WiiM Amp Streaming Amplifier Review

[dominikz]

Maybe Erin or Amir should overlay that load dependency plot on the same scale as a typical speaker plot, and it will be easy to see how quickly its lost in the other system variances.

I agree.

This is how the WiiM Amp frequency response into a dummy "complex load" looks in Erin's review (digitized), note the zoomed-in vertical scale:

This is how the same curve looks if I use the 50dB vertical scale normally used for loudspeaker reviews:

And this is how this curve would affect the anechoic response of a real loudspeaker (e.g. using ASR spinorama of Revel M16):

Note: in reality the response would look a bit different because Revel M16 impedance isn't identical to Erin's complex load - but hopefully this helps to very roughly illustrate the extent of such deviations.

 

[dzerig]

Nice unit. Perfect desktop one box solution.

 

[NTK]

Sadly it's not about 15Khz,it's more about the x-over area,2-4Khz where is the great differentiator between a speaker that's listenable and one that can make your ears bleed in 10 minutes or so.

EQ could be a solution if one has a speaker that's EQable (smooth on and off-axis,etc) and has the combined anechoic data of the speaker and amp so to correct properly.
Cause EQing the most sensitive area of a speaker,where our hearing can tell 0.5db deviations and separates actual nice speakers from torture devices is not trivial.At all.

View attachment 351343
(from Erin's review)

The bumps at 2.2 kHz for the "complex load" and 3.3 kHz for the "simple load" are about 0.6 dB with a Q of ~2. They are at best barely audible using "the most revealing" test signal in fast switching A/B tests.

Figure source: https://www.harman.com/documents/AudioScience_0.pdf

 

[Inertiaman]

I agree.

This is how the WiiM Amp frequency response into a dummy "complex load" looks in Erin's review (digitized), note the zoomed-in vertical scale:
View attachment 351352
This is how the same curve looks if I use the 50dB vertical scale normally used for loudspeaker reviews:
View attachment 351353

And this is how this curve would affect the anechoic response of a real loudspeaker (e.g. using ASR spinorama of Revel M16):
View attachment 351354

Note: in reality the response would look a bit different because Revel M16 impedance isn't identical to Erin's complex load - but hopefully this helps to very roughly illustrate the extent of such deviations.

Thank you, thank you, thank you for those plots.

Those dotted lines sure won't make my ears bleed. But give it 30 minutes and someone will claim that 50ohm portions of some speaker's impedance curve will shoot those dotted lines WAY off the solid lines.

If the load dependency dogma crowd really want to convince us, pick a speaker (or 2 or 3) and perform a microphone-measured speaker FR with a load-independent reference amp, and again with a Wiim amp. Then bring the plot with both lines so everyone can see a quantified representation of this behavior. I would expect negligible differences, but I'm more than happy to change that attitude if shown significant FR changes.

 

[Sokel]

So you're going to argue the semantics of what constitutes a wiggle?
Please help us all out here . . . what range of Q is a wiggle, and what shall the informal terms be for a distorted waveform of greater or lesser Q? A waggle? A woggle?

So "wiggle" is a scientific term?
I have admitted lots of times,English is not my language,but it doesn't sound like a scientific one to me.
Put simply,a wiggle to me is a high-ish Q deviation,a low Q ones will cover a broader area and to me is not a "wiggle" is more like a narrow tone control effect would have,or a "presence" slider in a DAW.

 

[Sokel]

The bumps at 2.2 kHz for the "complex load" and 3.3 kHz for the "simple load" are about 0.6 dB with a Q of ~2. They are at best barely audible using "the most revealing" test signal in fast switching A/B tests.

View attachment 351356
Figure source: https://www.harman.com/documents/AudioScience_0.pdf

A Q of 2 is not barely but clearly audible if I read that correctly.
Specially at the area where we are the most sensitive..

 

[dougq]

I bought this, and a pair of Revel M16 (used for 200 each) for near field desktop 2.1 setup (with an older SVS sealed sub) and it is sublime. Connected to my LG oled as a monitor it seamlessly switches between audio sources (Console, PC, Macbook). Nicely executed app with good configuration options. I'm really happy with this setup.

 

[Sokel]

Thank you, thank you, thank you for those plots.

Those dotted lines sure won't make my ears bleed. But give it 30 minutes and someone will claim that 50ohm portions of some speaker's impedance curve will shoot those dotted lines WAY off the solid lines.

If the load dependency dogma crowd really want to convince us, pick a speaker (or 2 or 3) and perform a microphone-measured speaker FR with a load-independent reference amp, and again with a Wiim amp. Then bring the plot with both lines so everyone can see a quantified representation of this behavior. I would expect negligible differences, but I'm more than happy to change that attitude if shown significant FR changes.

Someone already did it for us:

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

That’s a start, though the THD vs Frequency vs Power is very hard to read. All other graphs are not biased to 0 dB, making them also harder to read. He will improve no doubt :)

www.audiosciencereview.com

(thanks to @ctrl )

 

[NTK]

A Q of 2 is not barely but clearly audible if I read that correctly.
Specially at the area where we are the most sensitive..

