Weiss Saracon

stevekale

New Member
Aug 8, 2012
83
0
0
Hi. This is my first post in these forums. I was drawn to this thread as I was Googling about this criticism of Weiss Saracon. I guess I'm a little disappointed that the original discussion quickly faded. I'm a newbie to all of this stuff but would appreciate education by the experts here. I understand that DSD's higher sample rate of 2.8MHz provides a greater dynamic range capability but that noise shaping associated with 1 bit means that above 20kHz the signal is being swamped by noise. I also understand that 20kHz is generally regarded as the high frequency limit of human hearing and that as we age our ability to hear higher frequencies reduces. I guess what I don't understand is how the discussion about rolling off inaudible frequencies (if a filter roll-off of inaudible frequencies did not affect audible frequencies would I really care? my cat might be thankful for the peace and quiet) and the sample rate of the PCM->DSD conversion are related. Put another way, when I look at the graphs posted by Bruce in his initial post, doesn't the higher sample rate conversion benefit from the higher sample rate (the higher resolution) even if both roll-off inaudible frequencies in the same manner? Appreciate the help on this.
 

Bruce B

WBF Founding Member, Pro Audio Production Member
Apr 25, 2010
7,006
512
1,740
Snohomish, WA
www.pugetsoundstudios.com
Welcome Steve.... well the problem with this is twofold.. first, if you as a consumer were to purchase a 24/96 or 24/192 file and there is no content at all above 30k (regardless if it was just noise or not), wouldn't you feel cheated that you spent the extra money to get the added fidelity?
Second, because the filter is so close to the audible range, you will have smearing and phase shift. This can be measured and heard. Just as Redbook CD ends at 22.05k, the filters effect start way before that frequency. Some filters affect the signal way back at 12-14k. If you want to get rid of the noise of DSD, either record at DSD128fs where the noise is shoved up further away from the audible bandwidth or you can use filters that don't affect anywhere near the audible range. I use filters that start at 35 - 40k, just when the noise shaping starts to climb and use a gentle filter.
 

stevekale

New Member
Aug 8, 2012
83
0
0
Hi. To your first point, I wouldn't care about anything I can't hear or feel. So if there's "data" outside my hearing range being cut out I don't consider that "information" lost. (Given my age, I don't care about anything higher than about 14-15kHz.) I obviously do care about impact to the audible range.

What I would expect is that an analogue stream sampled at a higher rather than lower sample rate and then converted back into analogue sound would have more "resolution" (if I can use that term from a lay perspective) and sound better than one that was crudely sampled. As a understand it, "quality" in a conventional LPCM ADC is determined (amongst other things) by bit depth and sample rate. I understand bit depth is a key determinant dynamic range and must be sizeable enough to provide good range above the noise floor. Crudely, I think of sample rate as pixels per inch in a 2D equivalent sense (a measure of resolution) and number of bits per observation as the ability to distinguish shades of colour (yes I am a photographer). At some point extra bit depth and samples become redundant to my sensory experience (although higher values for each may be useful if they need to be transformed in the digital domain) but there's no doubt I generally prefer DVD-Audio over Red Book and SACD over DVD-Audio (while I recognise there's often a lot of variables affecting the quality of each, not the least being the analogue master that each started with).

Now, of course, we have an intermediary digital transform occurring when a DAC isn't DSD capable. I use a Theta Digital Casablanca III HD with Xtreme DACs in my audio system. It is not (currently) DSD capable and so the DSD to LPCM conversion is one that I'm interested in. With respect to SACD, I guess I am stuck with whatever process my Oppo player deploys (and of course the Oppo's LPCM is limited to a sample rate of 88.2kHz). However, DSD downloads are becoming increasingly available as is other DSD content (all of which I don't have control over the initial recording so your suggestion regarding modifying the way the initial recording to DSD is done is, I'm afraid, not applicable to me). I would generally expect a conversion of a DSD stream to 24 bit samples at a rate of 176.4kHZ to have more "resolution" than when the sampling is less frequent. (Otherwise why not just go back to 44.1kHz, unless if course the extra resolution up to 88.2kHz is discernible and the additional resolution of 176.4kHz is not.)

