Let’s talk DAT (Digital Audio Tapes)

[gregorio]

[1] DAT is LPCM which is different than PCM. PCM is pulse coded modulation, whereas LPCM is linearpulse coded modulation. Linear means that the values are linearly spaced - the values are directly proportional to the signal amplitude
[2] So, anything that has the meaning to record in LPCM can take huge advantage of Magnetism.
[3] Because the Amplitude are infinitely accurate rather than guesstimate modulated like PCM from PC or the modern PCM encoding.
[4] This was why I called DAT is a perfect hybrid technologies from the point of view of Digital.
[4a] Instead of typical PCM where the amplitudes and reconstructions of the square waves have to be modulated to become, the magnetism mechanic of DAT tapes already did the works.

1. Correct. However, what difference does it make? Sure, DAT is LPCM but so is CD and pretty much all lossless digital audio formats (with the exception of DSD)!

2. LPCM takes no more "advantage of magnetism" on a DAT tape than it does on a HDD and in fact there are better (faster and/or more reliable) media types which don't involve magnetism on which to store LPCM, CD or SSD for example.

3. No, there is no amplitude with digital audio, it's just zeros and ones (bits) that represent amplitude values and obviously we have a limited number of bits, 16bits in the case of CD and DAT. As you yourself quoted, with LPCM "the values are directly proportional to the signal amplitude", NOT "infinitely accurate"! However, we do not need infinitely accurate digital values in order to achieve audibly perfect analogue signal reconstruction, which is the whole point of digital audio, it wouldn't work otherwise! FURTHERMORE, as just mentioned, there is no "rather than", wav and aiff files ("PCM from PC or the modern PCM encoding"), redbook CD, AES3, DVD, BluRay, HDMI, DCP, LaserDisk, RF64 and DAT, are all LPCM!

4. Obviously that statement is incorrect. As just explained, DAT is NOT a "hybrid technology" it's exactly the same basic technology as CD and computer wav and aiff files, the only difference is that the zeroes and ones are stored on magnetic tape rather than magnetic platters (HDDs).
4a. No, there is no "reconstructions of the square waves" and the magnetic mechanism of DAT does not "already did the work". What you seem to be describing is an analogue conversion, NOT a digital to analogue conversion! Digital audio is just zeros and ones, the difference between a one and a zero is represented by a polarity difference in the magnetic field on DAT tapes, floppy drives and HDDs (or by pits and lands on optical media for example). The output of all the different media types is ultimately an electric signal with an alternating high and low voltage which the DAC chip converts back into "ones" and "zeros" respectively and THEN "does the work" of reconstruction. You seem to be getting confused between analogue and digital.

[1] I never said anything I posted were facts.
[2] I said that I observed the differences and it sounds better than what the new digital system has and that is that.
[3] All signals digital would have to come out as Direct Stream Digital first before it comes out to be analog, so Ofcourse it has 1 bit converter
[4] What else would it be ? PCM and digital amplifications instead was what you meant ? Similar to A class D amp ?
[5] The SV-3800 Pro-DAT features with a 1-bit, 64X oversampling A-D converter.

1. If that's true then I'm sorry but I don't understand much of what you posted. For example, when you stated "The idea of why I am interested in DAT tapes is that it is a hybrids between digital and analog. It doesn’t rely all on the algorithms to generate and reconstruct the pulses. It utilizes magnetism instead, and at the speed that it has, the density is pretty awesome. Just as you mentioned, DAT tapes has no wow/flutters." - If you never said anything you posted "were facts" then: "The idea of why you are interested in DAT tapes is that it is NOT a fact that DAT is a hybrid. And, you are not saying it's a fact that DAT doesn't rely on the algorithms, utilises magnetism instead or have speed or density this is pretty awesome" - Which I don't understand and seems to be the opposite of what you actually stated.

2. The only rational explanation is that there must have been some error in your observation because there is no difference, it's exactly the same PCM (LPCM) as found for example in wav files. The newer "digital systems" (such as computer wav files) are actually technically superior to DAT, because in addition to the 16bit 48kHz max limit of DAT, they also allow for LPCM with greater bit depths and sample rates, although this results in no audible improvement for consumer distribution.

3. Clearly that cannot be true. The storage/playback of Direct Stream Digital (DSD) first became available in 1999, 12 years after DAT was released! And as you yourself have quoted, the digital signal on DAT tapes is LPCM, not DSD. What "comes out" of a DAT tape is 16bit LPCM, then, depending on the DAC (either internal or external) that 16bit LPCM signal maybe converted by the DAC's converter chip into an oversampled delta/sigma (DS) signal.

