Beta-test: DeltaWave Null Comparison software

[pkane]

DeltaWave v2.0.22 is now the latest released version on https://deltaw.org

Download installer: https://deltaw.org/DeltaWaveSetup.zip

Changes in 2.0.22​

• Add: Option to automatically remove the effects of a minimum-phase DC filter. The automatic DC filter calculation and removal is configured under settings
• Add: Menu option to calculate the corner frequency of a minimum-phase DC filter (Process->Compute DC Filter Correction)
• Fix: Trim ends function could cause occasional phase errors at the end of the comparison
• Change: optimized and improved clock drift calculation

 

[pkane]

A quick v2.0.23 DeltaWave update was published adding a new result to easily compare to the results to those published in the GearSpace thread.

This adds a line to the results tab that will make it easier to compare to the results posted in that thread. Make sure to select Stereo mode if you want results for both, left and right channels. That line starts with text GS: under the results tab.

The result shows delay (in ms), level difference for L and R channels, then the RMS null result for L and R channels. Last, the same RMS null results A-weighted:

MOTU_16A_(2025)_Ch_1&2_Out_Ch_1&2_In_-5.0dB_Output_Gain
GS: -2.883703 ms, -6.1071 dB (L), -6.1136 dB (R), -44.3 dBFS (L), -45.7 dBFS (R) [-52.0dBFSA (L) -53.1dBFSA (R)]

MOTU 896 Mk3 Hybrid
GS: -1.998035 ms, -1.575 dB (L), -1.56 dB (R), -45.9 dBFS (L), -46.7 dBFS (R) [-52.1dBFSA (L) -50.9dBFSA (R)]

ART DPS II (bypass preamp -3dB).wav
GS: -1.827044 ms, -3.686 dB (L), -3.619 dB (R), -51.9 dBFS (L), -51.1 dBFS (R) [-59.0dBFSA (L) -55.4dBFSA (R)]

 

[bennetng]

Hello pkane,

I found that some files become bit-perfect or not bit-perfect by swapping the comp and reference order. I am using default settings as shown on the screenshots. Please check PM for the files in question, thanks.

 

[pkane]

Hello pkane,

I found that some files become bit-perfect or not bit-perfect by swapping the comp and reference order. I am using default settings as shown on the screenshots. Please check PM for the files in question, thanks.

View attachment 462956
View attachment 462957

I'll take a look, but there is a difference between the reference and comparison files, one is 24 bits the other 32. Since DeltaWave tries to adjust the comparison file to match the reference, using 24 bits reference should work fine and produce a bit-perfect result, but using 32 bit reference and trying to match 24 bit file to it will fail since there are more bits of precision in the reference than in comparison.

 

[bennetng]

If I do the same test using a 16/24 bit file pair the results are bit-perfect regardless of order.

 

[pkane]

If I do the same test using a 16/24 bit file pair the results are bit-perfect regardless of order.
View attachment 462982
View attachment 462983

The result you get from using 24 bit as a reference and 32 bit as comparison is due to imprecision in the gain matching operation -- a simple floating point multiply. The 32-bit file must be scaled to match the level of the 24 bit file. That is done in 64-bit floating point. The result is a difference of -270dB, which represents an error in the 44th bit, so well below the original bitness of either file. For all intents and purposes they are bit perfect at 32 bits, but the calculation to scale the file produced a tiny error of 0.00000000000003 which triggers the NOT BIT PERFECT message

 

[bennetng]

Thanks for the clarifications now I can trust the 24-bit and 32-bit files are bit-perfect.

 

[akafreak]

Can someone please help me understand these results.

Process : Recorded the same FLAC song (44100Hz stereo 1411kbps) using Focusrite Scarlett 2i2 Gen4 instrument inputs at 24 bit 48KHz into a WAV file.

First sample is iFi Zen One Signature DAC. Second sample is Fiio BTR5 Dongle DAC. Both are at 32 bits 48KHz playback in Windows.

Using DeltaWave, matched the two and listened to the "Delta Waveform" tab.


This is the Delta waveform

Is my process incorrect? Because I am getting a poor Fit Quality indication. The Delta sounds like a very high frequency version of the song.

This is the overall waveform.

Thanks a lot...this software is amazing.

This is the matched result.

 

[pkane]

Can someone please help me understand these results.

Process : Recorded the same FLAC song (44100Hz stereo 1411kbps) using Focusrite Scarlett 2i2 Gen4 instrument inputs at 24 bit 48KHz into a WAV file.

First sample is iFi Zen One Signature DAC. Second sample is Fiio BTR5 Dongle DAC. Both are at 32 bits 48KHz playback in Windows.

