DSD to PCM - dither now or later?
OddJobPeters
Thread Starter
#1
25th June 2009
Old 25th June 2009
  #1
Gear interested
 

Thread Starter
DSD to PCM - dither now or later?

Not sure if I'm posting this in the right area, so I apologize if I'm not.

I will be tracking some music using my Korg MR1000, but will be mixing in Logic Express, obviously using PCM. When I export the tracks to PCM, should I dither then or only when doing the final mix down? Korg's Audiogate provides the option not do dither, but I'm not sure if this will have an effect when combining tracks in the DAW. I will only be using the DAW, no outboard gear.

Also, when staying inside a DAW, is there a benefit to using higher sample rates? Should I try to work in 88.2/96k, or even higher if my system can take it? With this scenario it will only be a matter of converting differently from the DSD file, so if there is an advantage, I sure would prefer to apply it.

As for content, I intend record some full band rock songs, some solo acoustic and classical guitar, piano, maybe strings later.

Thanks!
#2
25th June 2009
Old 25th June 2009
  #2
Lives for gear
 
Teddy Ray's Avatar
 

Quote:
Originally Posted by BigJohn View Post
Not sure if I'm posting this in the right area, so I apologize if I'm not.

I will be tracking some music using my Korg MR1000, but will be mixing in Logic Express, obviously using PCM. When I export the tracks to PCM, should I dither then or only when doing the final mix down? Korg's Audiogate provides the option not do dither, but I'm not sure if this will have an effect when combining tracks in the DAW. I will only be using the DAW, no outboard gear.

Also, when staying inside a DAW, is there a benefit to using higher sample rates? Should I try to work in 88.2/96k, or even higher if my system can take it? With this scenario it will only be a matter of converting differently from the DSD file, so if there is an advantage, I sure would prefer to apply it.

As for content, I intend record some full band rock songs, some solo acoustic and classical guitar, piano, maybe strings later.

Thanks!

I think if you are doing a lot of editing/processing, the higher rates will be of some value.

dither/resample should be the final step.

never dither twice and always dither as the very last step
#3
26th June 2009
Old 26th June 2009
  #3
Lives for gear
 

Quote:
Originally Posted by Teddy Ray View Post
dither/resample should be the final step.

never dither twice and always dither as the very last step
This is incorrect, dither should theoretically be used whenever there is a truncation of the sample stream..

However the source may already have enough noise on it that the signal will already be decorellated from the noise, in which case dither may have no beneficial effect (you're trying to randomize something that's already random by adding some random noise).

So assuming the Korg developers know what they're doing, the answer to the OPs question is theoretically yes, you priobably should be using dither. In practice it may make no audible difference.

I'd agree with the higher rate though, if you have the disk space and processing power it won't hurt, and depending on various factors, it could help.
#4
26th June 2009
Old 26th June 2009
  #4
Gear nut
 
wheatus's Avatar
 

I work with Sonoma DSD....I find that all of the dithering settings in Weiss Saracon have a detrimental effect when converting DSD masters to PCM except the default tpdf setting, which is truly awesome....It stands up to the Super Bit Mapping feature in Sonoma.

So I would stay away from the big ditherators for as long as you can....In my experience, dithering always does something I don't really like...but that's just me. I have also found that the more work I do in Sonoma (in the DSD format) the better. I use Pyramix for PCM Mastering because it has a very transparent limiter on the buss tools feature.

Can I ask, is your Korg a Multi Track recorder? I thought the Korg Series DSD recorders were all stereo....How are you going to be mixing if it is?

I can master in DSD for you and down sample if you'd like....cheap...our Sonoma system is 8 track so some mixing is possible too. The Korg Audiogate software isn't the best for DSD to PCM conversion but it is decent.

Good luck,
bbb
wheatus.com
#5
26th June 2009
Old 26th June 2009
  #5
Lives for gear
 
Teddy Ray's Avatar
 

Quote:
Originally Posted by Jon Hodgson View Post
This is incorrect, dither should theoretically be used whenever there is a truncation of the sample stream..

However the source may already have enough noise on it that the signal will already be decorellated from the noise, in which case dither may have no beneficial effect (you're trying to randomize something that's already random by adding some random noise).

So assuming the Korg developers know what they're doing, the answer to the OPs question is theoretically yes, you priobably should be using dither. In practice it may make no audible difference.

I'd agree with the higher rate though, if you have the disk space and processing power it won't hurt, and depending on various factors, it could help.

everything I have ever read, and heard speaks to the opposite....also..dithering and truncating==two different processes. you dither to AVOID the errors caused by truncation.

given that most people stay at a high sampling rate/word length until they are ready to dither/resample to "redbook" , etc...my advice was pretty well on.


