PT HD and SACD [Archive] - Avid Pro Audio Community

Nika

01-25-2002, 02:10 PM

[QUOTE]Originally posted by dBHEAD:
It seems to me that Digi has taken an inherently astute step with the new HD system. This is speculation on my part (and, frankly, theirs), but I believe SACD with a 5.1 mix will end up taking the place of the conventional CD in a few more years.

It really depends on the record companies. Half of them are gravitating toward SACD. The others seem to be going after the PCM market, such as DVD-A

Now, as many of you know, the SACD has a sampling rate of 2.822 mHz --approximately 64 times the sampling rate on CDs. In a nutshell, the rationale for such a high sampling rate seems to have been that you can much more closely reproduce complex waveforms and therefor sonic detail. At least that seems to be the jixt of the technical FAQs I've read on it.

Not really. The benefit being touted is the absence of any filters in the signal path, combined with reproduction of higher frequencies. I haven't seen anyone talking about more accurately representing the signal because in many capacities it doesn't. It actually does a much worse job of representing the signal depending on the frequency of the signal itself.

It also allows for reproduction of high frequencies beyond the human hearing range

Yes, but oftentimes filters are still installed in those systems at around 50kHz. Then again, people who think that 192kS/s would be beneficial seem to be endorsing 100kHz responses anyway.

But why? Does it even make any actual sonic difference? In theory, it isn't supposed to but my understanding is that even though people cannot consciously tell a difference, research has shown the brain does react to those frequencies when present. There is a theory (just a theory) that one result is that people, subconsciously, actually "feel" better about music recorded with extended frequency responses.

There is a lot of reading to do on this subject. You're speaking here of the Oohashi study out of Japan.

A bunch of B.S.? Maybe, but it seems unlikely Sony and Phillips would have invested in a technology so heavily if it was just going to be a fad. Clearly upsampling will be involved, but I have to think 192k would sound better (er, okay, feel better) than 44.1k when upsampled.

Hmm. Why?

Anyway, I'd like to hear opinions on this issue because, frankly, I believe the whole controversy about higher sampling rates revolves around it. In essence, these are the questions: 1. Do you think there is a difference in a listener's internal reaction to improved sonic detail, even though they cannot consciously hear it? and 2. How high is high? Is the 2.822 mHz sampling rate of SACDs a marketing gimmick or will we be seeing PT 2.822 mHz in the future? Even though higher sampling rates seem certain, is 2.822 mHz ridiculous?

You're presenting DSD as though it's comparable to PCM. It is a whole different method of sampling, and is therefore not really accurate to compare to PCM in this manner.

Nika.

dBHEAD

01-25-2002, 05:03 PM

Zeus:

Your point about mics and pres not being able to put out those frequencies anyway was exactly my reaction when I first heard about SACD. However, I've since learned that, at least on the preamp end, some pres are designed to pass extremely high frequencies. The Avalon M5, for example, has a frequency response from 1 Hz to 120 kHz. The Grace Design pres go all the way up to 1 mHz.

Now mics, admittedly, are a different story. At least going by published specs, almost none exceed 20 kHz. Yet, even that figure is usually in the context of a 3 dB or 10 dB tolerance --so it doesn't mean they don't capture any of the really high frequencies.

But your point is well taken about all the different components in the chain, and your point about consumer gear not even being close to being up to the task is definitely a strong argument for the pointlessness of focusing on super high frequencies.

Nika:

My point about upsampling is based on the idea that you'd capture more of the high frequencies (beyond 20k) with 192k than 44.1k. If the Nyquist theory is applied you'd be able to capture frequencies up to 96 kHz. I was referring back to the point about higher frequencies possibly having that sort of subliminal effect of making you feel better.

As far as comparing PCM and DSD -- I have to admit some ignorance there. I thought sampling was sampling: periodically assesing amplitude and converting into a binary expression, but evidently there are some differences I'm unaware of. I'll have to try to read up on that one. But if you happen to have a link, I'd be thankful.

Nika

01-25-2002, 06:41 PM

[QUOTE]Originally posted by dBHEAD:
Zeus:

Your point about mics and pres not being able to put out those frequencies anyway was exactly my reaction when I first heard about SACD. However, I've since learned that, at least on the preamp end, some pres are designed to pass extremely high frequencies. The Avalon M5, for example, has a frequency response from 1 Hz to 120 kHz. The Grace Design pres go all the way up to 1 mHz.

As an owner of both Grace and Avalon boxes as well as microphones spec'd above 20kHz I feel it appropriate I respond here. The only reason the those devices are spec'd beyond 20kHz is to indicate their ability to reproduce proper phase below 20kHz. Any mic preamp has to have specs up in the xxxkHz range in order to sound anywhere near transparent at 10kHz. A slight deviation in its ability to reproduce high frequency data will greatly affect the phase at which audible data is transmitted, and this would result in poor results. They don't spec up to 500kHz because they anticipate any signal will be run through it, but rather to show that it will not affect phase of audible signal because it is flat up so high.

Microphones are the same way. Mics that capture up to 40kHz ACTUALLY capture 20kHz in better phase. That's all. Sure, it is CAPABLE of capturing HF data above our hearing range, but that is not where the audible effect is.

My point about upsampling is based on the idea that you'd capture more of the high frequencies (beyond 20k) with 192k than 44.1k. If the Nyquist theory is applied you'd be able to capture frequencies up to 96 kHz. I was referring back to the point about higher frequencies possibly having that sort of subliminal effect of making you feel better.

I think that most equipment has filters that insure that data up to 100kHz doesn't get through anyway. Many pieces of equipment won't allow any signal in above 50kHz so as to insure that these frequencies don't wreak havoc upon the analog components.

As far as comparing PCM and DSD -- I have to admit some ignorance there. I thought sampling was sampling: periodically assesing amplitude and converting into a binary expression, but evidently there are some differences I'm unaware of. I'll have to try to read up on that one. But if you happen to have a link, I'd be thankful.

You have a lot of reading to do! It is important to understand the foundation of the principles you're discussing!

For the quick explanation, DSD doesn't measure the voltage at unique and predictable increments of time. It rather measures whether the signal is higher or lower than the current quantization step. If the signal is higher then a "1" is transmitted, and the next sample goes at the next higher quantization step. If the signal is lower then a "0" is transmitted, and the next sample goes at the next lower quantization step. Then the analysis is done again. The faster this is done the higher the frequency it will be able to capture, depending on the amplitude of the waveform, and thus the slew rate of it. I haven't done the math myself, but apparently 2.28MHz is able to reproduce up to 100kHz or so at maximum amplitude.

Anyway, only a 1 bit stream is used in this case, but the data has to be turned into PCM in order to really process it, and the data is inherently noise as the signal is constantly going up-down-up-down-up-down, etc. when in PCM it would be able to stay at a fixed quantization step.

It's really two totally different art forms, each with a unique set of problems. DSD has a real problem with too much HF data and the inability to properly dither the signal without clipping. Comparing the two in this format is really misleading.

Nika.

Zep Dude

01-25-2002, 07:20 PM

but the data has to be turned into PCM in order to
really process it, and the data is inherently noise as the signal is constantly going
up-down-up-down-up-down, etc. when in PCM it would be able to stay at a fixed quantization step.

