Is recording at higher sample rates worthwhile if ultima

Discussion in 'Mastering' started by Rowan, Feb 7, 2003.

  1. Rowan

    Rowan Active Member

    I have a 24/96khz ad/da and maintain audio at 32 bits internally for processing and mixing prior to dithering back to 16 bits for burning to CD.

    Is there any advantage in recording at 88.2 or 96khz and then sample rate converting back to 44.1 at the end for CD? My guess is that any advantage will be completely lost due to both artifacts caused by the sample rate conversion process and the ultimate 22.5khz upper frequency limit and associated filtering. :confused:
     
  2. Ethan Winer

    Ethan Winer Active Member

    Rowan,

    > Is there any advantage in recording at 88.2 or 96khz and then sample rate converting back to 44.1 at the end for CD? <

    No advantage that I can see. Others may disagree, but my feeling is that any perceived improvement from recording at high samples rates - whether you convert later to 44.1 or not - is either imagined or else due to other variables. Like one system has better A/D/A converters, which renders the comparison invalid.

    --Ethan
     
  3. Kurt Foster

    Kurt Foster Distinguished Member

    First off to Ethan, Respect!! But ... I have done a/b comparisons using a 2" analog tape playback through a MCI 600 console and an Apogee PSX 100 at 96, 48 and 44.1 and the difference between 48 and 96 is dramatic! It doesn't take a golden ear to hear it. I concede that it may not be so noticeable in all but a direct a/b comparison. So when tracking in digital at higher rates and better high frequency response, the lower harmonics are excited by the upper ones. While you are going to down sample to a lower rate for Redbook, there is a difference in the end product. Fats
    ------------------------------------------------------------------------
    Tannoy, Dynaudio, Blue Sky, JBL, Earthworks, Westlake, NS 10's :D , Genelec, Hafler, KRK, and PMC
    Those are good. …………………….. Pick one.
    ------------------------------------------------------------------------
     
  4. audiowkstation

    audiowkstation Active Member

    Fats, this is true. Another reason to do things in hi-bit is you may want to release a DVD-A disc.

    Their are two consumer audio high resolution digital formats. Both of them are growing but their is a huge division between the users of the formats. It is like the VCR wars of the past between beta and vhs.

    The formats are:

    DVD-A (digital versatile disc/Audio format)

    SACD DSD (Super audio compact disc/Direct stream digital)

    DVD-A discs are played back through a DVD-A player. This is a different machine that the DVD machines for movies. All DVD-A compatible playback machines are also DVD players but DVD players may or may not have DVD-A compatibility. DVD-A's are multichannel (6 channel/5.1) and they must sum this downmix internally for 2 channel reproduction. They are using PCM (pulse code modulation) at 24 bits, 88.1, 96 or 192K. Most discs are 24/96 or 24/192. The authoring in mastering for them is slightly more expensive that redbook (40% more) but if you are prepared to release your material in DVD-A, upsampling an original 16/44.1 to the format is ill advised unless it is remastered professionally and authored professionally, including text, footnotes and video.

    For me to do a DVD-A, it cost MUCH more than redbook. I have the goods to do it..but the format is considered "new" and not many takers yet...really no takers. I have done a few for ME and it is a wonderful format. Head and shoulders above redbook.


    SACD/DSD

    This format seems to be the choice for the audiophiles. Direct stream digital samples at single bit 2.82MHZ! The format can be multichannel (5.1) stereo channels (2-bus..not summed) and contains the redbook layer as well. The days of an SACD only disc are soon to pass as all of the current releases on SACD are "dual layer hybrid" and can work on redbook or SACD players. At this time, the cats that are hand selected by Sony are butchering the format IMHO. Of the SACD's I own, I have redbook CD's that slay the SACD's. The guys in charge of remastering for SACD on "average" are missing the mark. Roger Waters "in the flesh" and "TOTO IV" are sad examples of this format. For good examples, look toward TELARC. They do it better. Sony has its proverbial head up its ass with this "superior" technology. It has been out now 4 years and their are only 800 titles. The one to watch for will be released 03/03/03. It is "dark side of the moon" by Pink Floyd. Unfortunantly, Alans original Quad mix was not used and he was not invited to remaster it (Head up ass syndrome again) so until this is officially released, we will have to wait and see if it can come close to the QS4 Quad album. To master to SACD means it basically will cost you an arm and two legs. Right now this boys club is doing them at 10 grand a pop and pressup is around 7.75 per unit. IMO, the format is not worth it due to the crappy mastering I have experienced. Every SACD I own could have been done much better by your truly. This is fact..not blowing smoke here. Michael has heard it too.

