A video on testing the audibility of recording preamps

[LTig]

Yea I know how to REW loopback my card and it's easy to have a preamp in the middle and run a sweep but what I want to try is, based on db of amplification what is the harmonic content generation and where do my preamps saturate or start to saturate and with what kind of knee...

Like here (click the spoiler)?

Marantz AV7705 Home Theater Processor Review

This is a review and detailed measurements of the Marantz AV7705 UHD Home Theater Audio Video Processor. It is on kind loan from a member. The AV7705 costs US $2,200 from Amazon with free shipping. From what I have read this unit is current until October of 2020. The design is typical of...

www.audiosciencereview.com

 

[frcake]

Like here (click the spoiler)?

Marantz AV7705 Home Theater Processor Review

This is a review and detailed measurements of the Marantz AV7705 UHD Home Theater Audio Video Processor. It is on kind loan from a member. The AV7705 costs US $2,200 from Amazon with free shipping. From what I have read this unit is current until October of 2020. The design is typical of...

www.audiosciencereview.com

Umm kinda yea, but also like Jim did and showed the 1:1 (linear) -> inf:1 (hard-clip).

I guess maybe I could feed different output volumes from the computer so they are very specifically set and then measure something inside REW for like X db step jumps. Also slew would be amazing to measure as well.. This is why I feel unqualified :/

 

[LTig]

Umm kinda yea, but also like Jim did and showed the 1:1 (linear) -> inf:1 (hard-clip).

I guess maybe I could feed different output volumes from the computer so they are very specifically set and then measure something inside REW for like X db step jumps. Also slew would be amazing to measure as well.. This is why I feel unqualified :/

I haven't worked with REW for quite a while but I think you can program it to measure different output levels in steps. Slew you could measure with rectangular signals and REWs scope, using the highest samplerate possible.

 

[frcake]

I haven't worked with REW for quite a while but I think you can program it to measure different output levels in steps. Slew you could measure with rectangular signals and REWs scope, using the highest samplerate possible.

Great, I'll do some digging, thank you for the information and advice!

 

[frcake]

Guys, help me figure out something. What if I run a signal 10x times using the same line amp and then feeding the signal back to it after I've recorded it in my computer. That's a difference in series though (where the line-amp characteristics get exaggerated), I don't know if it makes sense as a test in parallel, like if you run it 10x times does it equal the effect of having 10x the same line amps in your mix? Will this exaggerated "signature" sound be spread on your mix and have an effect?

 

[LTig]

Guys, help me figure out something. What if I run a signal 10x times using the same line amp and then feeding the signal back to it after I've recorded it in my computer. That's a difference in series though (where the line-amp characteristics get exaggerated), I don't know if it makes sense as a test in parallel, like if you run it 10x times does it equal the effect of having 10x the same line amps in your mix? Will this exaggerated "signature" sound be spread on your mix and have an effect?

You do it in series, not in parallel. And yes, this is a good idea as long as the ADC and DAC are transparent (which the usually are), as any diversion from perfect will be multiplied by the number of tests.

You could run 10 tests in loopback mode (without the amp) and then 10 tests with the amp. Then listen to both recordings to hear the influence of the amp, multiplied by 10.

@Blumlein 88 made such a test with a sound interface:

Listen and choose the 8th generation digital copy.

Each of these zip files will open into three wav files. One is labeled Reference, one is A and one is B. These are 30 second snippets of full songs. The originals were from CD so all 44.1 khz and 16 bit. Either A or B has been thru 8 generations of DA to AD copying. The other is identical...

www.audiosciencereview.com

 

[TK_IIVI]

Guys, help me figure out something. What if I run a signal 10x times using the same line amp and then feeding the signal back to it after I've recorded it in my computer. That's a difference in series though (where the line-amp characteristics get exaggerated), I don't know if it makes sense as a test in parallel, like if you run it 10x times does it equal the effect of having 10x the same line amps in your mix? Will this exaggerated "signature" sound be spread on your mix and have an effect?

