How can converting PCM to DSD be better?
ฝัง
- เผยแพร่เมื่อ 6 ก.ย. 2024
- PS Audio DACs first convert incoming PCM from CDs into DSD for better sound. But, how is that possible? Is there something being added or subtracted? Have a question you want to ask Paul? www.psaudio.com...
I am getting close to publishing my memoir! It's called 99% True and it is chock full of adventures, debauchery, struggles, heartwarming stories, triumphs and failures, great belly laughs, and a peek inside the high-end audio industry you've never known before.
I plan a few surprises for early adopters, so go to www.paulmcgowa... and add your name to the list of interested readers. There's an entire gallery of never before seen photos too.
I have a CD from a Metal band called Accept - Staying a Life, it shows on the Japanese liner that it is mastered in DSD. This CD is the best sounding CD I have ever heard, You can crank the music and it seems like it just doesn't fatigue my ears, distortion seems really hard to introduce...Also, the left and right channels are really defined and extremely clear...I can play this CD on a basic sound system in my car and it still sounds really good.
Got the same CD 😂 need to get it out now!
Got to be careful about that volume level mate. Tinnitus is a bitch. Once the hearing cells die they can't be resurfaced.
Here's a very brief engineering explanation: In PCM, you are encoding (the "C" for code) an analog wave into a math function (generally speaking, a "Delta-Sigma" Modulation, the "M".) That code is then flashed at a clock cycle across a medium, much like Morse code (hence the Pulse.) This is received and decoded on the other end, where the wave is regenerated by inverting the math function from the previous end. Alternately, in PDM, instead of using math to encode the wave into an equation, the modulation operation is a bit different. For PDM to work, you need a clock cycle like PCM, but instead of an agreed-upon math function, we get to leverage an implicit property of nature found in calculus called an Integral. This is the idea that a series of discrete points can be used to define a smooth curve. Now, instead of pulsing a Morse code, you're playing Flappy Bird. Each flash will represent an impluse, and this will boost the signal level. This blip will then decay. If you pulse it again, it'll raise it up a little more, and it'll decay a little more, just like tapping the screen in Flappy Bird makes it bolt up, then drop down, then bolt up and then drop down. If your clock speed is fast enough, and your impulse and decay rate are calibrated correctly, it will trace out an extremely accurate wave form, though it will be "stippled," much like an inkjet printer. If you apply a smoothing capacitor, you will have a pure analog wave. Interestingly, this is very similar to the technology used to make Class D power supplies. (PWM or Pulse WIDTH Modulation uses how LONG to pulse for, not how MANY pulses.)
Mr. Bennett Thanks for this, good explanation.
Thank you so much for this amazing explanation!
Why not use TWO bits instead of just one? That way the direction of the pulse could be controlled. Both bits ON could also be used for something.
@@getsideways7257I think you're basically talking about a timecode essentially
@getsideways7257 meaning how MADI is better than toslink ADAT
Great great question. But you didn’t actually answer it! We understand DSD is more true to life, but why converting PCM to DSD which your products do, is not explained in your answer. I use Roon to do this and it clearly sounds better and the sound stage is wider and the sound is more real.
Exactly what felt. Need more specific answers.
I understood it as, converting PCM to analogue can be quite complex and difficult to get perfectly, being a 2+ step process. The simplest PCM DACs (as explained on Technology Connections th-cam.com/video/pWjdWCePgvA/w-d-xo.html) passes through an array of resistors which can be vulnerable based on the accuracy of the resistors, and then though a LPF, whereas this guy is saying that DSD basically just needs to go through a LPF, therefore less room for the reproduction to go wrong.
Totally agree. Any chance to hear the answer?
I convert all PCM to DSD 64 with my LUMIN U1 and the difference in sound coming out of the DAC is incredible!
I have in my lifetime collection about 500 cd's, some of which i thought sounded pretty good untill about 4 years ago when I was interduced to DSD. I then started buying SACD's instead, ripping them and sending to my Aurender N10. The difference to me was very noticeable to me right from the start. Recently when I was unfortunately diagnosed with Covid 19, I decided I was going to give myself a project and convert all of my CD's to DSD. I purchased some software online and worked for 6 days on the conversion process. Before long I was hearing those CD's for the first time again! They sound so much better in DSD! Don't get me wrong, my Holo May KTE makes CD's sound very good but the converted DSD files sound more analog. I will also say I have some CD's that were just bad recordings, DSD doesn't help that. They are still bad recordings! Overall, I am very happy I converted my entire collection and it was definately worth the time and effort it took to complete. The only drawback on this is storage. The DSD files take up 3 to 4 times the space on your disc drives. I have a couple of Drobo's so to me it wasn't a big deal but if you have a lot of conversions, you'll need a space to store them.
