• 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!

DacMagicPlus Filter Mirror Images - Short Study

pma

Major Contributor
Joined
Feb 23, 2019
Messages
4,592
Likes
10,728
Location
Prague
DAC filters have been discussed in several threads now. I have decided to measure responses of filters in DacMagicPlus especially with respect to mirror images, that create beats at difference frequency and possible IMD tones in inevitably nonlinear components of the audio chain like speakers.

23 kHz sine sampled at Fs = 48 kHz was played through the DacMagicPlus and captured with 48 kHz bandwidth. Three filters of the DacMagicPlus were used, the Linear Phase filter, the Minimum Phase filter and the Steep filter. They are switchable by a button on the front panel. Then the frequency response to a white noise was measured to see the behavior in the passband and near Fs/2.


1. Linear Phase filter

filter_linphase.png

See 25kHz mirror image about 42dB below the original 23kHz signal

filter_linphase_white.png

Response to white noise see low attenuation above Fs/2 = 24kHz


2. Minimum Phase filter

filter_minphase.png

Again we can see almost same mirror image

filter_minphase_white.png

Response to white noise and again low attenuation above Fs/2


3. Steep filter

filter_steep.png

There is an attenuation of the 23kHz sine, however no mirror image

filter_steep_white.png

Response to white noise and good attenuation above Fs/2

Conclusion

The only filter that prevents mirror images is the steep filter. The decay below Fs/2 starts earlier than in the first two filters, however this would be inaudible. The benefit of good attenuation below Fs/2 is much higher than the small trade off in the passband.
 
Last edited:

Filtersweep

Member
Joined
Jan 21, 2021
Messages
8
Likes
8
Thanks for posting this and putting in the work to perform the measurements. I was thinking about dac filters reading the D30 pro review noting the attenuation at 22 kHz was quite low on all of the filters. I think a steep filter may have been promised for the final release.
I have the dacmagic plus myself and stick with the steep filter. I haven’t tried picking the filter differences blinded but the sounds of the high frequency components and overall separation of individual sounds is subjectively better to me on the steep filter.
Am I fooling myself or would the differences you’ve shown be audible in the way I described?
 
  • Like
Reactions: pma
OP
pma

pma

Major Contributor
Joined
Feb 23, 2019
Messages
4,592
Likes
10,728
Location
Prague
I have the dacmagic plus myself and stick with the steep filter. I haven’t tried picking the filter differences blinded but the sounds of the high frequency components and overall separation of individual sounds is subjectively better to me on the steep filter.
Am I fooling myself or would the differences you’ve shown be audible in the way I described?

IMHO you might be quite right, though it would be very difficult to make a proof in a DBT. I can just tell you that I use a setting to the Steep Filter as well and my impression on classical music is that it has the best definition and resolution of instruments sound especially when many of them are playing together.
 

bennetng

Major Contributor
Joined
Nov 15, 2017
Messages
1,634
Likes
1,692
OP
pma

pma

Major Contributor
Joined
Feb 23, 2019
Messages
4,592
Likes
10,728
Location
Prague
The price is only 80-ish attenuation depth, but both filters are only 80-ish anyway.

The parameter that counts is the attenuation below Fs/2, in our case 20-24kHz band and the steep filter is much better here, as we can see from mirror images. The absolute attenuation level in the rejection area is not that much of interest. Just race of numbers. On the other hand, attenuation at Fs/2 matters, even if some designers are not willing to admit.
 

bennetng

Major Contributor
Joined
Nov 15, 2017
Messages
1,634
Likes
1,692
as we can see from mirror images
Sure, why not.

@pma
Let's read how others think when they actually hear these images, rather than simply looking at graphs.
https://hydrogenaud.io/index.php?topic=120491.0

Not that I would extend this argument to something like MQA filters or NOS, but minor imaging like these half-band filters are not really that offensive even in the audible band.
 
Last edited:
OP
pma

pma

Major Contributor
Joined
Feb 23, 2019
Messages
4,592
Likes
10,728
Location
Prague
View attachment 112082

The price is only 80-ish attenuation depth, but both filters are only 80-ish anyway.

Different products have different implementations though.
https://www.audiosciencereview.com/...-d30pro-review-balanced-dac.20259/post-668061
https://www.audiosciencereview.com/...d90-balanced-usb-dac-review.10519/post-291483

Without DBT it is just another "looking at graphs" exercise.


The WM8740 plots are rather pointless here, DM+ uses Analog Devices ADSP21261 DSP digital filters. However it was not the main point here. My main point is to prevent mirror images.

1613139763815.png
 
Last edited:

bennetng

Major Contributor
Joined
Nov 15, 2017
Messages
1,634
Likes
1,692

mansr

Major Contributor
Joined
Oct 5, 2018
Messages
4,685
Likes
10,700
Location
Hampshire
Because you see it, right? Did you read the link I posted, in the case when the mirror images fall into the audible range?
If everything else is ideal, there is no problem. However, a 23 kHz tone can intermodulate with its 25 kHz image to produce a very audible 2 kHz tone.
 

bennetng

Major Contributor
Joined
Nov 15, 2017
Messages
1,634
Likes
1,692
If everything else is ideal, there is no problem. However, a 23 kHz tone can intermodulate with its 25 kHz image to produce a very audible 2 kHz tone.
That's why I pointed to a discussion about polynomial resampling which is popular in game consoles, samplers, expensive keyboard workstations and so on. If one must use a high amplitude, near Nyquist tone as test signal then things are pretty much expected. I wonder why very few people complained about such things because many published music are produced in this way, on the other hand listeners often complain things like loudness war and such since they can clearly see a squashed waveform.
 