Actually as of the AES show, Sony was demonstrating a functioning 8 track computer based DSD workstation that applied EQ and Compression without going to PCM. This was not a system they are marketing to the public but more a demonstration of the technology for licensing purposes. Still, the Sony dude I talked to felt it was a year or two before we'd likely see the "Alesis DSD workstaion" type of thing.

The one demonstration I attended at Sony Studios consisted of a live jazz band being recorded. The engineer switched between the the console direct output and then the ouput of the DSD recorder (being fed from the console output obviously). The LACK of difference between the two signals was the real beauty and won me over to DSD. A very non-digital sounding digital.

I've read a few audiophile reviews of DSD players where the reveiwer, previously a vinyl snob, stated that this was the first time he had heard digital approach the fidelity of analog.

I have yet to do a side by side shootout of DSD vs PCM and frankly, until we have DSD as a workstation option it isn't really relevant. Deep down though, I hope DSD gets through the door -whatever the specs say it sounds damn good!

Nika

01-25-2002, 07:53 PM

<BLOCKQUOTE>quote:<HR>Originally posted by Zep Dude:
I have yet to do a side by side shootout of DSD vs PCM and frankly, until we have DSD as a workstation option it isn't really relevant. Deep down though, I hope DSD gets through the door -whatever the specs say it sounds damn good!<HR></BLOCKQUOTE>

Albeit, it sounds good. I'm not convinced it sounds better than direct PCM can sound. Since most of DSD signals are unadulterated, untouched, straight to disk digital conversions, I'm not convinced that PCM can't do the same, but rarely do you get a straight to disk digital conversion of PCM that gets there without processing. I'm not sold yet.

Nika.

D Clementson

01-25-2002, 10:29 PM

<BLOCKQUOTE>quote:<HR>Originally posted by Nika:

For the quick explanation, DSD doesn't measure the voltage at unique and predictable increments of time. It rather measures whether the signal is higher or lower than the current quantization step. If the signal is higher then a "1" is transmitted, and the next sample goes at the next higher quantization step. (snip)

<HR></BLOCKQUOTE>

Yes, but one small point. DSD _does_ sample at discrete points in time, but at VERY high sample rates - usually 3.07MHz. And jitter is incredibly important on that clock.

The thing is, this is _exactly_ what is happening at the front end of a traditional PCM delta-sigma converter chip. The audio is sampled at 3.07MHz 1-bit, then converted to 48/96/192kHz at 24-bits.

In fact, the DSD workstations use Crystal PCM converter chips which use the same die as the 888|24's ADC chip. They simply bypass the 1-bit to 24-bit conversion filter.

One consequence of this is that it is really important that you drive the signal to the 3.07MHz sampler properly - for PCM or DSD. This is where Ed Meitner's wizardry comes in to play on the 192. Some people say that what makes DSD sound so good is Ed's converters. If that's true, the 192 I/O has inherited some of that from him.

DC

dBHEAD

01-26-2002, 12:30 AM

It seems to me that Digi has taken an inherently astute step with the new HD system. This is speculation on my part (and, frankly, theirs), but I believe SACD with a 5.1 mix will end up taking the place of the conventional CD in a few more years.

Now, as many of you know, the SACD has a sampling rate of 2.822 mHz --approximately 64 times the sampling rate on CDs. In a nutshell, the rationale for such a high sampling rate seems to have been that you can much more closely reproduce complex waveforms and therefor sonic detail. At least that seems to be the jixt of the technical FAQs I've read on it. It also allows for reproduction of high frequencies beyond the human hearing range (100 kHz - you could create a hi-fi system for bats at that range).

But why? Does it even make any actual sonic difference? In theory, it isn't supposed to but my understanding is that even though people cannot consciously tell a difference, research has shown the brain does react to those frequencies when present. There is a theory (just a theory) that one result is that people, subconsciously, actually "feel" better about music recorded with extended frequency responses.

A bunch of B.S.? Maybe, but it seems unlikely Sony and Phillips would have invested in a technology so heavily if it was just going to be a fad. Clearly upsampling will be involved, but I have to think 192k would sound better (er, okay, feel better) than 44.1k when upsampled.

Anyway, I'd like to hear opinions on this issue because, frankly, I believe the whole controversy about higher sampling rates revolves around it. In essence, these are the questions: 1. Do you think there is a difference in a listener's internal reaction to improved sonic detail, even though they cannot consciously hear it? and 2. How high is high? Is the 2.822 mHz sampling rate of SACDs a marketing gimmick or will we be seeing PT 2.822 mHz in the future? Even though higher sampling rates seem certain, is 2.822 mHz ridiculous?

CO2

01-27-2002, 12:45 AM

Hello Dave:

I am certain DSD has benefited from Ed Meitner's converters in the same way the 192 HD has. BUT I think DSD sounded better than PCM even before Ed got involved.

I am with ZepDude on this. DSD will hopefully gain acceptance because it sounds great.

Nika, you miss the point. The beauty of DSD is that it does not need to be adulterated, touched or decimated in any way.

The SACD is a consumer delivery format that plays the music exactly as the engineer heard it. No down converting necessary.

BUT, the dual layer SACD's also have a down converted 44.1 PCM layer available which is compatible with all CD players.

The DSD 2.8224 MHz sample rate easily down converts to 44.1 using Sony's Super Bit Mapping (SBM) retaining much of DSD's high resolution.

As Dave pointed out earlier, and Zep Dude alluded to as well, it will probably be a very LONNNNG time before a plug-in driven DSD DAW exists. But Sony and Philips have clearly raised the digital bar.

I think it would be sad if DSD ends up going the way of the 1/2" consumer Beta format, which was far superior to the VHS format. It would be a blow to the quest for better sounding digital.

Best Regards

Nika

01-27-2002, 06:27 AM

[QUOTE]Originally posted by CO2:
Nika, you miss the point. The beauty of DSD is that it does not need to be adulterated, touched or decimated in any way.

So you're saying not to record in DSD but just master to it? That doesn't seem to be what Sony is pushing.

The SACD is a consumer delivery format that plays the music exactly as the engineer heard it. No down converting necessary.

Now hold on. If I'm in a studio listening to converters at 44.1k how is it that the consumer is going to hear this EXACTLY as I hear it when an additional layer of non-random dither is added and a new filter is put in at 50k?? If I want them to hear it exactly as I hear it, why would I not deliver it in the same format that I'm listening to it in?

The DSD 2.8224 MHz sample rate easily down converts to 44.1 using Sony's Super Bit Mapping (SBM) retaining much of DSD's high resolution.

What does this mean? You're saying that you don't have to be at DSD in order to sound that good? Then why go to DSD in the first place? Why not just make really good sounding PCM?

I think it would be sad if DSD ends up going the way of the 1/2" consumer Beta format, which was far superior to the VHS format. It would be a blow to the quest for better sounding digital.

I think it would be a blow to the trends of large corporate manufacturing dictating to the public unnecessary solutions. The quest for better sounding digital will happen at the converter side of things and the processor side of things, not the storage format.

Nika.