First example
Suppose a 4 bit PCM-signal (4 bits per sample). 4 bits has 16 values (0-15). You can convert that to 16 samples in DSD (16 samples of 1 bit per sample). The value 12 converts to 12 ones (and 4 zero’s) (and you can convert it back by counting the ones). That requires a 4 times higher bit rate (16 bits in stead of 4).
Second example
Suppose a 16 bit PCM-signal (16 bits per sample). 16 bits has possible 65.536 values. You can convert one PCM sample to 65.536 samples in DSD (65.536 samples of 1 bit per sample). And that requires a 65.520 times (65.536 - 16) higher bit rate. So you can convert lossless a 16 bit at 44100 Hz PCM-stream to a 2.890.137.600 Hz DSD-stream (that is 2890 MHz).
Third example
Suppose a 24 bit PCM-signal (24 bits per sample). 24 bits has possible 16.777.216 values. You can convert one PCM sample to 16.777.216 samples in DSD (16.777.216 samples of 1 bit per sample). And that requires a 16.777.192 times (16.777.216 - 24) higher bit rate. So you can convert lossless a 24 bit at 176.400 Hz PCM-stream to a 2.959.500.902.400 Hz DSD-stream (that is 2.959.500 MHz or 2.959 GHz or almost 3 THz).
Fourth example
The other way around. Suppose a DSD-signal (64 times the sample rate of a CD). 64 pulses you can code with 7 bits. You can convert a DSD signal to a 7 bit at 44100 Hz PCM signal (just count the pulses).
Conversion from PCM to DSD and back can be done lossless (as shown). But it requires very large sample rates and very large bitrates in DSD format. So, in my opinion you better use PCM as a format and convert to DSD in hardware, in the DAC. And I think this is what most DAC’s do: converting a high resolution / low sample rate to a low resolution / high sample rate. And because the needed sample rate are too high the DAC’s convert it to a 4 bit signal in stead of a 1 bit signal. So, in my opinion, use PCM and let the DAC do the conversion (with the current DSD sample rates).
I would really like to hear my mic pre's through DSD. As an engineer who grew up with Neve, SSL and even DDA pre's, I have some issues with the PCM conversion in my current setup even with the Midas equipment. Would be cool if PS Audio could make a recording interface that supported DSD.
Thanks, this is very true. You can hear the difference. The only digital conversion i can think of that improves the output quality.
I had a cheap Topping DAC for a while that was buggy in converting PCM to DSD before reaching the DAC. But now I have a new DAC that does it beautifully and is a much better sounding DAC (Matrix mini i pro).
Why does my DAC produce better output when I upsample? Why does another DAC I have sound better when I stream DSDx2? The answer is, essentially, my software (I use JRiver) is good at doing this conversion, and the DAC is better at reproducing this input than the native 44.1khz/16bit. There may be any number of reasons for this, but the point is that DACs are not perfect, and that is why you do such conversions.
Thats really not answering the question.. but we have the audible fact that it actually does sound better.. For all the people that tried i know at least. The question why is just not conclusively answered yet.. On the recording/transmission line side of things there are enough facts that make it easy to explain why DSD sounds better.. so a pure DSD to storage and DSD back reproduction is by all measurable means better... But the interesting question why the DSD conversion of a PCM data storage is still sounding better has no really adequate answer ...yet.. And objective tests are difficult anyway..since any converter and analog backend is coloring the sound in one way or the other.. in minuscule ways that are supposed to be below the borders of human perception... but we hear differences never the less.. its really Not! like that that every device that holds certain specs would sound the same.. quite the opposite...everything sounds slightly different.. you hear different any day.. Is that just an illusion and subjective? or are there hard physical facts that interfere with your perception? Yes there are..Air pressure for example..hard physical fact that changes your listening experience just by weather changes..But look at physics..Do we know if there is audio perception on a quantum level yet? NO... we dont..can we exclude that? NO..rather the opposite is happening..research shows that our brains actually might be quantum computers... And is there a difference between a process where you probe and measure to a process where you dont probe and measure on a quantum level? Yes there is..thats where all this quantum craziness has its start point..where this branch of physical science is emerging from.. Call it a wild theory if you want..i call it an educated guess.. The answer why we experience such realtime transformations like DSD to please our Brains more than PCM will become explainable as more we learn about quantum physics and its relation to our brain function.. my guess.. You even can call it an uneducated guess if you want.. Since i know that i dont know.. But i believe what my ears tell me, and since there is a consistent perceiveable difference..over many years now.. its no illusion, and probably no voodoo either. There must be an explanation..its just still in the works..