I think the bigger question might be (to me at least) if there's a difference between running a correlated signal multiple times (in parallel or series) as opposed to multiple uncorrelated signals.

 

[LTig]

I think the bigger question might be (to me at least) if there's a difference between running a correlated signal multiple times (in parallel or series) as opposed to multiple uncorrelated signals.

You cannot run signals in parallel if you have only one interface (OK, if you have an 8 channel DAC and ADC you could run 8 signals in parallel). So if you run the same signal multiple times (N) through a DAC/ADC conversion and then add them up and divide by the N, you can get a reduction of noise by the square root of N (signal is multiplied by N, but stochastic noise by square toot of N)[1]. The result would then be better than a single test. So there is a difference between parallel and serial - parallel hides flaws, serial enhances flaws.

[1] In NMR and ESR this is called averaging. It's used to enhance SNR of small signals. N can be up to a million, to get an improvement of SNR by a factor of 1000.

 

[TK_IIVI]

You cannot run signals in parallel if you have only one interface (OK, if you have an 8 channel DAC and ADC you could run 8 signals in parallel). So if you run the same signal multiple times (N) through a DAC/ADC conversion and then add them up and divide by the N, you can get a reduction of noise by the square root of N (signal is multiplied by N, but stochastic noise by square toot of N)[1]. The result would then be better than a single test. So there is a difference between parallel and serial - parallel hides flaws, serial enhances flaws.

[1] In NMR and ESR this is called averaging. It's used to enhance SNR of small signals. N can be up to a million, to get an improvement of SNR by a factor of 1000.

I understand. My point was just that in a real-world scenario we're going to have multiple sources feeding each amp and they're all going to be uncorrelated from each other. If I recorded literally the exact same source 10 times and sum them then yeah, I'd need to make up for that by lowering the level to get back to the input level of one channel (assuming that was the desired level). But that's not a real world scenario.

If I'm recording a drum kit, bass, electric guitar and electric piano all at the same time those are all different signals. Once summed the total level obviously will be much louder, but it's all uncorrelated sources. Therefore a more interesting experiment would be uncorrelated sources going through amps at different settings vs not going through them and comparing the sums in each scenario to each other.

I don't brain good so this is obviously possibly not right, but one part of my brain says that if you overdrive a couple of channels, given that they're playing in the same key and might play similar notes, some frequencies may be emphasized. For example one instrument playing an A 440Hz and another A 880Hz will lead to both generating not only 880Hz but also the harmonic overtone series for each, both probably generating the first 5th for the 880Hz (being the second 5th for the 440Hz). So the question in my mind is if added harmonics from distortion can interact when you have multiple instruments/sources to the degree where once you lower the output of the summed signal it's still noticeable.

It's a different test, but to me it's the more interesting one. Unless I'm thinking about it wrong.

 

[LTig]

I understand. My point was just that in a real-world scenario we're going to have multiple sources feeding each amp and they're all going to be uncorrelated from each other. If I recorded literally the exact same source 10 times and sum them then yeah, I'd need to make up for that by lowering the level to get back to the input level of one channel (assuming that was the desired level). But that's not a real world scenario.

Amps are linear devices (as long as they do not clip) which means that the superposition of several signals does not create additional signals. Since linearity is not perfect we measure IMD or use multitone signals to see what a real amp produces.

If I'm recording a drum kit, bass, electric guitar and electric piano all at the same time those are all different signals. Once summed the total level obviously will be much louder, but it's all uncorrelated sources. Therefore a more interesting experiment would be uncorrelated sources going through amps at different settings vs not going through them and comparing the sums in each scenario to each other.

I don't think this to be a good idea. A much better approach is to use Multitone signals. The single tones are generated explicitly such that the harmonics of each tone to not cover other tones at higher frequencies, so that everything between the tones is distortion of any kind (HD, IMD, noise). Otherwise you wouldn't be able to distinguish distortion of a lower tone from a higher tone.

