When did recording studios switch to digital recording from tape?

Just for fun, I grabbed a couple of old Mix Magazine issues of the shelf, one from 1992, the other from 1998. In 92, analog decks were still featured in articles and advertisements, along with tape from BASF and Ampex. The brands that offered analog record decks tended to also have digital models, including Sony, Studer and Tascam, whom I didn't realize had a DASH 24 track. Mistubishi had their digital line up, including a 3rd generation 32 track. Alesis ADAT was featured prominently as I think it had just been introduced a year or 2 earlier.

By the 1998 issue, analog still gets a mention here and there, but clearly is not taking up near as much space. More digital audio workstation ads and software plug-in ads and articles are featured as even DASH tape based digital seems to be in decline by this point. ADAT had moved into ADAT II which was a 20 bit format, still using S-VHS tape. More mention of digital outboard gear, hard drive based recorders and full digital consoles by this time as well. Early mentions of DSD and SACD pop up a couple of times, one being a small article on Telarc and how they'd moved from early Soundstream recorders when things still came out on vinyl, to different PCM recorders and output on CD, to eventually SACD. I'm not sure what formats they currently support, if they're still around.

Anyway, it was interesting to see that magazine's focus change. Wish I had only kept more issues.

 

Funny. Then years later, Mits trounced nearly everyone in the rear-projection TV game....credit where credit is due.



Dan

 

I think country music was among the earliest users of digital tape. Alabama was using Mitsubishi machines by 1984 or 1985 if I remember correctly, and it seemed like almost all big names in country were using it by 1984. Classical music was already perfecting the use of digital, and some big pop names were using it. But somehow, Janet Jackson's Rhythm Nation 1814 I think was ADD or DAD. But I could be wrong. However, it took rock and metal ages to move to digital from what I can gather. Mid 90s, a lot of bands were still doing analog tape to record on. Lots of grunge and metal, specifically.

 

Agree on country music adapting early. I got the impression those Mitsubishi machines were more popular in Nashville than either coast. Dunno if that’s true or not, just a feeling I got at the time. I think there was (and continues to be) a lot of hybrid sessions. The liner notes of Pink Floyd’s Momentary Lapse of Reason states that the bass and drums were recorded analog. I seem to remember Metallica’s black album being tracked on both 2” and DASH in their A YEAR AND HALF IN THE LIFE OF videos. I think Motley Crue’s Doctor Feelgood was digital, so maybe Bob Rock specifically liked some of the things digital could do.

 

The way that I am writing the software -- requires a lot of CPU. Middle 1990's computer technology would likely have been overwhelmed trying to do full quality DA decoding. It might have been possible to approximate a DolbyA decodnig operation on a fast DSP or very fast computer of the time, but there is a LOT of math going on -- -a mind boggling amount of math. For an idea -- the main high level section of the code is pretty dense at 3400 lines, but 1/3 of those lines, where there might be an add/multiply or subtract operation, it looks like just one being done... Nope, it is actually doing 8 operations at a time (the wonders of SIMD programming.) Almost every instruction is done on EVERY sample, and things like 3072 tap Hilbert transforms (90 degree phase shifts) are CPU hogs. For the lowest quality modes, three are being done for each sample, the purpose being a slippery way of mitigating the IMD created by the

Just to follow up on a somewhat mellow DolbyA encoded material vs the decoded equivalent. Usually, just doing the -3, -5 or -6dB at 3kHz fixes things to be listenable instead of decoding. Of course, decoding usually (believe it or not, not always) sounds better though.
* Note, the only reason for using DHNRDS instead of a real DolbyA is that it isn't available/set-up right now. DHNRDS the comparison between undecoded and fully decoded is still operative. The EQ only version sounds 'okay', but still has hiss. Decoding DolbyA is VERY difficult -- esp difficult to sound good, even with DolbyA HW. I alledge that SOMETIMES there is undecoded DolbyA material distributed with EQ instead of decoding -- still 'acceptable' sound, but more convienient when it isn't obvious that decoding is actually necessary.

