Need a MultiTone test based on actual music

[SHB]

It wasn’t until I saw the testing done at ASR that I ever saw a multitone test published in an equipment review. Obviously, it’s extremely valuable in assessing the distortion characteristics of electronics, but there is one problem: it’s not based on music.

As it stands, there are two problems:
• There are too many tones in the bottom 3 octaves
• The energy of the multiple tones is not equal in each octave

Problem 1:
In any musical genre, there is never so much going on in the bottom three octaves as is done here. It’s cacophony. As a bass player, I’d never want to hear this played back.

Even with two bass lines going simultaneously, you’d have one an octave above the other.

Any percussion in this region is there and gone so fast that its harmonic content isn’t a large consideration. Also, even large kick drums can’t be tuned so low that their fundamental is much below 80Hz if you want a nice snap to their transient that cuts through the mix. I’ve heard kicks that had a substantial fundamental around 30Hz (a bass low B note), but I’m pretty sure the audio engineer was running a harmonizer or sub-bass synthesizer to get that.

Also, other than pipe organ and some very rare synth-based albums (or a few bass players a nuts as me), the lowest notes you’re going to find in music will be a low B (30.5Hz).

The solution:
The 1st octave should have a single 30Hz tone. The 2nd octave should have a single 60Hz tone. The 3rd octave should have a tone at 80Hz (about a low E on a guitar) and 110Hz (an A note).

Once you have that out of the way, you should continue with 3 tones per octave from 160Hz going up the scale.


Problem 2:
In both noise & modern music, every octave has twice the audio energy of the octave below it and half as much as the octave above regardless of where that octave is. That’s music, baby. 110Hz, 220Hz and the famous 440Hz are all A notes.

To have equal energy per octave, we use pink noise which has a -3dB/octave slope going up in frequency for testing.

I’ve played a lot of music through BlueCat’s FreqAnalyst real-time analyzer plugin and most music is basically engineered to follow that -3dB/oct. slope.

That’s the way humans perceive sound. It’s also the way audio is reproduced. Higher frequencies have higher energy. This is true for both sound and radio waves (yes, this is a simplified explanation).

The solution:
Create a multitone signal with a -3dB/oct. slope.

• 30Hz @0dB
• 60Hz @-3dB
• 80Hz @-4.5dB
• 110Hz @-6dB
• 160Hz @-7.5dB

Above 160Hz, you have 3 tones per octave (21 of them) so, with increasing frequency, every tone is 1dB lower than the one below it. That gives you equal energy per octave.

This will present the amp (thinking of this more for amps, but it’s appropriate anywhere) with an energy spectrum like music.

My final thought is that this multitone test be run at 1/2 or 2/3 an amp’s rated power. Current multitone tests are run at 5 watts. That’s appropriate for what a tweeter will be getting in its top octave, but has little to no bearing upon what an amp will be sending to a full-range speaker during a serious listening session.

The difference in amp power necessary to reproduce a 30Hz note with the same amount of perceived acoustic power as a 16kHz harmonic is at least 20dB (100x… maybe more). Again a simplification, but it’s generally true.

This would give readers a sense of what an amp will do when playing actual music.

 

[voodooless]

You’ve got this backwards. Real music has a downward-sloping spectrum — bass dominates energy. Higher frequencies do not carry more energy. And multitone tests are intentionally inharmonic and unmusical so IMD products land at new frequencies and can be seen clearly. Making the stimulus “music-like” would actually hide distortion, not reveal it.

The multitone was never meant to be music-like. It only looks that way in the time domain, precisely because it’s inharmonic, and therefore will not show repeating waveforms.

 

[SHB]

You’ve got this backwards. Real music has a downward-sloping spectrum — bass dominates energy. Higher frequencies do not carry more energy. And multitone tests are intentionally inharmonic and unmusical so IMD products land at new frequencies and can be seen clearly. Making the stimulus “music-like” would actually hide distortion, not reveal it.

The multitone was never meant to be music-like. It only looks that way in the time domain, precisely because it’s inharmonic, and therefore will not show repeating waveforms.

I stated pink noise has a -3dB/oct. slope. with increasing frequency.

So a 1 inch tweeter can produce as much bass as a 15 inch sub?

A speaker diaphragm hardly has to move and can be tiny to reproduce a 16kHz tone at 100dB, but to accomplish the same feat at 30Hz requires huge excursions, radiation area and power outlay.