CO2

01-27-2002, 01:56 PM

Hello Nika:

NOW you are really missing my point!

Please check out the link I have enclosed. It is written by Tom Jung. He is a very well respected engineer (even by other distinguished peers) in the music industry. He also is a contributing columnist to Pro Audio Review.

The mere fact that Tom launched DMP (Digital Music Products) as a totally digital record company (PCM at that time) speaks volumes on his committment to digital.

The fact that he has become such a strong advocate of DSD (along with Ed Meitner, Bob Ludwig, Mark Levinson, John Gatski and a list of others that is growing every day) speaks volumes about the quality of DSD.

The link addresses all the points I wanted to make about DSD.

• There is already a consumer delivery standard in place that allows consumers to experience pure DSD (SACD).
(The reference to hearing what the engineer heard relates to DSD comparing favorably with the analog mix bus feeding the DSD recorder. Plus not requiring down conversion when delivered on SACD)

• That DSD can down convert to 44.1 and retain a lot of its original resolution for those who do not own SACD players.

Please take a moment to read this link. Tom is an excellent writer and he is quite concise in his presentation (unlike me).

web page (http://www.dmprecords.com/technology.htm)

Best Regards

Nika

01-27-2002, 07:16 PM

[QUOTE]Originally posted by CO2:
Hello Nika:

NOW you are really missing my point!

Please check out the link I have enclosed. It is written by Tom Jung. He is a very well respected engineer (even by other distinguished peers) in the music industry. He also is a contributing columnist to Pro Audio Review.

I think you assume I am unfamiliar with DSD. This is not the case. I first heard DSD about four years ago, and have heard many demonstrations since.

I also believe that I understand the system well enough to understand it's flaws. Don't get me wrong - DSD is a fine DELIVERY format, but as a recording or editing format it is inherently flawed, and most DSD processing is now being done in various types of PCM anyway.

The fact that he has become such a strong advocate of DSD (along with Ed Meitner, Bob Ludwig, Mark Levinson, John Gatski and a list of others that is growing every day) speaks volumes about the quality of DSD.

What attracts people like this is the lack of filtering that gets done in order to use DSD. The problem is that any editing of any form requires filtering to be done in order to turn it into PCM code. Filtering must also be done at the output stage. Finally, editing has to be done in order to add dithering, as dithering cannot be added with only 1 bit. Therefore, editing must be done and so filtering must be done and therefore the benefits of this sampling method are in some capacity moot.

• There is already a consumer delivery standard in place that allows consumers to experience pure DSD (SACD).
(The reference to hearing what the engineer heard relates to DSD comparing favorably with the analog mix bus feeding the DSD recorder.

I don't use an analog recorder, nor do most people in this forum, I don't believe.

• That DSD can down convert to 44.1 and retain a lot of its original resolution for those who do not own SACD players.

All of us already have DSD down converting ability. This is called an A/D converter. Any A/D converter already uses a delta-sigma modulator running at (often times) 64x. The difference between DSD and PCM is that in DSD that signal goes directly to disk. In PCM that signal is filtered and downsampled for editing. But if you want to edit DSD you pretty much have to do this anyway, so the difference is at which stage you downsample.

PCM:
Analog filter -> DSmodulator -> filter & downsample -> storage -> editing & mixing -> dithering -> filter & upsample -> D/A conversion -> analog filter.

DSD:
Analog filter -> DSmodulator -> storage -> filter & downsample -> editing & mixing -> dithering -> filter & upsample -> D/A conversion -> analog filter.

It's the same thing. The only benefit here to DSD is that if you avoid the editing and mixing process you can also avoid the filtering and down/upsampling, but again, you have to have dither in this system, and dither can only be applied while it's in PCM, so some sort of filtering and conversion to PCM has to happen anyway.

Please take a moment to read this link. Tom is an excellent writer and he is quite concise in his presentation (unlike me).

I've read Tom's page many times before. I wish I could point you to some of the info I've got on DSD v PCM, but much of it is not publishable.

Nika.

JamerJ

01-27-2002, 08:19 PM

Excellent question!!!

My studies have shown any rate based on
48khz is a human friendly rate. 44.1 is like a flicker of sorts to that part of
us that feels music and is hence not able to transfer emotional body content. Intellectually
we get 44.1 but vibe wise??? Throw on
some laquer and feel the diff! Its' there.

As far a 96 and 192 go they are as well
able to register on the emotional body.
44.1 is dead to the emotional body. The
big contrast is perceived in vocals. They
sound 3d at 48 ect.

44.1 is an artificial timing base, nowhere in Nature. 48 is more of a natural flicker
rate. As the rates speed up past 48 there is the ability to perceive harmonics far above the program info. Any reeds recorded at higher rates will retain their quality and be less metalic sounding.

I am humorously convinced that 44.1 was actually chosen becuase then there would be a built in upgrade path for CD... lol

Really though, 48,96, and 192 support wavelengths which the emotional body can
receive. 44.1 is dead like plastic but
can be intellectualized. Again, try
some old lauquer, take an asprin, and
email me in the morning.

Join the "longing for the warmth" club...
images/icons/tongue.gif images/icons/tongue.gif images/icons/tongue.gif images/icons/tongue.gif

V76 junkie

01-27-2002, 08:57 PM

Back to the original topic..... images/icons/wink.gif

It is true that you cannot directly hear the frequencies above about 22khz, however, you can hear the way the frequencies affect the frequencies below them. At that point it's just simple waveform addition to see the difference those frequencies make on the overal waveform.

As for equipment, my V76's have phenomenal frequency response; well into the 60-70 Khz range.

images/icons/cool.gif ahhh, the beauty of vintage german tube gear.......

D Clementson

01-27-2002, 10:39 PM

Nika's comments are pretty much on-the-money as far as I know about DSD. I'm no major expert - I only saw Andreas Koch's presentation at Sony during a local AES meeting, I talk to Meitner alot, and I know how delta-sigma processing works.

One tiny point is that _editing_ of DSD is trivial. You can cut and paste anywhere, just like analog tape. It is the _processing_ (EQ, gain change, etc.) that gets really hard in DSD. In order to do that at all, you have to remodulate the DSD stream, which usually involves turning it back into PCM as an intermediate step. It may not be 24-bit @ 1xFS, though. Other bit depths and oversampling rates can be used.

DC

Nika

01-28-2002, 05:22 AM

[QUOTE]Originally posted by V76 junkie:
It is true that you cannot directly hear the frequencies above about 22khz, however, you can hear the way the frequencies affect the frequencies below them.

This is not true in any capacity.

At that point it's just simple waveform addition to see the difference those frequencies make on the overal waveform.

It is not waveform addition at all. We have hammered this out multiple times on this forum. You might try looking on this thread for additional info on this:
http://duc.digidesign.com/cgi-bin/ubbcgi/ultimatebb.cgi?ubb=get_topic&f=2&t=005834

As for equipment, my V76's have phenomenal frequency response; well into the 60-70 Khz range.

But not in correct phase. The reason your V76's have this response is so that the 20kHz waveform is closer to truer phase. Most of my preamps are spec'd to 500kHz or more. This isn't because it's passing a bunch of data up in the 300kHz range - it's so that the data at 20kHz is in correct phase. Once again, this has been hashed out here many times before.