OP
pma

pma

Major Contributor
Joined
Feb 23, 2019
Messages
4,592
Likes
10,728
Location
Prague
Because you see it, right? Did you read the link I posted, in the case when the mirror images fall into the audible range?
I even listened to the correct filters with even better steepness and attenuation than what you posted, I still hear no difference.
https://www.audiosciencereview.com/...le-review-measurements-also.11868/post-398661

Not because I see it, but because it is the right way to do so. To do so in order to achieve proper signal reconstruction. And it is not my problem what someone hears or not. I want the technical things to be done properly and I do not like excuses like “it does not matter”.

For example in case of these filters all of them are wrong.

63C1CA5D-3926-45E2-A7F7-3484FE438ED4.png
 
Last edited:

bennetng

Major Contributor
Joined
Nov 15, 2017
Messages
1,634
Likes
1,692
Not because I see it, but because it is the right way to do so. To do so in order to achieve proper signal reconstruction. And it is not my problem what someone hears or not. I want the technical things to be done properly and I do not like excuses like “it does not matter”.
I think you did not really carefully read the link I posted. People heard the differences when the imaging falls into the audible range, and some of them still prefer the sound with non DSP perfect algorithms. In the case of 44.1k+ materials I heard no difference. So the conclusion is when people really heard them they have mixed opinions, when people cannot hear them it is nothing more than bragging rights, just like people posting Monty's video to automatically get likes. Of course they also deleted the article with test samples using somethings like 25 and 28kHz (or some other combinations) creating IMDs at 3kHz to justify why 192kHz can degrade listening experiences, which is plain silly and unrealistic.
 

KSTR

Major Contributor
Joined
Sep 6, 2018
Messages
2,690
Likes
6,013
Location
Berlin, Germany
I think the best way to evaluate audibility of DAC filters is to off-load them to a software filtering & upsampling preprocessing stage with, for example the SoX utility, to 8x or 16x. Provided the DAC can handle 8x and 16x rates (352.8kHz and above). Or just use a upsampling plugin for foobar2000 etc. Some of them also have exotic algorithms available, including NOS. NOS usually requires some EQ'ing to be flat enough to 20kHz for a meaningful comparison. It would also allow to have "effective sample/filter frequencies" below 44.1kHz so that more artifacts fall into the audible range, and also the sinc ringing can be brought closer to normally audible ranges. Just to get a clue how it sounds like when exaggerated...
 

bennetng

Major Contributor
Joined
Nov 15, 2017
Messages
1,634
Likes
1,692
It would also allow to have "effective sample/filter frequencies" below 44.1kHz so that more artifacts fall into the audible range, and also the sinc ringing can be brought closer to normally audible ranges. Just to get a clue how it sounds like when exaggerated...
Exactly the same thing I talked about, in case some people did not read it:
https://hydrogenaud.io/index.php?topic=120491.0
 

AnalogSteph

Major Contributor
Joined
Nov 6, 2018
Messages
3,338
Likes
3,278
Location
.de
If everything else is ideal, there is no problem. However, a 23 kHz tone can intermodulate with its 25 kHz image to produce a very audible 2 kHz tone.
Thankfully, in actual music there generally is precious little going on up there. Which is why using half-band filters in DACs works as well as it does.
 
OP
pma

pma

Major Contributor
Joined
Feb 23, 2019
Messages
4,592
Likes
10,728
Location
Prague
Thankfully, in actual music there generally is precious little going on up there. Which is why using half-band filters in DACs works as well as it does.

The worse situation is with Fs = 44.1 kHz and the mirror images are same as shown here, so the issue is not limited to 23kHz/48kHz. The cure is proper filtering or higher sampling rate. I think that manufacturers want for their products look better so they extend the passband as high as possible at the expense of mirror images. Marketing vs. proper technical solution, as always.
 

solderdude

Grand Contributor
Joined
Jul 21, 2018
Messages
15,891
Likes
35,912
Location
The Neitherlands
The D30 filters are fine for 48kHz (and that's what they seem to be designed for).
So let's resample all 44.1kHz to 96kHz where the resampler has a proper 20kHz filter ? :D
 
OP
pma

pma

Major Contributor
Joined
Feb 23, 2019
Messages
4,592
Likes
10,728
Location
Prague
The D30 filters are fine for 48kHz (and that's what they seem to be designed for).
So let's resample all 44.1kHz to 96kHz where the resampler has a proper 20kHz filter ? :D

Is there any source that would show D30 filters response at @48kHz sampling rate? Amir's measurement is at @44.1kHz, so we cannot tell the filter response @48kHz based on Amir's measurement at @44.1kHz.

I agree, resampling from 44.1kHz to 96kHz with a proper re-sampler and resampling filters is the way to get better results and usually better sound as well ;).
 
Top Bottom