I have heard from many people that DSD->PCM conversions done with Weiss Saracon sound considerably better than Korg Audiogate. There's a rather significant price differential between the two and hence my interest in this discussion.

So back to my questions I guess. The general first.

I don't understand how one could say that based on the exclusion of inaudible data, buying the higher resolution LPCM file (176.4kHz) over the lower resolution file (88.2kHz) is a waste of money unless one thought that the extra resolution/sampling was wasted on the human ear in which case this would be true regardless of the conversion software. This is what I meant by not understanding the link between sample rate and a roll-off filter above the range of human hearing. Presumably a filter at 30kHz affects a file of frequency samples conducted at 44.1kHz, 88.2kHz and 176.4kHz in broadly the same manner. (If it were a hard cut such than everything over 15kHz was stored as 15kHz, there'd be no observations of sound over 15kHz in any of the bitstreams.)

I take it your view is that Weiss Saracon's filter deployment does affect the audible range/result enough to degrade quality? You say this can be measured as well. I guess the relevant chart (if any) to post is not the difference in what's outside the audible range but the difference to what's inside the audible range.

If one weren't to use Weiss Saracon or Audiogate, what software would you recommend, is it generally available and is it sensibly priced? (Bear in mind I'm not sure WS meets the last point!)

Thanks in advance. This is an interesting topic and I am keen to learn more.

Regards

Steve

(PS with respect to my question regarding software, I use a Mac)
 
Last edited:

Bruce B

WBF Founding Member, Pro Audio Production Member
Apr 25, 2010
7,006
512
1,740
Snohomish, WA
www.pugetsoundstudios.com
Hi. To your first point, I wouldn't care about anything I can't hear or feel. So if there's "data" outside my hearing range being cut out I don't consider that "information" lost. (Given my age, I don't care about anything higher than about 14-15kHz.) I obviously do care about impact to the audible range.

So does this mean you don't care about LFE (freq below 30Hz) when you're at the movies/home theatre since you can't hear the frequencies anyway?

I don't understand how one could say that based on the exclusion of inaudible data, buying the higher resolution LPCM file (176.4kHz) over the lower resolution file (88.2kHz) is a waste of money unless one thought that the extra resolution/sampling was wasted on the human ear in which case this would be true regardless of the conversion software.

The correlation I was trying to make with Saracon is that DSD files that are downsampled to PCM are exactly the same, whether they are 88.2, 176.4, 96, or 192 or whatever. ALL information is being cut off at ~35k. So paying more for a 192 file is a waste of money. In my view, paying money for any file that came from DSD done by Saracon is a waste of money. No matter what sample rate you choose, it creates the same exact file. I love Saracon for strictly PCM work, but there are other, more efficient and better sounding SRC out there for DSD. Audiogate is a great DSD converter. It allows you to select filters and dither. Audiogate is a free software that can be used in Windows or OSX.

We worked for months with HDtracks in choosing the best DSD->PCM conversions for all the SACD rips we were doing. We tried AudioGate, Saracon, SBM, Pyramix and even DiscWelder. In every case, the Saracon converters just seemed to take the life out of every file. I've posted samples here for folks to test for themselves. I even created a thread on what people should listen for when they compare hi-rez files to Redbook.
 

Vincent Kars

WBF Technical Expert: Computer Audio
Jul 1, 2010
860
1
0
Let’s assume you hearing is gapped at 15 kHz.
If we apply a brick wall filter at 15.1 kHz will you hear it?
The answer might be yes because the strong pre-ringing of this filter affects the frequencies below 15.1.
This is what happened to the first generation CD player. A brick wall filter at 20 KHz to filter out the aliases (not only your cat but your tweeters appreciates this as well).
Later they introduced oversampling. Take a 44.1 stream and repeat each sample 8 times at a frequency of 352.8. This of course won’t add any information but now the aliases starts at 176.
Even if we use a brick wall filter at 176, the pre-ringing will never reach the audible range or we might decide to use a smooth filter starting somewhere at 50 kHz and slowly rolls of.