Nika.

Nika

01-28-2002, 06:32 AM

[QUOTE]Originally posted by JamerJ:
Excellent question!!!

My studies have shown any rate based on
48khz is a human friendly rate.

Your "studies"? Can you point us to these? You're making some very broad statements that counter most of my understanding of digital audio. Would you please explain these "studies" and quote your sources? Remember that we're not hearing the sampling points, anyway. We're hearing the results of them turned back into analog, then passed through an analog filter. The final sampling rate we end up hearing is generally 8x the base rate, so around 352kHz. Are you saying that 352kHz is not suited to our biorhythms either? There is just about NOTHING that you hear that is at 44.1kS/s.

44.1 is like a flicker of sorts to that part of
us that feels music and is hence not able to transfer emotional body content.

Whew! What a relief, because we don't actually listen to anything at 44.1kS/s. Please specify for us exactly which frequencies are prone to activate our proper biorhythms, please? Your neurological experience and expertise will come in most handy here.

Intellectually we get 44.1 but vibe wise???

I'm trying to find this term in the AES preprints, or in the American Physiological Association works and can't find it anywhere. Can you point us to the correlation between Alpha ElectroEncephelograms and "vibewiseness"?

As far a 96 and 192 go they are as well
able to register on the emotional body.
44.1 is dead to the emotional body. The
big contrast is perceived in vocals. They
sound 3d at 48 ect.

Yeah, back to pulling out those studies, please?

44.1 is an artificial timing base, nowhere in Nature. 48 is more of a natural flicker rate.

Now this is where it REALLY gets interesting. Do explain these concepts to us, please. This is really helpful.

This post is about the least beneficial and most misguided post I've seen on this topic in a long time. I found this most counterproductive in this thread where we're searching for real answers and not wizardry and artistic 'hypothesis' of an uninformed scientific situation.

Nika.

dBHEAD

01-28-2002, 08:59 AM

Okay, I spent a little time reading up on DSD, both the pros and cons, and I have to say I came away feeling that it's probably a better way to record and re-create the wave. First, it's a much more straightfoward method of capturing minute details in a complex wave.

I realize there are problems. A wave could increase or decrease in amplitude in 1/2,822,000 of a second more than the single bit could accurately convey, but with so many samples the code would eventually "catch up" with the trend. I feel there's absolutely no doubt that in most instances (most, not all) DSD would more accurately reproduce a wave than PCM would -- unless, of course, you went to extremely high sampling rates with PCM.

One of the most telling comparisons I read was the ability of each method to reproduce a square wave. As it turns out, because of all the averaging typically done, a square wave fed into a PCM system is likely to look like a sine wave when reproduced. That isn't the case with DSD - it can follow the contours of a square wave pretty accurately.

There's no question that DSD is popular with audiophiles; many hace raved about the sound, comparing it to state-of-the-art anaolg in many cases. This was kind of the point I was getting at. Clearly, the dynamic range is no better than 24 bit, and the audible range for humans doesn't change. So how can it really sound "better?" I think the answer has to be either an ability to more accurately reproduce complex waveforms or the ability to reproduce higher frequencies. Maybe both. There certainly seems to be a healthy school of thought out there that higher frequencies (beyond 20K) do in fact "color" the sound we can hear.

Corey Shay

01-28-2002, 09:31 AM

There is something I don't quite understand concerning the nature of DSD or "Pulse Density Modulation" as it is referred to.

The concept of outputting a 1 to signal an increase in air pressure or voltage in most cases and a zero for a decrease seems relatively straightforward, but something is bothering me. I downloaded a PDF from somewhere (apparently written by Sony) titled Super Audio Compact Disc: A Proposal. It has a few illustrations in it to explain DSD's workings. One shows a sine wave. Below the sinewave is a bitsream. We se lots of ones crowded together at the top of the sine wave's full positive cycle, very few ones at the bottom of its negative cycle, and what looks like alternating ones and empty space (zeroes) at the zero crossing points.

This is what throws me off. While this seems to fit with the term "Pulse Density Modulation", it appears to clash with the one for up, zero for down explanation. Shouldn't we see the most ones at the waveform's positive (upward) zero crossing point and the least ones at the waveform's negative (downward) zero crossing pont? Shouldn't we see alternating ones and zeroes at both the full positive and full negative extremes of the cycle? (because this is where the waveform is actually not rising or falling).

An analogy... In high school physics we measured velocity and acceleration with ticker tape. However, greater distance between ticks was a direct measurement of velocity, not acceleration. In the above explaination of DSD, I would think that the bits would in a way represent both vector and "velocity" similar to the ticker tape phenomenon. However the illustration I alluded to implies the bits represent the amount of "acceleration" along with vector. At both the peak and the trough of the sine wave there is the most "acceleration", but the least amount of "velocity". Just like a pendulum. It has no velocity at its maximum swing, but is fastest at its middle point. However, it is not accelerating at its middle point, but has the most acceleration at its maximum swing.

I hope I'm making a little sense here... Pohlmann is a bit difficult to understand at times. Dave Clementson or Nika?

BTW, thanks Nika and Dave. Very well written posts here.

Corey Shay

01-28-2002, 09:52 AM

Actually now that I think about it a little more the illustration might simply be representing bits for the waveform's position relative to the zero cross line, and not acceleration. To further add to dBHead's comments, I wonder what the bits would look like if the waveform were square.

Slope distortion? Hmm...

DBHead, in regards to your comments concerning "a much more straightfoward method of capturing minute details in a complex wave", I think I already know what Nika is about to say because it has already been discussed before. That PCM accurately captures all the subtleties of a waveform below the Nyquist limit. And that higher frequencies (that we don't hear) that affect lower frequencies (that we do hear) are already sampled because the affected lower frequencies are still below the Nyquist limit.

CO2

01-28-2002, 10:11 AM

Hello dB:

Thank you. Very well stated.

To me, it is the absence of brickwall filtering.

Hello Nika:

I am not trying to start a digital range war with you. Clearly, you suffer from extreme technical sophistication (a compliment to be sure) and I pale in comparison.

Personally, I am more interested in advancing the quality of digital audio by whatever means accomplishes it than arguing technical issues.

Since I am a major fan of the warm vintage analog tube sound, I welcome any digital format (DSD or PCM) that maintains the integrity of that sound.

The fact that this forum is bursting with threads about attempting to recreate the sound of analog in digital, clearly proves we are not there yet.

Do not get me wrong. I love digital for its editing, duplication and storage capabilities.

Ideally, I believe in using vintage analog audio gear to capture performances and using digital gear to edit and store them.

"Audiophiles" have always impressed me with their desire to obtain recordings which mirror the original performances.

Throughout audio history, many innovators and visionaries have pushed the quality envelope (hence my respect for current innovators such as Tom Jung, Mark Levinson, Bob Ludwig and Ed Meitner).

So, in the interest of improving the quality of recorded sound (and realizing that digital <in some form> is the future of audio), I only suggest that we continue to keep an open ear (and mind) to the possibilities that help us accomplish that goal.