    Ok, sorry for the book but really, if you want to use one of these emerging formats, you should be recording at least 24/192 or 24/96..

    SACD/DSD mastering stations are still basically too expensive to make money with right now and DVD-A is extreamly affordable, time consuming (multichannel) and can be burned in house without being a slave to Sony. Their is only one pressing plant for SACD's in the USA at this time. The total now is up to 4. Anyone with 500 dollars can burn a DVD-A.
     
  5. Ethan Winer

    Ethan Winer Active Member

    Fats,

    > First off to Ethan, Respect!! <

    Hey, all of this is with great respect for each other, and a desire to learn and understand. And also with the knowledge that what works for one person doesn't have to work for another.

    > I have done a/b comparisons using a 2" analog tape playback through a MCI 600 console and an Apogee PSX 100 at 96, 48 and 44.1 and the difference between 48 and 96 is dramatic! <

    Just so I'm clear, you are saying that you did two identical mixdowns, one right after the other, where nothing changed except the sample rate? And I assume you recorded into a computer DAW?

    If the mixes were identical and a difference was clearly audible, I wonder what actually changed from one pass to the other. I don't believe anyone can hear much past 20 KHz., or even that our hearing is influenced by supersonic content. But some people do claim to hear a meaningful difference between 44.1/48 and higher sample rates. So my interest is finding out what is changing in the audible band when the sample rate is switched.

    > I concede that it may not be so noticeable in all but a direct a/b comparison. <

    If it's clearly audible then it's definitely worth understanding!

    --Ethan

    PS: Why is it that whenever thus stuff comes up, the other person is too far away for me to drive over so I can listen for myself? :D
     
  6. audiowkstation

    audiowkstation Active Member

    Ethan, you are right that doubting anyone can hear past 20K; ALTHOUGH... the point is it is not that you can hear that high, it is what the presence of the freqencies above 20K do to the shape of the waves we do hear. Take a 30KHZ frquency sine wave...pulse it from -20 to -6dB in slow (about 3 seconds between) fade ups and fade downs While you have a steady state 1KhZ @ -4dB tone running. You see how the presense of the 30khz wave changes the shape of the 1khZ tone. Do this with loudspeakers making sure A, they can get up there and B, you are point source from the tweeters as loudspeakers dispersion at 30 K is very narrow. Providing your mic pres and the microphones can do it, the exact same thing happens on playback only more so. The presence of the signal changes the shape electrically and the presence of the signal "in the air" changes the shape acoustically. When I put a 50khZ wave through the system at 85dB reference to a 1KhZ signal..my Little Dogs head will twist and turn looking at the speakers. It is there.

    Now, like I say, it is not that we can hear these freqencies but it is there presence that changes the shape of the ones we do hear and that...is clearly audible.

    When downconverting, the changes in the shapes of audible freqencies are retained.

    Flat response to 100KHZ is the goal for high fidelity reproduction IMHO
     
  7. Ethan Winer

    Ethan Winer Active Member

    Bill,

    > the point is it is not that you can hear that high, it is what the presence of the freqencies above 20K do to the shape of the waves we do hear. <

    I don't see why that would matter. Yes, if you have material with audible content and also supersonic content, the supersonic content will be visible on a 'scope and will come out your speakers if they can reproduce that high. But since nobody can hear the supersonic portion, I don't see how it's presence affects the audible portion.

    For example, if you strike a cymbal or orchestra triangle, frequencies far beyond 20 KHz. are generated. You can then capture that with a microphone that responds to super high frequencies. I have done this, and watched the result on a hardware spectrum analyzer and seen content all the way out to 50 KHz. But though the content past 20 KHz. is present and visible, it contributes nothing to the perceived sound.

    Years ago a friend who's an engineer at Hewlett-Packard brought over some high end test gear, including the spectrum analyzer mentioned above. He also brought over a sweepable low-pass filter, and we played with that too using a set of car keys jingling in front of a good mike. This is the test where we saw content out to 50 KHz. We then lowered the filter cutoff frequency until we could hear a difference - the knob read 18 KHz. or something close to that.

