The Audio Path In Consumer-Grade Products

[PENG]

The rated power A-weighted specifications are so suspect that I probably wouldn't believe the calculation, but if you start with a one watt measurement and calculated the rated power number I'd believe that.

I am talking about the 1W to rated power numbers only, not the weighted ones.

 

[Irvrobinson]

Can you post an example of such a plot?

Here's a link to a TI document that contains a chart like that on page 2:

tidu016.pdf

Edit: This paper is better:

http://www.electro.fisica.unlp.edu.ar/temas/pnolo/p1_AN-104.pdf

 

[<eargiant]

IMO, the answer is: it depends. For example, for amplifiers, it is very important to know the power level that the SNR is measured relative to. (This is why Stereophile's measurements are not directly comparable to Benchmark's specs.) Most manufacturers like Benchmark, including ATI incidentally, quote SNRs for amplifiers relative to rated power output, while John Atkinson (and Gene) measure SNR relative to one watt. I dislike SNR measurements referenced to rated power, for a lot of reasons, not the least of which is that it's a marketing trick for producing a higher number as a figure of merit, just like A-weighting does, but it's worse because every amp has different gain and rated power, and because amps typically don't run at rated power. That's why one watt unweighted SNRs for amps make the most sense - because one watt is a realistic power level for actual use, and every amp is measured at the same level.

Noise, unlike distortion, is mostly a constant value regardless of output, and most absolute measurements I've seen are in microvolts, and IMO are best plotted in a microvolt scale across a frequency spectrum. That's cumbersome, so broad-spectrum unweighted SNRs are the convenient figure of merit.

So, for unweighted 1W SNRs, -75db is a very good measurement. I think about -97db is the best I've ever seen (not that I've looked much lately). I've seen amps with measurements as low as -57db. Is there an audible difference between amps at -57db, -75db, and -97db? Maybe. It might depend on the speakers.

If you use very sensitive speakers, let's say 95db/2.83v/1meter, since noise is mostly a constant in the amp, and sensitive speakers need much less output, in this case 10-12db less power than a typical speaker, it is possible that a better noise spec could make an audible difference in quiet passages. I have to admit I doubt it even at -57db, but I'd bet $100 that at -75db it was never more audible than -97db. Would I rather own the -97db@1W amp? Absolutely. Is such a low noise spec audible? I really doubt it, especially with more typical speakers.

Good stuff.

Can you speak to the various input Vrms when measuring SNR and the effect of using different Vrms if any?

Also, can you elaborate on the differences between "Wideband", "20Hz-20kHz" and "Dynamic" when comparing SNRs?

Lastly, would the SNR increase if the pre had a lower gain setting?

 

[Irvrobinson]

Can you speak to the various input Vrms when measuring SNR and the effect of using different Vrms if any?

I think you want to use whatever input voltage is necessary to achieve one watt of output. Or perhaps I don't understand your question.

Also, can you elaborate on the differences between "Wideband", "20Hz-20kHz" and "Dynamic" when comparing SNRs?

Wideband = unweighted, so I think that produces the most informative reading. I don't know what "Dynamic" means.

Lastly, would the SNR increase if the pre had a lower gain setting?

I don't know.

 

[<eargiant]

I think you want to use whatever input voltage is necessary to achieve one watt of output. Or perhaps I don't understand your question.



Wideband = unweighted, so I think that produces the most informative reading. I don't know what "Dynamic" means.



I don't know.

Thanks, for the fast response. Not to dwell on SNR but since it was brought up in post #4 and is still with us, I want to be sure we are all talking about the same apple.

Your response makes sense to me and you seem to be very knowledgeable so I'm asking you to clarify this for me.

I am confused by what AcuDefTechGuy said at the start of this thread. In post #6 while commenting on Benchmark's SNR he said that "The Denon X3300's SNR of -90dB unweighted is also excellent."

Then in post #14 I said "You conveniently left these parts out:
"Referenced to the AHB2's clipping power of 100W, this is equivalent to a dynamic range of 128.5dB, which is close to the specified 132dB.""

Then in post #16 he replied:
"Are you saying SNR is the same thing as DYNAMIC RANGE ?

