that matter, speaker impedance will vary much differently than a complex dummy load under high power. Am not convinced that simulated complex loads will be more revealing than a 2 ohm resistive load for that matter. Did Erin build a complex load that can handle more than 200 watts? If so, would like to see it.

As for the audibility of 1 dB differences, still sure it will depend on conditions. @ctrl shows that he can hear it under certain conditions and have no reason to doubt him. If you train long enough, am confident others may too. This still does not mean most people will be able to discern. Notably, in a room with stereo speakers, it is much harder to do. Even if an amplifier is perfectly flat under all conditions, even the best speakers are not even close to +/- 1 dB over any wider bandwidth.

 

[VintageFlanker]

Like that @VintageFlanker used an actual speaker in his measurements of this amp. Might be hard to do at high power.

That why I used 25-years-old Boston CR8s. Not that afraid to damage them, honestly. I could also try to get a pair of Focal Chora 806s at a broken price. SoundStage uses this for "complex" impedances, but I believe with a simulated load of it. Anyway, at least we know their impedance curve...

Please note that I only plugged the speakers for FR captures at 1 and 5 Watts max (and it was already painful standing close to these, sweeping at 5 Watts).

 

[Rick Sykora]

That why I used 25-years-old Boston CR8s. Not that afraid to damage them, honestly. I could also try to get a pair of Focal Chora 806s at a broken price. SoundStage uses this for "complex" impedances, but I believe with a simulated load of it. Anyway, at least we know their impedance curve...

Please note that I only plugged the speakers for FR captures at 1 and 5 Watts max (and it was already painful standing close to these sweeping at 5 Watts).

Lol, know what this is like in just testing speakers. The occasional level setting mistake can get very scary!

Having built several high power dummy loads, am still leery of even short burst testing. It does not take much to heat up power resistors and have tried cooling with fans, oil baths or both. As you approach clipping, many of the amps start to squeal a bit just to add to the tension. Controlling the conditions and trying to do so consistently is NOT trivial. I burned up a QA451 more than once before I built more robust loads.

My heart races whenever I test some of the higher power Hypex amps and the fire extinguisher is always close by too!

 

[ctrl]

That why I used 25-years-old Boston CR8s. Not that afraid to damage them, honestly. I could also try to get a pair of Focal Chora 806s at a broken price. SoundStage uses this for "complex" impedances, but I believe with a simulated load of it. Anyway, at least we know their impedance curve...

Can you show us an impedance measurement of your speaker load (Boston CR8)? I suspect that the speaker does not present a very challenging load.

This is evident from the comparison of your load-dependent frequency response measurements with those of Erin (same scaling of both diagram). The constant loads of 4 and 8 ohms align quite well in the range of 30-15,000 Hz. However, I'm not sure at what voltage Erin conducts the load-dependent frequency response measurements.

The bass reflex tuning of your Boston CR8 seems to correspond to the 20 ohm difference of Erin's complex dummy load, but in the frequency range above 200Hz, the impedance seems to be relatively uniform, exceeding 8 ohm?

Please note that I only plugged the speakers for FR captures at 1 and 5 Watts max (and it was already painful standing close to these sweeping at 5 Watts).

Cover speakers with thick blankets (check for changed impedance response) and/or wear ear protection (I can recommend 3M E-A-Rsoft FX)

 

[Deleted member 48726]

As more folks can afford audio analyzers, am hoping the vendors focus more on better performance. Like that @VintageFlanker used an actual speaker in his measurements of this amp. Might be hard to do at high power. For that matter, speaker impedance will vary much differently than a complex dummy load under high power. Am not convinced that simulated complex loads will be more revealing than a 2 ohm resistive load for that matter. Did Erin build a complex load that can handle more than 200 watts? If so, would like to see it.

As for the audibility of 1 dB differences, still sure it will depend on conditions. @ctrl shows that he can hear it under certain conditions and have no reason to doubt him. If you train long enough, am confident others may too. This still does not mean most people will be able to discern. Notably, in a room with stereo speakers, it is much harder to do. Even if an amplifier is perfectly flat under all conditions, even the best speakers are not even close to +/- 1 dB over any wider bandwidth.

The 1 dB variance has not so much to do with the frequency response of your speaker per se, but the alteration of that.