Using DeltaWave, matched the two and listened to the "Delta Waveform" tab.


This is the Delta waveform
View attachment 470639


Is my process incorrect? Because I am getting a poor Fit Quality indication. The Delta sounds like a very high frequency version of the song.

This is the overall waveform.
View attachment 470641


Thanks a lot...this software is amazing.

This is the matched result.
View attachment 470643

Hard to tell without seeing the files, but the RMS difference error is fairly large, so the match does appear poor. You can check other plots for some insight as to why, delta of spectra would be a good one to start. Gain error would be another that can tell you the approximate time where the largest differences in the file exist. It appears the two clocks are not synced well at all, and large jitter or large clock variation on one or both devices could explain this mismatch (see the large jitter value and the bad fit quality).

 

[DrCWO]

I like to share some findings regarding VB-HiFi-Cable.
(not off topic please continue reading )

VB-HiFi-Cable is touted as a lossless virtual audio cable for Windows. With it you can connect the Audio Output of a Windows App to the Audio Input of another Windows App. For example, one of its use cases might be recording audio data from an audio playback application into a recording application.

After discovering VB-HiFi-Cable, I wanted to test the promise of lossless transmission.

As recording application on windows, I use the freeware Windows App CD-Wave-Editor. It allows 24/96 recording and also can split a recording in single (FLAC) tracks. I normally use this feature for digitizing LPs.

I used Roon as playback application because I am familiar with it.

After the installation of VB-HiFi Cable I see and enabled it as an Audio Device in Roon:

In CD-Wave-Editor it looks like this:

I recorded a track and compared the original with the recorded one using DelataWave. Result: NOT Bit Perfect!
This was disappointing as I expected Bit Perfect transfer by VB-HiFi-Cable. Looking at the Spectrum Output of DeltaWave I saw that the recording was cut off above approx. 20 kHz.

I tried Foobar as playback app with same result.
Instead of VB-HiFi-Cable I next tried S/PDIF output of my ADI-2 Pro as output, routing it to its S/PDIF input and recorded this. Result: BIT PERFECT!

Digging a bit deeper I finally found the settings of VB-HiFi-Cable. You have to open Systemcontrol/Sound. On the playback tab you see HiFi Cable Input in the list of devices. Now select it and open its properties:

At the “Extended” tab you can select the default format. And guess what: It was set to 24Bit 44.1kHz by default! I changed this setting to 24/96.

Same thing I did at the recording Tab for the HiFi-Cable Output.

Hoping I now will be Bit perfect I again recorded from Roon via HiFi-Cable and analyzed the result with it with DeltaWave.

Now the recording was BIT PERFECT!
So don't believe you get Bit Perfect with HiFi-Cable by default. You have to adjust its settings first!

Dear @pkane, thank you for developing DeltaWave. It's a truly versatile tool that helped me test and ultimately utilize HiFi-Cable's Bit Perfect promise.

Best DrCWO

 

[akafreak]

Hard to tell without seeing the files, but the RMS difference error is fairly large, so the match does appear poor. You can check other plots for some insight as to why, delta of spectra would be a good one to start. Gain error would be another that can tell you the approximate time where the largest differences in the file exist. It appears the two clocks are not synced well at all, and large jitter or large clock variation on one or both devices could explain this mismatch (see the large jitter value and the bad fit quality).

Thank you for your very quick response. I repeated the same process for another song and was able to get a "Very Good" fit with equalized loudness.

I am able to hear a distorted version of the song in the "Delta of Spectra" page with +22dB. My guess would be this is the difference between the two DACs. Although I am not able to audibly differentiate between them when using the "Comparator" function.

Another silly question but the delta would be the overall delta between the Reference & Compare files. Is there a way to know if it is Reference - Compare or Compare - Reference.

What I am trying to do is compare the two DACs. I can see that there is a miniscule difference between the two even if I can't hear it under normal listening volume. But don't know if it is DAC1 - DAC2 or DAC2 - DAC1.

Again, thanks a lot for this amazing program.

 

[danadam]

What are those strange time labels on the screenshots above ("001:000:000", "001:500:000", etc)?

 

[danadam]

Another silly question but the delta would be the overall delta between the Reference & Compare files. Is there a way to know if it is Reference - Compare or Compare - Reference.

You could put some file into "Reference" and a digital silence file into "Compare" and see if the polarity of the delta is the same as the reference or inverted.

