Is ASR missing the point? Why still focus on Dacs?

[pos]

I have done this a few times. It always shows what you can predict from the rest of the measurements since it is a linear parameter.

I would not agree on that.
EQ hardware needs to be tested with EQ applied.
In this situation a typical IIR EQ hardware will introduce both linear distortion compared to the theoretical target (EQ cramping when approaching Nyquist), and non-linear distortion (32bit float rounding errors).
Both are problematic and should be measured.

 

[amirm]

Both are problematic and should be measured.

Measured how? Anything that impacts the spectrum can cause errors in the measurement. Say you put a high pass filter. Now all low frequency noise goes away, causing measurements like SINAD to be exaggerated. Conversely, if you boost the generator frequency, you would be creating more distortion. Much of our electronics measurement relies on flat frequency response.

If you mean signal processing errors, then that is beyond the scope of what I do unless it manifests itself as is.

 

[Keith_W]

In this situation a typical IIR EQ hardware will introduce both linear distortion compared to the theoretical target (EQ cramping when approaching Nyquist), and non-linear distortion (32bit float rounding errors).

Umm, those are mathematical errors and not device errors.

 

[Sokel]

Umm, those are mathematical errors and not device errors.

They may be, but if a piece of gear presents flawless results while another presents a mess down low (with the same filters applied) it's device's error too.

 

[pos]

Measured how? Anything that impacts the spectrum can cause errors in the measurement. Say you put a high pass filter. Now all low frequency noise goes away, causing measurements like SINAD to be exaggerated. Conversely, if you boost the generator frequency, you would be creating more distortion. Much of our electronics measurement relies on flat frequency response.

If you mean signal processing errors, then that is beyond the scope of what I do unless it manifests itself as is.

Two things need to be measured:
1- linear distortion: how does the EQ response (both magnitude and phase) compare to the ideal theoretical target. This could be done by inversing the response for a given set of EQs, or simply by overlaying the responses of several EQ points over their theoretical curves.
2- non-linear distortion: does applying an EQ (within the headrooms limits) add distortion. This can even often be tested by applying two EQ cancelling each other resulting in a flat response.

This is very much the responsibility of the device to produce accurate EQs, matching the target and not introducing non linear distortion.
Some device do this properly, some do not.

To mitigate point 1 most devices rely on a higher sample rate (eg 96kHz), but this can also be addressed by generating "decramped" biquad parameters. Both solution have their pro and cons and can be combined to approach the theoretical target.
Point 2 can be mitigated by either using a higher bitdepth (64bit float, 48bit fixed, etc.) or by choosing a proper biquad algorithm.
Both points can also be addressed by using FIR instead of IIR when applicable,n introducing other forms of linear distortion on the other side of the spectrum.

All this to say that yes this can be measured, yes this is the responsibility of the device because yes this can be addressed in many different ways, and yes the differences between devices in this regard can be major and should be measured.

 

[DanielT]

I have only done one. I am not setup to test them. With it raining here most of the year, outdoor testing is not a good approach for me to say nothing of the weight of these beasts. I am also not a fan of current protocols for measuring them as I feel they lack any foundation in proper listening tests. I have some ideas to deal with this but nothing concrete yet.

I think that others should do the subwoofer measurements, so you can concentrate on other things.

You don't need a Klippel for subwoofer measurements. You just need a few willing hands carrying subwoofers, an open field and good weather. Like this:
(plus measuring equipment, knowledge to measure)

'I measured 10 subwoofers.'

Oct 28, 2023

Hi, it's been a while!
I would like to thank the people who made this project possible on a scale never before imagined in South Korea. I measured a total of 10 subwoofers, including my own DIY subwoofer(L26RO4Y), and measured them to the CEA-2010 standard using a Klippel TBM module. Below are photos and a list of the subwoofers I measured.

(In order from left to right)

- Arendal 1961 1V

- Arendal 1961 1S

- Neumann KH750

- Bowers & Wilkins DB4S

- SVS SB-3000

- SVS SB-2000 pro

- SVS SB-1000 pro

- Polk audio...

• Nuyes
• arendal 1961 1s arendal subwoofer bowers and wilkins sub woofer cea-2010 diy subwoofer neumann subwoofer sb1000 sb2000 sb3000 svs subwoofer
• Replies: 330
• Forum: Speaker Reviews, Measurements and Discussion

• 

 

[DanielT]

I do not have a large backlog of speakers. Fewer come in these days that used to. I also used to buy a lot of them but almost all turned into duds, sitting here gathering dust.

If people sent in more, you would see more of them measured.

But what is wrong with people, Americans, who are active here at ASR and have a golden opportunity to send you speakers and DON'T do it?!

With big heavy floor-standing speakers there I see logistical send in packages problems and challenges with lifting them up on Klippel. But small bookshelf speakers, more tests of such.

Edit:
Okay, even ASR members living outside the US can send small speakers to you, even though it's more cumbersome, but there should be enough people interested in doing so living in the US.

 

[pos]

Umm, those are mathematical errors and not device errors.