In terms of distortion, there is technically no distortion produced by resampling beyond what any other filtering would produce (round-off error). However, if the band-limiting performed by the resampling filter is inadequate, aliasing will occur. Usually, resampling filters are chosen so that the aliased components will be small enough to be beyond the limits of the number system being used.

In order to decide at what point in the mixing/mastering process to perform resampling, we will again return to Fig. 1. As mentioned earlier, it is possible that resampled data can achieve peak levels above the those of the original signal. This can occur whether the final sampling rate is higher or lower than the original. For this reason, it is usually advisable to run peak-limiting algorithms after resampling. Otherwise, the resampled data will most likely exceed the levels set in the limiting stage. However, the changes produced by resampling can have a disastrous effect on any dither that has been applied. Therefore, it is best to leave dithering until after resampling has been performed. Dither itself can also change peak levels, but only by a few LSBs for most types of dither. For 16-bit audio, at full-scale, one LSB is about one ten-thousandth of a dB. A signal that is limited to -0.1 dB will thus still be below digital full scale even after dithering. The recommended order of final processing would then be: resampling ‡ peak limiting ‡ dither. If desired, limiting could be performed before resampling, but in that case the signal would have to be renormalized to the desired peak level before dithering.
#6
26th June 2009
Old 26th June 2009
  #6
Lives for gear
 

Quote:
Originally Posted by Teddy Ray View Post
everything I have ever read, and heard speaks to the opposite....also..dithering and truncating==two different processes. you dither to AVOID the errors caused by truncation.
Last part right, first part wrong.
Truncation (or rounding) is a form of quantization and so leads to an error which is correlated to the input signal, which is distortion. Dithering is the process of adding noise to the input signal in order to randomize that distortion, spreading it out into noise (usually white noise, but it can also be shaped).
Quote:
given that most people stay at a high sampling rate/word length until they are ready to dither/resample to "redbook" , etc...my advice was pretty well on.
Your advice was probably not damaging because in the case of a 24 bit signal the noise floor from the original sampled analogue input is such that any quantization is probably going to be pretty well decorellated anyway, but as a theory and a policy it's contrary to the facts of how signals work. If for example the OP was dropping down to 16 bits, it would be essential to dither.
Quote:
In terms of distortion, there is technically no distortion produced by resampling beyond what any other filtering would produce (round-off error). However, if the band-limiting performed by the resampling filter is inadequate, aliasing will occur. Usually, resampling filters are chosen so that the aliased components will be small enough to be beyond the limits of the number system being used.

In order to decide at what point in the mixing/mastering process to perform resampling, we will again return to Fig. 1. As mentioned earlier, it is possible that resampled data can achieve peak levels above the those of the original signal. This can occur whether the final sampling rate is higher or lower than the original. For this reason, it is usually advisable to run peak-limiting algorithms after resampling. Otherwise, the resampled data will most likely exceed the levels set in the limiting stage. However, the changes produced by resampling can have a disastrous effect on any dither that has been applied. Therefore, it is best to leave dithering until after resampling has been performed. Dither itself can also change peak levels, but only by a few LSBs for most types of dither. For 16-bit audio, at full-scale, one LSB is about one ten-thousandth of a dB. A signal that is limited to -0.1 dB will thus still be below digital full scale even after dithering. The recommended order of final processing would then be: resampling ‡ peak limiting ‡ dither. If desired, limiting could be performed before resampling, but in that case the signal would have to be renormalized to the desired peak level before dithering.
I don't know who wrote that, but I'm afraid they're strictly speaking incorrect, though it's not something they should have to worry about, dithering if necessary at the appropriate points is the job of the guy writing the algorithms, not the user, unless the developer wants to give the option to save processing power if the signal allows it.

Note also that dither only works when applied before a quantization, so if that resampling involved dropping to 16 bits, you're screwed if it wasn't properly dithered. You can't dither after the fact.
OddJobPeters
Thread Starter
#7
26th June 2009
Old 26th June 2009
  #7
Gear interested
 

Thread Starter
Quote:
Originally Posted by wheatus View Post
I work with Sonoma DSD....I find that all of the dithering settings in Weiss Saracon have a detrimental effect when converting DSD masters to PCM except the default tpdf setting, which is truly awesome....It stands up to the Super Bit Mapping feature in Sonoma.
Well, the Korg software supports TDPF dither, so that is definitely good to know.