Carpenters -- Reason To Believe
DolbyA encoded: Dropbox - 04 - Reason to Believe-Encoded-000-055.mp3 - Simplify your life
DolbyA encoded, but EQ to sound 'okay': Dropbox - 04 - Reason to Believe-Encoded-EQonly-000-055.mp3 - Simplify your life
Decoded by the soon to released DHNRDS decoder: Dropbox - 04 - Reason to Believe-V0.9.5C-Decoded-000-055.mp3 - Simplify your life

 

Yes and no. Fleetwood Mac were using digital back on Tusk in '79. Madonna's Like A Virgin was all-digital. Sting went digital fairly early on. Not sure about harder rockers than those two, but I'm betting quite a few were digital recordings as well. Tunnel of Love was noted as being recorded on Sony Digital, and the mastering was done on Neve's digital console per the liner notes.

 

It gets even more complicated, since by the mid-'80s a lot of stuff was going thru samplers like the Fairlight or Synclavier, even things like the lead vocals. And a lot of effects processors - everything from special effects to just the basic ambience settings, reverb and echo - were done digitally. So you could be recording to "analog" tape, but literally everything being put down on that analog tape could have gone thru a sampler, digital reverb, echo and compression, an effects gadget like an Eventide Harmonizer (which had been around since the '70s), or potentially all of them.

So just how "analog" is that recording? The short answer is, it isn't. I think that makes it almost impossible to tell if high-res remasters of pop/rock/dance material recorded from '82 onwards are taken from an "analog" master or not, and it may be irrelevant. I'm guessing on a lot of that material, the only ultrasonics are noise and the bias signal, since virtually everything else went thru a 44.1kHz (or less!) digital stage prior to ending up on that master, and any ultrasonic material would have been filtered out long before it got to the final master.

 

At the risk of seeming like I talk about the same thing all of the time -- I'd suspect that a lot of the satisfying excursions above 20kHz would be from the NR system (of whatever type, including DolbyA) producing IMD splats. You can start with pristine, 20kHz frequency response limited material -- run it through certain NR systems, then lo+behold, so called 'music' extending above 20kHz!!!

Of course, for recordings made in digital format, the NR system wouldn't be used during production, but when producing the final tape used for making vinyl albums -- just might include analog tape with DolbyA/SR? I do NOT know the process, but I have seen a suspiciously large amount of signal up to the 30+kHz range where the recording might have been done in primitive/low sample rate digital, and also most microphones might not have a lot of response up that high?

This is WAY WAY outside of my area of exprtise, but distortion products sometimes being what is thought of as 'music' seems to be plausible, considering material being produced at 44.1k, yet the existance of 'stuff' well above 20kHz -- of course, coherent signals (tones/28.8kHz for example) are often up there also. What do you think?

John

 

Yup, I think the vast majority of crap people find up in the ultrasonic range on "high-res" recordings are all kinds of electronic artifacts and noise. I suspect little to none of it is actual signal, especially on pop/rock/country recordings once you get into the '80s. Too much digital in the signal chain for that stuff to be music.

Even before the digital era, I suspect a lot of ultrasonic content would have been filtered out along the way, even intentionally by the equipment itself, in order to avoid noise and interference. You wouldn't want your $15,000 recording ruined because some microphone was making a loud ultrasonic noise and saturating your tape...

I've tried shifting the ultrasonic stuff I've seen on some high res recordings down into the audible spectrum to see if it sounds like something made by cymbals or whatever, but it always just sounds like noise of some sort to me. IMD would be a good guess.

 

True, by the mid-80s, these were the only guys still recording all analog!

 

For pop/rock stuff I first took notice around 1981 with the Rickie Lee Jones Pirates album. My friends father was a Classical music lover and I remember seeing the Soundstream Digital Banner/Logo thingy on his albums a few years before that. I saw it as a back and forth thing in pop/rock until Protools came out, then a major shift to digital.