I don't brain good so this is obviously possibly not right, but one part of my brain says that if you overdrive a couple of channels, given that they're playing in the same key and might play similar notes, some frequencies may be emphasized. For example one instrument playing an A 440Hz and another A 880Hz will lead to both generating not only 880Hz but also the harmonic overtone series for each, both probably generating the first 5th for the 880Hz (being the second 5th for the 440Hz). So the question in my mind is if added harmonics from distortion can interact when you have multiple instruments/sources to the degree where once you lower the output of the summed signal it's still noticeable.

Harmonics of the 440 Hz tone do add up with harmonics of the 880 Hz tone, but not as one might think. There is also the phase of the harmonics, so they can add up or cancel each other out.

It's a different test, but to me it's the more interesting one. Unless I'm thinking about it wrong.

IMV this is a test where the result is difficult to interpret. You need to do a null test to see what has changed or been added, and possible need to listen to it, if the signals are music.

 

[TK_IIVI]

Amps are linear devices (as long as they do not clip) which means that the superposition of several signals does not create additional signals. Since linearity is not perfect we measure IMD or use multitone signals to see what a real amp produces.

I wasn't talking about doing an actual measurement of the amp per se.

I don't think this to be a good idea. A much better approach is to use Multitone signals. The single tones are generated explicitly such that the harmonics of each tone to not cover other tones at higher frequencies, so that everything between the tones is distortion of any kind (HD, IMD, noise). Otherwise you wouldn't be able to distinguish distortion of a lower tone from a higher tone.

That might be interesting if you're measuring an amp the way you're thinking about it, but the fact remains that if we're talking about music the signals are going to be correlated because they're all being limited to existing in the same key.

Harmonics of the 440 Hz tone do add up with harmonics of the 880 Hz tone, but not as one might think. There is also the phase of the harmonics, so they can add up or cancel each other out.

Sure, and it's possible that some cases would be more "noticeable" than others, if at all.

IMV this is a test where the result is difficult to interpret. You need to do a null test to see what has changed or been added, and possible need to listen to it, if the signals are music.

At the most basic level what is difficult is probably setting up the test in a way that's "fair". To me, within the context of the video in the OP, the only thing that matters is if / how many people can tell a difference with some confidence.

In other words I'd say that if a test was set up reasonably then doing ABX testing or some other sort of blind testing on a decent sample size of people would be of value. If people continuously notice that there is a difference then the idea that "it sounds the same" falls apart.

 

[LTig]

In other words I'd say that if a test was set up reasonably then doing ABX testing or some other sort of blind testing on a decent sample size of people would be of value.

It is a typical approach of subjectivists to ask for measurements and auditions with real music. From a layman viewpoint this seems to make sense but psychoacoustic research has shown that human hearing is much worse in detecting flaws with music than with simple signals. Distortion for example is easier to detect with sine waves than with music. Non flat frequency response is easier to detect with pink noise than with music.

If people continuously notice that there is a difference then the idea that "it sounds the same" falls apart.

Absolutely yes. The point however is that no one has yet come up with a proper controlled listening test which demonstrated audible differences when measurements told that differences are inaudible. It is a fact that we can measure differences several orders of magnitude lower than human hearing can resolve. Distortion of any kind at -130 dB and frequency response deviations of 0.001 dB are possible to detect with an AP like @amirm has. Even the RME ADI-2 PRO can measure far better than human hearing is able to resolve.

 

[TK_IIVI]

First let me correct myself: I think I used the word "correlation" a couple of different ways. What I meant to say was that feeding a test the same signal multiple times in parallel or series is going to be different from feeding it different signals. In that sense I meant that different signals are uncorrelated. I think I then said that because music sources play in the same key the signals are correlated, but by that I simply meant that the harmonic series in all sources will be related to each other. A bit sloppy use of the word there. But I'm guessing you understood that.

Absolutely yes. The point however is that no one has yet come up with a proper controlled listening test which demonstrated audible differences when measurements told that differences are inaudible.

Well, I can't disagree with that since you just created something that can't be refuted; if it's inaudible of course no test will show that it's audible. I just feel like you're missing the point because you're arguing for something that is ultimately complimentary.

If you conduct a test that yields results that say "differences are inaudible" then what I propose would be a waste of time, assuming those test conditions can be reasonably applied to the real world.