Nice pictures from Portugal in the background ;)
what is the best wv to DSD converter? my music sound dull/flat when converted from .WV (wave) to DSD & the bass completely disappears
So well explained!! Thank you!!
Alright, this got me curious so i proceeded to convert a well recorded 44.1kHz/16-bit wav/pcm file to DSF/DSD 2.8Mhz using the Tascam hi-res editor software. I am very familiar with this song and I played a specific sequence of each file over and over and over using a DSD capable DAC and for the life of me i can't tell them apart. At this point i need to be proven wrong if i state that DSD is no better at decoding to analog a PCM file that was converted to DSD than PCM is at decoding to analog the original PCM file.
new to all this Hires DSD discution...
I was able to get a protable DAM/AMP decent enough for day to day listening (Fiio Q1 Mark II).
Looking at your video I decide to look at ways to make mi Windows use my DAC in ASIO DSD256 as output.
End up using the chain:
1. HIFI-Cable & ASIO Bridge (windows sound to ASIO proxy)
2. ASIO PRoxy setup to output all samplerates to DSD256 in DSD mode
3. Use Fiio ASIO driver to connect my device and sent DSD bitsteam to the DAC.
This wal all your windows audio will be sent as DSD bitstream to the DAC, no need to convert or transcode anything; and cloud services like deezer, youtuve, tidal, soundcloud will be sent as DSD stream.
the only BUT is latency added by the buffers in the diferent steps; this solution needs tunning if you where to use it in gaming for example.
Still Waiting for your debate with Ethan Winer.
.......hold your breath
I emailed Paul about that. It ain't happening haha :)
@@KarlHamilton Yeah, figured as much. Still, it was a public invitation and basic manners dictate that he has to decline publicly. Otherwise we might arrive to the wrong conclusions :P
@@lupahole I think when Ethan copped a shit attitude it guaranteed he wouldn't get a response. He could have went about it in a better way.
@@KarlHamilton You may be right, but we all know why this debate isn't happening...Still, dont hold me on that, i hate predicating things.
An interesting article, thanks Paul. I'm going through my record collection and digitizing with a view to converting to DSD128. Would it be best to capture at 176.4kHz because that is an integer multiple of DSD128 sampling frequency? Would there be any advantage/disadvantage in making the original capture at 192kHz, which is the highest rate my ADC will do?
The main difference between PCM and DSD(PDM) is speed of dynamic change in time domain. PCM can do dynamic change from 0 to max depending on bit depth use, in one slice of time frame. Where does DSD can only do increment of 1 bit per time slice, although DSD usually process in smaller time frame.
So DSD is more near 'analogue' because analogue media like tape and vinyl take time to change dynamic. Which also make sense once you transfer PCM to DSD then back to PCM it will force PCM to record DSD's slow change in dynamic i.e. make it more analogue sounding.
As DSD more near real sound? I have my question especially for faster modern music.
A LEE this is best answer. I am satisfied.
You seem to be under the impression that the sample value is absolute, i. e. a bigger number produces higher output voltages. That is not the case. Be it 1, 65,536 or 16,777,216 - they all produce maximum voltage in their respective "bitness" (2, 16 and 24 bits). Except DSD can generate maximum volume at pretty much RADIO frequencies (at every sample or every next sample at over 11 MHz).
After hearing Paul talk about DSD I got my hands on some music files.....
WOW..!! They sound amazing, I'm now an instant DSD convert..!!
Jules 327 where can I get DSD files?
@@zoltar808 A mate of mine gave me them on a USB drive, but try a search on Google..
There very large files though even bigger than FLAC files....
Make sure to listen to them on a system with a DAC that supports DSD, otherwise most PCs/MACs/Smartphones will transform the output to PCM anyways.
I convert all PCM to DSD 64 with my LUMIN U1 Mini and the difference in sound coming out of the DAC is incredible!
DCS Audio discovered this back in the early 90s and knew it was the future of digital audio. No manufacture at the time made a dac that could handle it so they made their own convert chips specifically to work with DSD which they still use today
IIn 98 they came out with the Elgar converter and Purcell upsampler that took all rates and converted them to DSD. They sold this dac for 10 years before replacing it. In my opinion they still sound better than most new equipment and still sell for over 3,000$ for a 20 year old digital source on the used market Better still, DCS will still service and repair them. Not many companies can say that about their digital devices
You definitely didn't answer the question. The guy asked a really technical question, the only way to answer it is to get into the weeds of technicalities. You avoided that when you didn't want to get into the details of PCM reproduction, but that's exactly the point of the question. Do you know the answer to the question?