Quote:
Originally Posted by wheatus View Post
Can I ask, is your Korg a Multi Track recorder? I thought the Korg Series DSD recorders were all stereo....How are you going to be mixing if it is?
You are correct, the MR1000 is stereo only. I will simply be tracking on a multitracker in parallel while recording with the Korg, then lining up the DSD tracks afterwards using the other tracks as a reference. Of course, I will be converting the DSD sources to PCM before which is where the dither question comes in.

Quote:
Originally Posted by wheatus View Post
I can master in DSD for you and down sample if you'd like....cheap...our Sonoma system is 8 track so some mixing is possible too. The Korg Audiogate software isn't the best for DSD to PCM conversion but it is decent.
Thank you for the offer - I will keep it in mind. My initial goal is to get my recordings down and do a mix to the best of my abilities (part of the fun). Then if the song(s) are good and I want to take them further, I have full resolution tracks as a base and I can move them on to a studio in whatever format is required and let the pros take it from there. That is, if my mix already isn't considered to be good enough.
#8
26th June 2009
Old 26th June 2009
  #8
Lives for gear
 
Teddy Ray's Avatar
 

Quote:
Originally Posted by Jon Hodgson View Post
If for example the OP was dropping down to 16 bits, it would be essential to dither.


.
I dont recall saying that it is not necessary to dither when going to 16 bits?

I simply said...save it until last.
OddJobPeters
Thread Starter
#9
26th June 2009
Old 26th June 2009
  #9
Gear interested
 

Thread Starter
Teddy Ray and Jon Hodgson, thanks for your input. From what I am understanding so far, if I'm going to be converting to 24bit there are probably not going to be any issues either way, but I should dither to be on the safe side. Correct?
#10
26th June 2009
Old 26th June 2009
  #10
Lives for gear
 

Quote:
Originally Posted by Teddy Ray View Post
I dont recall saying that it is not necessary to dither when going to 16 bits?

I simply said...save it until last.
Perhaps I misunderstood your wording, however I read it as meaning that you should not do any dithering until the final stage, no matter what was done before that.

The reality is that in fixed point processing dither should ideally be used whenever there is a reduction in word length, and that includes truncation of intermediate results inside dsp algorithms. Though with a 24 bit sample stream, especially one taken from an analogue source, I'd say that most of the time at least it's not going to be discernable.
#11
26th June 2009
Old 26th June 2009
  #11
Lives for gear
 

Quote:
Originally Posted by BigJohn View Post
Teddy Ray and Jon Hodgson, thanks for your input. From what I am understanding so far, if I'm going to be converting to 24bit there are probably not going to be any issues either way, but I should dither to be on the safe side. Correct?
Well the problem with giving a definitive answer (though I suspect the correct one is yes) here is that I don't know what the Korg guys have done, especially since I've never seen the software.

For example is the dither shaped in any way? I've never taken the time to actually work out what effect this will have on subsequent processing, but my first intuition is to avoid noise shaped dither in intermediate stages, quantization noise spread evenly through the spectrum seems like the safest bet, especially when it's a 24 bit signal (where there would be little to gain from shaping it anyway)

To be honest I don't know why they made it switchable (except for choosing dither shape), signal processing theory says you should do it, and in an off line process the processor load is not an issue.
OddJobPeters
Thread Starter
#12
26th June 2009
Old 26th June 2009
  #12
Gear interested
 

Thread Starter
Quote:
Originally Posted by Jon Hodgson View Post
Well the problem with giving a definitive answer (though I suspect the correct one is yes) here is that I don't know what the Korg guys have done, especially since I've never seen the software.

For example is the dither shaped in any way? I've never taken the time to actually work out what effect this will have on subsequent processing, but my first intuition is to avoid noise shaped dither in intermediate stages, quantization noise spread evenly through the spectrum seems like the safest bet, especially when it's a 24 bit signal (where there would be little to gain from shaping it anyway)

To be honest I don't know why they made it switchable (except for choosing dither shape), signal processing theory says you should do it, and in an off line process the processor load is not an issue.
Thanks for your input. I'll use the TDPF dither for converting for the time being.
#13
26th June 2009
Old 26th June 2009
  #13
Lives for gear
 
jinksdingo's Avatar
I got an MR1000 as well and have used it to track like you detailed earlier. It is a fiddly process aligning the tracks!
I don't dither during the conversion as it wasn't the last step. I didn't notice anything non optimum.
Also, it is not downsampling a PCM format but a conversion from DSD to PCM so I am not sure if the same rules apply regarding dithering or if they should?
OddJobPeters
Thread Starter
#14
26th June 2009
Old 26th June 2009
  #14
Gear interested
 