That is the very definition of higher energy at higher frequencies.

 

[voodooless]

You’re now talking about speaker efficiency and diaphragm excursion, not signal energy. That’s a different domain.

An amplifier driving a load delivers electrical power independent of whether the downstream transducer is a tweeter or a woofer. A 30 Hz and a 16 kHz tone at the same RMS voltage deliver the same electrical power into the amp’s load.

The fact that bass requires more cone excursion to produce equal SPL is a speaker efficiency issue, not evidence that higher frequencies contain more energy and it’s irrelevant when discussing amplifier distortion tests.

I can perfectly well design a 100dB@1W 30 Hz subwoofer. It will produce the exact same SPL as the tweeter with the exact same input power. It’s just going to be big and ugly The amp will never know, though.

 

[solderdude]

The solution:
Create a multitone signal with a -3dB/oct. slope.

• 30Hz @0dB
• 60Hz @-3dB
• 80Hz @-4.5dB
• 110Hz @-6dB
• 160Hz @-7.5dB

See my thread here (I started 9 months ago):

multitone spectrum proposal

I know my proposal is not a standard and maybe even nonsensical but here it goes. Did some dgging into spectra of music in AES papers and other research (also from MQA Bob) and found quite a lot of data on the spectra of music. Even across various genres which do have some differences but aside...

www.audiosciencereview.com

This would be the most 'music alike' multitone.
You can even use it on speakers as it has a 'pink noise' spectrum.

The current multitone, however, paints the 'worst case scenario' for electronics.
Not recommended for speakers but can be used with headphones but would look really messy.

 

[Sokel]

REW Fast Subband Adaptive Filtering (FSAF) measurement

I just discovered this new measurement option in REW. AV Nirvana Thread REW Guide I have not been able to study enough yet to understand what it is for and what advantages it can provide compared to the sweep. It would seem that it can produce valid TD+N measurements for complex signals such...

www.audiosciencereview.com

 

[solderdude]

Real music has a downward-sloping spectrum — bass dominates energy.

That's what OP states too.

 

[kemmler3D]

I think the point of the multi tone currently in use is to make it easy to see IMD and other types of distortion with a complex signal. I don't think the point is to see how the device behaves with actual music.

For that we could use pink noise or actual music, and Deltawave.

 

[Skeptical_Aus]

Audio Science Review - the name says it all. Making devices to reproduce music is actually making them to reproduce sound. Music is only one kind of acoustic phenomenon. The methodology employed by Amir is valid.

 

[Pow1]

I think one thing that has not been addressed by the OP (or I might have missed) is what the “more musical” multi tone test attempts to do or fix over the current test.

Does it reveal something that the current test doesn’t? Is the current test too harsh on equipment that the testing methodology needs to change?

At the moment I’m not sure what the “more musical” multitone test would reveal that the current one doesn’t. And the current test has the advantage of having been in use for some years now across multiple equipment. Unless the test is flawed in some way, I’m not sure if “not musical enough” is sufficient reason to change it. One might argue for it to be added to the existing suite of tests, though again I’m not sure what we might gain from it that the current multitone test doesn’t show us

 

[RexrothPigeon]

I think one thing that has not been addressed by the OP (or I might have missed) is what the “more musical” multi tone test attempts to do or fix over the current test.

Does it reveal something that the current test doesn’t? Is the current test too harsh on equipment that the testing methodology needs to change?

At the moment I’m not sure what the “more musical” multitone test would reveal that the current one doesn’t. And the current test has the advantage of having been in use for some years now across multiple equipment. Unless the test is flawed in some way, I’m not sure if “not musical enough” is sufficient reason to change it. One might argue for it to be added to the existing suite of tests, though again I’m not sure what we might gain from it that the current multitone test doesn’t show us

This is the actual rub: using audio that most people consider music* for these tests would be less revealing and informative than the using test signals.

*(after all, some music consists of nothing more than an unaccompanied sine wave)

 

[Hayk]

In solid state audio ampliers where the output stage carry high currents, unlike tube amps, electromagnetic radiation causes unwanted feedback which can be positive or negative. The rails particularly are half wave rectified causes the main trouble.
If 2 waves are used to measure the IMD, it doesn't show any wrong but multiwaves can.
There are many songs that sound dirty, although the amp measure extremely low THD, IMD. The simplest I use is the Happy together by the Turtles and the most difficult is the Dies Irae of Verdi.
It is possible to reduce the multi tone to 3 frequencies, a pair with 1khz difference in high range and a third very low about 10 hz with amplitude more the the sum of the other 2.