If you conduct a proper blind listening test of real-world content that yields results where listeners continuously can tell A from B then differences must be audible, which means the other test has a different utility.

Therefore you can do either.

It is a fact that we can measure differences several orders of magnitude lower than human hearing can resolve.

I'm aware of that.

 

[Phos]

Keep in mind that Jim also compared what the Nieve sounded like when distorting compared to a purely digital hard cut off style of artificially induced distortion and to my ear they sound the same. It's in the video at 17:43. I think that's why he didn't feel the need to do an exhaustive survey of intentionally hot preamp usage.

 

[TK_IIVI]

Keep in mind that Jim also compared what the Nieve sounded like when distorting compared to a purely digital hard cut off style of artificially induced distortion and to my ear they sound the same. It's in the video at 17:43. I think that's why he didn't feel the need to do an exhaustive survey of intentionally hot preamp usage.

But again, that Neve preamp isn't the one people gush over when they're talking about vintage Neve, they're talking about class-A amps, typically also with an EQ circuit. It's simply not the same.

 

[drumphil]

I'm putting this on the essential videos list for all our sound production students. The world of music and sound production is disappointingly susceptible to the same kind of nonsense that audiophiles get sucked into, with people who should be better placed to reject such magical thinking.

The same guy does other interesting videos like "Where does the tone come from in a guitar amp" and "Where does the tone come from in a microphone", and "Where does the tone come from in a guitar amp. isolating and testing individual components, and combinations of components to get to the bottom of why things sound like they do, and what things do and don't matter for getting that result.

Well worth watching. I especially like the bit with the guitar where he gets down to a bridge, headstock, pickup, strings, and just open air in between!

 

[Redacted]

Yup. The guitar tone thing is essentially a myth. Fancy pickups are a myth (but singles can sound different to humbuckers). Tonewood is a myth. Tone does not 'come from the fingers' (but good playing does).

Any guitar with a bit of EQ and the right gain can be made to sound like any other - it's not magic, mystery or special sauce, it's really just simple coils and vibrating wires.

This is precisely why low-cost Chinese guitars have taken the market by storm, and the traditional manufacturers who seem to spend most of the cost of a guitar on marketing are getting very worried and litigious. Even details like quality hand-finishing on frets is better on Chinese guitars. I don't necessarily mean a $99 guitar from a grocery store of course. The bubble has burst!

I say hats off to the young players buying these guitars, not because they are cheaper, but also because they will tell you they are better made, and these young players are (frequently) not stuck with the old tosh their grandfathers believe in.

You may find their are some parallels in the world of HiFi gear....

 

[dfuller]

The guitar tone thing is essentially a myth. Fancy pickups are a myth (but singles can sound different to humbuckers).

This is just demonstrably not true... Though most of it is pretty easily measured with an LCR meter.

Guitar pickups are not super far off a speaker motor with a blocked voice coil. All the things that play into that have an effect here too. It makes complete sense to me that they can sound different based on changes in construction, wire gauge, number of wraps, how it's wrapped on the bobbin, bobbin dimensions, wire insulation, even baseplate material and whether or not there's a metal cover can have an effect. This is pretty much all reflected in the impedance curve, which, again, not linear. This all plays into the frequency response coming out of the instrument.

There's no magic. People just don't understand it well enough to understand that it isn't magic.

This is precisely why low-cost Chinese guitars have taken the market by storm

It's because they're cheap and good enough.

 

[Redacted]

The whole point of a pickup is that the inherent LCR of the thing forms resonant circuits - easily and provably recreated with EQ.... Simple enough electronics here!

 

[dfuller]

The whole point of a pickup is that the inherent LCR of the thing forms resonant circuits - easily and provably recreated with EQ.... Simple enough electronics here!

Yeah, but the way you manipulate things physically to change that resonant circuit is tricky. That's where the "magic" is. But what you're saying is there's no difference between two different humbuckers, which demonstrably isn't true.

And remember, EQ with guitar circuits is at best rudimentary and broad stroke.