    Which brings us back to why a recording made at a high sample rate sounds different than 44.1.

    --Ethan
     
  8. audiowkstation

    audiowkstation Active Member

    I guess I was not clear...put it in other terms.

    The interference patterns that the presence of ultra high frequencies create add to the color of the ones we do hear. Removing them and the inteference is removed. Buy having them there in the first place, the interference that they caused will hold over on the audible freqencies. If they were never there to begin with, the intererence would not be there.

    If you drop a large rock into a pond and watch the waves, dropping a small pebble within those waves "interferes" with the large waves. Had the small pebble not been dropped, it would not have happened. Now if you have the fundamental of the longer wave intact with the interference of the pebble saved, you would not need said pebble to duplicate the event. The smaller pebbles signature is already "set in stone" (pun intended).

    This make any sense at all?

    By going back to 44.1K you are not reshaping the lower waves, the interference is still there, you are simply removing the source of the interference.

    Perhaps I need to do some screen shots and show you what I mean.

    It will take time though..and really today, I have paint work to finish.
     
  9. Ethan Winer

    Ethan Winer Active Member

    Bill,

    > The interference patterns that the presence of ultra high frequencies create add to the color of the ones we do hear ... dropping a small pebble within those waves "interferes" with the large waves ... This make any sense at all? <

    I did understand you the first time, but I'm afraid it does not make sense. Using your analogy, fast ripples correspond to the high frequency content and slow ripples to the audible content. But unless there is a nonlinearity in the system - distortion - one does not affect the other. And one can be filtered out completely without affecting the other in any way.

    I know it may seem that this would be audible, but it's really not. Again, in order for frequencies to interact there needs to be a non-linearity somewhere. And then you'll have IM distortion which creates sum and difference products. But in modern digital systems IM and other distortion components are extremely low - likely below the -96 dB. noise floor of 16 bits. A distortion figure of 0.01% means the distortion components are 80 dB. below the signal. I think most modern A/D converters can manage even better than that. Further, anything that's -80 below the music, and also masked by the music in addition to being so soft, is inaudible as far as I'm concerned.

    > Perhaps I need to do some screen shots and show you what I mean. <

    No, I already know what high and low frequencies at the same time looks like. :)

    --Ethan
     
  10. audiowkstation

    audiowkstation Active Member

    You will always get wave interference when you pass two soundwaves through a medium called "Air"
     
  11. Ethan Winer

    Ethan Winer Active Member

    Bill,

    > You will always get wave interference when you pass two soundwaves through a medium called "Air" <

    Yes, but not in the way you are thinking! Again, the interference must be nonlinear in order for one frequency to "interact" with another. Without nonlinearity the two frequencies are merely combined, and so can be easily separated again later with simple filtering. One does not affect the other in any way; they merely coexist at the same time on the same piece of wire, or in the same volume of air.

    Let's back up a few messages...

    > The interference patterns that the presence of ultra high frequencies create add to the color of the ones we do hear. <

    What sort of color do they add? What does it sound like? Most important of all, how do you know this is true, and what tests have you performed to arrive at this conclusion?

    > having them there in the first place, the interference that they caused will hold over on the audible freqencies. <

    Again, the same questions: What sort of interference specifically? What does that interference sound like? How do you know this is really the case?

    I'm sorry if this sounds confrontational, because that's really not my intent. :) But you can't just say there's "interference" and so it must be audible.

    --Ethan
     
  12. audiowkstation

    audiowkstation Active Member

    I need to paint today..but I can demonstrate the shift in the shape of 1KhZ when 50K is applied and this is at 24/192 then I can resample this to 16/44.1K and the 1K wave is still shifted, even with the absense of 50K.

    No system is absolutely linear.

    I suppose that manufactures made a huge mistake in developing higherbit recording technologies?

    What test have I done,

    Many. Their would be no point whatsoever for any sampling frqency above 32K if we could only hear to 16KHZ? IS that wat you are saying?

    So, test one involves 3 computer towers running at 3 different sampling frqencies. Tower one was running 24 bit/192K. Tower 2 at 24/96K and tower 3 running at 16/44.1K when I recorded a classical concert in the year 2001. I did this so I could burn 3 separate 16/44.1K and put this to rest for me. Hell, I don't like rendering a 4gig file anymore than the next cat.

