“The Superiority of Analog Audio Tape”

[Beck]

The nerve of these people! RMGI get permission from Dr. Seibt to use this article when he obviously just compiled a number of my posts from this forum since 2003. Oh well, they gave me free samples too so I won’t make a big deal of it.

http://www.rmgi.eu/pdf/TestReportjune2007.pdf

He apparently started drinking (or smoking) something a bit passed halfway into the article, and by the end was well-drunk. But overall some good info worthy of follow-up research nonetheless.

 

[SteveM]

Cool, I'll get back to that. It's 5am. What am I doing up?

 

[MCI2424]

Love this line:
"....This is aggravated by the fact that, as a rule, older tape recorders will be defective...."


?????????????????????

 

[cjacek]

Love this line:
"....This is aggravated by the fact that, as a rule, older tape recorders will be defective...."


?????????????????????

Yeah, I got stuck at that precise line and thought.. "what the hell is this fool talking about?" .... Yeah, a mind bender to be sure. Was the guy drunk or something? Geez.....

 

[cjacek]

...anyway, will finish up the article a bit later... when I'm fully awake.

 

[evm1024]

I have no doubt....

....that some day digital will be an acceptable recording media.

Ages ago I started with machine that had 128 kbytes of storage. I am now sitting on 100 TB of data in raid 5 arrays. That is about a 780 billion to one increase in storage of my life. Ditto, CPU speeds and network speeds. And lets not forget the 8 GB thumb drives.

I can easily envision having 32 TB thumb drives with automatic bad sector replacement as well as 100 PB "disk drives" for music storage. Longevity in digital storage comes about by replication. We can expect that to be solved for consumers at some time in the future.

Moving over to the actual A/D - D/A process I think that when we go to 32 bit floating point encoding we will achieve a usable dynamic range. I pick 32 FP with a 5 bit exponent. Using a 1 volt reference full scale would give us around 32 volts. With the 27 bit mantissa we have around 162 dB encoded of which we need to around 60 dB as guard bands for both the noise floor and hard clipping.

What is an interesting problem is deciding on a sampling rate. When we sample the information between any to sampled points is lost. As the article points out current digital assumes that this is a sine wave and thus introduces distortions. Phase, frequency and harmonic relationships are lost and the distortions create artifacts which are not harmonically related and are thus very "harsh" to our internal signal processing. It has been shown that 7th harmonics are quite apparent and objectionable to most listeners. What is very interesting is that they can be heard/noticed at 70 dB below the fundamental and that the fundamental does not need to exist as that the brain will deduce it.

Thus the sampling rate should be selected to reduce the distortions and artifacts to a level at least 70 dB below the momentary signal level. Sampling rate and guard band work together here. I have not done those calculations but I suspect that the rate lies somewhere above 256 K samples per second.

At 32 bits per channel and 256 KHz sampling rate we have about 1 MB per channel per second of data to store. (not going in to digital compression schemes at this time). So, a typical 5 minute song in stereo will take up 112 MB. This works out to 6.7 GB per hour. If we assume 40:1 compression we end up with 168 MB per hour. Or, 6 hours per GB. This is about 2.1 GB/hour for a 24 track recorder. Well within the capacity I have here at work ignoring that I cannot get that must data on or off the disks at a realtime rate.

Network bandwidth is another problem that I think will be solved. There are some real limits, the speed of light being one. If we assume a 1 Gbps network backbone we can transfer the 32 b, 256 kHz, 2 channel audio in real time. It works out to 16 Mbit per second. An hour of recorded audio could be transferred in about a minute. USING the ENTIRE backbone bandwidth.

Thus network speeds will need to increase. Even going to 10 gbps backbones will not solve the problem on a regional scale. Witness the problems that iPhones have created on many university networks.

There are other encoding schemes that will be useful. compressing the data stream does not help as that compressing a compressed file resultes in a larger file.

One such scheme is to increase the data density by moving from binary to trinary sometime in the distant future. Bit rate and trit rate would be the same but trits encode more data for the same clock rate. (2^32 is roughly equal to 3^20 thus only 20 clock states are needed rather than 32) But that is so blue sky....