4. Sorry I don't understand, what "digital amplifications" and I don't see the similarity/connection with analogue signal amplifiers?

5. As far as I'm aware, pretty much ALL professional ADCs from the beginning of the 1990s onwards initially converted the analogue input signal to digital using one or a handful of bits at very high sampling (oversampled) rates. However, this is NOT what the ADC chips actually output, after initial digitization the signal is converted into LPCM (at the selected bit depth/sample rate) in a process called "decimation". Of course there could be no other option until 1999 because there was no way to store a 1bit 64x oversampled signal and, this is exactly how professional PCM (LPCM) ADCs still work today, although 256x or 512x oversampling became the norm in the 2000's, which provides marginally better noise performance. So, the difference you describe does not exist, it's exactly the same basic technology as today's technology but with slightly worse performance, not better.

Again, why don't you just ask if you don't know, instead of making incorrect factual statements that you then say were not factual, which is extremely confusing and misleading? So for example, try asking: "What's the difference between the digital data stored on a DAT tape compared to the digital data stored in a wav file on a computer drive?" instead of incorrectly stating that "DAT is a different, hybrid technology compared to wav files on a computer drive" and then later that "you don't really know how DAT works and want to learn".

G

 

[Whitigir]

3. No, there is no amplitude with digital audio, it's just zeros and ones (bits) that represent amplitude values and obviously we have a limited number of bits, 16bits in the case of CD and DAT. As you yourself quoted, with LPCM "the values are directly proportional to the signal amplitude", NOT "infinitely accurate"! However, we do not need infinitely accurate digital values in order to achieve audibly perfect analogue signal reconstruction, which is the whole point of digital audio, it wouldn't work otherwise! FURTHERMORE, as just mentioned, there is no "rather than", wav and aiff files ("PCM from PC or the modern PCM encoding"), redbook CD, AES3, DVD, BluRay, HDMI, DCP, LaserDisk, RF64 and DAT, are all LPCM!

I never claimed anything I posted were facts....I was posting as debating to learn And was stating that pretty clearly. anyways, You are right

 

[originalsnuffy]

DAT is LPCM which is different than PCM

PCM is pulse coded modulation, whereas LPCM is linearpulse coded modulation. Linear means that the values are linearly spaced - the values are directly proportional to the signal amplitude

So, anything that has the meaning to record in LPCM can take huge advantage of Magnetism. Because the Amplitude are infinitely accurate rather than guesstimate modulated like PCM from PC or the modern PCM encoding.

This was why I called DAT is a perfect hybrid technologies from the point of view of Digital. Instead of typical PCM where the amplitudes and reconstructions of the square waves have to be modulated to become, the magnetism mechanic of DAT tapes already did the works.

I am not convinced that this is correct. As far as I know, PCM and LPCM are one and the same. You could rip a CD to DAT and vice versa; non destructively. Just different technology for laying down the digital info; but CDs and DATs are virtually the same. Though I think DAT offered both 16-44 and 16- 48.

 

[Whitigir]

I am not convinced that this is correct. As far as I know, PCM and LPCM are one and the same. You could rip a CD to DAT and vice versa; non destructively. Just different technology for laying down the digital info; but CDs and DATs are virtually the same. Though I think DAT offered both 16-44 and 16- 48.

what I have learned so far is that

LPCM is uniformly quantized and PCM is non uniform. The PCM have to follow logarithm to process, and hence the differences between them are quantization errors and accuracy. There are many steps within when the Binary encoded on a drive are being processed, for example, an electrical signals have to pull the information out, and during this time any fluctuations in power supplies or noises induced into the supplies would bring in the errors + the errors from PCM and quantization, then being over sampled through Sigma Delta are just going to multiply the errors and results in a more degradations. However one look into debating it, it could be a pros to retrieve better details with higher dynamic range and less storages, or higher noises and higher power consumptions to process and filters the ultra-range frequencies.

The advantages of DAT and or Optical eyes in CD (with Linear PCM) are so the eyes with photosensor or magnetic heads can basically pull out the accurate informations without the power supplies errors. There will be errors counter as mentioned above or REED to recover the informations losses, but at the expenses of details or higher frequencies....typically it is ok and as human we don’t realize it existence until it get so bad and totally distort and became unintelligent Or stuck repeating. Hence periodic cleaning of these mechanism needs to be performed. You can see this photo of square wave digital as per Raise time for 1 string and 0 for 0 string To represent the informations


The more accurate quantizing steps = the higher the accuracy to stay to the original sinuous waves. The less accurate the quantizing steps would requires the higher sampling in order to get that accurate originally waves, but this will result in higher quantization noises which would then require a more aggressive Digital filterings.