Best Regards

Nika

01-28-2002, 01:00 PM

[QUOTE]Originally posted by dBHEAD:
Okay, I spent a little time reading up on DSD, both the pros and cons, and I have to say I came away feeling that it's probably a better way to record and re-create the wave. First, it's a much more straightfoward method of capturing minute details in a complex wave.

Simple? Perhaps, though inherently flawed. Think for a moment of passing DC through a DSD system. It is impossible for it to pass DC. It instead pass a series of "up" oops, too high "down" oops, too low "up" oops, too high... Think for a moment about how PCM handles this - let's say at the zero crossing - "zero" and still "zero" and still "zero"...

Now let's project this into a 50Hz square wave. Since this signal is almost completely DC, and since PCM is capable of passing DC and DSD is not, which one looks more like the original waveform?

I realize there are problems. A wave could increase or decrease in amplitude in 1/2,822,000 of a second more than the single bit could accurately convey,

No, this is how the maximum frequency is established - what frequency, at the highest "q" or "slope" or "slew rate" at the zero crossing can still be accurately transmitted within 1/2822xxx/second.

I feel there's absolutely no doubt that in most instances (most, not all) DSD would more accurately reproduce a wave than PCM would -- unless, of course, you went to extremely high sampling rates with PCM.

You have a lot more reading to do, I'm afraid.

One of the most telling comparisons I read was the ability of each method to reproduce a square wave. As it turns out, because of all the averaging typically done, a square wave fed into a PCM system is likely to look like a sine wave when reproduced. That isn't the case with DSD - it can follow the contours of a square wave pretty accurately.

Read a few paragraphs above.

I think what you're having a hard time with is the concept of something like a 15k squarewave, where it essentially gets converted as a sinewave. This, however, is irrelevant, as it is unncessary to capture the "Remainder" of the components in a square wave, as they all represent unnecessary frequencies that are above the human hearing capabilities. A 15k square wave is composed of 15k, 30k, 60k, 120k, etc. The ear can only utilize 15k. 44.1k PCM can only capture 15k. DSD can only capture up to ~100k, so it still doesn't yield a "square wave". It just yields the first three frequencies of a square wave. Again, this is irrelevant, because at 15k the ear doesn't hear a square wave. At 15k, the ear hears almost exclusively sine waves, and a square wave played into the ear at 15k is precisely a 15k sine wave. So PCM actually more closely aligns with what the ear can hear.

I think the answer has to be either an ability to more accurately reproduce complex waveforms

Yes, but if those complex waveforms are above our hearing range then this point is moot, right?

or the ability to reproduce higher frequencies. Maybe both. There certainly seems to be a healthy school of thought out there that higher frequencies (beyond 20K) do in fact "color" the sound we can hear.

"healthy"? Where is this school of thought? Where can I read from the experts in this regard? I've read a LOT of studies of higher sampling frequencies, and am on a committee of the AES working to standardize listening protocols for higher sampling rate material, and have NEVER run across a valid school of thought discussing the merits of reproducing higher frequencies. Please, do explain.

Nika.

Nika

01-28-2002, 01:01 PM

<BLOCKQUOTE>quote:<HR>Originally posted by Corey Shay:
DBHead, in regards to your comments concerning "a much more straightfoward method of capturing minute details in a complex wave", I think I already know what Nika is about to say because it has already been discussed before. That PCM accurately captures all the subtleties of a waveform below the Nyquist limit. And that higher frequencies (that we don't hear) that affect lower frequencies (that we do hear) are already sampled because the affected lower frequencies are still below the Nyquist limit.<HR></BLOCKQUOTE>

images/icons/smile.gif

Nika

01-28-2002, 01:07 PM

<BLOCKQUOTE>quote:<HR>Originally posted by CO2:
Hello Nika:

I am not trying to start a digital range war with you. Clearly, you suffer from extreme technical sophistication (a compliment to be sure) and I pale in comparison.

Personally, I am more interested in advancing the quality of digital audio by whatever means accomplishes it than arguing technical issues.

Since I am a major fan of the warm vintage analog tube sound, I welcome any digital format (DSD or PCM) that maintains the integrity of that sound.

The fact that this forum is bursting with threads about attempting to recreate the sound of analog in digital, clearly proves we are not there yet.

Do not get me wrong. I love digital for its editing, duplication and storage capabilities.

Ideally, I believe in using vintage analog audio gear to capture performances and using digital gear to edit and store them.

"Audiophiles" have always impressed me with their desire to obtain recordings which mirror the original performances.

Throughout audio history, many innovators and visionaries have pushed the quality envelope (hence my respect for current innovators such as Tom Jung, Mark Levinson, Bob Ludwig and Ed Meitner).

So, in the interest of improving the quality of recorded sound (and realizing that digital <in some form> is the future of audio), I only suggest that we continue to keep an open ear (and mind) to the possibilities that help us accomplish that goal.

Best Regards<HR></BLOCKQUOTE>

Co2,

I appreciate your post. Thank you. I, too, am anxiously working at improving the state of digital to more closely represent reality (not analog, but reality). This is part of the reason that I just purchased some new converters, and part of the reason that I am a contributor to threads in here about the best way to transparently use the tools that Digidesign provides us.

I might encourage, however, before you jump on a bandwagon spearheaded by a multinational marketing machine in full force, that you take it upon yourself to do more extensive research than just what is readily available, and to force yourself to suffer from "extreme technical sophistication" if this is what it takes. You might try reading the works of Stanley Lip****z in the regards to SACD, or befriending it's opponents well enough to get the inside scoop on the breakdown of the system. It's just not quite as rosy as the big press machine makes it sound.

I just encourage people to not jump so strongly on a bandwagon without first determining where the wagon is going.

In the end, hopefully we will find a solution that provides people with the most accurate reproduction of reality. DSD, in my humble opinion, and with the research I have done, is not anywhere near being the proper answer.

Nika.

Nika

01-28-2002, 01:26 PM

[QUOTE]Originally posted by dBHEAD:
[QB]Yes, that seems to be the PCM gospel, and they've been saying it for years. I'm not a technical expert, but I've never believed it since seeing a stretched out waveform on a monitor almost a decade ago.

You might want to become more a "technical expert" then before denouncing the system, then.

When I stretch out a track's waveform on my PT system, I still can't believe it. These are not perfect sine waves --which I think could easily be represented by a limited number of samples. Two samples to represent a 20 kHz wave? Four (well, @ 4.5) samples to represent a 10 kHz wave?

This is because this is a visual representation of HALF of the digital process. It is taking a snapshot partway into a process, bastardizing it visibly, and then presenting it as though this is what comes out.

What actually comes out of a digital system is nothing like the representations you've seen on your monitor. That is merely a tool to make it visibly easier to edit. What actually comes out of your system are perfectly sinusoidal waveforms, increasingly so at up to 20kHz. The D/A converter does not just give you a spotty "dot to dot" conversion of each sampling point, essentially spitting out a series of square or triangle waves.

My position, though, is that they cannot properly represent the slop of the wave at high frequencies. And I'm not trying to be stubborn but I just can't see how it's possible.

Just because you can't see it, or because it's difficult to fathom does not mean that it is unable to do this. In fact, digital PCM is perfectly capable of representing the exact slope more accurately than DSD is capable of. As Corey Shay pointed out, it has to do with the filtering used.