The late Julian Dunn phrased it nicely:
A direct effect of the higher sampling rate is that for an identical filter design the time
displacements will scale inversely with sample rate. Hence an improvement can be
made just from raising the sample rate - even for those who cannot hear above
20kHz.
http://www.nanophon.com/audio/antialia.pdf

If we play a sine at 16 kHz you won’t hear it.
Obvious if we play a sine at 16.5 you won’t hear it either.
But the intermodulation distortion when the two are played simultaneously might again map into the audible range.

What is above the upper limit of our hearing we can’t hear.
But all above this threshold that maps into our audible range we do hear.
 

stevekale

New Member
Aug 8, 2012
83
0
0
So does this mean you don't care about LFE (freq below 30Hz) when you're at the movies/home theatre since you can't hear the frequencies anyway?

That's why I said "hear and feel". Does Saracon filter away low frequencies as well?



The correlation I was trying to make with Saracon is that DSD files that are downsampled to PCM are exactly the same, whether they are 88.2, 176.4, 96, or 192 or whatever. ALL information is being cut off at ~35k. So paying more for a 192 file is a waste of money. In my view, paying money for any file that came from DSD done by Saracon is a waste of money. No matter what sample rate you choose, it creates the same exact file.

Yes you've said the first part of this already. But, again, how can it be "the exact same file" if the frequency of samples is vastly different? That's like saying, to use my 2D photo analogy, a 10x8in image sampled at 180ppi is the same as one sampled at 360ppi. Taken to the extreme, regardless of bit depth, one file could have two observations to describe the scene and the other 176,000. You are saying these are one and the same. I don't understand how a snippet of sound (let's say one second) described by 88,200 observations can be a worse description of the analogue as the same second of sound described by 176,400 observations. Surely, the 2nd description has more detail, more resolution (subject to Nyquist limits). (I recognise that increasing the sample rate in my 10x8in image beyond a certain point for a given viewing distance will not improve my perception of the image and that there is a limit to improvement in perception as we increase sample rate in audio.)

When it comes to where to limit the range of each observation, do I really care that my observing device (or my transformation from one digital format to another) hits (or imposes) a limit so long as this is outside my ability to perceive differences at that point and there are no flow-on affects to other observations where I can perceive differences? (I could record an image's "colour" for any given pixel into the infrared band but that doesn't help me record my vision of the image. As far as I'm concerned, because I am interested in the image for what it looks like, the fact that certain parts of it are emitting invisible light is of no relevance, albeit it might have some relevance to my health!) What if the original analogue signal never exceeded the boundaries of human hearing, wouldn't a digital representation of that signal with more observations per second be more accurate (up to the Nyquist frequency)? And if the signal did exceed the boundaries of human hearing at certain points but these observations were truncated to lower levels still outside the boundaries of my hearing would I care then?

I still don't understand your principle criticism that a 176.4k samples per second file from Saracon is exactly the same as an 88.2k samples second rendition because Saracon imposes a filter not far enough outside the audible frequency range. How can they be exactly the same when one has twice the density of observations? Doesn't this double density of observations add value whenever any of those observations are within the audible range? I must be missing something very fundamental if this isn't the case.

It sounds to me like your criticism of Saracon is not that the files are the same but that Saracon's filter discolours in some way the sound that we can hear, that the effects of the filter at this level are discernible versus if the DSD were converted directly to analogue or, more to the point, versus the rendition that can be achieved with other software. But surely this has nothing to do with the impact to our perception of observations outside the audible range - hence my point that one needs to look for "damage" within the audible range and not outside. (And surely this isn't directly related to whether the output has twice the samples of the other.) If I played you a file of a single tone (at any sample rate you chose) where all information was cut off or capped at 35kHz and another where it was cutoff or capped at 50kHz, could you tell me which was which?