I am focusing only on SNR because it seems the article by Benchmark mentions SNR about 4 or more times, making it sound like SNR of the Marantz PrePros is LOW when in fact it is -130dBA, which is a lot higher than the -108dBA of the Benchmark amp!

Again, just looking at SNR (because the article significantly talks about SNR), not Dynamic Range, even a $300 AVR has a higher A-Wt SNR than the Benchmark!"

So herein lies my question. When I look at the SNR that Audioholics has posted for the AVR-3300 they state 90db unweighted Dynamic Range. So wouldn't the accurate comparison be to compare that to Benchmark's 128.5 dB Dynamic Range as specified by John Atkinson?

If Audioholics had measured the AVR-3300 @1 watt, wouldn't the SNR be much worse?

http://www.audioholics.com/av-receiver-reviews/denon-avr-x3300w-1/measurements

Here's another amp that Audioholics measured at a SNR of 90db but @ 1watt. Which I believe translates into an SNR that is markedly better than the AVR-3300.

http://www.audioholics.com/amplifier-reviews/yamaha-a-s801-amplifier-review/yamaha-a-s801-measurements

What are your thoughts?

 

[PENG]

I am focusing only on SNR because it seems the article by Benchmark mentions SNR about 4 or more times, making it sound like SNR of the Marantz PrePros is LOW when in fact it is -130dBA, which is a lot higher than the -108dBA of the Benchmark amp!

Again, just looking at SNR (because the article significantly talks about SNR), not Dynamic Range, even a $300 AVR has a higher A-Wt SNR than the Benchmark!"

So herein lies my question. When I look at the SNR that Audioholics has posted for the AVR-3300 they state 90db unweighted Dynamic Range. So wouldn't the accurate comparison be to compare that to Benchmark's 128.5 dB Dynamic Range as specified by John Atkinson?

If Audioholics had measured the AVR-3300 @1 watt, wouldn't the SNR be much worse?

http://www.audioholics.com/av-receiver-reviews/denon-avr-x3300w-1/measurements

Here's another amp that Audioholics measured at a SNR of 90db but @ 1watt. Which I believe translates into an SNR that is markedly better than the AVR-3300.

http://www.audioholics.com/amplifier-reviews/yamaha-a-s801-amplifier-review/yamaha-a-s801-measurements

What are your thoughts?

I don't think the Marantz AV8802's SNR is that high. It looked higher because S&V 's figures are typically weighted. If unweighted, at 100mv input, it would probably be 110-115dB, but I am guessing.

The Yamaha A-S801 is not "markedly better" than the avr-x3300w in the AH tests, it seems to me they are comparable.

 

[<eargiant]

The Yamaha A-S801 is not better than the avr-x3300w in the AH test, probably worse because it was for 200 mv input vs 100 mv input used in the Demon's test.

I'm not sure about the input, that's why that was the first question I asked Irv. Seems like your not sure either. Irv's answer seemed logical but if incorrect why would Gene use different inputs?

So it would appear that the Denon's SNR of 90db is being measured in the same manner as the Benchmark's 128dB?

As for the A-S801, the SNR is @1 watt and the Denon 3300 is Unweighted AES17 Dynamic Range which based on my understanding will always yield a higher number. Aren't those two different protocols with the @1 watt being more stringent and yielding a lower number? So if the @1 watt method was used on the 3300, wouldn't it be worse (unless of course we get confirmation on the input question).

Unless I'm missing something, it doesn't seem correct to compare one amps unweighted dynamic SNR to anothers @1 watt SNR number.

Maybe Gene can clarify this part for us?

 

[PENG]

I'm not sure about the input, that's why that was the first question I asked Irv. Seems like your not sure either. Irv's answer seemed logical but if incorrect why would Gene use different inputs?

What is your first question? Irv also said "Or perhaps I don't understand your question."

So it would appear that the Denon's SNR of 90db is being measured in the same manner as the Benchmark's 128dB?

In Gene's he did it with 100 mV input drive, 1V rms output and he measured the preamp output. How did Benchmark measured theirs?

