Is Bit Perfect DSD possible through Foobar2000 to Topping D50iii

[Scytales]

@AUDACC, have you contacted the Topping support service to tell them about the "352.8 kHz not supported" error message?

 

[DVDdoug]

IMO - "Bit perfection" is mostly "insurance" that the digital audio isn't getting altered or degraded... Nothing wrong with that. But, most audio problems/limitations are on the analog side. And many people use digital volume control (or EQ) which, of course intentionally alters the "bits". And there are no "bits" in the analog-output coming out of a DAC!

I am not sure, just trying to find a way to listen for myself and decide.

If you DON"T hear a difference, fine.

If you DO hear a difference you need to confirm with a Controlled Audio Blind Listening Test (40 minute video) and see What is a blind ABX test?

 

[Chrispy]

So what particularly caused you to go down the dsd rabbit hole to begin with? What do you believe it can do?

 

[AndyZ]

My Topping D50 iii plays DSD 64 fine from Roon, although I think my Audalytic D70 sounds better with DSD. Using fast filter with both. The sound identical to my ears playing PCM.

 

[terrys999]

I believe this DAC uses an ESS chip, which do not support bit perfect DSD. They just process it like PCM, sending it through the same modulators. Topping also doesn't believe in implementing DSD Direct, even when it's available on AKM chips.

The list of DACs that support bit perfect DSD is actually pretty short, and affordable options are basically SMSL D1, D200, and if you can find it, the old D-6. I own the D200 and the D-6 and they both perform well with DSD, there's an audible difference from PCM. If you don't need balanced outputs then the D1 would suffice.

Like drewdawg999 said.
Ess chips theres some sort of processing going on.
I use either mini pc or pi 5 streaming my dsd collection to eversolo t8 to either my adi 2dac fs akm , dsd direct or to topping dx9 akm dsd bypass turned on.
There is a slight difference between my adi 2/4 pro and topping d70 pro ess chips.
Difference is tiny. I now use any of em.

 

[Sokel]

@AUDACC, have you contacted the Topping support service to tell them about the "352.8 kHz not supported" error message?

With 99% certainty, it is because of the use of the DSP, which by Topping is limited to 192kHz (or is it 96kHz?) .
Plus the use of volume control.

That, if driver, etc are correctly installed.

 

[Scytales]

With 99% certainty, it is because of the use of the DSP, which by Topping is limited to 192kHz (or is it 96kHz?) .
Plus the use of volume control.

That, if driver, etc are correctly installed.

I think it is 192 kHz, as shown on the the relevant picture on this page.

 

[AUDACC]

@AUDACC, have you contacted the Topping support service to tell them about the "352.8 kHz not supported" error message?

Yes, did email Topping service noting the 352.8khz error. Attached is a word file with screenshots of the setup they recommended that still gave the 352.8khz error. The file is a 59 Mbyte DSD type file with a .dsf suffix.

 

[AUDACC]

So what particularly caused you to go down the dsd rabbit hole to begin with? What do you believe it can do?

To me bit perfect means that all the information available in the original DSD type recording makes it through the playback decoding process to the analog output. That should apply to re-coding to PCM.

My D50 wouldn’t play a DSF (DSD128) file. The file would play only if it was streamed to the D50 as 16 bit PCM 44.1 or 88.2. That is a lower resolution than the 24 bit 88.2khz or 96khz that would be certain to capture all the audible information in the DSD stream. I became curious as to why the D50 wouldn’t accept DSD files as advertised. And that started the process of sorting through fb2k components to fix that incompatibility. That process is still alive, but so far unsuccessful.

The science says there will be no audible difference when a DSD file is played back through different signal paths done correctly. The difference that might be heard would probably be caused by the implementation of the playback decoding hardware and software. I want to satisfy myself that there is “no difference” when the file is played through the D50 either natively or re-coded to PCM with sufficient resolution.

 

[Chrispy]

To me bit perfect means that all the information available in the original DSD type recording makes it through the playback decoding process to the analog output. That should apply to re-coding to PCM.

My D50 wouldn’t play a DSF (DSD128) file. The file would play only if it was streamed to the D50 as 16 bit PCM 44.1 or 88.2. That is a lower resolution than the 24 bit 88.2khz or 96khz that would be certain to capture all the audible information in the DSD stream. I became curious as to why the D50 wouldn’t accept DSD files as advertised. And that started the process of sorting through fb2k components to fix that incompatibility. That process is still alive, but so far unsuccessful.