At 20kHz there are only two points there to represent a single sine wave, as we have pointed out. The visual representation on the screen makes it "look" like a triangle wave. What actually happens is that the two sample points are fed into a digital filter that filters those two points into the only possible shape of a "legal" waveform that can pass through those two points - a sine wave. There is only one possible way that a sinewave can be drawn through two points, so the filter's job is easy.

Now the FIR reconstruction filter doesn't COMPLETELY do this. It pretty much filters everything out between 22.05kHz and 385kHz or so, creating a series of square waves that are much closer to the sine wave than the original points. Now we have 16 points to represent the sine wave instead of 2, and it looks like a Mayan temple - stairstep up and down. Then the D/A process happens, turning it into a waveform with a strong 20kHz component and a bunch of distortion above 385kHz.

Then an analog filter is applied, filtering everything out above 385kHz, so that all that is left is a very smooth waveform up and down - perfectly sinusoidal, and with no components at all above 20kHz at all.

Now think about this for a second. What does the frequency response curve of a sine wave look like? What about a square wave? What about a triangle wave?

OK, now if I tell you that these 2 sampling points yield a result that has absolutely NO components above 20kHz then what kind of waveform does that become? Does that answer your question?

Now, let's take an example. A 10 kHz wave which has a "step" a bit above the zero crossing line. Let's, for the moment, assume we can capture that detail in recording. I know the argument would be that momentum of the speaker diaphragm pushing foward would override that impulse anyway. That's true, but that does not mean it would have no audible effect. It could cause the diaphragm to slow down or speed up in ways which -if it could always get that detail- would add a subtle warmth to the music.

I think you're assuming that just because the screen only shows you a sampling point at 1 quantization step above the zero crossing that the amplitude of the resultant signal is only going to be that loud. This is simply not the case. The sinewave itself is redrawn to the correct amplitude by the effects of the reconstruction filters in the D/A converter. Music would sound HORRIBLE if it was done as you're putting forth. Fortunately it is done properly, and as Nyquist intended - with the use of filters to reconstruct the waveform.

I may be dead wrong but ...

Umm... images/icons/smile.gif

Nika.

dBHEAD

01-28-2002, 02:58 PM

Corey:

I remember that thread -- it was a good one. However, it was more focused on the visual representation on the screen. I'm not sure it totally answered my question.

Umminger:
"Given enough processing power and latency one can make an arbitrarily good approximation to a perfect brick wall filter and get an arbitrarily perfect analog reconstruction of the sine wave while using the same digital signal."

I guess the word "arbitrarily" kinda throws me off there.

Nika:

You bring up quite a few points. I don't really have time to respond to them all -- and some would require some research on my part. But there are a few I'd like to bring up.

"You might want to become more a "technical expert" then before denouncing the system, then."
I didn't think I "denounced" PCM at all. I only stated my opinion about what I felt was a limitation. I openly admitted that experts would disagree. I know the wave we see when we zoom in on PT is not an exact representation of the waveform, but even looking at sound on an o-scope it quickly becomes obvious that real sound produces some very oddly shaped waves. The reason I don't believe PCM can actually reproduce those waveforms with total accuracy is:

"At 20kHz there are only two points there to represent a single sine wave, as we have pointed out. "

Exactly. That's why I have a hard time fathoming it -- and I'll be the first to admit I do. If a 1kHz wave is represented by 44 points, isn't the detail of the wave going to be more accurately represented than the detail of a 20 kHz wave which is represented by only two?

That, admittedly, is the point I'm getting hung up on here. I don't doubt for a second that you can reproduce a sinusoidal wave of exactly the same amplitude of the original sound. But can you reproduce a wave of exactly the same shape as the original sound? You seem to be saying we can -and I know your technical knowledge is far superior to mine so I want to pick your brain here. Because if what you're saying is correct, couldn't we, in theory, replicate a waveform at any frequency with two samples? I mean, even if we're sampling at 44.1 k, could we, for example, throw out 439 samples of the 441 samples used to represent the wave? If all we need is peak to peak representation, that would seem to be the case.

"I think you're assuming that just because the screen only shows you a sampling point at 1 quantization step above the zero crossing that the amplitude of the resultant signal is only going to be that loud. This is simply not the case. The sinewave itself is redrawn to the correct amplitude by the effects of the reconstruction filters in the D/A converter. Music would sound HORRIBLE if it was done as you're putting forth. Fortunately it is done properly, and as Nyquist intended - with the use of filters to reconstruct the waveform. "

You might not have understood me on that one. I was adressing the issue of the most accurate way to reproduce the original analog waveform - regardless of how it was done. I was not endorsing any particular sampling scheme.

Thanks for the responses. I'm learning!

images/icons/wink.gif

Nika

01-28-2002, 03:38 PM

db,

Let's try it this way. Without pencil and paper this get's pretty difficult.

First you recognize that the ear can only hear material up to 20kHz, and that anything above 20kHz can be filtered out, right? OK, even if you don't agree with that it's a point we have to run with for the sake of discussion. We can battle about "psychoacoustics" later, but for now run with me that anything higher than 20kHz is not necessary to capture or reproduce.

Now it is true that waveforms are much more complex in nature than simple sinewaves, but if we filter out all of the garbage that is above what we need the waveforms get much more organized, and have certain, predictable properties. Take a 20kHz waveform that is "complex". What this means is that the waveform will cross the zero crossing 20,000 times per second, but there will be a lot of "harmonic content" added to the fundamental sine wave on either side of the zero crossing. This is basic wave theory stuff. What we're talking about is the "wiggles" in the waveform.

So if I take a 20kHz waveform it is a pure sine wave. As I had additional frequencies it changes the sinewave into other forms. Ya' add a little 35k and it gives it some "wiggles". Add a little 64k and it adds some more. Pretty soon it look much less like a sine wave, but is actually the result of adding a bunch of sinewaves together at a bunch of different frequencies. The bottom line - everything is a sinewave, but adding them together results in more complex things.

If I am very organized about how I create a complex waveform I can actually make a complex waveform look very "uncomplex". A square wave is a specific formula for adding together sine waves that results in a "complex waveform", but is just a very well organized one. It is (someone correct me if I'm wrong, here) equal values of all frequencies of even harmonics from the fundamental. This means that it's an equal amount of frequency X, frequency 2X, frequency 4X, frequency 6x, etc. This means that a 20kHz squarewave is an equal amount of 20kHz, 40kHz, 80kHz, 120kHz, etc.

This means that if I filter out everything above 20kHz, all I am left with is a very simple sine wave again, right? There's no 3kHz in a 20kHz square wave. It doesn't fit the specific formula.

Just as a segue, you can't hear a "square wave" if it is in the upper register of your hearing. A square wave at 15k has it's first harmonic at 30kHz, and since you can't hear that, all you can hear is 15kHz, and since it's only one frequency it's a perfect sinewave.