As for the A-S801, the SNR is @1 watt and the Denon 3300 is Unweighted AES17 Dynamic Range which based on my understanding will always yield a higher number. Aren't those two different protocols with the @1 watt being more stringent and yielding a lower number? So if the @1 watt method was used on the 3300, wouldn't it be worse (unless of course we get confirmation on the input question).

You cannot compare the two directly, that's why I said they "seemed" to be comparable. "seem" because their figures provided are not directly comparable but based on my other observations, my educated guess is that they are likely comparable.

Unless I'm missing something, it doesn't seem correct to compare one amps unweighted dynamic SNR to anothers @1 watt SNR number.

You are the one who try to compare them and said

Here's another amp that Audioholics measured at a SNR of 90db but @ 1watt. Which I believe translates into an SNR that is markedly better than the AVR-3300." right?

And I disagree as I see no evidence to support your "markedly" better comment.

Those tests do give us some idea, for example Gene commented that "The AVR-X3300W preamp output exhibited a low noise floor (90dB) unweighted with 100mV input drive." and "With a 200mVrms input, I measured > 89dB at 1 watt output and >93dB with A-wt filter engaged. This is a great result and demonstrates why I felt the noise floor on this product was completely inaudible."

Maybe Gene can clarify this part for us?

I agree, this is a good question for Gene. I read pretty much all of his reviews that include bench tests and I do feel there is room to improve in terms of consistency in his test protocols.

I have read a lots of bench test results from different sources and I can tell you with certainty that in general, integrated amps and AVRs within the same price range are comparable and quite often mid range AVRs such as the Denon 3000 and 4000 series beat their integrated counterparts. As others mentioned before, perhaps economy of scale is a factor. I also believe while putting everything in one box put them in some disadvantages, there are also benefits in terms of synergy and compatibility, i.e. easier to optmize/match components in the complete signal path, as well as keeping the path short and direct.

 

[AcuDefTechGuy]

How did Benchmark measured theirs?

Yeah, but can we trust them to measure their own products, instead of independent 3rd party measurements?

What do you think of the Stereophile measurement of the Benchmark amp?

"All measurements were taken with balanced input signals...The voltage gain at 1kHz into 8 ohms depended on the rear-panel switch position. With the switch set to a sensitivity of "9.8V/22dBU," the gain was 9.2dB...With the switch set to "4V/14.2dBU," the AHB2's gain was 17dB; with it set to "2V/8.2dBU," the gain was 22.9dB. With the amplifier used in bridged-mono mode, the gain in the "2V/8.2dBU" condition was at 29dB, an input signal of 100mV resulting in an output power of 1W into 8 ohms."

"The wideband S/N ratio, measured in the high-gain, stereo condition with the input shorted to ground and ref. 1W into 8 ohms, was very high, at 89.3dB. Reducing the measurement bandwidth to 22Hz–22kHz increased the ratio to 106dB, while switching an A-weighting filter into circuit increased it further, to 108.5dB."

 

[Goliath]

I have read a lots of bench test results from different sources and I can tell you with certainty that in general, integrated amps and AVRs within the same price range are comparable and quite often mid range AVRs such as the Denon 3000 and 4000 series beat their integrated counterparts. As others mentioned before, perhaps economy of scale is a factor.

Absolutely. Economies of scale of mass production is probably the number one reason for the high price/performance ratio of AV receivers.

Looking through Genes reviews I've noticed far less expensive AV receivers often outgunning integrated amps in 2 channel mode (power-wise) and offer either comparable or better technical performance overall.

Still, bench test figures are a little misleading when you consider that swept sine waves (~3 dB crest factor) as often used on the bench are many times more stressful to power supplies/heatsinks and output transistors than typical musical signals that have very high crest factors ( ~ 6/9 dB minimum for highly compressed up to 20 dB+).

Like THD+N figures, below a certain point it really doesn't matter because noise and distortion will be too soft for anyone to hear so it then becomes a numbers game. Just like very high damping factors.

 

[PENG]

Yeah, but can we trust them to measure their own products, instead of independent 3rd party measurements?

What do you think of the Stereophile measurement of the Benchmark amp?