 

[pkane]

Thank you for your very quick response. I repeated the same process for another song and was able to get a "Very Good" fit with equalized loudness.
View attachment 470701

View attachment 470704


I am able to hear a distorted version of the song in the "Delta of Spectra" page with +22dB. My guess would be this is the difference between the two DACs. Although I am not able to audibly differentiate between them when using the "Comparator" function. View attachment 470703

Another silly question but the delta would be the overall delta between the Reference & Compare files. Is there a way to know if it is Reference - Compare or Compare - Reference.

What I am trying to do is compare the two DACs. I can see that there is a miniscule difference between the two even if I can't hear it under normal listening volume. But don't know if it is DAC1 - DAC2 or DAC2 - DAC1.

Again, thanks a lot for this amazing program.

The Delta waveform is Reference - Comparison.

What are those strange time labels on the screenshots above ("001:000:000", "001:500:000", etc)?

That's a sample count instead of time. The selector at the top allows simple time display, display with milliseconds and microseconds, or showing sample count. Zoom in on the X-axis in the plot if you want to see more detail:

 

[akafreak]

Thank you again. I have been at it interchanging Reference & Comparison for the last hour or so and my grand conclusion is.......I can't hear a difference on my headphones.

iFi Zen One Signature has SINAD of ~95 while Fiio BTR5 has SINAD of ~83 on unbalanced. And I can't hear a damn difference in A/B test or ABX test.

Thank you @pkane for this amazing app which lets us do proper A/B comparison when volume matched. You just saved me ~200 USD which I was about to splurge on another "better" DAC. If I can't differentiate 83 and 95 SINAD on headphones, when actively listening for it, and within human hearing limits. I for sure can't tell a difference between 95 and 120 SINAD when casually listening on speakers.

And all that talk of iFi warm & natural, organic sound with Burr Brown chipset : I can't hear it and I can't see it when quantified against ESS DAC in Fiio. So, I don't believe it.

This is an amazing tool and I will urge every one I come across in Audio forums to try for themselves for DAC/Amp comparison. It lays bare all the audio marketing jargon for what it is...nonsense.

 

[Rantapossu]

DeltaWave plays a major role in the GoldenSound's (The Headphone Show) new video:

At least 80,000 people now know what DeltaWave is...

 

[pkane]

DeltaWave plays a major role in the GoldenSound's (The Headphone Show) new video:

At least 80,000 people now know what DeltaWave is...

He actually used DeltaWave in his MQA testing some years ago. From the looks of it, he also used Multitone in this video

 

[Froller]

Thank you again. I have been at it interchanging Reference & Comparison for the last hour or so and my grand conclusion is.......I can't hear a difference on my headphones.

iFi Zen One Signature has SINAD of ~95 while Fiio BTR5 has SINAD of ~83 on unbalanced. And I can't hear a damn difference in A/B test or ABX test.

Thank you @pkane for this amazing app which lets us do proper A/B comparison when volume matched. You just saved me ~200 USD which I was about to splurge on another "better" DAC. If I can't differentiate 83 and 95 SINAD on headphones, when actively listening for it, and within human hearing limits. I for sure can't tell a difference between 95 and 120 SINAD when casually listening on speakers.

And all that talk of iFi warm & natural, organic sound with Burr Brown chipset : I can't hear it and I can't see it when quantified against ESS DAC in Fiio. So, I don't believe it.

This is an amazing tool and I will urge every one I come across in Audio forums to try for themselves for DAC/Amp comparison. It lays bare all the audio marketing jargon for what it is...nonsense.

By only focusing on measurements in the audible range, we miss what's going on in the ultrasonic region above 20-22 kHz—things like the impulse response, which impacts how transients are reproduced. I've noticed this many times: the high frequencies we can hear might measure as rolled off, but after a while, I get listening fatigue or even tinnitus (and I listen at low volumes). When I investigated further, I found a huge spike at 28 kHz. We can't hear it, but we can sense it, and it's unpleasant.

 

[pkane]

By only focusing on measurements in the audible range, we miss what's going on in the ultrasonic region above 20-22 kHz—things like the impulse response, which impacts how transients are reproduced. I've noticed this many times: the high frequencies we can hear might measure as rolled off, but after a while, I get listening fatigue or even tinnitus (and I listen at low volumes). When I investigated further, I found a huge spike at 28 kHz. We can't hear it, but we can sense it, and it's unpleasant.

Where did the 28k peak come from? A proper 44.1k or a even a 48k recording shouldn't have any content at 28k. I very much doubt you can hear or sense 28kHz unless you're under 20 years old, and even then it would be pretty unusual.

 

[Thomas Christensen]

Hi.
Can Deltawave be set to only compare 2 files down til -40 db ?
As standard the Deltawave can find differences the human cant hear, so if it could be set to compare from highest peak and down to -40 would be nice