Anyway, if you take any waveform and filter out everything above 20kHz then the waveform becomes very predictable. It becomes increasingly "sinusoidal" as you approach 20kHz. Sure, a 15kHz complex waveform has more room for a LITTLE bit of variation, as it might have a little bit of 16.3kHz, 17.42kHz, and whatever else, but there won't be any "wiggle" on your 15kHz waveform that has a "tighter slope" than a 20kHz waveform. You can't have any "wiggle" in any waveform that has a "sharper edge or corner" than qualifies as a 20kHz waveform. Any sharp corner represents that higher frequencies were added together - a conversation for a later day.

Anyway, recognizing this waveforms that have been filtered are much more predictable, there is a mathematical principle that we need to throw in: a sinewave can only pass through greater than two given points in one possible way. If there are more than 2 sampling points per cycle in a graph then there is only one possible way to pass a sinewave through them. When I say "greater than 2 sampling points per cycle" I'm talking about 2.00000000001 sampling points per cycle. So since 20kHz has greater than 2 sampling points per cycle at 44.1kS/s there is only one possible way to draw the sinewave through those points. Further, by the time you have 3 or more sampling points it becomes even further defined and easier to see that there is only one possible way to draw your sinewave through these poitns.

Now, recognizing this, lets see how it happens:

What leaves the digital environment at 20kHz with the samping rate of 44.1kS/s is just two sampling points. These, in essence, represent a square wave or a triangle wave, would you agree? But we now know that a square wave or a triangle wave at 20kHz is actually indicative of 20kHz PLUS a bunch of other frequencies that are higher than 20kHz, right? So we put a filter in (called a "reconstruction filter, as it "reconstructs" the original waveform) and it removes all of the frequencies above 20kHz. And since we no longer have all of the frequencies that make the sine wave into a square (or triangle) wave, all that is left is our 20kHz, and a waveform that is only 20kHz can only possibly be one thing - a sinewave.

So let's visit a complex waveform at 20kHz being sampled. All that a good ear is going to hear is 20kHz, so all the ear will hear is a 20kHz sine wave, and none of the other material is necessary for the ear to hear this waveform.

So with our 20kHz waveform, first, the anti-aliasing filter filters out all of the "complexities", leaving only a sine wave. Then we sample any greater than two sampling points on that waveform. Then we send it through a DAW and out to the D/A as a couple of square (or triangle waves). Then the reconstruction filter removes all of the additional frequencies that are turning our sinewave into a square (or triangle) waveform by again filtering out everything above 20kHz. And all that is left is a 20kHz sine wave - exactly what the ear was able to hear in the first place. And there is only one possible way for the reconstruction filters to recreate that sinewave, because there is only one possible way that that sinewave could be drawn through the "greater than two" sampling points that it was given.

Does this help?

Let me know. Pictures and graphs come next.

Cheers!
Nika.

CO2

01-28-2002, 03:59 PM

Hello Nika:

Thank you for your tolerance.

I am pleased to see that you are also crusading to improve the state of digital.

Just for the record, I have not jumped on ANY bandwagon yet. I still have my MIX3 system and intend to hold on to it until which time I see just cause for jumping.

I do confess to being swayed by the opinions of credible engineers whom I trust to be objective in their research.

When the time arrives to jump on the next digital bandwagon, I want it to be an awe inspiring ride!

Best Regards

dBHEAD

01-28-2002, 04:51 PM

Nika:

Absolutely excellent my man! When things get explained that clearly, even I can understand them!

Now, in fact, I understand something I've had a misconception about for years: that what I see on an O-scope when any kind of real audio is running through it is not a single, isolated wave display but rather the cumulative result of all the waves interacting. I now see that audio on the 0-scope was never really explained right to me in the first place.

It really does make perfect sense now. I was aware of the filtering process used in digital audio (well, PCM anyway), but I actually thought the wrinkles in the wave were part of the waveform itself -- something that gave each wave a totally distinct "character." As I think you can see, THAT was what I thought DSD could do better -- imitate those wrinkles.

Thanks for your patience. You should consider writing a book -- your explanation was crystal clear and within the grasp and within the grasp of the average intellect.

images/icons/grin.gif

Nika

01-28-2002, 05:45 PM

dbHead and CO2,

Whew! I'm glad that all came through OK. If you have any more questions let me know.

As you can imagine, I have to tackle this and related subjects quite frequently. There is SOOOO much misinformation, presumptions, heavy marketing, half-understandings, and the like that I get called on like this on a nearly daily basis to address the public perception. It's an uphill battle trying to convince everyone not to waste their money, but I try where possible to answer people's questions and give them whatever information is missing from their equations. You really do have to have a complete understanding of all of this in order to understand why the current paradigms work so well.

Anyway, I'll see you in the next thread.

Nika.

king mob

01-28-2002, 08:12 PM

just wanted to say, nika, I have learned so much from your postings here. thank you for all of the wonderful information. you have really helped one beginning engineer to learn the physics behind all this stuff. i look forward to more postings. thank you.

CO2

01-28-2002, 11:31 PM

Hello Nika:

It occured to me to ask you one last question.

Who are you?

I am assuming you are either a highly credentialed educator or engineer.

If I knew a bit more about your qualifications, I might be inclined to add you to my list of respected engineers.

Just curious. Feel free to disregard the question if it is too invasive.

Best Regards

dBHEAD

01-29-2002, 12:07 AM

"DBHead, in regards to your comments concerning "a much more straightfoward method of capturing minute details in a complex wave", I think I already know what Nika is about to say because it has already been discussed before. That PCM accurately captures all the subtleties of a waveform below the Nyquist limit. And that higher frequencies (that we don't hear) that affect lower frequencies (that we do hear) are already sampled because the affected lower frequencies are still below the Nyquist limit. "

Yes, that seems to be the PCM gospel, and they've been saying it for years. I'm not a technical expert, but I've never believed it since seeing a stretched out waveform on a monitor almost a decade ago. When I stretch out a track's waveform on my PT system, I still can't believe it. These are not perfect sine waves --which I think could easily be represented by a limited number of samples. Two samples to represent a 20 kHz wave? Four (well, @ 4.5) samples to represent a 10 kHz wave? C'mon!

They CAN represent subtle moment-to-moment changes in amplitude -- absolutely no argument there whatsoever. My position, though, is that they cannot properly represent the slop of the wave at high frequencies. And I'm not trying to be stubborn but I just can't see how it's possible.

And, yes, I think those slopes are very important. After all, they determine how the speaker diaphragms will move.

Now, let's take an example. A 10 kHz wave which has a "step" a bit above the zero crossing line. Let's, for the moment, assume we can capture that detail in recording. I know the argument would be that momentum of the speaker diaphragm pushing foward would override that impulse anyway. That's true, but that does not mean it would have no audible effect. It could cause the diaphragm to slow down or speed up in ways which -if it could always get that detail- would add a subtle warmth to the music.

And that's the only point I'm trying to make here. I may be dead wrong but at least I'm trying to make sense.

I found this study pretty useful:
http://student-kmt.hku.nl/~arjan1/dsd/dsd.html

I was particularly impressed with comparisons of DSD and PCM when dealing with a square wave -- it's a good illustration of what I mean when I talk about being able to reproduce waveform complexities.

images/icons/grin.gif

Corey Shay

01-29-2002, 12:39 AM

I had the same concerns before over the waveform representation weirdness, but after some people with a much higher education level in electrical engineering actually explained the reason for the waveform weirdness, and why we don't actually hear this weirdness, it didn't take me very long to relent.