"All measurements were taken with balanced input signals...The voltage gain at 1kHz into 8 ohms depended on the rear-panel switch position. With the switch set to a sensitivity of "9.8V/22dBU," the gain was 9.2dB...With the switch set to "4V/14.2dBU," the AHB2's gain was 17dB; with it set to "2V/8.2dBU," the gain was 22.9dB. With the amplifier used in bridged-mono mode, the gain in the "2V/8.2dBU" condition was at 29dB, an input signal of 100mV resulting in an output power of 1W into 8 ohms."

"The wideband S/N ratio, measured in the high-gain, stereo condition with the input shorted to ground and ref. 1W into 8 ohms, was very high, at 89.3dB. Reducing the measurement bandwidth to 22Hz–22kHz increased the ratio to 106dB, while switching an A-weighting filter into circuit increased it further, to 108.5dB."

I like what I saw, enough to make me to "want" one but no I have too many amps already. I love great specs and measurements even when I know I can't hear the differences between gear that exceed the audible thresholds.

 

[AcuDefTechGuy]

I like what I saw, enough to make me to "want" one but no I have too many amps already. I love great specs and measurements even when I know I can't hear the differences between gear that exceed the audible thresholds.

Definitely not audible and we are just PLAYING with numbers here.

I think the SNR numbers (89dB, 106dB, 108dB) looked good, but not awesome.

If the wide-band Unweighted SNR were 120dB or the A-Wt SNR were 140dB, now that would be awesome. The Marantz 8802 already has an A-Wt SNR of 130dB.

 

[Irvrobinson]

I don't think the Marantz AV8802's SNR is that high. It looked higher because S&V 's figures are typically weighted. If unweighted, at 100mv input, it would probably be 110-115dB, but I am guessing.

I think an unweighted 110db SNR is remarkably high, in the I-don't-believe-it category.

 

[Irvrobinson]

Definitely not audible and we are just PLAYING with numbers here.

I think the SNR numbers (89dB, 106dB, 108dB) looked good, but not awesome.

If the wide-band Unweighted SNR were -120dB or the A-Wt SNR were -140dB, now that would be awesome. The Marantz 8802 already has an A-Wt SNR of -130dB.

Well, let's see. JA measured the Emotiva XPA Gen 3 at 70db unweighted. The pricey full-custom Levinson 536 at 71db. The Boulder 2150 at 78.5db. (The 2150 costs $99K/pair, I might add.) Nelson Pass's esoteric First Watt J2 at 84.1db. The $22K/pair Krell 575 was at 78.2db.

The $12K Levinson 585 integrated amp (which looks like another full-custom Harman design built by Mack Technologies) was measured at 72.75db. The Parasound Halo integrated amp achieved 73db. JA doesn't measure AVRs (because they would horrify Stereophile's readers, I suppose), so we can't compare them.

So the Benchmark amp is looking pretty good by comparison.

 

[Irvrobinson]

BTW, I just noticed an anomaly in this discussion, which I have traced back to Sound and Vision... signal to noise ratios are positive numbers, not negative numbers. ADTG was quoting SNRs from Sound and Vision, and he correctly quoted negative SNRs from the reviews, which got the rest of us doing that, but if you think about it, the ratio of signal to noise has got to be positive. The signal is xxdb higher in amplitude than the noise. I also noticed every other test report source uses positive ratios, so I wonder what Sound and Vision is thinking? I went back and edited a few of my posts to correct them, and I won't let myself get side-tracked again.

 

[AcuDefTechGuy]

Well, let's see. JA measured the Emotiva XPA Gen 3 at 70db unweighted. The pricey full-custom Levinson 536 at 71db. The Boulder 2150 at 78.5db. (The 2150 costs $99K/pair, I might add.) Nelson Pass's esoteric First Watt J2 at 84.1db. The $22K/pair Krell 575 was at 78.2db.

The $12K Levinson 585 integrated amp (which looks like another full-custom Harman design built by Mack Technologies) was measured at 72.75db. The Parasound Halo integrated amp achieved 73db. JA doesn't measure AVRs (because they would horrify Stereophile's readers, I suppose), so we can't compare them.

