• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Multitone test revisited: Level vs SINAD graphs?

ayane

Active Member
Forum Donor
Joined
Dec 15, 2018
Messages
182
Likes
684
Location
NorCal
Howdy! This question isn't a request or directed to anyone; it's just food for thought.

Most of the DACs tested in the past couple of years here all show the output of the APx555's 32-tone test signal in the review, which is fantastic because such a test immediately reveals flaws in the digital filter and various nonlinearities in the implementation. However, the test as it is still is pretty static since it only shows the output at one amplitude level. I was wondering if anyone has done this test swept through different levels, a la the SMPTE or CCIF IMD tests.

It'd be really interesting to see how the DACs perform not just at one output level, but at various different output levels. I think such a test would give greater insight into the performance and behavior of the DAC than the simple IMD test which only tests with two tones, and might carry over better to real-world performance since music signals are pretty complex and varied in amplitude. The X-axis can be the peak level of the signal, and the Y-axis can be SINAD - since this is not just a simple wave, the distortion products would include IMD, THD, and other kinds of distortions too. In essence, it's a worst-case scenario of SINAD.

Is such a test complex to do since it would require 32 notch filters, or is it feasible? If it is feasible, has anyone done this kind of test? Which products have been measured that way?
As a follow-up question, what is ASR's standard for measuring IMD? Is it SMPTE, CCIF (or what I call the "Bruno Putzeys IMD test" [Edit: it was actually Bruce Hofer!] with 18.5 kHz + 19.5 kHz), or something else entirely?
 
Last edited:
OP
ayane

ayane

Active Member
Forum Donor
Joined
Dec 15, 2018
Messages
182
Likes
684
Location
NorCal
Part of my motivation/rationale for asking this question is to get a deeper understanding of the infamous "IMD hump" present in many implementations of ESS DAC chips. How would this hump look if the IMD test was a multitone signal instead of a simple two-tone signal?
 

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,521
Likes
37,050
Amir does a 32 tone test in all of his DAC reviews. Like this one from a recent SMSL dac.

index.php
 
OP
ayane

ayane

Active Member
Forum Donor
Joined
Dec 15, 2018
Messages
182
Likes
684
Location
NorCal
@Blumlein 88 I'm already aware of that:
Most of the DACs tested in the past couple of years here all show the output of the APx555's 32-tone test signal in the review
and those graphs are irrelevant to my question. I'm asking the following:
It'd be really interesting to see how the DACs perform [on the multione test] not just at one output level, but at various different output levels.
 

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,521
Likes
37,050
You also could download pkanes Multitone software and do your own tests to see what happens.

 

restorer-john

Grand Contributor
Joined
Mar 1, 2018
Messages
12,579
Likes
38,274
Location
Gold Coast, Queensland, Australia
I don't think anything interesting is to be found at lower levels to justify doing them in every review.

Considering THD in digital falls off a cliff as the level goes down, it's kind of important. THD at 0dBFS is an absolute best case number. Most content people actually listen to is 16bit and Amir doesn't even look at that- he's feeding the D/A converters 24bit.
 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,368
Likes
234,381
Location
Seattle Area
Part of my motivation/rationale for asking this question is to get a deeper understanding of the infamous "IMD hump" present in many implementations of ESS DAC chips. How would this hump look if the IMD test was a multitone signal instead of a simple two-tone signal?
Why would you need another test when the current IMD test shows the problem clearly?

But no, there is no way to compute a value from 32-tone graph. You would have to write your own program to run the test and gather such data.
 

restorer-john

Grand Contributor
Joined
Mar 1, 2018
Messages
12,579
Likes
38,274
Location
Gold Coast, Queensland, Australia
Why would you need another test when the current IMD test shows the problem clearly?

Your AP's standard SMPTE IM test is 60Hz/7kHz at 4:1, a 12dB difference unless you've changed the respective levels.

Is it SMPTE, CCIF (or what I call the "Bruno Putzeys CCIF" with 18.5 kHz + 19.5 kHz), or something else entirely?

The CCIF style test is used for bandwidth limited devices where the distortion products would otherwise lie outside the measurement bandwidth.
 
OP
ayane

ayane

Active Member
Forum Donor
Joined
Dec 15, 2018
Messages
182
Likes
684
Location
NorCal
Why would you need another test when the current IMD test shows the problem clearly?

But no, there is no way to compute a value from 32-tone graph. You would have to write your own program to run the test and gather such data.
I guess I was wondering if the rise in IMD is creating patterns which depended on frequency. What if the hump goes away at some pairs of frequencies but comes back worse at others?

