Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Weyeswon]

Am trying to convert 2.1 GB and 2.6 GB .dsf files to FLAC. Have tried various compression settings. Files exist on a NAS server. Smaller (.6 GB) files convert fine. Does the file converter check for file size?

Running MC 25 build .104 on Win10 64 bit.

Thanks!

[dtc]

2 channel or multi-channel?

[Weyeswon]

2 channel or multi-channel?

2 channel. Temp file of zero size is created.

[Matt]

DSD has an extremely high sample rate, so you might need to set down-conversion of the sample rate for conversion.

[Weyeswon]

DSD has an extremely high sample rate, so you might need to set down-conversion of the sample rate for conversion.

But why do smaller files convert with no issues? BTW, these are DSD 128 files.

[DJLegba]

I just converted a 2.7GB dsf file to 24-bit FLAC (MC 25.0.104 64-bit)

[bob]

I just converted a 2.7GB dsf file to 24-bit FLAC (MC 25.0.104 64-bit)

What did you use for conversion settings?

[Matt]

I just tried to convert a 35 minute DSD file to FLAC.  It didn't work because the sample rate wasn't supported.

Then I tried converting the same file to APE and it didn't work because we were not using an int64 to pass the maximum size.  So I've got a fix coming for that in a future build, so at least APE will work.

[DJLegba]

What did you use for conversion settings?

I didn't change the default FLAC Encoder options, so:
Compression Level 6
Add 4k padding block
Add seek table

I did, however, apply DSP:
Parametric Equalizer +6 dB
Output Format: Input Greater than 768,000 Hz converted to Output 176,400 Hz (all others "No change")

The FLAC file is 24-bit 176.4 kHz, 507,652 kB
The original was DSD256, 2,790,582 kB

The other track on that album is 6,713,678 kB. I can try converting that as well if it will help.

[dtc]

A DSD file gets converted to PCM by MC  at factor of 8  to 1. So, a 1x DSD (2.822MHz) converts to a 352 KHz/24 bit file. A 2x goes to 705 Khz and a 4x goes to 1,410 Khz. The 4x conversion should have Output Format set to downsample to a lower PCM rate, like 176 Khz or 352 KHz.  You may be doing that - that wanted to point it out.

[Weyeswon]

I didn't change the default FLAC Encoder options, so:
Compression Level 6
Add 4k padding block
Add seek table

I did, however, apply DSP:
Parametric Equalizer +6 dB
Output Format: Input Greater than 768,000 Hz converted to Output 176,400 Hz (all others "No change")

The FLAC file is 24-bit 176.4 kHz, 507,652 kB
The original was DSD256, 2,790,582 kB

The other track on that album is 6,713,678 kB. I can try converting that as well if it will help.

I am unable to convert any DSD128 files using basically your settings except I set DSP Output Format : Greater the 768K to convert to 384K. Only a temp file of size zero was created. DSD file sizes were generally .5 or .6 GB. DSD64 files seem to convert no problem.

[DJLegba]

I am unable to convert any DSD128 files using basically your settings except I set DSP Output Format : Greater the 768K to convert to 384K. Only a temp file of size zero was created. DSD file sizes were generally .5 or .6 GB. DSD64 files seem to convert no problem.

When converting DSD128 with the settings as above the process finishes immediately, leaving a temp file of size 0. However, if I change all sample rates to 176.4 (instead of just greater than 768k) the conversion is successful. Try that.

[Weyeswon]

When converting DSD128 with the settings as above the process finishes immediately, leaving a temp file of size 0. However, if I change all sample rates to 176.4 (instead of just greater than 768k) the conversion is successful. Try that.

Yes! It is working! But why? Someone previously in this thread said that MC will convert DSD to 8x the sample freq. I kind of don't get the math. Is there math involved with this solution?

[DJLegba]

Yes! It is working! But why? Someone previously in this thread said that MC will convert DSD to 8x the sample freq. I kind of don't get the math. Is there math involved with this solution?

Here's the math:

A DSD file gets converted to PCM by MC  at factor of 8  to 1. So, a 1x DSD (2.822MHz) converts to a 352 KHz/24 bit file. A 2x goes to 705 Khz and a 4x goes to 1,410 Khz. The 4x conversion should have Output Format set to downsample to a lower PCM rate, like 176 Khz or 352 KHz

I was just too lazy to check if I needed both 705kHz and 768kHz set to downsample to 176, so I set MC to downsample everything. But it works if you select 705kHz only. I can't tell you why it fails with no error message if you don't downsample. Perhaps fixing that could be added to the "too easy" thread.