Thread Starter
Quote:
Originally Posted by jinksdingo View Post
I got an MR1000 as well and have used it to track like you detailed earlier. It is a fiddly process aligning the tracks!
I don't dither during the conversion as it wasn't the last step. I didn't notice anything non optimum.
Also, it is not downsampling a PCM format but a conversion from DSD to PCM so I am not sure if the same rules apply regarding dithering or if they should?
Yeah, I have done the alignment fiddling in the past, many years ago. It's fiddly for sure, but it's doable. The funny thing is, back then when I finally got a decent 8-track, I only used it for a solo guitar project, go figure.
#15
26th June 2009
Old 26th June 2009
  #15
Lives for gear
 

Quote:
Originally Posted by jinksdingo View Post
I got an MR1000 as well and have used it to track like you detailed earlier. It is a fiddly process aligning the tracks!
I don't dither during the conversion as it wasn't the last step. I didn't notice anything non optimum.
To be honest I wouldn't expect you to notice anything, the quantization steps are 144dB below full scale and the noise on the original signal is going to be more than that, as is the noise on your output system.
Strictly speaking it should be done, practically speaking, it is most probably unneccessary.
Quote:
Also, it is not downsampling a PCM format but a conversion from DSD to PCM so I am not sure if the same rules apply regarding dithering or if they should?
Actually there's no "conversion" to do, it's a straight filter and resample like any other downsampling. There is no difference in the process of going from 192kHz to 96kHz than from DSD to 44.1kHz, all that changes is the filter coefficients and the amount of downsampling.
#16
26th June 2009
Old 26th June 2009
  #16
3 + infractions, forum membership suspended.
 

Quote:
Originally Posted by BigJohn View Post
You are correct, the MR1000 is stereo only. I will simply be tracking on a multitracker in parallel while recording with the Korg, then lining up the DSD tracks afterwards using the other tracks as a reference. Of course, I will be converting the DSD sources to PCM before which is where the dither question comes in.
hi,

dsd and pcm are two different things [although both digital audio].

dsd is one bit, and pcm is multi-bit.

dither is something you add when you do word length reduction. however, you do not do word length reduction going from one bit to multi-bit.

i think people generally use pyramix to convert dsd to 384kHz pcm, which could then be sample rate converted to whatever you are aiming for.

weiss saracon is a good sample rate converter. everyone seems to recommend it as the best. i believe there is a version that will also do the dsd to pcm conversion.



"ΔΣ modulation (SDM) is inspired by Δ modulation (DM), as shown in Fig.*2. If quantization was homogeneous (e.g., if it was linear), the following would be a sufficient derivation of the equivalence of DM and SDM:
1. Start with a block diagram of a Δ-modulator/demodulator.
2. The linearity property of integration () makes it possible to move the integrator, which reconstructs the analog signal in the demodulator section, in front of the Δ-modulator.
3. Again, the linearity property of the integration allows the two integrators to be combined and a ΔΣ-modulator/demodulator block diagram is obtained.
However, the quantizer is not homogeneous, and so this explanation is flawed. It's true that ΔΣ is inspired by Δ-modulation, but the two are distinct in operation. From the first block diagram in Fig.*2, the integrator in the feedback path can be removed if the feedback is taken directly from the input of the low-pass filter. Hence, for delta modulation of input signal u, the low-pass filter sees the signal

However, sigma-delta modulation of the same input signal places at the low-pass filter

In other words, SDM and DM swap the position of the integrator and quantizer. The net effect is a simpler implementation that has the added benefit of shaping the quantization noise away from signals of interest (i.e., signals of interest are low-pass filtered while quantization noise is high-pass filtered). This effect becomes more dramatic with increased oversampling, which allows for quantization noise to be somewhat programmable. On the other hand, Δ-modulation shapes both noise and signal equally.
Additionally, the quantizer (e.g., comparator) used in DM has a small output representing a small step up and down the quantized approximation of the input while the quantizer used in SDM must take values outside of the range of the input signal, as shown in Fig.*3.

Fig.*3: An example of SDM of 100 samples of one period a sine wave. 1-bit samples (e.g., comparator output) overlaid with sine wave where logic high (e.g., + VCC) represented by blue and logic low (e.g., − VCC) represented by white.
In general, ΔΣ has some advantages versus Δ modulation:
• The whole structure is simpler:
• Only one integrator is needed
• The demodulator can be a simple linear filter (e.g., RC or LC filter) to reconstruct the signal
• The quantizer (e.g., comparator) can have full-scale outputs
• The quantized value is the integral of the difference signal, which makes it less sensitive to the rate of change of the signal."