    I burned 3 CD's of each recording directly and I had another output going to a stand alone 1" machine (16track @15ips) .

    The files at 192K that I burned 3 CD's from were the best. They were alive, cleaner, clearer. Their are members of this board and even moderators of this board that have heard this CD. One moderator heard both the 24/192K transfer and the 16/44.1K transfer.

    That was the test, it conclusively proved that their is a reason to go to higher sampling frequencies.

    Second. Why is their any equipment that goes above 44.1K? Why are DVD-A selling? Why are SACD's Selling?

    Because 44.1K is a lousy idea from the get go. Only reason it was settled on was technology was not up to par at that time and video recorders were being used to transfer and edit audio digitally in the old days.

    Perhaps my loudspeakers and my room and my equipment and my ears can account for most of the differences and why I master in higher bitrates. Fats has heard the difference. Rick Hammang has heard the difference and depending on the set-up, it is there. The difference is *it sounds closer to the actual performance*

    I can take the 9CD's and 100% blindfolded in even the car stereo, I tell you which is which. Ever heard of ringing? What happens to the frequency response as you approach nyquist? The filter is so sharp that when you use it in recording, use it in mixdown, use it in mastering and use it in pressing..all this adds up.

    Perhaps my ears are playing huge tricks on me. You do not have to believe. But if you want to see if I know what I am doing or not, you can go over to the Audio Projects site and look at the comments of artist I have remastered their wares and see what they think of the work.

    We are moderators here. To tell me I do not know what I am talking about will meet resistance, especially when I can prove it time and time again.
     
  13. Ethan Winer

    Ethan Winer Active Member

    Bill,

    Let me address your last point first.

    > But if you want to see if I know what I am doing or not ... To tell me I do not know what I am talking about will meet resistance <

    This is what I'm trying to avoid! My goal is to stay on topic and discuss the issues only. If I implied anything otherwise I apologize as that surely was not my intent.

    > I can demonstrate the shift in the shape of 1KhZ when 50K is applied and this is at 24/192 then I can resample this to 16/44.1K and the 1K wave is still shifted, even with the absense of 50K. <

    Yes, this is exactly what I'm getting at. Though I'd rather see a spectral analysis than a picture of waveshapes, because waveshapes can be changed dramatically with no audible affect by, for example, phase shift.

    > No system is absolutely linear. <

    Agreed, and that's why I made the point that nonlinearities in modern digital systems are low and further masked by the program itself. So this is one avenue worth persuing: independent of ultrasonic content, are components that are 80 dB. or whatever below the program audible?

    > Their would be no point whatsoever for any sampling frqency above 32K if we could only hear to 16KHZ? IS that wat you are saying? <

    Almost. I am not saying that higher sample rates cannot sound different than standard 44.1. But what I do question is the reason why they sound different. If it turns out that the difference is due to avoiding ringing near the filter cutoff, then perhaps the ultimate solution is to make better filters, or oversample at an even high rate, to avoid the waste of resources used by high sample rates yet still retain the improved fidelity.

    > Hell, I don't like rendering a 4gig file anymore than the next cat. <

    Right, and this is precisely my interest in getting to the bottom of what matters, what doesn't, and why.

    --Ethan
     
  14. SonOfSmawg

    SonOfSmawg Well-Known Member

    OMG this is so funny! MODERATORS in debate! Nobody's getting rude or disrespectful, it's all good, so I aint toughing this! Keep it clean, guys, because this is great! FINALLY, two guys that can have a good debate without going below the belt!
    BTW ... even though I only have a little home studio, and seldom participate in the studio threads, even I know there is no grey area on this question. One of you is right, and one of you is wrong, but I'll be damned if I'll say a word! I wanna watch this! ROTFLMFAO!
     