So if you have made it this far you can ask if I am full of it or if I'm on drugs. And, just what is the point.....which is that digital has a long way to go but I am sure that we will get there someday.

Till then tape rules.

 

[Beck]

...anyway, will finish up the article a bit later... when I'm fully awake.

Yeah, read it while wide awake for sure. English probably isn’t Dr. Artur Seibet’s first language, but many technical points are right on in this article. At times he rambles a bit, but it doesn’t take away from his technical insight overall. He certainly has the credentials, so as I said some of the points he brings up are worth further research.

Some of us have already said on this forum much of what he’s said in this article… as much as we could be put in laymen’s terms.

 

[The Ghost of FM]

Love this line:
"....This is aggravated by the fact that, as a rule, older tape recorders will be defective...."


?????????????????????

I think all he's really referring to is that older decks will stray from their original factory fresh specs because of mechanical wear and the slowly declining slope of parts such as electrolytic caps.

While a modest degree of analog tape users will get heads re-lapped and guide posts rotated or replaced, very few of us will recap an entire machine and instead wait for parts to outright fail and only then replace the failed parts alone.

It's sort of like religion in that it's one thing to know the rules and another to actually follow them.

Cheers!

 

[cjacek]

I think all he's really referring to is that older decks will stray from their original factory fresh specs because of mechanical wear and the slowly declining slope of parts such as electrolytic caps.

No doubt that's what he meant but he surely has problems expressing that properly..

 

[cjacek]

Yeah, read it while wide awake for sure. English probably isn’t Dr. Artur Seibet’s first language, but many technical points are right on in this article. At times he rambles a bit, but it doesn’t take away from his technical insight overall. He certainly has the credentials, so as I said some of the points he brings up are worth further research.

Some of us have already said on this forum much of what he’s said in this article… as much as we could be put in laymen’s terms.

I agree. A good read indeed and yeah, you can tell he did some "research" here, on HR.com/bbs

 

[Tim Gillett]

Ethan,


You didnt mention that we decide the sampling rate on the needed bandwidth.

Yes the information between the two sampled points is lost. But having decided a sample rate based on the required bandwidth, the information not captured is outside the bandwidth specifications and so irelevent.
All equipment has a specification, whether analog or digital. To criticize a piece of equipment because it performs to its specifications but not beyond them is just silly.
Yes the equipment assumes the points at the limits of its sample rate is a sinewave because that's what the specification means. So when the specification for an analog machine says it has a response up to say 20khz within a certain tolerance, it's talking about sinewaves too!

Your discussion about harmonics is interesting. Without mentioning actual frequencies it makes little sense. 7th harmonic of which frequency?

If harmonic distortion must not be allowed then what about the known harmonic distortion in analog tape recording? Why do manufacturers quote a harmonic distortion factor at say 400hz (where that harmonic distortion is well within the hearing range ) but you seem to be worrying about digital induced distortion of frequencies which are much higher and therefore less likely to be heard anyway? Any harmonics of say 20khz will be inaudible regardless of whether induced by a digital or an analog recorder.

Do we need sample rates of 256khz? That's like saying do we need 100 khz bandwidth. If so, analog tape is sunk too. It cant capture anywhere near that specification.


Tim.

 

[evm1024]

Deeper undestanding needed

Hi Tim,

Psycoacoustics is not the study of engineering specs of audio equipment. It is the study of how sound interacts with humans and how we "hear". It has been shown that human processing of sound is very complex, Humans find it very easy to listen to a harmonic series that does not contain the fundamental and deduce the fundamental. It is the 7th harmonic of that (present or deduced) fundamental that we find quite offensive and noticable at low levels.

Between each sample point of a digital recording is data that is harmonically related to what is being recorded. This data is lost. As was pointed out in the paper digital recording assumes that we are recording a sine wave.

By assuming a sine wave we end up with differences between the program source and the (eventual) digital reproduction. These differences are distortions that are random and not part of the harmonic structure.

One of the points of what I wrote is to say that selecting the sampling rate based on required bandwidth is an engineers answer that ignores the human element. We need to select it to reduce the digitally introduced artifacts to an unperceived level.

Oh, Yes by the way. Some do say that we do need a 100 kHz bandwidth....