"The process of creating a DSD signal is conceptually similar to taking a 1-bit delta-sigma analog-to-digital (A/D) converter and removing the decimator, which converts the 1-bit bitstream into multibit PCM. Instead, the 1-bit signal is recorded directly and in theory only requires a lowpass filter to reconstruct the original analog waveform. In reality it is a little more complex, and the analogy is incomplete in that 1-bit sigma-delta converters are these days rather unusual, one reason being that a 1-bit signal cannot be dithered properly: most modern sigma-delta converters are multibit.
Because of the nature of sigma-delta converters, one cannot make a direct comparison between DSD and PCM. An approximation is possible, though, and would place DSD in some aspects comparable to a PCM format that has a bit depth of 20 bits and a sampling frequency of 96 kHz [2]. PCM sampled at 24 bits provides a (theoretical) additional 24 dB of dynamic range."



right.
#17
26th June 2009
Old 26th June 2009
  #17
Lives for gear
 

Quote:
Originally Posted by oky**** View Post
hi,

dsd and pcm are two different things [although both digital audio].

dsd is one bit, and pcm is multi-bit.

dither is something you add when you do word length reduction. however, you do not do word length reduction going from one bit to multi-bit.
Yes you do.

The process of going from a DSD stream to a multi-bit one is...

Firstly you apply a filter, this results in a stream of samples which exceed the word length of your target stream, you therefore need to quantize those samples down to your target word length, a reduction in word length.

Then you subsample that stream.
#18
26th June 2009
Old 26th June 2009
  #18
Lives for gear
 

Quote:
Originally Posted by oky**** View Post
hi,

dsd and pcm are two different things [although both digital audio].

dsd is one bit, and pcm is multi-bit.
The number of bits is irrelevant to what it is.

Having thought about it I've realized I was right to start with, semantically and practically. DSD is 1 bit PCM, the secret to accepting that however is to understand what signal the DSD is a PCM of.

1 bit sigma delta converters are an extreme example of noise shaped PCM converters, which work using a feedback loop to add an error compensating signal to the input signal, the result is increased accuracy in the desired frequency range in exchange for higher noise outside of it.

If you do a frequency analysis of the DSD bitstream (in the same way as you would any PCM stream) and compare it to a frequency analysis of the original signal, this becomes clear, you see the same frequency content in the audible band, and a whole lot of noise building rapidly above that.

A clue to this is given in the text you quoted

The demodulator can be a simple linear filter (e.g., RC or LC filter) to reconstruct the signal

If simply removing the upper frequencies demodulates the signal this tells us that the desired signal already exists in the lower bands.

So, in summation, the DSD stream is a 1 bit PCM of a signal that combines the audio signal with a compensating signal in the frequencies above it, returning to the original signal is just a case of filtering out the error compensating component.
#19
27th June 2009
Old 27th June 2009
  #19
Banned
 

I've yet to see the advantage of DSD. You have to convert it to PCM for any little move you make.

If you actually want to start manipulating the files as in a mix or editing then the whole exercise seems fruitless.

I really liked the sound of DSD the one time I used it, but did not hear any significant difference over 192kHz.

I archive for a living, so I'm pretty used to doing ABX tests. DSD just does not justify the hoops you have through to get something that sounds pretty much like an HD PCM file.

imho, ymmv etc...
#20
27th June 2009
Old 27th June 2009
  #20
3 + infractions, forum membership suspended.
 

Quote:
Originally Posted by Jon Hodgson View Post
Yes you do.

The process of going from a DSD stream to a multi-bit one is...

Firstly you apply a filter, this results in a stream of samples which exceed the word length of your target stream, you therefore need to quantize those samples down to your target word length, a reduction in word length.

Then you subsample that stream.
hi,

uh, uh. you are saying just what i said. the dsd [1 bit] has to be converted to multibit first, and dither is not required [or useful] until you start reducing the word length.

what you wrote is kind of vague, but you appear to be saying that you convert the 1 bit dsd into some unpecified type of sample stream [not pcm?], with an unspecified word length [apparently greater than 24 bits?]. and then you do a bit depth reduction to 24 or 16.

that's very different than what you said earlier in the thread i.e. "There is no difference in the process of going from 192kHz to 96kHz than from DSD to 44.1kHz".

according to what you are now saying, there is clearly a difference in that there is a need to convert from 1 bit dsd to "extended multi-bit", before going to 24 or 16 bit pcm.

anyhow, saracon is a good program to do all that, i think. but honestly, with all that voodoo, it may be smarter to simply resample the analog output of the dsd boxes.


right.
#21
27th June 2009
Old 27th June 2009
  #21
Lives for gear
 

Quote:
Originally Posted by MarkRB View Post
I've yet to see the advantage of DSD. You have to convert it to PCM for any little move you make.