  15. audiowkstation

    audiowkstation Active Member

    A better filter. Well they should have figured out that at least 350K sampling frequency would allow for shallow slope filters to be used but you have to be at least 70dB under the source level at the sampling frequency. 22.5K means 70dB per octave. and the therum is more like 83dB. Filters that cut off that steep will ring. No doubt about it. It is audible to even this 44 year olds ears. The ringing starts in the 12K region (9th order harmonics) and is clearly audible. Now if you have the same situation done 5 times over (recording, mixing, editing, mastering, in the CD player) you cannot possibly say that this is inaudible. The reason for 192K is to get you at least flat to 50K before the ringing at 65K starts. Filters are filters. The steeper the slope, the more they will ring. When I designed loudspeaker crossovers commercialy, I was adimate that the tweeters be beefy enough in construction but light enough for transient response that I could do 1st order because it sounds so much better and intergrates with other components so much better. I actually came up with 3dB/octave slopes using a battery of bypass resistors to cut the sharpness of even a 6dB/octave slope down some.

    Nope, filters are the culpret and the only way to overcome the ringing filter syndrome (not to mention leaving the upper harmonics intact whether you hear them or not..I DO... I know when they are there or not instantly) is to raise the sampling frquency to astronimical levels and employ less steep curves to the shelf of the filter. A multishelved filter (as philips employes in my A/D D/A) is fully adjustable and audible. I got this piece of gear from a corsortum of manufactures because of my involvement in testing, calibration and reviewing in the Higher fidelity realum.

    BIO


    R1

    R2

    R3

    Others


    I can simply say this. At no time will digital ever be able to give flat freqency response that is necessary to humans to achieve the best the format can offer without limitation, if steep filters are employed in the audible spectrum. Many folks have the ability to hear the absense of upper harmonics. I for one can every day , even with a headcold. I have been employing wave shaping techniques to digital mastering now for 20 years...yes since late 1982. I know that going back to analog and repairing the losses and putting it back together does restore the build-up of all these filters and it is not so bad of it happens once, instead of multiple times.

    Ok Smawg, IF I am wrong, Fire me. I got other things to do than beat this tired old horse.
     
  16. Doug Milton

    Doug Milton Active Member

    I don't want to step into the middle of this and incur anyone's wrath, but…..there is actually a question at the end of a bit of rambling.

    When asked about sample rates by people who know nothing about technology, I draw them an analog sine wave. Think of sample rate as taking with a camera "snapshots" of that sine wave. At 44.1k we really have 44,100 snap shots per second. If we were to zoom in, we would see a stair step effect as there was signal happening between snap shots. Simple logic (as works in my simple mind) would seem to indicate that by doubling or tripling the number of snap shots taken in that same amount of time, would result in smaller "steps", getting us closer to the pure smoothness of our original analog signal. It would seem to me that regardless of output, starting with the highest possible resolution would have to result in a smother (truer to analog, less digitized) sound.

    Q: Does that not seem logical regardless of what frequencies I can hear?
     
  17. Ethan Winer

    Ethan Winer Active Member

    Doug,

    > It would seem to me that regardless of output, starting with the highest possible resolution would have to result in a smother (truer to analog, less digitized) sound. <

    Yes, it would seem that way, but that's not really how digital audio works. The filtering by the D/A converter smooths out the steps so as to completely restore the original waveform.

    What results when the bit rate is low is distortion that can be directly computed based on the number of bits. Assuming, of course, a D/A converter that is high quality and thus limited by the number of bits and not limitations in its own circuitry. Even 16 bits, when recorded full scale, can yield very acceptable distortion values.

    And what results when the sample rate is low is a reduction of the highest frequency that can be captured. Theoretically, 44.1 KHz. can capture at least to 20 KHz., though as you can see there is some disagreement as to the audibility of other factors, such as filter ringing and other artifacts.

    --Ethan
     
  18. audiowkstation

    audiowkstation Active Member

    The earth is flat.....the earth is flat.......

    Oh $*^t!
     
  19. audiokid

    audiokid Staff

    hehe . That is funny :w:
     
  20. Kurt Foster

    Kurt Foster Distinguished Member

    Once again,....
    I know this is true. Higher sample rates in tracking sound better when downsampled. Not dithering , that's bits but sample rates. I have heard this in A/B tests in a controlled environment on a good monitor system. Believe me "the earth ain't flat!". It's not like I have an agenda regarding this. I would love to be able to say don't worry about it, it makes no difference, save your money. But I can't. it makes a difference. Fats
    ------------------------------------------------------------------------
    Tannoy, Dynaudio, Blue Sky, JBL, Earthworks, Westlake, NS 10's :D , Genelec, Hafler, KRK, and PMC
    Those are good. …………………….. Pick one.
    ------------------------------------------------------------------------
     

Share This Page