So, when reading live an encoded PCM or LPCM stored on a SS, while processing live on sigma delta modulations and DSP altogether, the errors just becomes even more involved and further degrading the reconstructed sound performances. This is why carefully designing DAC section from a good power supply vs a weak design of supply would results in a variety of different performances giving the same DAC-IC being used. To mitigate this, the files are better offline processed with DSP applications and Sigma Delta modulated into DSD or DSF files. This will only ask for a Low pass filter to become analog and less live processing and errors could be involved. The order, the timing, and the buffering, latency of pulling informations on solid drives would also result in a variety of different characteristics. This is where firmwares and or different playback apps can have different signatures as well. But of course, once you only thinking 1 and 0 are all the same 1 and 0 then, there is no need to get any further

Delta Sigma applied to a PCM or LPCM is to process the digital information Into 1 bit stream. The 1 bit stream itself is a DSD signals or direct stream digital, and this 1 bit needs exactly 1 bit DAC as an interface to transfer it into Lowpass Filters for becoming analog. DSD as categorized started out as minimum or as low as DSD64, and that is 2X over sampling over the original Nyquist theory. The Non over sampling 1 bit stream was never called DSD. DSD64 and higher would need an interfaces fast enough and capable enough to handle the stream, which was not available then.

The older chips were able to directly generate the digital wave form from 16 bits LPCM into analog form as an all in one IC. However, there were and still are machines that has 1 bit DAC such as 75ES Sony DAT and Panasonic SV3800

 

[gregorio]

[1] LPCM is uniformly quantized and PCM is non uniform. The PCM have to follow logarithm to process, and hence the differences between them are quantization errors and accuracy.
[2] There are many steps within when the Binary encoded on a drive are being processed, for example, an electrical signals have to pull the information out, and during this time any fluctuations in power supplies or noises induced into the supplies would bring in the errors + the errors from PCM and quantization, then being over sampled through Sigma Delta are just going to multiply the errors and results in a more degradations.
[2a] However one look into debating it, it could be a pros to retrieve better details with higher dynamic range and less storages, or higher noises and higher power consumptions to process and filters the ultra-range frequencies.
[3] The advantages of DAT and or Optical eyes in CD (with Linear PCM) are so the eyes with photosensor or magnetic heads can basically pull out the accurate informations without the power supplies errors.
[3a] There will be errors counter as mentioned above or REED to recover the informations losses, but at the expenses of details or higher frequencies....
[3b] typically it is ok and as human we don’t realize it existence until it get so bad and totally distort and became unintelligent Or stuck repeating.
[4] Hence periodic cleaning of these mechanism needs to be performed.

1. Again, there is effectively no difference between PCM and LPCM because whenever the term PCM is used in audio applications, it is always LPCM and therefore, "the differences between them" do not apply!

2. Unfortunately, this is yet again incorrect! Fluctuations or noise in power supplies does NOT cause errors and there are no errors "from PCM and quantisation" because it's LPCM and not PCM!
2a. What "higher dynamic range" and "ultra-range frequencies"? You again seem to be confusing an analogue signal with a digital signal, despite the fact that you yourself actually posted a representation of a digital signal! As your posted image states: The signal has "two distinguishable levels", that's it, nothing more, there are NO different frequencies and there is NO dynamic range except the difference between the two states. The signal is completely unaffected by noise or any other type of interference, unless the interference is so great that the two different levels/states cannot be distinguished. A noisy "one" is still just a "one" and is therefore exactly the same as a noiseless "one". This is the whole point why digital audio transfer was developed in the first place.

3. Correct, although the same is true of all other digital data storage devices.
3a. This statement is obviously false, there cannot be a loss of "details" or "higher frequencies" because there are no details or higher frequencies to loose, ONLY zeros and ones, NOTHING else. Reed-Solomon error correction looses nothing, the error corrected zeros and ones are absolutely identical to the original zeros and ones. The only time we could have errors is when there are so many consecutive errors that the error correction cannot engage. Therefore:
3b. Correct, as humans (or in fact anything else) we cannot "realize it's existence" because there's absolutely no difference to "realize", unless, as you say, there are so many errors the error correction cannot engage/function and we get a dropout or total distortion.