If you actually want to start manipulating the files as in a mix or editing then the whole exercise seems fruitless.

I really liked the sound of DSD the one time I used it, but did not hear any significant difference over 192kHz.

I archive for a living, so I'm pretty used to doing ABX tests. DSD just does not justify the hoops you have through to get something that sounds pretty much like an HD PCM file.

imho, ymmv etc...
DSD made sense for certain applications when it came out, these days I find it a bit silly.

Avoiding the decimation stages in an ADC makes sense if you're either never going to process it in the digital domain, or if you want to be able to take advantage of any improvement in filter designs in the future. So if you're Sony and you have an analogue archive you want to make digital backups of, it makes sense to take the best ADC architecture of the day and bypass the decimation stages.

The reason I find it a bit silly today though is that 1 bit converters are not the best ADC architecture we have today, the industry has moved on to multi-bit sigma-delta, due in part to problems identified as inherrent in the 1 bit modulator. In fact some "DSD" converters are actually multibit sigma-delta ADCs, followed by a 1 bit digital sigma-delta modulator, so if you later convert to PCM you've actually gone from A to B via C, going from A to B directly would have given better results.
OddJobPeters
Thread Starter
#22
27th June 2009
Old 27th June 2009
  #22
Gear interested
 

Thread Starter
Quote:
Originally Posted by MarkRB View Post
I archive for a living, so I'm pretty used to doing ABX tests. DSD just does not justify the hoops you have through to get something that sounds pretty much like an HD PCM file.

imho, ymmv etc...
My personal decision to go with the Korg was based on cost and flexibility. I got the recorder B-stock for the same amount as any decent computer audio interface. I get a great portable recorder instead that allows me to convert anything I record to whatever maximum PCM resolution a studio might have, or my own future resources may allow. Can't speak for those working with DSD for the whole chain, but that was my justification.

Regardless, it was not my intention to start another discussion of DSD vs PCM. I read through plenty of those here before buying my recorder .
#23
27th June 2009
Old 27th June 2009
  #23
Lives for gear
 

Quote:
Originally Posted by oky**** View Post
hi,

uh, uh. you are saying just what i said. the dsd [1 bit] has to be converted to multibit first, and dither is not required [or useful] until you start reducing the word length.

what you wrote is kind of vague, but you appear to be saying that you convert the 1 bit dsd into some unpecified type of sample stream [not pcm?], with an unspecified word length [apparently greater than 24 bits?]. and then you do a bit depth reduction to 24 or 16.
Yes, except that the intermediate stream is PCM.
Filtering involves multiplication and addition, now if we're going straight from the DSD stream to say a 24 bit PCM stream then your filter coefficients would sensibly be at least 24 bits, which means that your accumulated values will be more than 24 bits, requiring truncation to 24 bits.

In practice you might well not do the whole decimation in one stage, but the same rules apply, you'll be needing to truncate.
Quote:
anyhow, saracon is a good program to do all that, i think.
I'm under the same impression
#24
27th June 2009
Old 27th June 2009
  #24
Lives for gear
 
Teddy Ray's Avatar
 

Quote:
Originally Posted by MarkRB View Post
I've yet to see the advantage of DSD. You have to convert it to PCM for any little move you make.

If you actually want to start manipulating the files as in a mix or editing then the whole exercise seems fruitless.

I really liked the sound of DSD the one time I used it, but did not hear any significant difference over 192kHz.

I archive for a living, so I'm pretty used to doing ABX tests. DSD just does not justify the hoops you have through to get something that sounds pretty much like an HD PCM file.

imho, ymmv etc...
DSD is in its place in one (1) application: Mastering from analogue. When the recording chain is completely analogue, you can feed the audio from the analogue mastering into a DSD A/D converter and cut that signal straight onto a disc without any further processing. It is in this application that DSD can be viewed as pretty transparent. When you convert the signal twice, however (such as when using a DSD recorder as the tracking medium), the second conversion is no longer transparent, due to the HF noise present in the source signal hitting a second analogue deltasigma modulator.
#25
27th June 2009
Old 27th June 2009
  #25
Banned
 

Quote:
Originally Posted by Teddy Ray View Post
DSD is in its place in one (1) application: Mastering from analogue. When the recording chain is completely analogue, you can feed the audio from the analogue mastering into a DSD A/D converter and cut that signal straight onto a disc without any further processing. It is in this application that DSD can be viewed as pretty transparent. When you convert the signal twice, however (such as when using a DSD recorder as the tracking medium), the second conversion is no longer transparent, due to the HF noise present in the source signal hitting a second analogue deltasigma modulator.
Say again, less slanty?

