I just made program to demonstrate mixing two 1bit dsd64 bitstreams without any conversion to multibit or anything else, this is fully lossy free format, this makes mix in pure 1bit dsd. The only frequency is changed to x2 because we have two bitstreams so mix file will become dsd128 but this will not change purity of your dsd64 tracks since two bitstreams on dsd128 will have two dsd64 bit flows, so two tracks is unchanged. I'm a bit hardcoded things here and you will need to rename your two DsDiff files to dothat.dff and traore.dff, it must be in DSD64 in DFF format. Double click .exe and after 10-20 secconds tracks will become mixed out in new file caled mix.dff and thats your mixed 1bit dsd128 track. Let me know your opinion.
Last edited:
I think I have done similar as this:
https://www.nativedsd.com/news/mixing-in-pure-dsd-no-pcm-allowed/
https://positive-feedback.com/interviews/mastering-engineer-tom-caulfield/
in my case no atenuation or efects is aplied, both songs play unchanged at the same time. I will test how two dsd microphones record will sound in mix as a stereo track. I'm also researching how to aply equaliser on 1bit dsd bitstream.
https://www.nativedsd.com/news/mixing-in-pure-dsd-no-pcm-allowed/
https://positive-feedback.com/interviews/mastering-engineer-tom-caulfield/
in my case no atenuation or efects is aplied, both songs play unchanged at the same time. I will test how two dsd microphones record will sound in mix as a stereo track. I'm also researching how to aply equaliser on 1bit dsd bitstream.
Last edited:
With this technique you can mix 32 mono records made in 32 rooms for example, each record can be fully thrown in or trown out without breaking any single bit of other records, fully reversible to the absolut original state!
it is interesting that when two the same dsd64 tracks is mix, they sound like one but now it become dsd128 : )
it is interesting that when two the same dsd64 tracks is mix, they sound like one but now it become dsd128 : )
Last edited:
There is some possibly related literature about what can be done with DSD, but its pretty limited what can be done properly.
Other than that its possible to improperly chop up DSD bits into smaller pieces, but doing so is likely to create a lot of digital artifacts that can down-convert into the audio band. Thing is, in the discrete time domain there are infinite aliases for a set of sample points. Its not hard with DSD and or DSD dacs to create mixing (more or less discrete time intermodulation) that appears in the audio band as garbage.
Other than that its possible to improperly chop up DSD bits into smaller pieces, but doing so is likely to create a lot of digital artifacts that can down-convert into the audio band. Thing is, in the discrete time domain there are infinite aliases for a set of sample points. Its not hard with DSD and or DSD dacs to create mixing (more or less discrete time intermodulation) that appears in the audio band as garbage.
Last edited:
I've seen that document, it's interesting, but not very detailed. I'm trying to figure out how a specific tone at specific frequency can be isolated from a 1-bit dsd bistream, how to determine that, how to know where that part of interest is. Or for example how to cut all tones >10KHz for example.. Mean in software way. An example would be need to understand. I am able to produce echo efect when two identic 1bit dsd bitstreams is mixed so that seccond one is started from an specific offset : ) There is some opensource players which have volume control on 1bit dsd bitstream, I should look it sometime when i have nothing to work.
Last edited:
The thing about a 1-bit bitstream is that its probably gone through a dithered modulator. That means the bit pattern for one cycle of your test tone isn't going to be the same for each cycle. The only way to get your test tone back is probably LPF the DSD to get the average signal level over the short-term that would correspond to one PCM sample (IIRC @xx3stksm spoke in more detail as to the actual required process in another thread). Unfortunately it will have had some noise added to it when it was dithered. That's the conventional sort of approach, something more or less like that. Then you do whatever you're going to do with the PCM, then remodulate it back to DSD. If you only do it once, and if you use enough mathematical precision, then it can work pretty well. The LPF should be really high quality, and maybe 128-bit or 256-bit math. The example math precision numbers I suggested are based on other audio processing software reputed to be of very high sound quality, not because of a more theoretically principled reason. IOW, you could try it and see what you think.
Last edited:
Thasts not interesting for me, I want lossy free format, no want any conversion to pcm. If 1bit dsd is converted to pcm there is no way to get them back to apsolut original dsd material. Pcm not interest me because of decimation and interpolation makes things lossy. Dsd is definitely something promissing in near future, can you imagine how it will become good to sample an analog signal in 1bit dsd format for example in dsd2048, it will be brutally good! What I'm searching by now is mathematic examples about volume control or equaliser on 1bit dsd bitstream, in order to understand principle, but seems I am not lucky to find it on the net or maybe not enought searching I have made. What I know right now HQPlayer have all that things allready (links in post #4)
PCM is interesting to me because most digital recordings are in PCM.
Regarding mixing, most modern ADCs operate internally on what might be termed Raw-mode/DSD-wide/PCM-narrow. IOW, its neither 1-bit DSD nor 24-bit PCM. It has to be converted internally in the ADC chip or else externally (such as in an FPGA) to the desired digital audio output format. So, you are probably stuck with a conversion whether you like it or not.
That being that case, usually best to convert the ADC output to PCM which is much easier to mix/edit/effect/etc. DSD is only needed at the dac either because its a DSD-only dac, or because the dac sounds better in DSD mode.
Also, the Comtrue ASRC which IIRC you are using to convert to DSD512 is what I would consider to be not good enough to use. What is good enough is the FPGA-based Simple DSD Converter project which converts PCM to DSD256. With version 3 firmware, galvanic isolation, and reclocking, its actually quite good. I don't see any way an ASRC-based converter could compete at that level.
Regarding mixing, most modern ADCs operate internally on what might be termed Raw-mode/DSD-wide/PCM-narrow. IOW, its neither 1-bit DSD nor 24-bit PCM. It has to be converted internally in the ADC chip or else externally (such as in an FPGA) to the desired digital audio output format. So, you are probably stuck with a conversion whether you like it or not.
That being that case, usually best to convert the ADC output to PCM which is much easier to mix/edit/effect/etc. DSD is only needed at the dac either because its a DSD-only dac, or because the dac sounds better in DSD mode.
Also, the Comtrue ASRC which IIRC you are using to convert to DSD512 is what I would consider to be not good enough to use. What is good enough is the FPGA-based Simple DSD Converter project which converts PCM to DSD256. With version 3 firmware, galvanic isolation, and reclocking, its actually quite good. I don't see any way an ASRC-based converter could compete at that level.
Last edited: