Looking beyond the single headline SINAD value...

[-Matt-]

Which amps offer the best distortion linearity?

What I mean by distortion linearity (possibly the wrong term) is the most consistent (and lowest) distortion vs power across all frequencies.

Here is the graph from the NCx500 review:

The maximum SINAD before clipping (@15Khz) is about -75dB.


Here is the same graph from the Nord Three SE 1ET400A (Purifi) review:

Again, maximum SINAD is about -75 to -80dB for 15kHz.


Now here is the graph from the review for the Soncoz SGP1 (class AB):

Apart from the 20Hz curve, which was apparently fixed, the performance seems very well behaved (about -85dB max SINAD @15Khz).


One that fares a bit worse is the MiniDSP SHD Power (just for example):

A max SINAD of -55dB @15Khz.


...and of course AVRs also tend to also score quite poorly in this, for example the Denon AVR-X8500H:

Although it behaves predictably SINAD is up to -65dB @15Khz.


So if I'm picking a power amp to improve on this aspect of the AVRs internal amp performance, which models would be best?

The other requirement is to preferably exceed the AVRs 2 channels driven power of 225W (into 4 Ohms).

Edit: Is it fair to compare amplifiers by these graphs? What is the effect of different amplifier gain? Some amps are probably measured in low gain mode which wouldn't be suitable for use with an AVR.

 

[DVDdoug]

Distortion is rarely audible unless you push the amp into clipping. It can be helpful to see where it starts clipping compared to the published power rating.

IMO - The noise component is more important, although it's not necessarily audible either. And that can depend the sensitivity of your speakers, how close you are to the speakers, and the acoustic noise in the room.

Edit: Is it fair to compare amplifiers by these graphs? What is the effect of different amplifier gain? Some amps are probably measured in low gain mode which wouldn't be suitable for use with an AVR.

Amir measures SINAD at 5 Watts which is "fair" because the signal part of SINAD is constant.

 

[-Matt-]

As it would likely be the first suggestion, here is the same graph for the Benchmark AHB2:

About -95dB at high power for 15kHz. But this is in low gain mode. Do we have this data for the AHB2 in high gain mode?

 

[-Matt-]

Distortion is rarely audible unless you push the amp into clipping. It can be helpful to see where it starts clipping compared to the published power rating.

If you look at the graph for the AVR above you will see that the SINAD at 15kHz is poor at all power levels (not only near clipping). It is much higher than Amir's requirement for audio transparency.

I currently have the thought that the often stated sensation that sounds sometimes seem mushed together (especially at higher frequencies) could be related to this sort of frequency dependent performance.

If we focus too much on just the headline SINAD measured at 1kHz we could make it easy for manufactures to "game the system".

 

[wwenze]

Can you even hear the second harmonic of 15kHz

 

[Blumlein 88]

If you look at the graph for the AVR above you will see that the SINAD at 15kHz is poor at all power levels (not only near clipping). It is much higher than Amir's requirement for audio transparency.

I currently have the thought that the often stated sensation that sounds sometimes seem mushed together (especially at higher frequencies) could be related to this sort of frequency dependent performance.

If we focus too much on just the headline SINAD measured at 1kHz we could make it easy for manufactures to "game the system".

The 2nd harmonic of 15 khz is 30 khz. If it produced any audible distortion it likely will be IMD. The non-linearity that causes harmonic distortion will cause IMD as well. The saving grace would be that levels at 15 khz are usually low so levels of IMD would be low as well. The opposite point is they won't necessarily be masked, but more than likely will be almost all the time with music.

 

[-Matt-]

Can you even hear the second harmonic of 15kHz

Nope!

...but 10kHz and 5kHz maybe (which are also high in SINAD).

 

[JSmith]

If it produced any audible distortion it likely will be IMD.

Yes IMD is not discussed enough here IMO;

Intermodulation Distortion

Intermodulation Distortion (IMD)

sound-au.com

JSmith

 

[AnalogSteph]

HF THD is tricky. In AB amps, it tends to be correlated to IMD that we do care about. The same in a Class D (PWM) amplifier, rather not. For an apples to apples comparison, multitone is more useful.

 

[Killingbeans]

Is it fair to compare amplifiers by these graphs?

Nahh. Measurements with 45kHz bandwidth, meaning that most of the ugly stuff probably takes place above 20kHz.