Why would you master from an analogue chain to DSD then back to analogue?

Edit: Sorry, I didn't realise you meant SACD.

But yeah, my point still stands. Another unnecessary expensive step in getting waveforms from musicians, to bits to consumers with zero advantage to anyone.
#26
27th June 2009
Old 27th June 2009
  #26
3 + infractions, forum membership suspended.
 

Quote:
Originally Posted by Jon Hodgson View Post
The number of bits is irrelevant to what it is.

Having thought about it I've realized I was right to start with, semantically and practically. DSD is 1 bit PCM, the secret to accepting that however is to understand what signal the DSD is a PCM of.

1 bit sigma delta converters are an extreme example of noise shaped PCM converters, which work using a feedback loop to add an error compensating signal to the input signal, the result is increased accuracy in the desired frequency range in exchange for higher noise outside of it.

If you do a frequency analysis of the DSD bitstream (in the same way as you would any PCM stream) and compare it to a frequency analysis of the original signal, this becomes clear, you see the same frequency content in the audible band, and a whole lot of noise building rapidly above that.

A clue to this is given in the text you quoted

The demodulator can be a simple linear filter (e.g., RC or LC filter) to reconstruct the signal

If simply removing the upper frequencies demodulates the signal this tells us that the desired signal already exists in the lower bands.

So, in summation, the DSD stream is a 1 bit PCM of a signal that combines the audio signal with a compensating signal in the frequencies above it, returning to the original signal is just a case of filtering out the error compensating component.

hi,

you're on your own with all that. its a very artificial and strained way of trying to concoct an argument to support a conclusion that you previously desired to justify. that type of thing is generally seen as disingenuous.

i'm not an expert myself, but every single thing i have read, other than your arguments, is in conflict with your conclusion.

it is generally acknowledged that pcm and dsd are different, although there are obvious similarities between digital processes. it think it is safe to say that your opinion has been considered and rejected by the experts [not that i am one of them, so what do i know].

i would simply note the obvious [and i believe others have noted it also]. what you consistently fail to address is the fact that dsd is all about taking one bit and simply documenting whether an amplitude increases or decreases. there is no attempt to document the value of the sample [amplitude] with that one bit. conversely, pcm, by its very nature, is an attempt to use several bits to describe the value [amplitude] of each sample directly [not just whether that value has increased or decreased, but the actual magnitude of the amplitude].

if you do not understand that those two things as fundamentally different, then, with all due respect, it is difficult to believe anything you say.

i have no idea why you have made the issue your raison d' etre on gearslutz all of a sudden. what is your motivation for wanting to promote your theory in conflict with what all the authorities appear to have concluded?

anyhow, if you want try to establish some basis i guess you could write a paper and submit it to aes for consideration or something like that. have it peer reviewed, and see if the scientific community thinks there is any merit to your argument. short of that, making off-the-cuff statements seems improper. you should at least acknowledge in your unqualified posts that what you are stating is only your own theory and argument, and that it is not widely held [or held at all] by others.


right.
#27
27th June 2009
Old 27th June 2009
  #27
Lives for gear
 
jinksdingo's Avatar
Thanks for the explanations, but I am more confused now about the dither question even with the info Hehe! I wonder how the OP is getting on with his question.

It looks like using it to capture audio and then convert to PCM to some is an additional waste of time. Logic being to just record to PCM on the first place.

Using it as a mixdown device but only really justified if your have outboard processing and sending the resultant DSD file for mastering may not be supported without conversion to PCM by your ME.

If you mix in the box why would you add a DA AD processes to your file just to get the mix to DSD?

Might put mine up for sale before it becomes anymore of an embarrassment
to one laying around idle and put the funds to a better computer.