The science says there will be no audible difference when a DSD file is played back through different signal paths done correctly. The difference that might be heard would probably be caused by the implementation of the playback decoding hardware and software. I want to satisfy myself that there is “no difference” when the file is played through the D50 either natively or re-coded to PCM with sufficient resolution.

How are you equating the different bit depth/sampling rates?

 

[drewdawg999]

Yes, did email Topping service noting the 352.8khz error. Attached is a word file with screenshots of the setup they recommended that still gave the 352.8khz error. The file is a 59 Mbyte DSD type file with a .dsf suffix.

59 Mb is awfully small for a DSD128 file, only a couple seconds. But I see you have a whole song cued up, I don't think you have the whole file, or it may be corrupted.

As for your SACD settings, it should work if you select output type as just DSD. Also under Input, the preferable area should be set to stereo (you have it on none).

 

[AUDACC]

How are you equating the different bit depth/sampling rates?

A quick listening test showed that the file conversion of DSD128 to 24 bit 96khz PCM by an online transcoder was not sufficient to capture all the audible information in the original file. The instruments and voice were subdued and less present in the 96khz file. After doing the Grok search below, it appears that conversion to 176khz or 352khz at 24 bits would have been a better choice.

"Yes, DSD audio recordings can be transcoded to PCM without adding any meaningful distortion or noise in the audible band (20 Hz to 20 kHz), provided the conversion uses a high-quality decimation filter. In practice, well-implemented conversions are effectively transparent in the human hearing range.

archimago.blogspot.com
Why Conversion is Needed and How It WorksDSD (Direct Stream Digital) is a 1-bit sigma-delta modulated format with an extremely high sample rate (e.g., 2.8224 MHz for DSD64). It pushes quantization noise far into the ultrasonic range (typically rising sharply above ~20–30 kHz). To convert to multi-bit PCM (e.g., 24-bit/88.2 kHz, 176.4 kHz, or higher), the process involves:

• Applying a digital low-pass filter (decimation filter) to remove the ultrasonic noise.
• Downsampling to the target PCM rate.
• Increasing bit depth (with proper dithering if reducing to 16/24 bits).

This is a standard digital signal processing step. Poor filters can allow noise to alias (fold back) into the audible range or introduce minor artifacts, but modern, high-quality implementations minimize this to negligible levels—far below what any DAC, amplifier, or human ear can detect.

audiopraise.com
Performance in the 20 Hz–20 kHz Band

• Distortion and noise added: With good software (e.g., Weiss Saracon, JRiver, or well-tuned open-source tools like SoX with DSD support), measurements show no audible-range distortion or noise elevation beyond the original recording's limits. Independent analyses conclude that DSD-to-PCM conversion is transparent—the audible output matches what a high-end DAC would produce from native DSD, minus the ultrasonic content.
  
  archimago.blogspot.com
• The ultrasonic noise in DSD (which can reach -85 dB or higher above 20–50 kHz) gets filtered out. This filtering does not introduce distortion or raise the noise floor in the 20 Hz–20 kHz band when done correctly.
• DSD64 has noise rising near the upper audible limit, so a gentle filter (e.g., starting around 30–50 kHz) is standard. Higher-rate DSD (DSD128/256) allows even gentler filtering with even less impact.
• Resulting PCM (especially at 176.4 kHz or 352.8 kHz) often retains full resolution in the audible band, with dynamic range and low distortion equivalent to or better than the DSD source in practice.

Many DSD recordings originate from DXD (high-rate PCM) masters anyway, so the "conversion" back to PCM is particularly benign.

audiosciencereview.com
Practical Considerations for Best Results

• Target sample rate — Use multiples of 44.1 kHz (e.g., 88.2 kHz, 176.4 kHz, 352.8 kHz) for cleaner math and fewer artifacts. Avoid awkward rates like 96 kHz if possible.
• Filter choice — A 24–30 kHz low-pass or 50 kHz cutoff (per SACD recommendations) works well. Stronger filters remove more ultrasonic noise; gentler ones preserve more "air" if your downstream gear handles it.
• Bit depth — Output at 24-bit or 32-bit floating point, with TPDF dithering if needed.
• Software examples — High-quality options include Weiss Saracon, JRiver Media Center, AuI ConverteR (in optimized modes), or foo_input_sacd in foobar2000 with multistage settings. Avoid cheap or poorly filtered tools that might let noise alias back.
• Real-world audibility — Controlled tests and measurements consistently show differences are inaudible in the 20 Hz–20 kHz range. Any perceived differences usually stem from level matching, ultrasonic content interacting with equipment, or placebo.