From the original Toole and Olive paper, Q = 1 peaks are just audible with pink noise at amplitudes of 0.5 dB. So Q = 2 peaks at 0.6 dB are clearly audible when listening to music which is the use case? The figure description also clearly states that "audibility is similar at all frequencies except, perhaps, at very low frequencies".

 

[375HP2482]

No, it's really many orders of magnitude more pronounced than a "magical" capacitor. FR error is in the order of .5 to 1 dB, which can certainly be in the realm of audibility.

I have yet to meet the mortal who can detect a 1 dB change at 20 kHz.

 

[voodooless]

I have yet to meet the mortal who can detect a 1 dB change at 20 kHz.

But it it’s not only at 20 kHz… issues can start at 2 kHz already, depending on your speaker load.

 

[Sokel]

From the original Toole and Olive paper, Q = 1 peaks are just audible with pink noise at amplitudes of 0.5 dB. So Q = 2 peaks at 0.6 dB are clearly audible when listening to music which is the use case? The figure description also clearly states that "audibility is similar at all frequencies except, perhaps, at very low frequencies".

View attachment 351375

I would like to know if it's not only similar but also proportional to our sensitivity to the defined freqs (which sounds logical).
ctrl's test with fixed 1db is maybe low or high,and that's the whole matter and my objection,it's really random and hard to fix without data of the combined (speaker and amp) performance as we already know that stuff up there should be fixed with anechoic data.

The "load dependency dogma crowd" which I obviously belong is not as vocal as it is for no reason.
Engineers have struggled to fix stuff like this in the early days,it's somehow funny to get back to it.

 

[martin900]

Low SINAD + dominant 3rd harmonic, this will not sound nice at all.

 

[Endibol]

Afaik, WiiM offers replacements for early Amp revisions suffering from PS hum/whine:

My WiiM Amp tests

As I got yet another interface - UCX II, quick test to verify if something has changed in terms of sub out performance. No, no change :cry:

forum.wiimhome.com

This was fixed with PCB_version 5.

Again, great service!

 

[staticV3]

Low SINAD + dominant 3rd harmonic, this will not sound nice at all.

You're kidding yourself if you think that HD3 at -95dB will be even remotely audible.

 

[Inertiaman]

Someone already did it for us:

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

That’s a start, though the THD vs Frequency vs Power is very hard to read. All other graphs are not biased to 0 dB, making them also harder to read. He will improve no doubt :)

www.audiosciencereview.com

(thanks to @ctrl )

View attachment 351357 View attachment 351358

The example you provided isn't even remotely what I was suggesting. Your example is a cherry-picked piece of music for sample A, then sample B is the same music with a contrived filter applied via two sliders on a graphic equalizer, raising the entire ~ 2kHz to 4kHz region by 1dB.

No one here is arguing that the difference between two such audio clips isn't audible.

My suggestion was to measure the FR of speaker(s) with both a flat FR amp and a load-dependent FR amp, and plot them both together.

 

[staticV3]

Statement was that it will not sound nice at all. Nothing more, nothing less.

???

Statement was:

Low SINAD + dominant 3rd harmonic, this will not sound nice at all.

That reasoning is nonsense and deserves to be called out as auch.

 

[Inertiaman]

Statement was that it will not sound nice at all. Nothing more, nothing less.

Bolding above by me. Isn't that equivalent to saying it will be VERY audible?

 

[Inertiaman]

So "wiggle" is a scientific term?
I have admitted lots of times,English is not my language,but it doesn't sound like a scientific one to me.
Put simply,a wiggle to me is a high-ish Q deviation,a low Q ones will cover a broader area and to me is not a "wiggle" is more like a narrow tone control effect would have,or a "presence" slider in a DAW.

No its not a scientific term, that was my entire point, made in response to you literally stating that the shapes do not qualify as wiggles!

Now you've gone on to limit the term "wiggle" to high-Q shapes. You are trying to make it a scientific term for ch$%#'s sake.

FWIW, the definition of wiggle is "to move back and forth or up and down". I used the word as a convenient informal term in reference the deviations from the line in Erin's plots, never in a million years imagining that someone would try to invalidate my position by asserting that what I've called wiggles don't qualify as wiggles.

 

[NTK]

I would like to know if it's not only similar but also proportional to our sensitivity to the defined freqs (which sounds logical).
ctrl's test with fixed 1db is maybe low or high,and that's the whole matter and my objection,it's really random and hard to fix without data of the combined (speaker and amp) performance as we already know that stuff up there should be fixed with anechoic data.

The "load dependency dogma crowd" which I obviously belong is not as vocal as it is for no reason.
Engineers have struggled to fix stuff like this in the early days,it's somehow funny to get back to it.

I totally agree with you that these days there isn't a good excuse not to implement PFFB, especially when it can be implemented without a significant BOM cost increase.