Even if a speakers FR varies wildly you would still hear a difference between the two files. The speaker doesn't have to be absolutely flat for one to hear the difference. That's not the point.

 

[Mnyb]

Frequency response measurement with real speakers does not have to be performed at full power to see the effect of the amps impedance, so it can be done for alla amps at the same level 1 or 5 watt ?

 

[pma]

@ctrl is right, the load impact to FR is to be simulated. Using one specific speaker is not enough.

As a reminder, A07 output impedance. Quite typical for this class of amplifiers. And they are continuing to be happily recommended in the reviews here, as the recommendations are purely subjective.

 

[grogi.giant]

However, I'm not sure at what voltage Erin conducts the load-dependent frequency response measurements.

Voltage or Power levels are mostly irrelevant for this test. The transfer function will be almost identical if you swipe at 1W, 5W or 20W. It is only close to maximum power that some differences will appear.

 

[grogi.giant]

I'm not sure how you intend to calculate the difference in frequency response deviation of the load-dependent amplifier with a load of, for example, 20 ohms versus 60 ohms if you don't know the load-dependent behavior of the amplifier.

If there is no PFFB loop, it not very complex (actually you need complex numbers though )

I don't have the exact (mathematical) tools needed, but I am pretty certain if you know the function that describes the output impedance of the amplifier and input impedance of the speaker, you can calculate a function that describes the filtering characteristics (amplitude and phase shift) of the complete system.

 

[grogi.giant]

@ctrl is right, the load impact to FR is to be simulated. Using one specific speaker is not enough.

As a reminder, A07 output impedance. Quite typical for this class of amplifiers. And they are continuing to be happily recommended in the reviews here, as the recommendations are purely subjective.

View attachment 351176

Did you measure the phase as well?

 

[dominikz]

If there is no PFFB loop, it actually not very complex (actually you need complex numbers though )

I don't have the exact (mathematical) tools needed, but I am pretty certain if you know the function that describes the output impedance of the amplifier and input impedance of the speaker, you can calculate a function that describes the filtering characteristics of the complete system.

It is indeed possible to calculate the response deviations - it boils down to a simple voltage division circuit.

In this post (link) I tried to illustrate how this can be done, and compared the calculation result with an actual measurement with the same amp/speaker combination. Unfortunately my amplifier output impedance measurement didn't contain phase information which caused small differences.

 

[antcollinet]

I'm not sure how you intend to calculate the difference in frequency response deviation of the load-dependent amplifier with a load of, for example, 20 ohms versus 60 ohms if you don't know the load-dependent behavior of the amplifier.
Wouldn't you need to simulate the entire amplifier circuit to make a prediction about the frequency response with a specific load profile?

But I have limited knowledge of electrical circuits; if you have the answer, please share it.

Good point well made - I had in mind only the the impact of output (and speaker cable) impedance, not amp load dependency due to pre filter feedback.

 

[voodooless]

I think you can assume that the pre-filter feedback would strive to keep the output impedance low over a wide bandwidth. Given that, the largest contributor is the low pass filter after it. I would guess that you can come really close by simulating this and combining it with a real speaker load to come up with a compensation PEQ.

 

[Talisman]

Just one consideration, correcting the deviation from the FR based on the impedance curve is like equalizing the anechoic response of a speaker, it's just an exercise that makes little sense once the speakers are positioned in the room.
It is much more convenient to directly and overall correct the frequency response errors that can be recorded with rew and microphone in the listening environment.
As far as we know, the response error due to the load that may occur could randomly compensate for a response in an environment complementary to that type of variation, so we would not be interested in equalizing it

 

[voodooless]

Just one consideration, correcting the deviation from the FR based on the impedance curve is like equalizing the anechoic response of a speaker, it's just an exercise that makes little sense once the speakers are positioned in the room.
It is much more convenient to directly and overall correct the frequency response errors that can be recorded with rew and microphone in the listening environment.
As far as we know, the response error due to the load that may occur could randomly compensate for a response in an environment complementary to that type of variation, so we would not be interested in equalizing it

Exactly: https://www.audiosciencereview.com/...streaming-amplifier-review.52372/post-1890619

Except that, for these measurements, the room influence is tricky and is best not factored in too much, so better to use nearfield measurements.