Of course he can't.
All available information in the original source is there and nothing can be done to make it sound 'better'.
This is a Snake-Oil supporting channel but why point out the obvious.
@@christianholmstedt8770 Definitely true.
I left this channel time ago because two reasons: first, there is a lot of self advertising in it. Second, there is a lot of "Snake Oil" in it (the praise for absurd and hugely expensive cables is the quintessence of this point).
using (DS-DAC-10R) or any DSD&PCM interface (ADC)
What would be better ?
-analog signal --> PCM 24-192
-analog signal --> DSD128 --> PCM 24-192
You explained why DSD is better than PCM but wasn't the question how can something already recorded as PCM sound better when upsampled to DSD ?
A DSD recording played back as DSD can sound close to analogue but how can converting a PCM recording to DSD have the same effect ?
John He did answer it. When PCM is converted to DSD, the conversion back to analog does not require the sharp roll-off filter, just a gentle one, so the analog output suffers little to no damage compared to PCM’s DA conversion.
@@doowopper1951 / Thanks, so PCM to DSD conversion is not considered harmful its just the filter used in converting PCM back to analogue ? I've not thought of it in this way before as I thought the improvement was related to upsampling .
Should we rip our cds to wav or dsd?
So, If I can't have a Directsream Dac due to my economic limitations, but have a DSD decoding dac, can I use a software to convert all my music into my HDDs or rip my cds to DSD files and then send that via USB to my dac and get better sound due to better and simpler decoding?
Sounds like adc converters are not up to snuff for pcm.
Now am curious how that process works, converting pcm to dsd,,, what i do know is that an fm & psg soundchip produces pure analogue audio and it only uses digital co’ordinated signals to tell what and how to produce sound,, and it is possible to even simulate pcm audio on a fm & psg chip (depending on those rule restrictions wich differs per chip)by mixing all the soundchannels together and then quickly alternate the volume to create wave forms, thus achieving a low bit pcm audio in pure analogue form,because those pcm samples were beforehand converted to co,ordinated signals to control those analogue chips,,,
Also a pwm audio chip works also similar like an fm chip in that it workds with square waves but in this case it only works with square waves,, so a pcm has to be converted into a co’ordinate signal at how to control these pwm square waves to generate a pure 100% analogue version of itself,,
I would be not amezed if a dsd system works in a similar fashion like this.
So the upsampling in your DAC will take a PCM recording and convert it to DSD, then to analog? Or do I have to re-buy all my PCM disks, including SACD and get the DSD version? Many, I don't want to do that.
Yes, the DS upsamples everything, converts it to DSD, then outputs DSD into a simple analog filter to get the audio.
Paul is confusing the source media format with the DAC conversion principle of modern DAC chips using a delta-sigma architecture. Your PCM or DSD source material sounds as good as the bit rate allows. PCM can easily be better than DSD if more data is used for your music. If your DAC chip internally somewhere on the silicon is converting your music to DSD to better translate to analog with less complex filtering, it’s just a hidden trick but has nothing to do with the media you play.
Exactly what I was thinking. He didn't even define what the dynamic range, bit rate, KHZ range of the PCM source was he's speaking of.
@@salviabloom3662 Very good point. He said DSD encoding is better than PCM because of the superior sampling rate of DSD/PDM compared to PCM...but...what is a PCM track is recorded in DXD 384kHz, is it still inferior to DSD???
Paul, where is the DirectStream Memory Player?
www.psaudio.com/directstream-memory-player/
@@arthurwatts1680 , yeah... ive seen that. remember, my post was a _year_ ago. thanks tho.
thankfully i got one of the last with the Oppo transport.
Did Paul teleport himself from Axpona?
You answered ok for me!
PCM upsampled and converted to DSD and then played back sounds far better than the original PCM track. Check Audirvana Plus (a well reviewed music player). It sound better, because, as you said, once the data is encoded in DSD the digital to analog conversion is easy and non damaging.
upsampled? so like running a compressed MP3 through noise filters, and then printing a Red book CD? Please clarify ... it sounds contradictory.
@@salviabloom3662 I do not know the details but Audirvana+ converts PCM to DSD (64/256) when connected to an external DSD-compatible DAC. And yes, even MP3s sound better. Their algorithm must decode the MP3 into an uncompressed/decoded PCM stream and from there upsample to DSD. The difference is not small.
@Hex Omega All of the above.
A bit messed up
How do you do error correction in DSD? A feedback loop?? Or is DSD a totally immaculate process without fault? (Insert eye-roll here.)