In short: Yes—proper DSD-to-PCM transcoding adds no audible distortion or noise in the 20 Hz–20 kHz band. It's a mature, well-understood process, and the result is sonically indistinguishable from the original DSD for all practical purposes in the audible spectrum. If you're archiving or playing on PCM-only gear, high-rate 24-bit PCM is an excellent choice."

 

[AUDACC]

59 Mb is awfully small for a DSD128 file, only a couple seconds. But I see you have a whole song cued up, I don't think you have the whole file, or it may be corrupted.

As for your SACD settings, it should work if you select output type as just DSD. Also under Input, the preferable area should be set to stereo (you have it on none).

The 59Mbyte file was the whole DSD128 file converted to 24 bit 96khz. Already tried the DSD only output without success. I am waiting for Topping's reply to this last fail. Will report when I know more. Thanks for following and offering suggestions.

 

[Chrispy]

A quick listening test showed that the file conversion of DSD128 to 24 bit 96khz PCM by an online transcoder was not sufficient to capture all the audible information in the original file. The instruments and voice were subdued and less present in the 96khz file. After doing the Grok search below, it appears that conversion to 176khz or 352khz at 24 bits would have been a better choice.

"Yes, DSD audio recordings can be transcoded to PCM without adding any meaningful distortion or noise in the audible band (20 Hz to 20 kHz), provided the conversion uses a high-quality decimation filter. In practice, well-implemented conversions are effectively transparent in the human hearing range.

archimago.blogspot.com
Why Conversion is Needed and How It WorksDSD (Direct Stream Digital) is a 1-bit sigma-delta modulated format with an extremely high sample rate (e.g., 2.8224 MHz for DSD64). It pushes quantization noise far into the ultrasonic range (typically rising sharply above ~20–30 kHz). To convert to multi-bit PCM (e.g., 24-bit/88.2 kHz, 176.4 kHz, or higher), the process involves:

• Applying a digital low-pass filter (decimation filter) to remove the ultrasonic noise.
• Downsampling to the target PCM rate.
• Increasing bit depth (with proper dithering if reducing to 16/24 bits).

This is a standard digital signal processing step. Poor filters can allow noise to alias (fold back) into the audible range or introduce minor artifacts, but modern, high-quality implementations minimize this to negligible levels—far below what any DAC, amplifier, or human ear can detect.

audiopraise.com
Performance in the 20 Hz–20 kHz Band

• Distortion and noise added: With good software (e.g., Weiss Saracon, JRiver, or well-tuned open-source tools like SoX with DSD support), measurements show no audible-range distortion or noise elevation beyond the original recording's limits. Independent analyses conclude that DSD-to-PCM conversion is transparent—the audible output matches what a high-end DAC would produce from native DSD, minus the ultrasonic content.
  
  archimago.blogspot.com
• The ultrasonic noise in DSD (which can reach -85 dB or higher above 20–50 kHz) gets filtered out. This filtering does not introduce distortion or raise the noise floor in the 20 Hz–20 kHz band when done correctly.
• DSD64 has noise rising near the upper audible limit, so a gentle filter (e.g., starting around 30–50 kHz) is standard. Higher-rate DSD (DSD128/256) allows even gentler filtering with even less impact.
• Resulting PCM (especially at 176.4 kHz or 352.8 kHz) often retains full resolution in the audible band, with dynamic range and low distortion equivalent to or better than the DSD source in practice.

Many DSD recordings originate from DXD (high-rate PCM) masters anyway, so the "conversion" back to PCM is particularly benign.

audiosciencereview.com
Practical Considerations for Best Results

• Target sample rate — Use multiples of 44.1 kHz (e.g., 88.2 kHz, 176.4 kHz, 352.8 kHz) for cleaner math and fewer artifacts. Avoid awkward rates like 96 kHz if possible.
• Filter choice — A 24–30 kHz low-pass or 50 kHz cutoff (per SACD recommendations) works well. Stronger filters remove more ultrasonic noise; gentler ones preserve more "air" if your downstream gear handles it.
• Bit depth — Output at 24-bit or 32-bit floating point, with TPDF dithering if needed.
• Software examples — High-quality options include Weiss Saracon, JRiver Media Center, AuI ConverteR (in optimized modes), or foo_input_sacd in foobar2000 with multistage settings. Avoid cheap or poorly filtered tools that might let noise alias back.
• Real-world audibility — Controlled tests and measurements consistently show differences are inaudible in the 20 Hz–20 kHz range. Any perceived differences usually stem from level matching, ultrasonic content interacting with equipment, or placebo.