 

I knew that Fleetwood Mac experimented with digital, but they only did one album, I think. Wasn't good enough to compare with analog tape back in 79-80. And I honestly had no clue Madonna or Sting used digital then. Always thought all their stuff till the 90s was analog. I knew Sting went digital a little earlier than that, but I always thought they were analog. Then again, I don't listen to either one, so I've got no clue

 

I think the analog drums and bass are what give MLOR such a great sound. It has a good bottom end but everything else does sound digital. I think Bob Rock liked it because it was easier to edit than analog, but for sound quality, analog was the way to go till the early to mid 90s when it became high enough quality for the price a lot of these bands could afford to use.

 

"Nothing Like The Sun" from 1987 was Sting's first DDD album. "The Dream of the Blue Turtles" from 1985 was ADD.

 

Right, they mixed down to digital. Quite a few records did that. Apparently Touch by Eurythmics was ADD as well, at least in part.

 

This isn't even remotely true.

 

Depends on what you mean by "recording digitally". Sony moved a lot of those outboard PCM units. I'm betting any decent-sized studio had one sitting around by the middle of the decade (or the equivalent from another manufacturer). Quite a few acts purchased their own and were recording with them at home as I recall, so the studios needed the things to play back tapes cut elsewhere.

In fact, wasn't Tusk partly recorded at Lindsey's house using the first Sony PCM converter and a U-matic or Betamax VCR?

 

Y'all can say what you want about early digital recordings, but IMHO "Nightshift" by the Commodores is one of the best-sounding records of the early 1980's and it was at least mixed digitally, according to the liner notes. Vinyl was mastered at Motown by John Matousek, assuming from the digital masters. It sounds great, even on vinyl.

 

In my opinion the analogue mutitrack recording mixed down to 192/24 digital trounces the original full digital recording which sounds more dinamic and noise free but also over-analitical, harsh and artificial. This is a soundtrack I've always loved but hated its Sony digital origins, I really really dislike how the original 1979 soundtrack sounds, the 2012 analogue remix to digital was a real blessing. And beware, there's very little tape hiss on the analogue recording as it was recorded with Telefunken's TelCom NR that gets a better S/N ratio out of tape than Dolby A.

 

I have to got hand it to the marketing department at Sony: That was a brilliant ad! LOL!

 

I hear that Mitsubishi made a great little pickup truck. According to Sony's advertisement for their DASH machine. Nashville chose Mitsubishi and the Pro-Digi format over Sony DASH, by the way. Jimmy Bowen of what was then MCA Records, the big proponent of digital recording in the day. Do remember, that tapes recorded on one of these two formats, could not be played or recorded on the other, they were incompatible with each other.

 

Interesting you bring up other NR systems -- I think you might be talking about Telcom C4. We are just restarting that project (it got delayed by about 1month.) We have a participant who has done all of the reverse engineering on C4, and I have a pretty good picture on how it works and what it does.

It actually does have good advantages over DolbyA, but oddly (really -- it is a bit of a mystery to me), that DolbyA is better at encode/decode cycles, and just might be better on transients. The C4 is better on SNR -- and they did some things REALLY correctly on C4 -- really good design, and smartly used a 1.5 compression/expansion ratio and not 2:1 (like DBX did, and in a very limited way -- DolbyA did in a degenerate way.) 1.5:1 actually fits the dynamic range better, and doesn't create as much demand on the recorder. We have also looked at the Burwen unit -- his unit -- not many were sold, was 3:1, which I believe to be too aggressive. The Burwen unit was an attempt to do a lot better -- tall order, very difficult to do better than DolbyA at the time. C4 (I think) could have been much better all around -- but IMO, SNR improvement is only part of the picture, but a very important part.