Have you ever discussed just how many digital to digital, and analog to digital, and digital to analog, and analog to analog conversions are implicit the typical audio stream, CD or LP? How many times is the digital signal decoded, re-encoded, transcoded, and run through Pro Tools as well as outboard plugins? Maybe a nice Pultec or two. It boggles the mind, not to mention the poor ear. And if we could just do all this with shining, glowing vacuum tubes -- oh, the hardware you could sell to audiophiles. Smooth out all the digital jaggies with electron transit time. Ah, like velvet to the ears. "Tube sound."
Modern commercial recordings are to the ears what CGI is to the eyes. Basically, an artifact of a musician's effort. In reproducing that, be careful what you wish for. The serious production money is for Lexus radios and Beats headphones.
This is like comparing two different things that should not be compared.. I could argue that if I took the PCM data and fed it into one of my fpga's and interpolated the data between the samples then put it out on a eg. simple DAC8550 or preferably even faster dac. I could probably output data at 1Mhz or so, a bit of cic and low pass filter on that also .. so that is what I do in my diy synthesizer but x8 (one for each voice) .. So I was gonna say that this obsession about dsd is bullshit but I get it , I may have another angle of view due to working with electronics and programming fpga's. I'm gonna say, they do it that way because it's CHEAP ,good and they can sell it EXPENSIVE, too bad but that's very common in this world. Meanwhile the local stores are piling up pure SHIT that you cant even use hearphones with without hearing some artifacts of poor dac's.. so YES ,compared to that pile of local hifi crap, everything else is going to sound like magic and it's all probably the chipmakers fault because very few decent dac's exist anymore from my point of view..
Well no one likes this but I have to disagree, if the output equals the input there isn't any loss of information. Just search "D/A and A/D | Digital Show and Tell (Monty Montgomery @ xiph.org)" Correct me if I'm wrong but used encoding/decoding method should not be the issue here, but the playback stage?
I belongs To ELECTRICAL ...
PCM @Time pulse (this is parallel Transmition of Digital signal ...so easy to store and edit & large file Transfer )
DSD/PDM@ time (this is SERIAL Transfer of DATA ...so easy to Tranfer ...simple to convert Analog ...but NOT DEPTH as PCM ...PCM Canbplay 32 bits per secon ...but DSD can play ...1 bit per secon or it need to play more bits per secon....)
Overall DIGITAL PCM OR DSD ...
Both give same. CLARITY ...(DIGITAL dont loose Data ...)
While DSD uses less circuit COMPONENTS ..(while in DSD ...ANALOG-to-DIGTAL DSD -to-File Transfer -to-AMP DSD Decoding -to-ANALOG ......DSD decoding is not complex as PCM)
PCM uses LOGIC Gates& Ladder DOCODING Circuit ...
Logic Gates are Complec But ..
Never reported ...as in-efficient
ANALOG=time T & Magnitude
DIGITAL= Frequency 1/T & Magnitude 0 or 1 & Bus Width ( else defaults 2 bit)
DSD =Frequncy 1/T (constant 5.2Mhz or 11 Mhz ...no need to consider ) & Magnitude (Default 0 or 1 no need consider ) & Bus width (1 line or 1 wire only with Ground point ...No need to consider )
PCM =Frequency 1/T (based on File 192khz or any ) & Magnitude ( depends on file type 16 bit or 24 or 32 bit ...gives levels of magnitude ) & Bus width ( Based on file type 16/24/32 )
From ABOVE ....DSD is simple & UNCOMPLCATED it only depends upon 1 wire transmission
Demirits ....
@ multy channel ...un recommended ..
@ manipulation not possible
@ speaker calibration is difficult ...by this DSD format ..
Upto stereo ..superb format ...& DSD uses LESS complex circuitary
that's why u use tascam hires ediotor :D and convert my cd's to dsd. very usefull tool
Can you tell the difference fom the original PCM files after converting them to DSD? I cannot.
It's degradation.
i think it would be better to honestly explain that DSD (PDM) codec is not new and was already used in phone applications. It would also be better to be honest and stop claiming that envisioning signal has anything to so with analog signal, and it would finally be time to admit that analog to digital and back does not degrade signal if one keeps in mind Nyguist theory. So PCM is enough and enough is enough. Either that or simply keep silent on the subject and use regular wide, air, clear, breath, warm, detailed and whatever else words you use to sell your products. And if someone asks you about DSD ot PCM use word DXD as the rest of the industry does and amaze the gullible. If DXD word was not enough to amaze the subject keep away from him and look for new Hi-Fi purist.
Horse manure pile