In short, the waveform weirdness gets filtered out. Here is a link to the thread that convinced me...
http://duc.digidesign.com/cgi-bin/ubbcgi/ultimatebb.cgi?ubb=get_topic&f=2&t=004386

You can actually see exactly when I saw it.

Nika

01-29-2002, 03:59 AM

<BLOCKQUOTE>quote:<HR>Originally posted by CO2:
Hello Nika:

It occured to me to ask you one last question.

Who are you?

I am assuming you are either a highly credentialed educator or engineer.

Just curious. Feel free to disregard the question if it is too invasive.

Best Regards<HR></BLOCKQUOTE>

CO2,

No bother. I'm noone important, and though you can find albums with my name on them it's never for the engineering. You're welcome to respect me if you wish *S*.

My name is Nika Aldrich and I am a sales engineer/consultant at Sweetwater. I have taught at the college level in my spare time and when requested to, though I would not say I am "highly credentialed" (though the universities apparently thought that I was ?)

I do a bit of remote classical recording, and have a decent compilation of mics and the whatnot assembled for that, as well as a Protools system back home for editing and mixing.

I have a 1 year old son, a wife, a boxer, and another child on the way in 15 months (thought my wife doesn't know yet).

My own education is a couple of years at the Edmonds Community College, north of Seattle, a semester at the University of Washington, crash course training at Sweetwater, and 10 years at the school of hard knocks.

I guess that's it? Anything else you wanted to know?

Nika.

Jeff_Francis

01-29-2002, 06:59 AM

Nika,

Thanks for spending the time to educate. It is much easier to see thru myths and marketing hype when you understand the basic principles of digital audio.

A small correction on square waves...

A square wave is made up of ODD harmonics with decreasing amplitude. The amplitude of a harmonic is 1/n, where n is the harmonic number. So a 20kHz square wave is made up of a 20kHz sine wave, a 60kHz sine wave of 1/3 the amplitude, a 100kHz sine with 1/5 the amp, 140kHz with 1/7, ad infinitum. Still an endless spectrum.

This puts the overtones of a 20kHz square wave even further out of the range of human hearing.

Thanks again,

Jeff

Nika

01-29-2002, 07:14 AM

Jeff,

Thanx much! I always get confused between that and Triangle waves (which must be the even harmonics, then). I think I might have skipped the day we did Fourier Transforms in junior high images/icons/smile.gif

Nika.

CO2

01-29-2002, 05:16 PM

Hello Nika:

Thank you for your reply.

Best Regards

Curve Dominant

01-29-2002, 08:54 PM

Nika,

CLAP CLAP CLAP CLAP CLAP CLAP CLAP CLAP CLAP CLAP CLAP CLAP CLAP CLAP CLAP CLAP CLAP CLAP CLAP CLAP CLAP CLAP CLAP CLAP CLAP CLAP CLAP CLAP CLAP CLAP CLAP CLAP CLAP CLAP CLAP CLAP CLAP !!!

I was standing, by the way.

Eric images/icons/smile.gif

JamerJ

01-30-2002, 01:25 AM

"This post is about the least beneficial and most misguided post I've seen on this topic in a long time. I found this most counterproductive in this thread where we're searching for real answers and not wizardry and artistic 'hypothesis' of an uninformed scientific situation.

Nika. "

Future science will prove it my intellectualized friend. Sorry to
fall short of your expectations.
Condemnation without investigation is a sign of ignorance...

A question for your prejudging mind:
Does 44.1k divide into a second equally?
What is the second ultimately a division of?
Yes, the period of rotation of earth. Its okay neo, some choose to take the blue pill...

I'll leave my factual sources out of
it and let you flap around a while...
You seem to have it all worked out anyway,
who am I to interfere? images/icons/cool.gif images/icons/cool.gif

Disco_Doctor

01-30-2002, 04:18 AM

Hey Nika - I have a question:

The one thing that it seems to me that higher sample rates will greatly improve, as I presently understand it all, is stereo imaging/phase coherence between phase related tracks.

For example: at 44.1K, if you have a stereo drum overheads track, and you shift the left channel forward or backward one sample, it is plainly audible: the whole stereo image shifts.

If you have a stereo pair of microphones set up, either in X/Y or M/S, and you're running at a higher sample rate, I'm thinking that you will much more accurately capture the timing differences between the waves arriving at the X vs the Y mic (or the M vs the S mic). I imagine that this will make the stereo soundstage much more crisp and detailed.

Also, everyone keeps talking about whether or not we will hear higher frequencies at higher sample rates, but that seems like a non-issue to me. Phase cancellations between phase related sources cancels a lot of high frequencies out anyway, so I hypothesize that capturing at a higher sample rate will more accurately reproduce frequencies in the audible range (due to more accurate capture of phase related waves at the source) and improve the stereo imaging and phase coherence between related stereo pairs (or between multi-mic sources such as an entire drum kit, string section, etc).

What do you think? Does this make sense? To me, this seems like the one thing that will really audibly benefit from higher sample rates.

images/icons/smile.gif

Nika

01-30-2002, 07:49 AM

[QUOTE]Originally posted by Disco_Doctor:
Hey Nika - I have a question:

The one thing that it seems to me that higher sample rates will greatly improve, as I presently understand it all, is stereo imaging/phase coherence between phase related tracks.

Hmm. I don't follow.

For example: at 44.1K, if you have a stereo drum overheads track, and you shift the left channel forward or backward one sample, it is plainly audible: the whole stereo image shifts.

When you shift a left or right channel forward or back by a sample in a DAW it will indeed cause a phase shift of 1/44100/s, and this would be the same effect as moving one of the microphones 7mm in the acoustical space. If you are working at 96kS/s then shifting a track versus another by one sample would only be the effect of shifting one of the microphones by 3mm in acoustical space.

If you have a stereo pair of microphones set up, either in X/Y or M/S, and you're running at a higher sample rate, I'm thinking that you will much more accurately capture the timing differences between the waves arriving at the X vs the Y mic (or the M vs the S mic). I imagine that this will make the stereo soundstage much more crisp and detailed.

I don't think you're looking at it quite right. 44.1k digital audio is FAR more accurate at capturing phase than anything in the analog world, and more accurate than you might imagine. Any waveform in digital audio is accurate in phase to something like a thousandth of a degree. Sure, we only sample the waveform twice per cycle, but WHERE we sample it tells us everything about the true phase of the signal. In fact, digital audio is more accurate at reproducing phase than it is amplitude, but don't get too hung up over that.

Certainly, if you want to adjust phase later, especially with higher frequencies, higher sampling rates would be more beneficial, but I'm not sure what the practical implications of this are?

...so I hypothesize that capturing at a higher sample rate will more accurately reproduce frequencies in the audible range (due to more accurate capture of phase related waves at the source) and improve the stereo imaging and phase coherence between related stereo pairs (or between multi-mic sources such as an entire drum kit, string section, etc).

I'm afraid not, but don't really understand your angle well enough to explain why this isn't the case.

Nika.