4. Periodic cleaning of DAT players/recorders was necessary. Deposits from the tape coating or dust/dirt could build up on the tape heads and certain metal oxide tapes could physically wear the heads both of which cause the misreading of the data, eventually to the point of an unrecoverable number of errors. However, this isn't the case with other storage media types (such as CD or HDD for example), because there is no physical contact between the media and the reading/writing mechanism.

[1] The more accurate quantizing steps = the higher the accuracy to stay to the original sinuous waves.
[1a] The less accurate the quantizing steps would requires the higher sampling in order to get that accurate originally waves, but this will result in higher quantization noises which would then require a more aggressive Digital filterings.
[2] So, when reading live an encoded PCM or LPCM stored on a SS, while processing live on sigma delta modulations and DSP altogether, the errors just becomes even more involved and further degrading the reconstructed sound performances.
[3] This is why carefully designing DAC section from a good power supply vs a weak design of supply would results in a variety of different performances giving the same DAC-IC being used.
[4] To mitigate this, the files are better offline processed with DSP applications and Sigma Delta modulated into DSD or DSF files.
[5] The order, the timing, and the buffering, latency of pulling informations on solid drives would also result in a variety of different characteristics.
[6] This is where firmwares and or different playback apps can have different signatures as well.
[7] But of course, once you only thinking 1 and 0 are all the same 1 and 0 then, there is no need to get any further
[8] Delta Sigma applied to a PCM or LPCM is to process the digital information Into 1 bit stream. The 1 bit stream itself is a DSD signals or direct stream digital, and this 1 bit needs exactly 1 bit DAC as an interface to transfer it into Lowpass Filters for becoming analog. DSD as categorized started out as minimum or as low as DSD64, and that is 2X over sampling over the original Nyquist theory. The Non over sampling 1 bit stream was never called DSD. DSD64 and higher would need an interfaces fast enough and capable enough to handle the stream, which was not available then.
The older chips were able to directly generate the digital wave form from 16 bits LPCM into analog form as an all in one IC. However, there were and still are machines that has 1 bit DAC such as 75ES Sony DAT and Panasonic SV3800.

1. Unfortunately, this is also incorrect! It would be correct if it were not for "dither" but dither is required and has ALWAYS been applied by all ADCs. The dither process effectively converts ALL inaccuracies to white noise, so what we end-up with, at ANY bit depth (number of quantisation steps), is effectively perfect accuracy plus a fixed amount of white noise. In the case of 16bit, this white (dither) noise is inaudible at any reasonable listening level and not even reproducible by the vast majority or reproduction systems.
1a. In the most extreme example, just 1bit encoding (only two available quantisation steps), then we still get effectively perfect accuracy upon reconstruction but such a large amount of white (dither) noise that it's almost entirely obscured. However, this dither noise can be "shaped" - moved to parts of the audio spectrum that are inaudible, thereby revealing the effectively perfect accuracy. SACD (which is 1bit) only works because of this fact!

2. There is no "reading live" LPCM from an SSD or "processing live on delta/sigma and DSP altogether" and there are no errors when reading data from an SSD or HDD because it's all error corrected. The only exception is if the storage media/device has a serious fault and large portions of the data are unreadable, in which case you get no data (a dropout).

3. No, either the power supply supplies the appropriate power for the DAC chip to work perfectly or the DAC chip won't work at all. There is the potential question of isolating the power supply from the reconstructed analogue signal but even cheap DACs manage this to levels well below audibility.

4. This is incorrect because there is nothing to mitigate!

5. There cannot be ANY different characteristics, let alone a "variety of different characteristics". The "order" of the zeros and ones cannot be different because this would be corrected by error correction. The timing and latency are irrelevant as all the data has to be buffered, because SSDs read data at several Gigabits per sec, while 16/44 for example only requires 1.4 Megabits per sec. The same is of course true of HDD: Although significantly slower than SSDs, HDDs still transfer data many times faster than required by audio and also has to be buffered. Surely you're not suggesting audible differences between audio data sourced from an SSD and HDD or other types of storage device?

6. Playback apps can have different sound signatures though, some apply compression or other processing to the LPCM data.

7. Of course one has to think that, because if one didn't, no digital device would ever work! All digital devices work precisely because there are only two possible states (one or zero), which can be stored and transferred without error.