I love Saracon for strictly PCM work, but there are other, more efficient and better sounding SRC out there for DSD. Audiogate is a great DSD converter. It allows you to select filters and dither. Audiogate is a free software that can be used in Windows or OSX.

I'm glad you like Audiogate. The problem is, I've heard completely the opposite about its quality from many people. The fact that it doesn't do multichannel is also not in its favour. Given it's free I will give it a go but experimenting takes time, a commodity I have very little of. What settings in Audiogate do you prefer?

We worked for months with HDtracks in choosing the best DSD->PCM conversions for all the SACD rips we were doing. We tried AudioGate, Saracon, SBM, Pyramix and even DiscWelder. In every case, the Saracon converters just seemed to take the life out of every file. I've posted samples here for folks to test for themselves.

Did HDTracks take your advice? What software do they employ?

Lastly, by "here" do you mean in this section of the forum? I will take a dig around.

I even created a thread on what people should listen for when they compare hi-rez files to Redbook.

But we are comparing different methods of DSD to PCM conversion.

Incidentally, does anyone know how the Oppo converts DSD to LPCM on the fly? Filters?
 

stevekale

New Member
Aug 8, 2012
83
0
0
What is above the upper limit of our hearing we can’t hear.
But all above this threshold that maps into our audible range we do hear.

But of course! Which is why I make the point that one must look at what's happening within the audible range and not what's happening outside it even though the affect in the audible range may be caused by interjection outside the range. The graphics posted at the beginning of this thread all focus on 20k and up.

And how can a filter outside the audible range make a file of observations be exactly the same as a file with half the number of observations?
 

stevekale

New Member
Aug 8, 2012
83
0
0
It would be great if we could hear more from Daniel.



PS: Bruce, your studio sounds awesome! (A few words on the rest of my system - all speakers are Egglestonworks, Andra IIIs up front, Andra centre, Rosa for the surround; Theta CB III HD as mentioned before; Krell FPB 200 driving the front and 2 Krell KAV 150a amps driving the rear and centre; Synergistic Research Element Copper speaker cabling up front.)
 
Last edited:

Bruce B

WBF Founding Member, Pro Audio Production Member
Apr 25, 2010
7,006
512
1,740
Snohomish, WA
www.pugetsoundstudios.com
That's why I said "hear and feel". Does Saracon filter away low frequencies as well?

No... but you can also feel ultrasonic frequencies. Ever been to a dentist?


But, again, how can it be "the exact same file" if the frequency of samples is vastly different?

Because the files will null to -70 to -80dB

It sounds to me like your criticism of Saracon is not that the files are the same but that Saracon's filter discolours in some way the sound that we can hear, that the effects of the filter at this level are discernible versus if the DSD were converted directly to analogue or, more to the point, versus the rendition that can be achieved with other software. I'm glad you like Audiogate. The problem is, I've heard completely the opposite about its quality from many people. The fact that it doesn't do multichannel is also not in its favour. Given it's free I will give it a go but experimenting takes time, a commodity I have very little of. What settings in Audiogate do you prefer??

Yes, that's one reason I don't prefer Saracon for DSD work.

Did HDTracks take your advice? What software do they employ? ?

We had a custom filter written for us that starts out at ~60k and goes out to 100k

Lastly, by "here" do you mean in this section of the forum? I will take a dig around.

Yes

But we are comparing different methods of DSD to PCM conversion.

But this can also be applied to 192k files vs. 88.2 files.


This whole thing came about when the "Audacity Cowboys" started checking Rolling Stones tracks and were upset that there was no content above 35k even though they had purchased 176.4 files. We then checked and found the filters were engaging in the same place no matter what sample rate we chose. The Rolling Stones files were done by a 3rd party using Saracon.


I can offer this... put up a file or two using 3 different sample rate converters and let folks decide which one they like best.
 