 

[Skeptical_Aus]

Including John Cage 4'33" - sorry, couldn't resist. Get the all acoustic elements optimised and the music takes care of itself. Audio is more than music anyway - watch a movie and pay attention to what you hear.

 

[voodooless]

If 2 waves are used to measure the IMD, it doesn't show any wrong but multiwaves can.

Any amp measurement done on ASR uses multi-tone measurements. If there were such a thing it would surely show up there.

There are many songs that sound dirty, although the amp measure extremely low THD, IMD. The simplest I use is the Happy together by the Turtles and the most difficult is the Dies Irae of Verdi.

You can just record the songs from the amp terminals and use DeltaWave to compare to the original. That should tell you where the differences can be found.

 

[oleg87]

I’m not really sure what this tells you that Amir’s usual multitone doesn’t - tweaking the amplitude of the components seems like adding a variable to the test for no particular reason. Why not just assume an amp might be called to reproduce the more demanding signal?

Also, you don’t want the components to be harmonics of each other, for obvious reasons in a harmonic distortion test.

 

[Hayk]

Any amp measurement done on ASR uses multi-tone measurements. If there were such a thing it would surely show up there.

You can just record the songs from the amp terminals and use DeltaWave to compare to the original. That should tell you where the differences can be found.

Thank you very much for the Deltawave, very useful tool specially to measure transient character using musical instruments sounds as piano or drums instead of sine waves. Unluckily, my windows based PC, tablet are 32 bit.
Standard way of measuring IMD is listed bellow.
The problem with electromagnetic radiation distortion can be resolved very easily as it is 100% PCB issue. The rail tracks, I solder upon them magnetic wires. Short ones for chips I use capacitor leads and for long tracks, acoustic guitar, brass wound strings.

 

[solderdude]

I think one thing that has not been addressed by the OP (or I might have missed) is what the “more musical” multi tone test attempts to do or fix over the current test.

Does it reveal something that the current test doesn’t? Is the current test too harsh on equipment that the testing methodology needs to change?

At the moment I’m not sure what the “more musical” multitone test would reveal that the current one doesn’t. And the current test has the advantage of having been in use for some years now across multiple equipment. Unless the test is flawed in some way, I’m not sure if “not musical enough” is sufficient reason to change it. One might argue for it to be added to the existing suite of tests, though again I’m not sure what we might gain from it that the current multitone test doesn’t show us

One advantage could be that one can use the 'more musical' version to test loudspeakers without blowing up the tweeters.
Another difference could possibly be seen in the multitone IM products being lower (as they would be with music) as most IM garbage happens at higher frequencies and as these are lower the IM products created by higher frequencies will be lower too.

The current multitone is more IM-products revealing for electronics and that's what we want to know.

 

[solderdude]

It is possible to reduce the multi tone to 3 frequencies, a pair with 1khz difference in high range and a third very low about 10 hz with amplitude more the the sum of the other 2.

Sure possible to make but it might only reveal something when the output stage reaches near the max. output voltage and only for amplifiers that do not have a not regulated power supply with the driver stage not being regulated.
That will also be captured by the currently used multitone though.... at clipping levels.

Many other test signals could be conceived to test for certain aspects in certain conditions b.t.w.

 

[voodooless]

One advantage could be that one can use the 'more musical' version to test loudspeakers without blowing up the tweeters.

At the regular 5W, I see very few issues with blowing up tweeters. The 5W is over the whole spectrum, so the tweeter will get a fraction of that. SPL will be around 96 dB/1m (86 dB efficiency assumed), which will already be loud as f*ck Erin certainly has little issue using a standard multitone on speakers, and neither does Joseph Crow, though visualization is slightly different. The latter looks similar to what we have for electronics:

Amir doesn't do it for speakers because he feels that the data will just raise more questions than it answers. I still think it is a very nice additional data set to have.

 

[solderdude]

It is an unrealistic level for tweeters in music conditions.

For the same reason (but arguably there is a lot more power then) one tests speakers or adjusts EQ using pink noise and not white noise.
White noise at high levels also isn't particularly pleasant to be around the same goes for multitone at high levels.

Yes, for testing at 5W the regular multitone is fine (tweeter would see around 0.5W).
Would not recommend to test multitone with speakers near full power in >50W amps though as that tone is continuous where music isn't.