8. I'm not sure what you're trying to say or what audible effect it could have. I will mention that even the very first consumer digital audio devices (CD players in 1983) used oversampling to reconstruct the analogue signals.

G

 

[Whitigir]

I don’t think you really know how a PCM is being oversampled and applied with sigma delta as you don’t seem to grasp the “oversampling and error corrections with noises and yet you talk about dithering”...

You are right again speaking of all the facts and not made up or assumptions

thank you

 

[Whitigir]

However, this dither noise can be "shaped" - moved to parts of the audio spectrum that are inaudible

yet you keep on saying that I/you don’t know what ultra range frequencies I mentioned about.

We are talking about preserving the original signals and yet you are talking about how to correct and add noises and make the errors irrelevant.

 

[originalsnuffy]

Pretty simpleminded on my part. I can transfer my DAT tapes to my computer losslessly using my coax input for the audio card. At that point, bits is bits.

 

[gregorio]

We are talking about preserving the original signals and yet you are talking about how to correct and add noises and make the errors irrelevant.

No, I am not talking about making quantisation errors "irrelevant", I am talking about converting quantisation error into inaudible white noise, so that they don't exist. And obviously, "the original signals" with no error would be, by definition, perfectly "preserved". However, as we also have some additional inaudible noise, then we are limited to stating that "the original signals" have been preserved audibly perfectly.

G

 

[sander99]

@Whitigir: I stumbled into this old thread by coincidence and I think I can clarify the PCM vs LPCM confusion a bit.
PCM is simply the collective name for various forms of PCM. LPCM is more specific collective name for all forms of PCM that use linear quantization. But it happens to be so that almost all PCM is LPCM.
There indeed also exists forms of PCM that use non-linear quantization. For example it was used by some DV video cassette recorders, quote from
https://en.wikipedia.org/wiki/DV:

Audio can be stored in either of two forms: 16-bit Linear PCM stereo at 48 kHz sampling rate (768 kbit/s per channel, 1.5 Mbit/s stereo), or four nonlinear 12-bit PCM channels at 32 kHz sampling rate (384 kbit/s per channel, 1.5 MBit/s for four channels). In addition, the DV specification also supports 16-bit audio at 44.1 kHz (706 kbit/s per channel, 1.4 Mbit/s stereo), the same sampling rate used for CD audio.[8] In practice, the 48 kHz stereo mode is used almost exclusively.

On a side note: Although the statement "PCM is LPCM" is not always correct, it is almost always correct (in the context of audio applications, as @gregorio mentioned), and next to that I think you can safely trust everything else @gregorio wrote in this thread.

 

[Chops]

Did somebody mention DAT and analog? The answer is yes.

This is when I had the Panasonic SV-3700 (purchased new) in the main system about 6 months ago. I now am in the process of building up a "vintage" system in the spare room, so it's in there right now, which that room is a total wreck at the moment, so no pics of that hot mess.

I also have a stack of TOTL cassette decks behind me in the main system, connected to the Freya+ via 30' long XLR's and a Rane Balance Buddy. Two Nakamichi decks, a Yamaha and a JVC (all purchased new except the Yamaha). There's a specific reason why I have those particular decks (including the DAT) as well, but that's for another thread.

 

[originalsnuffy]

Nice rig. What's with all the DBX equipment? Do you have a lot of DBX compressed music?

I moved recently and sold off my last Nakamichi piece. Still have a Tascam DAT player but it gets used like zero. Almost all my music just goes through the HTPC into my receiver. Sometimes I feed music from an older Dune media player and sometimes from my Amazon Fire Cube. But most of the time its HTPC with either Foobar or JRiver.

 

[Chops]

Nice rig. What's with all the DBX equipment? Do you have a lot of DBX compressed music?

Thanks.

The two top dbx pieces are 400x route selectors, just to route to and from the various decks, as well as the big old Pioneer reel to reel when I get that up and running. But it's huge and heavy, hence why it's not in the system yet.

The bottom dbx with the meter is a 224x noise reduction unit, kind of like Dolby NR, but a lot better. The Yamaha cassette deck also has dbx NR built into it, along with Dolby B/C & HX Pro.

No dbx compressors/limiters/expanders in the system.

 

[originalsnuffy]

When I was trying to be a completist, I purchased a copy of Broken Barricades on vinyl that had DBX encoding. Of course I never had a DBX encoder so never played it. I am sure it was rare but I got tired of carrying it around when I moved as a college student. So I sold it.