Possible the MR2000s is a bit better spec'd for those who outboard mix.
#28
27th June 2009
Old 27th June 2009
  #28
Lives for gear
 
Teddy Ray's Avatar
 

Quote:
Originally Posted by MarkRB View Post
But yeah, my point still stands. Another unnecessary expensive step in getting waveforms from musicians, to bits to consumers with zero advantage to anyone.

the advantage is the better sound..the pursuit of which I hope you find worthy?
#29
27th June 2009
Old 27th June 2009
  #29
Lives for gear
 

Quote:
Originally Posted by oky**** View Post
hi,

you're on your own with all that. its a very artificial and strained way of trying to concoct an argument to support a conclusion that you previously desired to justify. that type of thing is generally seen as disingenuous.

i'm not an expert myself, but every single thing i have read, other than your arguments, is in conflict with your conclusion.

it is generally acknowledged that pcm and dsd are different, although there are obvious similarities between digital processes. it think it is safe to say that your opinion has been considered and rejected by the experts [not that i am one of them, so what do i know].

i would simply note the obvious [and i believe others have noted it also]. what you consistently fail to address is the fact that dsd is all about taking one bit and simply documenting whether an amplitude increases or decreases. there is no attempt to document the value of the sample [amplitude] with that one bit.
The problem with this "fact", is it is not a fact, it is a misconception, what you have described is a delta converter, not a sigma delta converter.

Working with signals is one of the things I do, I started studying it over two decades ago, I've serviced audio hardware, I've developed audio and video compression codecs, I've worked on 3d sound algorithms, and I've developed a soft-synth that some people here may be familiar with. This isn't guesswork on my part.

Quote:
anyhow, if you want try to establish some basis i guess you could write a paper and submit it to aes for consideration or something like that. have it peer reviewed, and see if the scientific community thinks there is any merit to your argument. short of that, making off-the-cuff statements seems improper. you should at least acknowledge in your unqualified posts that what you are stating is only your own theory and argument, and that it is not widely held [or held at all] by others.
If someone would like to send me some DSD files, I'll write a pcm converter for them... would that convince you that I understand what's in them? There's nothing to submit to the scientific community because nothing I've said is controversial, a sigma delta sample stream (be it one bit or 24 bit) is not a PCM of the original input signal, it is a PCM of that signal with an error compensating component (which is going to be much smaller in the case of the 24 bit conversion). From the DSP programmer's point of view this is key to understanding how to work with it, from the user's point of view it is key to understanding that DSD doesn't challenge Shannon-Nyquist in any way, understanding it requires a greater understanding of signals that the simplistic one that word with means accuracy and sample rate means bandwidth and never the twain shall meet (something which you are at least aware of I know), but it's not a different technology or set of concepts as some believe.

In normal conversation and descriptions however it is understandable and in many ways beneficial to refer to use the term PCM to mean non noise shaped PCM (the one everybody knows where when you look at the samples on the computer screen they look pretty much like the original waveform), and DXD, DSD, DSD wide etc to refer to those particular formats and the signal characteristics involved. It's like fruits and tomatos, normally when people talk about fruits they don't think of tomatos, you don't put tomatos in a fruit salad, tomato ice cream is probably something only Heston Blumenthal would think of, but sometimes it's important to understand that tomatos are actually fruits.

(Incidentally, I've stated this a couple of time in Bruno Putzey's forum in threads where Bruno has been active, and he's never disagreed with me, do you at least accept that Bruno understands about converters?)
#30
27th June 2009
Old 27th June 2009
  #30
Lives for gear
 

Here's an explanation if sigma delta converters by a company that makes, them...

http://pdfserv.maxim-ic.com/en/an/AN1870.pdf

The interesting (and useful) thing about this description is that it shows how you get to a single bit sigma delta converter from a multi-bit linear PCM converter (drop to 1 bit, oversample, noise shape) and what it does to the output stream signal content.

Anybody who understands their DSP (notably downsampling) can probably get what I am talking about just by looking at figure 9, you'll notice that the process for going from 1 bit highly oversampled to a lower sample rate is identical to any other downsampling in PCM.

That's also why the output of DSD is simply a one bit DAC into a low pass filter, if you understand how filters work then you have to realize that the 1 bit signal coming out of the DAC must contain all the information you want in the ratios you want, the filter can only remove stuff you don't want.
New Reply Submit Thread to Facebook Facebook  Submit Thread to Twitter Twitter  Submit Thread to LinkedIn LinkedIn  Submit Thread to Google+ Google+ 
 
Thread Tools
Search this Thread
Search this Thread:

Advanced Search
Similar Threads
Thread
Thread Starter / Forum
Replies
kinghand / So much gear, so little time!
18
blissmusic / So much gear, so little time!
78
bleak orange / Mastering forum
19
DSD_Mastering / Gearslutz Secondhand Gear Classifieds
2

Forum Jump
 
Register FAQ Search Today's Posts Mark Forums Read

SEO by vBSEO ©2011, Crawlability, Inc.