[dtc]

Actually, for a smaller 4X DSD file, if I put 352 KHz in the >768 line, I get a 352 KHz file. And putting 176 KHz there produces a 176 KHz file. I am not sure why it is different for what you are seeing.  I am using 25.0.108.  I have always thought that putting all the sample rates up to 4xDSD into Output Format would be a good idea. But, fortunately, you can usually get it to work as  is.

I believe the factor of 8 conversion is arbitrary. The original author decided to, loosely, turn 8 0s and 1s into a PCM number.  He could have decided to use 16 instead of 8. I do not know the whole history, but there was one public source set of code to do this that many people used originally. A 2.822 MHz file turns into a 352 KHz PCM file, which gives people plenty of room to decide how they want to resample the PCM to the final sample sample.

One more thing to consider is the filter that has to be applied after a DSD to PCM conversions. Due to the nature of DSD, with a PCM conversion, there is ever increasing noise at higher frequencies. It should not be allowed into the electronics.  In original SACD, Sony used a 48 dB/octave filter low pass filter at 50 KHz. That is a steep enough filter to remove the noise. The JRiver default for that filter is 48 dB/octave at 24 Khz.  There are also options for 24 dB/octave at 30 Khz and 50 KHz.  The default is very effective, but quite conservative in terms of frequency. The higher frequency ones are, unfortunately, not quite step enough to really eliminate the noise. You can turn the filters off and set your own Sony like filter (48 dB/octave at 50 KHz) in DSP Studio Parametric Equalizer, although I do not know the quality of the filter.  The standard filter options are in Tools - Audio - Advanced - Input Plugins - DSD. You can also increase the sound level by +6dB in that section, if desired.  Some people do, but it runs the risk of clipping on occasions.

[Weyeswon]

Actually, for a smaller 4X DSD file, if I put 352 KHz in the >768 line, I get a 352 KHz file. And putting 176 KHz there produces a 176 KHz file. I am not sure why it is different for what you are seeing.  I am using 25.0.108.  I have always thought that putting all the sample rates up to 4xDSD into Output Format would be a good idea. But, fortunately, you can usually get it to work as  is.

I believe the factor of 8 conversion is arbitrary. The original author decided to, loosely, turn 8 0s and 1s into a PCM number.  He could have decided to use 16 instead of 8. I do not know the whole history, but there was one public source set of code to do this that many people used originally. A 2.822 MHz file turns into a 352 KHz PCM file, which gives people plenty of room to decide how they want to resample the PCM to the final sample sample.

One more thing to consider is the filter that has to be applied after a DSD to PCM conversions. Due to the nature of DSD, with a PCM conversion, there is ever increasing noise at higher frequencies. It should not be allowed into the electronics.  In original SACD, Sony used a 48 dB/octave filter low pass filter at 50 KHz. That is a steep enough filter to remove the noise. The JRiver default for that filter is 48 dB/octave at 24 Khz.  There are also options for 24 dB/octave at 30 Khz and 50 KHz.  The default is very effective, but quite conservative in terms of frequency. The higher frequency ones are, unfortunately, not quite step enough to really eliminate the noise. You can turn the filters off and set your own Sony like filter (48 dB/octave at 50 KHz) in DSP Studio Parametric Equalizer, although I do not know the quality of the filter.  The standard filter options are in Tools - Audio - Advanced - Input Plugins - DSD. You can also increase the sound level by +6dB in that section, if desired.  Some people do, but it runs the risk of clipping on occasions.

Now using 25.0.108. Successfully converted my 3 DSD128 files to 352kb/24 bit files and the Id now adds them to the library (although the Id’s tag field for bit depth for all my 24 bit hires files shows 16 bits). I still think MC has a bug when adding large DSD files but I’ve done what I wanted to do - at least I can play the music I bought from the Id even though not in the native format ☹️.

Thanks for your help.

[bob]

Now using 25.0.108. Successfully converted my 3 DSD128 files to 352kb/24 bit files and the Id now adds them to the library (although the Id’s tag field for bit depth for all my 24 bit hires files shows 16 bits). I still think MC has a bug when adding large DSD files but I’ve done what I wanted to do - at least I can play the music I bought from the Id even though not in the native format ☹️.

Thanks for your help.

Does the same converted file show 24 bits in the BitDepth field in Windows? Sounds like there is a bug in here somewhere...

[Weyeswon]

Does the same converted file show 24 bits in the BitDepth field in Windows? Sounds like there is a bug in here somewhere...

Yes. Just looked at random hires files and Windows MC report the correct bit depth of 24 bits while the Id reports 16.