Last edited:

Vincent Kars

WBF Technical Expert: Computer Audio
Jul 1, 2010
860
1
0
I must be missing something very fundamental if this isn't the case.
Probably nothing fundamental but something pragmatic.
If the sample rate is fs then the highest frequency allowed is fs/2

If you convert a DSD to PCM and the software used applies a filter at 35 kHz than 2x35=70 kHz
One might argue that using this software to convert DSD to PCM with a sample rate > 70 is nonsense as there is no frequency > 35 in the output due to this filter.

Edit: missed Bruce’s answer when writing this
 

stevekale

New Member
Aug 8, 2012
83
0
0
No... but you can also feel ultrasonic frequencies. Ever been to a dentist?

That's a vibration I prefer not to feel ;-)


This whole thing came about when the "Audacity Cowboys" started checking Rolling Stones tracks and were upset that there was no content above 35k even though they had purchased 176.4 files. We then checked and found the filters were engaging in the same place no matter what sample rate we chose. The Rolling Stones files were done by a 3rd party using Saracon.

This is the bit I am failing to grasp. I would be interested in the 176.4k samples per second not because I want to know that it has samples with frequencies above 35kHz but because every second of audio I can hear will be represented more accurately.

A hypothetical example: we record a track with lovely tonality all within the audible range and for the last 10 seconds after the song has finished the instruments let rip a constant tone at 100kHz. If I sample this at 44.1k samples per second and then at 88.2k samples per second (no filters), which would sound better? Surely the second and not because one records the 10 second blast at the end better than the other.

I know that the Nyquist-Shannon sampling theorem "states that perfect reconstruction of a signal is possible when the sampling frequency is greater than twice the maximum frequency of the signal being sampled" (to quote Wiki because I'm lazy). So we care about sample rate relative to frequency range. Put another way, increased sample rate allows us to "perfectly reconstruct" a wider frequency of signal. And because of filtering and noise its beneficial to oversample. But do I really care that I can accurately reconstruct inaudible frequencies? Do I even care whether the 10 second blast in my example is rolled off at 80kHz or recorded at all?

In photographic printing we talk of "pixel peeping" - burying one's nose deep into an image from a very close distance or with the aid of a loupe, well beyond what is useful or important with respect to viewing the print at a sensible distance. While it can be useful if we're trying to find the culprit for something that manifests itself at the appropriate viewing distance (e.g. colour casts in B&W printing with a regular colour-inked digital printer) it is completely irrelevant if there is nothing wrong with the image in normal viewing conditions.

Why would I want or care to peep into an LPCM file to find observations well above the audible range? Does finding such observations prove that the sample rate is indeed 176.4? Not at all. Does the absence of them prove a lower sample rate? Not at all. What I do care about is whether I got the benefits 176.4k sampling provides (greater resolution, lower noise, aid to avoiding anti-aliasing etc). I don't care whether inaudible samples were more accurately rendered. But, yes, I do care if they've been rolled off in a manner which would affect the audible parts of the recording. If a filter at circa 35kHz were to roll-off frequencies above that level and not impair that below it then its happy days. So I shouldn't be upset that there's no content above 35kHz - regardless of the sample rate - and only be unhappy a filter removed such "information" in a manner which was detrimental to the audible range. This could presumably happen regardless of the sample rate.

At the end of the day, I agree a hearing test (blind software A against software B) is the true one but I've really had trouble understanding some of the arguments put forth in here against Saracon. I wonder if Daniel will now take the floor.
 

mep

Member Sponsor & WBF Founding Member
Apr 20, 2010
9,481
17
0
This whole thing came about when the "Audacity Cowboys" started checking Rolling Stones tracks and were upset that there was no content above 35k even though they had purchased 176.4 files. We then checked and found the filters were engaging in the same place no matter what sample rate we chose. The Rolling Stones files were done by a 3rd party using Saracon.