Other differences is that the C4 feedback is more parametric and not-so-much audio -- that is a good thing (very good thing) for more direct software emulation. C4 also has different attack/release for each band, but the attack/release scheme seems not to have the finesse that DolbyA has. The big evil thing that DolbyA did that C4 didnt -- the 3k-20+k and 9k-20+k overlap -- it makes decoding a little more problematical, even in HW. I *think* that I really like the C4 design. C4 does have a kind of overlap -- it seems to be pretty interesting -- I can't wait to actually listen to encoded material.

The big worry for me (the person -- fool -- writing the software) -- it looks like the higher frequency bands have very fast attack/release (I haven't read the most recent reverse engineering report yet -- I might be wrong) -- the super fast attack/release could cause some complications for the software emulation. I am IMD phobic, so that is going to be 'interesting' when I get the bare bones decoder working (hopefully in a few weeks.)

*One reason why C4 is going to be almost plug-in easy (well, not quite), is that my software infrastructure parameters are already set up for 2 channels of 4 bands. That means that all of the filters structures, signal paths, etc will be able to mostly be reused. I wrote the DA decoder so that it is not dependent on a lot of the symmetry in the DolbyA design (for example, the design isn't really dependent on the LF/MF bands and HF0/HF1 bands having the same attack/release characteristics.)

I dont own the reverse engineering for the C4 project, so I feel a little more reserved in discussing more of the deep details. Anyone really curious how to do DolbyA decoding -- I am pretty much willing to talk about a lot of the details. Some of the stuff is not for those who dislike DSP technqiues.)

John

 

When I need Dolby A, I don't use software emulations which purport to be. I use Real Dolby A in hardware, with a properly aligned, and set up tape machine. Same thing when I need Dolby B or dbx Type 1 or Type 2. These units are also like the tape machines used with them, maintained to standard. John's discussion is also off topic. Digital recording didn't use either. Analog tape machines did if tape was encoded and transferred to digital. Digital recording itself is not dependent on NR systems. Unless absolutely necessary, I use NO NR on analog tape machines. NR is a PITA.

 

Yes -- most earlier DolbyA emulations don't work quite 'right'. Of course, you should use what is good for you. The DHNRDS is less a DolbyA HW replacement, but more of a tool for archivists. Rather than 'just decode' DolbyA material at the level of the hardware, the DHNRDS uses very advanced technqiues to remove/mitigate the effects (damage) of decoding the material. DolbyA doesn't work in the nice encode/decode symmetry that is implied in the documentation and common-knowledge, but the decoding process is where the trouble is. So there is a lot more 'distortion free' material on the recording than what is normally accessed.
Until the DHNRDS type decoder is available and heard (and compared) -- it is incredibly difficult to explain the lossage of normal decoding processes (vocal chorus type things are the most astounding to me.) For day-to-day, the DolbyA HW is fine. The losses for normal DolbyA HW don't accumulate much after the first encode/decode. Once the first decode happens, the losses don't accumulate much.
I have been amazed about the lossage from DolbyA HW decoding when comparing the two. There are a few ways that the DHNRDS will surprise (much less grainy -- but with slightly higher NR), certain kinds of percussion sound SLIGHTLY different from the original -- probably cannot distinguish without careful A/B on headphones. That was actually a tradeoff, and singing chorus are infinitely more clean -- vocals are much more distinct, much more like the original.

The DHNRDS is NOT perfect -- but comes pretty close. There are a few ways that the DolbyA HW does better.
The first real realease is in the hands of my project partner right now -- going through a 2nd phase of testing/verification. It is NOT a loose approximation at all.
One thing for sure -- the DHNRDS will not replace the normal DolbyA workflow -- it cannot do it from a physical standpoint, but it can do the extraction of the most clean possible copy of the material (really.)