So the Benchmark amp is looking pretty good by comparison.

Oh, I didn't realize the unweighted SNR were usually like that.

Gene measured the unweighted SNR of the Denon AVP-A1 as 95dB.

Gene measured the Denon AVR-X3300W SNR as 90dB unweighted with 100mV input drive, as PENG mentioned.

So why is the unweighted SNR of a lowly Denon 3300 better than the Benchmark and all those other expensive amps you mentioned?

 

[Irvrobinson]

Oh, I didn't realize the unweighted SNR were usually like that.

Gene measured the unweighted SNR of the Denon AVP-A1 as 95dB.

Gene measured the Denon AVR-X3300W SNR as 90dB unweighted with 100mV input drive, as PENG mentioned.

So why is the unweighted SNR of a lowly Denon 3300 better than the Benchmark and all those other expensive amps you mentioned?

I'm not sure. One difference could be the relative quality of the test environments. Gene's could be better. I admit that I tend to only compare measurements from the same tester; it is difficult to know if the test factors are really equivalent, especially at very low noise-signal voltages as you have with 90db SNRs. Or the Denon may have an exceptionally low noise level. Since, as I mentioned, JA never tests equipment like the Denon we may never know. I know they both use Audio Precision test equipment.

 

[AcuDefTechGuy]

I'm not sure. One difference could be the relative quality of the test environments. Gene's could be better. I admit that I tend to only compare measurements from the same tester; it is difficult to know if the teat factors are really equivalent, especially at very low noise-signal voltages as you have with 90db SNRs. Or the Denon may have an exceptionally low noise level. Since, as I mentioned, JA never tests equipment like the Denon we may never know. I know they both use Audio Precision test equipment.

Excellent point.

I'm sure nothing is perfect, but I think we can agree that Gene's measurements are done very well.

Perhaps we can't really prove that the Denon X3300 RECEIVER has a better Unweighted SNR than the Benchmark amp or all those other high-end amps. But I think we can agree that an Unweighted SNR of 90dB is a superb measurement result and inaudible.

 

[Irvrobinson]

Excellent point.

I'm sure nothing is perfect, but I think we can agree that Gene's measurements are done very well.

Perhaps we can't really prove that the Denon X3300 RECEIVER has a better Unweighted SNR than the Benchmark amp or all those other high-end amps. But I think we can agree that an Unweighted SNR of 90dB is a superb measurement result and inaudible.

On your last point, agreed.

Regarding Gene's versus JA's test lab environments and methodologies, the facts are that we just don't know. JA has been testing amps for years, and he sure has tested a lot of them. Does that mean the quality of the measurements are high? Not necessarily. We just don't have enough information to judge.

For example, Gene has not tested the same Emotiva amp that JA did, Gene tested the XPR-1 in 2013. In Gene's test report (oddly) he did not state a wide-band SNR, but the unweighted SNR at one watt with a 20KHz low-pass filter was 89db. That's in the range of what JA measured with the Benchmark amp, but JA didn't mention a low-pass filter, so I suspect Gene's result would have been worse without the filter. My point is that quibbling about a few db of measurement difference at these noise levels might be more due to measurement methodologies than what the true performance differences are.

 

[AcuDefTechGuy]

On your last point, agreed.

Regarding Gene's versus JA's test lab environments and methodologies, the facts are that we just don't know. JA has been testing amps for years, and he sure has tested a lot of them. Does that mean the quality of the measurements are high? Not necessarily. We just don't have enough information to judge.

For example, Gene has not tested the same Emotiva amp that JA did, Gene tested the XPR-1 in 2013. In Gene's test report (oddly) he did not state a wide-band SNR, but the unweighted SNR at one watt with a 20KHz low-pass filter was 89db. That's in the range of what JA measured with the Benchmark amp, but JA didn't mention a low-pass filter, so I suspect Gene's result would have been worse without the filter. My point it that quibbling about a few db of measurement difference at these noise levels might be more due to measurement methodologies than what the true performance differences are.

Oh, I agree.

My point was never to debate a few dB difference. It was only to say that AVRs like the Denon X3300 can have excellent measurements.