Those are implementation-implied errors

 

[Sokel]

Those are implementation-implied errors

Yep, we have an example thread:

M

'DSP Measurements and Rising Noise Floor'

Feb 25, 2023

In several threads related to DSP platforms I've seen requests for measurements with filters engaged. I thought it unlikely that implementing filters in a DSP would have measurable effect on noise / distortion other than the obvious change in the fundamental, however @Sokel recently mentioned that he returned his miniDSP Flex after observing noise / distortion issues with filters applied. This piqued my interest as a long time DSP user and I decided to explore it for myself.

The first issue is what filters to use and what tests to run. I decided to use filters for my LXmini + sub...

• mdsimon2
• Replies: 200
• Forum: General Audio Discussions

• 

This very thread had miniDSP apply a partial correction which addressed some issues.

 

[Mnyb]

Those are implementation-implied errors

+1 several hardware devices show this, compared to theoretical results or decent PC software who does this on par with the theory as far as anyone can measure ?

@Sokel mentions just one tread it's common in many DAC's and Streamers and standalone DSP units ? I think mini DSP alleviated most of it by some improved firmware

Amir could apply his roon* EQ then implement the same in the test hardware and compare ?

*roon sems to have a decent implementation of EQ and filters .


All in all sloppy design is still possible so keep on testing mfg should be kept accountable . There are still surprises in many of Amirs and others reviews off thing going sideways where mfg should know better .

 

[jkim]

Two things need to be measured:
1- linear distortion: how does the EQ response (both magnitude and phase) compare to the ideal theoretical target. This could be done by inversing the response for a given set of EQs, or simply by overlaying the responses of several EQ points over their theoretical curves.
2- non-linear distortion: does applying an EQ (within the headrooms limits) add distortion. This can even often be tested by applying two EQ cancelling each other resulting in a flat response.

This is very much the responsibility of the device to produce accurate EQs, matching the target and not introducing non linear distortion.
Some device do this properly, some do not.

To mitigate point 1 most devices rely on a higher sample rate (eg 96kHz), but this can also be addressed by generating "decramped" biquad parameters. Both solution have their pro and cons and can be combined to approach the theoretical target.
Point 2 can be mitigated by either using a higher bitdepth (64bit float, 48bit fixed, etc.) or by choosing a proper biquad algorithm.
Both points can also be addressed by using FIR instead of IIR when applicable,n introducing other forms of linear distortion on the other side of the spectrum.

All this to say that yes this can be measured, yes this is the responsibility of the device because yes this can be addressed in many different ways, and yes the differences between devices in this regard can be major and should be measured.

+1 several hardware devices show this, compared to theoretical results or decent PC software who does this on par with the theory as far as anyone can measure ?

@Sokel mentions just one tread it's common in many DAC's and Streamers and standalone DSP units ? I think mini DSP alleviated most of it by some improved firmware

Amir could apply his roon* EQ then implement the same in the test hardware and compare ?

*roon sems to have a decent implementation of EQ and filters .


All in all sloppy design is still possible so keep on testing mfg should be kept accountable . There are still surprises in many of Amirs and others reviews off thing going sideways where mfg should know better .

+2 Exactly. The quality of actual DAC products cannot be separated from their implementation of DSP of some sort. In fact, the phenomenon of DRE-induced nonlinear distortion is nothing but digital processing artifacts. As I mentioned earlier:

DAC is a solved problem only if it is properly implemented. More and more DAC products keep adopting the Cirrus Logic CS431xx and other custom chips that exhibit peculiar distortion behavior if not properly dealt with in a firmware design. Industry is beginning to realize the issue and provide solutions.

And more and more products bundle PEQ with DAC, but based on my testing so far, their PEQ performance is often suboptimal with respect to accuracy and nonlinear distortion.

So, at least to me, there are reasons why we want to keep testing DACs.

The current increasing trend of adding onboard PEQ to DAC-contained devices (DACs, streamers, etc.) should be seriously considered in measurements-based reviews.

 

[Power Pop 23]

I have only done one. I am not setup to test them. With it raining here most of the year, outdoor testing is not a good approach for me to say nothing of the weight of these beasts. I am also not a fan of current protocols for measuring them as I feel they lack any foundation in proper listening tests. I have some ideas to deal with this but nothing concrete yet.

Consider the pros and cons of trying out a different method of procuring gear for measurement. By that I mean arranging for a 'roundup' of same (or very similar) audio gear to be measured at scheduled times of the year. I recall Stereo Review would occasionally supplement their equipment reviews with a similar focus.

This could allow for lead time for member input as to which gear to include in a roundup and a quicker turnaround time for member submissions to a roundup.

Perhaps planning for a roundup of subwoofers to measure (without bodily injury) in late Summer 2026 would allow enough lead time to develop protocols for measuring subwoofers.

off-topic - all my talk of 'roundup' has me now listening to a stream of Frankie Laine's cover of Rawhide by Dimtri Tiomkin and Ned Washington - with a conductor credit to 'Johnny Willams' aka composer John Willams