I bought one of the Rolling Stones supposed 24/192 albums from HDtracks and the sound was horrid. I emailed them and told them I was bummed at how crappy the album sounded and to their credit, they let me download any other album I wanted. I knew something was dead wrong and I didn't have to be blindfolded and hear ten passes to figure that out.
 

stevekale

New Member
Aug 8, 2012
83
0
0
Because the files will null to -70 to -80dB

But this ignores all the additional observations in a 176k file versus 88k file that are within the audible range (or under the effect of the filter)

I thought it interesting to run a spectrogram of the first 30 seconds of a DSD file processed to 24/176 PCM first by Saracon and second by Audiogate (soft roll-off) (no dither in either, 24/176). The effect of the differing filters above 35k is clear but I don't think one can draw any conclusions with respect to the quality of the audible content from such charts. I would also expect the 24/88 Saracon chart to show a similar cap in frequency but again there would be a lot less observations, both in total and within the audible range, and presumably less resolution as a result. If I looked at the Saracon chart I also couldn't conclude that the sample rate was half the Audiogate one and that as a result I had been ripped off. Is it a lower resolution or a filter?







Again, Saracon first and Audiogate second.




 
Last edited:

Bruce B

WBF Founding Member, Pro Audio Production Member
Apr 25, 2010
7,006
512
1,740
Snohomish, WA
www.pugetsoundstudios.com
The effect of the differing filters above 35k is clear but I don't think one can draw any conclusions with respect to the quality of the audible content from such charts. I would also expect the 24/88 Saracon chart to show a similar cap in frequency but again there would be a lot less observations, both in total and within the audible range, and presumably less resolution as a result. If I looked at the Saracon chart I also couldn't conclude that the sample rate was half the Audiogate one and that as a result I had been ripped off. Is it a lower resolution or a filter?

This is indeed the problem. Too many people are trying to judge audio quality by looking at graphs and specs. I've been preaching for years to listen with your ears and not your eyes. You can never judge audio quality in this way. Just like everything else, this is just one tool out of dozens that you can use. The only time I ever use a spectrogram, like the one above, is for restoration work. I much rather use the FFT graph that I usually post in these forums. It gives you much more information. After using Saracon, you could never tell the provenance of a file to determine if it came from a DSD recording.
 

stevekale

New Member
Aug 8, 2012
83
0
0
My point exactly. Just because the spectrogram/FFT shows >40k (or whatever) is rolled off doesn't say anything about quality of the audible product nor its resolution.

BTW, was the custom filter you mentioned being used with Audiogate or some other software?

Also, I noticed you use an Oppo as (a small) part of your setup. Do you by chance have any insight as to what they do with their DSD -> 88kHz LPCM conversion?
 

Bruce B

WBF Founding Member, Pro Audio Production Member
Apr 25, 2010
7,006
512
1,740
Snohomish, WA
www.pugetsoundstudios.com
My point exactly. Just because the spectrogram/FFT shows >40k (or whatever) is rolled off doesn't say anything about quality of the audible product nor its resolution. BTW, was the custom filter you mentioned being used with Audiogate or some other software?
Also, I noticed you use an Oppo as (a small) part of your setup. Do you by chance have any insight as to what they do with their DSD -> 88kHz LPCM conversion?

That's for sure....

The filter is used with MatLab and Pyramix.

Our Oppo was modified by Exempler and I only use it for surround playback via analog outs. I don't have a clue how it gets there!
 

About us

  • What’s Best Forum is THE forum for high end audio, product reviews, advice and sharing experiences on the best of everything else. This is THE place where audiophiles and audio companies discuss vintage, contemporary and new audio products, music servers, music streamers, computer audio, digital-to-analog converters, turntables, phono stages, cartridges, reel-to-reel tape machines, speakers, headphones and tube and solid-state amplification. Founded in 2010 What’s Best Forum invites intelligent and courteous people of all interests and backgrounds to describe and discuss the best of everything. From beginners to life-long hobbyists to industry professionals, we enjoy learning about new things and meeting new people, and participating in spirited debates.

Quick Navigation

User Menu

Steve Williams
Site Founder | Site Owner | Administrator
Ron Resnick
Site Co-Owner | Administrator
Julian (The Fixer)
Website Build | Marketing Managersing