Also, human hearing sensitivity decreases radically above 10kHz, and no music has much information worth preserving at those frequencies anyway.

Also, at high SPL masking curves become less symmetrical and dig a lot more into the frequencies above.

Combine that with the fact that humans generally suck at hearing harmonic distortion unless they have specific training in doing so, and there really isn't much to worry about.

My 2 cents: Pick any well designed amp with no audiophile gimmicks, enough power to avoid clipping, low enough residual noise to make tweeter hiss a non-issue, and then just let go of any nagging doubt.

 

[-Matt-]

Yes IMD is not discussed enough here IMO;

Intermodulation Distortion

Intermodulation Distortion (IMD)

sound-au.com

JSmith

Thanks, this was a good read.

By this metric would it be fair to conclude that the Soncoz SGP1 beats the Benchmark AHB2 by a small margin?

Benchmark AHB2:

Soncoz SGP1:

Nahh. Measurements with 45kHz bandwidth, meaning that most of the ugly stuff probably takes place above 20kHz.

That is a very good point about the measurement bandwidth. However, as you can see from the IMD graphs above there are a lot of components in the human hearing range, and these are comparable in magnitude to the ultrasonic components that you refer to.

 

[wwenze]

Actually does a square wave and a bandwidth-limited square-ish wave (brick wall filter at 20kHz) sound the same

Probably a fun experiment

 

[-Matt-]

Neither the Soncoz SGP1 nor the Benchmark AHB2 are class D. For comparison then, here is the Topping TP RA3 which is class D and scores an excellent -105dB SINAD on the single value metric:

In this case, the higher SINAD at high frequencies does seem to correllate with increased IMD.

 

[-Matt-]

Actually does a square wave and a bandwidth-limited square-ish wave (brick wall filter at 20kHz) sound the same

Probably a fun experiment

Isn't this the claim of proponents of super-tweeters, that they need the ultrasonics?

From a mathematical point of view, the higher frequency components that are present, the sharper the square wave can be.

But surely the ultrasonics are filtered out by the ear's lack of sensitivity above the human hearing range? If so, I don't think the ultrasonics do anything useful and so the bandwidth limited one would sound the same to us (but not necessarily to a microphone).


But my original question (clarified by others) was not about ultrasonic components. It was about the general sensation of muddying, mushing together of sounds (especially with complex music that has a lot going on at different frequencies all at once). Could this be the result of IMD in the audible range? I currently suspect that it could be.

 

[Sokel]

It's a shame we don't also have Benchmark's 15Khz chart,for us visuals charts like this are a joy to look at.

(inaudible,I know but we live in a mostly visual world)

 

[-Matt-]

It's a shame we don't also have Benchmark's 15Khz chart,for us visuals charts like this are a joy to look at.

(inaudible,I know but we live in a mostly visual world)

Yes, and I'd also like to see those measurements made in the high gain mode.

 

[welwynnick]

Actually does a square wave and a bandwidth-limited square-ish wave (brick wall filter at 20kHz) sound the same

I did this test several years ago and found that they sound distinctly different. There's a slightly different pitch that's easy to spot. I'd encourage everyone to try it.

The Soncoz is brilliant, isn't it?

 

[ZolaIII]

Ufff SINAD at 15 KHz? SINAD is THD+N in Db scale. This doesn't apply. Second harmonic of 15 KHz is 30 KHz. So it's desirable that the noise stays low but THD is not relevant. Future more there will be very little there above 12~12.5 KHz in recorded materials and our auditory system is most sensitive in mid frequencies (vocal range) and less in highs and least in bass. In music main tones stop at little under 4 KHz (picoilo - half sized flute) for everything except pipe organs (8 KHz) and there for SINAD over 5 KHz or 10 KHz as uper theoretical range is of little to no importance. As shown more like 4 & 8 KHz. However increasing noise is (somewhat) important and load dependance and damping factor more so. This also explains why SINAD is most important at mids (most informations and highest sensitivity there) and declared measured there.

 

[Blumlein 88]

I did this test several years ago and found that they sound distinctly different. There's a slightly different pitch that's easy to spot. I'd encourage everyone to try it.

The Soncoz is brilliant, isn't it?

How did you create the two waves and at what frequency?

 

[Talisman]

McIntosh MC462