I guess I could play with such things and post my findings here using my RME ADI-2 Pro, but I don't have any DACs to test it with =P
 
OP
ayane

ayane

Active Member
Forum Donor
Joined
Dec 15, 2018
Messages
182
Likes
684
Location
NorCal
Your AP's standard SMPTE IM test is 60Hz/7kHz at 4:1, a 12dB difference unless you've changed the respective levels.
[...]
The CCIF style test is used for bandwidth limited devices where the distortion products would otherwise lie outside the measurement bandwidth.
Thank you!
 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,368
Likes
234,381
Location
Seattle Area
I guess I was wondering if the rise in IMD is creating patterns which depended on frequency. What if the hump goes away at some pairs of frequencies but comes back worse at others?
I can change the frequency pairs to any value.
 

Soniclife

Major Contributor
Forum Donor
Joined
Apr 13, 2017
Messages
4,499
Likes
5,417
Location
UK
I guess I could play with such things and post my findings here using my RME ADI-2 Pro, but I don't have any DACs to test it with =P
You can use the pro to measure itself, see if anything interesting presents in a quality product. I'm sure you have some other product in the house with a DAC inside it that could be tested, an old phone for instance.
 

boXem

Major Contributor
Audio Company
Joined
Jun 19, 2019
Messages
2,014
Likes
4,852
Location
Europe
OP
ayane

ayane

Active Member
Forum Donor
Joined
Dec 15, 2018
Messages
182
Likes
684
Location
NorCal
I'm fascinated by the amount of discourse that has already happened/is happening on this topic, but I'm not surprised considering this is ASR, after all, haha!
Thank you to all who've replied! I know what rabbit hole I'm going down this weekend!
 

morillon

Major Contributor
Joined
Apr 19, 2022
Messages
1,336
Likes
263
I'm going to sound stupid but I've only just listened to this famous ap multitones ...
I just realize that if it is really a most useful approach... the sound produced,( necessarily close to an organ sound), is not so complex...
I should have realized it :-0
will try to produce others thanks to the famous multitone analyzer ( ;-) ) or rew ... more complex ...
you have the experience of the interest of multitones at more than 100 tones ?
advice, comments
?
;-)
 

pkane

Master Contributor
Forum Donor
Joined
Aug 18, 2017
Messages
5,626
Likes
10,202
Location
North-East
Howdy! This question isn't a request or directed to anyone; it's just food for thought.

Most of the DACs tested in the past couple of years here all show the output of the APx555's 32-tone test signal in the review, which is fantastic because such a test immediately reveals flaws in the digital filter and various nonlinearities in the implementation. However, the test as it is still is pretty static since it only shows the output at one amplitude level. I was wondering if anyone has done this test swept through different levels, a la the SMPTE or CCIF IMD tests.

It'd be really interesting to see how the DACs perform not just at one output level, but at various different output levels. I think such a test would give greater insight into the performance and behavior of the DAC than the simple IMD test which only tests with two tones, and might carry over better to real-world performance since music signals are pretty complex and varied in amplitude. The X-axis can be the peak level of the signal, and the Y-axis can be SINAD - since this is not just a simple wave, the distortion products would include IMD, THD, and other kinds of distortions too. In essence, it's a worst-case scenario of SINAD.

Is such a test complex to do since it would require 32 notch filters, or is it feasible? If it is feasible, has anyone done this kind of test? Which products have been measured that way?
As a follow-up question, what is ASR's standard for measuring IMD? Is it SMPTE, CCIF (or what I call the "Bruno Putzeys IMD test" [Edit: it was actually Bruce Hofer!] with 18.5 kHz + 19.5 kHz), or something else entirely?

You can use my Multitone software for this. There is a sweep option that uses an arbitrarily complex signal, including multitone to do the sweeping. For example, you could use a square or triangular wave, plus you can construct a custom test using any combination of frequencies and amplitudes.

While Multitone doesn't compute IMD for a 32 tone signal, it does do it for up to 5-6 tones. It also computes the TD+N for any multitone signal, from 2 to 1000. Here's a sweep with 32 tone test signal:

1655322276114.png


And here's one with 5 tones, showing IMD and TD+N vs level:
1655322642516.png
 

dc655321

Major Contributor
Joined
Mar 4, 2018
Messages
1,597
Likes
2,235
It also computes the TD+N for any multitone signal, from 2 to 1000.

Are you just putting the tones in the center of each fft bin, then subtracting the tones’ (known) energy from each bin on analysis ?
 
Top Bottom