(Rupert Neve comes to mind http://www.prosoundweb.com/chat_psw/transcripts/rupert.php)

 

[Tim Gillett]

So Ethan, they are the demands you would place on a digital recording system.

Now where is the analog tape system that, I take it, is yet to be invented that will also meet those same stringent psychoacoustic and engineering specifications?

Tim

 

[MCI2424]

Hi Tim,

Psycoacoustics is not the study of engineering specs of audio equipment. It is the study of how sound interacts with humans and how we "hear". It has been shown that human processing of sound is very complex, Humans find it very easy to listen to a harmonic series that does not contain the fundamental and deduce the fundamental. It is the 7th harmonic of that (present or deduced) fundamental that we find quite offensive and noticable at low levels.

Between each sample point of a digital recording is data that is harmonically related to what is being recorded. This data is lost. As was pointed out in the paper digital recording assumes that we are recording a sine wave.

By assuming a sine wave we end up with differences between the program source and the (eventual) digital reproduction. These differences are distortions that are random and not part of the harmonic structure.

One of the points of what I wrote is to say that selecting the sampling rate based on required bandwidth is an engineers answer that ignores the human element. We need to select it to reduce the digitally introduced artifacts to an unperceived level.

Oh, Yes by the way. Some do say that we do need a 100 kHz bandwidth....

(Rupert Neve comes to mind http://www.prosoundweb.com/chat_psw/transcripts/rupert.php)

Rupert made an error some years ago when he designed and built a console for Geoff Emerick. Geoff complained that there was something wrong with some of the pre-amps. Rupert found that some transformers were out of phase (I think, it has been some time) and the frequencies affected were in the 56KHz range.

Rupert never let go of this and has repeated it a few times since. It has been shown that the problem with the pre-amps also affected some lower (within human hearing range) frequencies and Geoff's ears were so good, he could tell there was something wrong when most others could not. This is one of those great urban legends out there.


Here is some info on a more plauseable theory:
http://recforums.prosoundweb.com/index.php/m/78031/0/#msg_78031

And for the really brainy people out there:
http://jn.physiology.org/cgi/content/full/83/6/3548

Enjoy, it is a real bore.

 

[evm1024]

Remember I am proposing a digital system to surpass Analog

So Ethan, they are the demands you would place on a digital recording system.

Now where is the analog tape system that, I take it, is yet to be invented that will also meet those same stringent psychoacoustic and engineering specifications?

Tim

Hi Tim,

The same demands I would place on any system - that it does not introduce any artifacts at all.

This is not possible with any current system - analog or digital.

The sampling rate and bit depth that I suggested was one that I think has the potential to record and reproduce original source with artifacts reduced to a level below perception.

Some would argue that we have that at this time but I think that there is ample evidence to show that we are not there yet.

Any artistic colorations are just that. Something that we put in, not something that the system puts in. And not part of this discussion.

Regards

 

[evm1024]

Thanks for the references

Rupert made an error some years ago when he designed and built a console for Geoff Emerick. Geoff complained that there was something wrong with some of the pre-amps. Rupert found that some transformers were out of phase (I think, it has been some time) and the frequencies affected were in the 56KHz range.

Rupert never let go of this and has repeated it a few times since. It has been shown that the problem with the pre-amps also affected some lower (within human hearing range) frequencies and Geoff's ears were so good, he could tell there was something wrong when most others could not. This is one of those great urban legends out there.


Here is some info on a more plauseable theory:
http://recforums.prosoundweb.com/index.php/m/78031/0/#msg_78031

And for the really brainy people out there:
http://jn.physiology.org/cgi/content/full/83/6/3548

Enjoy, it is a real bore.

I'm not sure that I would put that in the arena of Urban Legend. But you may of course.

As I recall the 3 channels output transformers were missing a bypass cap
and that Neve could not (at first) hear the difference. Emerick could. (Even my deaf great aunt could tell if they were out of phase so I think we can dismiss that as the problem)

The point being not that Emerick was hearing a 3 dB rise at 50+ kHz but rather he was hearing the results of harmonic mixing that fell in the audible range.