In short: Yes—proper DSD-to-PCM transcoding adds no audible distortion or noise in the 20 Hz–20 kHz band. It's a mature, well-understood process, and the result is sonically indistinguishable from the original DSD for all practical purposes in the audible spectrum. If you're archiving or playing on PCM-only gear, high-rate 24-bit PCM is an excellent choice."

The start with quick listening test lost me completely. What a bunch of nonsense!

 

[drewdawg999]

The 59Mbyte file was the whole DSD128 file converted to 24 bit 96khz. Already tried the DSD only output without success. I am waiting for Topping's reply to this last fail. Will report when I know more. Thanks for following and offering suggestions.

I see, so that's a flac. Usually these conversions result in a 6dB reduction in volume, unless you can boost it. Maybe that's why it sounds subdued to you, but from what I've heard, DSD is less detailed in the treble, the rolloff results in the perception of a warmer and less fatiguing presentation.

Also the ASR section of the Grok transcription suggests different sample rates, in multiples of 44.1. Can you change this in the converter? 24-88.2 should suffice, though 24-176.4 would be better.

 

[Chrispy]

dsd is so largely bullshit....

 

[AUDACC]

I see, so that's a flac. Usually these conversions result in a 6dB reduction in volume, unless you can boost it. Maybe that's why it sounds subdued to you, but from what I've heard, DSD is less detailed in the treble, the rolloff results in the perception of a warmer and less fatiguing presentation.

Also the ASR section of the Grok transcription suggests different sample rates, in multiples of 44.1. Can you change this in the converter? 24-88.2 should suffice, though 24-176.4 would be better.

I adjusted the volume of both files up and down several times so that one or the other seemed louder and I still perceived that the flac 96khz file was more subdued. I guess that was my reverse confirmation bias The sample rates offered by this converter were 44.1, 48, 96khz. If I continue down this path, I will try another converter at the Grok suggested higher multiples of 44.1khz.

Topping Service recommended trying other files. If other files fail to run native, they recommend uploading the file to Topping Service through Google Cloud for analysis.

In the interest of not wasting the time of all the posters on this thread, I will refrain from posting until a solution is found. Thanks to all for your helpful suggestions!

 

[Scytales]

dsd is so largely bullshit....

That's a bit of an exaggeration. It is a perfectly valid format. But, as anything technical, it has its pros and cons, that's about it.

The most efficient sigma-delta modulator I know of for producing plain old DSD at 64x rate is one of those described by Erwin Janssen, a Philips engineer, in a thesis he published in 2010 (see figure 12.7). The mathematically derived signal-to-noise ratio in the audio band of this modulator is equivalent to the theoretical resolution of a 32-bits PCM signal...

This is just to illustrate that DSD is a format whose potential performance is not yet obsolete and, in any case, superior to anything practically realizable in the analogue continuous time domain.

There is therefore no need to get upset every time someone wants to listen to music that is released in this format.

 

[Chrispy]

That's a bit of an exaggeration. It is a perfectly valid format. But, as anything technical, it has its pros and cons, that's about it.

The most efficient sigma-delta modulator I know of for producing plain old DSD at 64x rate is one of those described by Erwin Janssen, a Philips engineer, in a thesis he published in 2010 (see figure 12.7). The mathematically derived signal-to-noise ratio in the audio band of this modulator is equivalent to the theoretical resolution of a 32-bits PCM signal...

This is just to illustrate that DSD is a format whose potential performance is not yet obsolete and, in any case, superior to anything practically realizable in the analogue continuous time domain.

There is therefore no need to get upset every time someone wants to listen to music that is released in this format.

LOL upset no, mystified somewhat is all.

 

[AndyZ]

LOL upset no, mystified somewhat is all.

Most of my SACD's sound noticeably better than the original CD and Tidal's dynamically compressed supposedly high definition streams.