Regards

PS I'll give those references a look see once I get rested.

 

[Tim Gillett]

Hi Tim,

The same demands I would place on any system - that it does not introduce any artifacts at all.

This is not possible with any current system - analog or digital.

The sampling rate and bit depth that I suggested was one that I think has the potential to record and reproduce original source with artifacts reduced to a level below perception.

Some would argue that we have that at this time but I think that there is ample evidence to show that we are not there yet.

Any artistic colorations are just that. Something that we put in, not something that the system puts in. And not part of this discussion.

Regards

Ethan, interesting that you say neither digital nor analog tape can do it at present.

If you raise the bar to a 100khz bandwidth then that rules analog out for starters. It can do the high frequencies (far higher than 100khz) but then it goes wobbly on the lows. Cant do both at the same time.

Ethan, by implication at least you have condemned analog tape to your own dustbin, because it never did and, I venture, never will reach the dizzy heights of fidelity you claim are needed.

By contrast, digital , to my knowledge, has already well and truly passed (1) your bandwidth requirement (however ridiculous I think most would regard it as) and (2)your signal to noise requirement is just absurdly beyond analog equipment's capabilities. I mean things like microphones, preamps, power amps etc.

Eventually we have to come back to the real world.

regards Tim

 

[Tim Gillett]

I'm not sure that I would put that in the arena of Urban Legend. But you may of course.

As I recall the 3 channels output transformers were missing a bypass cap
and that Neve could not (at first) hear the difference. Emerick could. (Even my deaf great aunt could tell if they were out of phase so I think we can dismiss that as the problem)

The point being not that Emerick was hearing a 3 dB rise at 50+ kHz but rather he was hearing the results of harmonic mixing that fell in the audible range.

Regards

PS I'll give those references a look see once I get rested.

Interesting. And if the Neve desk that had been installed had not had that super ultrasonic response, the alleged beat frequency you seem to be referring to would never have been produced in the first place, because neither of the two frequencies that mixed together to produce the third, would have been passsed in the first place either... Perhaps one of the reasons why much good audio gear is designed to filter out supersonic frequencies before they create such problems? But I digress...


regards, Tim. All the best for your trip. Yes, drop by Oz on your sail.

 

[MCI2424]

I'm not sure that I would put that in the arena of Urban Legend. But you may of course.

As I recall the 3 channels output transformers were missing a bypass cap
and that Neve could not (at first) hear the difference. Emerick could. (Even my deaf great aunt could tell if they were out of phase so I think we can dismiss that as the problem)

The point being not that Emerick was hearing a 3 dB rise at 50+ kHz but rather he was hearing the results of harmonic mixing that fell in the audible range.

Regards

PS I'll give those references a look see once I get rested.

I am referring to the original time. When this happened, the news headlines were "Geoff Emerick hears 54KHz".

Of coarse it was really something in the audible range, but people were amazed at the time.

 

[evm1024]

Don't focus on 54 kHz

Ethan, interesting that you say neither digital nor analog tape can do it at present.

If you raise the bar to a 100khz bandwidth then that rules analog out for starters. It can do the high frequencies (far higher than 100khz) but then it goes wobbly on the lows. Cant do both at the same time.

Ethan, by implication at least you have condemned analog tape to your own dustbin, because it never did and, I venture, never will reach the dizzy heights of fidelity you claim are needed.

By contrast, digital , to my knowledge, has already well and truly passed (1) your bandwidth requirement (however ridiculous I think most would regard it as) and (2)your signal to noise requirement is just absurdly beyond analog equipment's capabilities. I mean things like microphones, preamps, power amps etc.

Eventually we have to come back to the real world.

regards Tim

Hi Tim,

Don't focus on 54 kHz. It is a non starter and just an example.

The major difference between analog (tape) and digital today is in the artifacts and distortions they both create. With tape the distortions are often pleasing to the human ear whereas digital distortions are not. Perhaps that is why we tend to call them artifacts.....

I think that the folks here on analog only use tape for a number of reasons including that we like "vintage" equipment.

And yes, some future generation of digital will capture and reproduce audio without audible artifacts and will become the norm. The 1024 is not in my screen name because I hate digital....

Regards