John

 

You are being very pessimistic -- probably from actual experience with poor emulations. If you knew how complex it is to properly emulate a DolbyA (pretty much within HW unit/version variations), then it is becomes EASY to accept why such emulations are so difficult and normally poorly done. Frankly, on normal orchestral type music and soft pop, it is probably difficult to tell the difference between DHNRDS and A HW. Also, the DHNRDS is actually a DA decoder. I am not legally as careful as I should be when talking technically -- I never do sales/marketing -- NEVER.) There are so many complex things going on in a REAL decoder that it almsot saturates a 4 core intel at 96k when running at full quality (the HW quality doesn't even have a chance when running in that mode) (no rate conversion -- everything dynamically adapts up to 96k, beyond that, perfect decimation/interpolation.) If at least one Intel Core cpu isn't significantly utilized, I doubt that a clean/accurate decoder can do its job. Mine does a lot more. In the lowest quality mode, it just about uses up one core -- recent improvements in the front-end detector required quite a bit of CPU -- the lowest quality mode used to use about 1/2 core for realtime at 96k.

* I am NEVER meaning to disparage R Dolby or Dolby in general. Anyone who REALLY understands the DolbyA design (I do), they'd see deeper genius than just a set of simple JFET feedback compressors. There is some REAL magic in that design. I have been amazed every time I have done a really deep review into economies of the DolbyA design. One really, really, really bad thing for HW implementation is the necessary component selection on the cat22 type design. Because of the variations in JFETs, and to a lesser extent semiconductor diodes, tedius and accuracy-challenged component selection is needed, and probably one reason for the variations in DolbyA behavior.

The most 'bad' thing about the DolbyA architectural concept is that the compressors in the feedback loop don't 'decode' as well as intuition might suggest. Think about this -- when using a straightforward gain stage, it has a certain response, but also a typically forgotten propagation delay. When that amplifier with delay is placed in a feedback loop, the results are NOT the same as a simple feedback formula might imply. Then, add the varying dynamics of each of the compressors -- the 'expander' emulation by placing compressors in the feedback loop are an approximation (it is a rougher than normally realized 'approximation'.) The DHNRDS does all of the feedback calculations EXPLICITLY rather than IMPLICITLY with errors, so that explict design opens up many possiblities for quality improvement. (True quality improvement, not tweaking or 'boosting'.) The reason why the explicit decoder design has not been done in the past: 1) Mathematically, it is exceedingly complex to undo the very highly dynamic compression 2) this exceedingly complex math requires huge amounts of CPU -- probably more CPU than existed in the US in the middle 1980s. Other than truly proprietary code in my places of employment, I have NEVER been as careful about source code security as for this project. The decoding technology is *special* and very likely first time that it has been accurately done.​

DBX is trivial, I don't know why you might have had bad experience with DBX emulations unless it was just a bad emulation. If done correctly -- the dbx detector is pretty easy (the key is -- dbx 'RMS' is not the same as EE-style 'RMS'), and the filters are not simple 1st or 2nd order.
B and C are coming up, but C4 is the main one being worked on right now. For C4, I wont' have to do the tedious reverse-engineering myself (acutally, got help from a recording pro -- a LOT of help for DA.) Also, DBX does DBX pretty well -- there are few problems with DBX other than the transient issues (endemic) or noise modulation (endemic.) I am NOT sure about C and B -- if there is any more quality possible from proper decoding. C4 seems promising for some improvement, and SR -- geesh!!! I can visualize all of the distortion sidebands swirling around -- but R Dolby was a very bright man, and he might have done some kind of magic cancelling that I dont see yet.

SR is hell to emulate, but I do have ideas for superior methods for decoding. Might take me a year or so longer of thinking to even start it. So, after doing C4, then stalling during development of dbx I/II and B/C type emulations -- then MAYBE I can really get my act together on SR. I do have EVERYTHING necessary to emulate SR, but that doesn't mean that I have yet figured out the most practical math.

If you don't want to use the DHNRDS -- that is cool. It is being 'sold' for nuisance fee only (I am NO WAY involved in that -- another party is doing that.) There is nothing but altruism in the work -- AT ALL.

My biggest motivations are the poorly decoded/poorly mastered material reaching the consumer -- we have NEVER been diluded into thinking that there is money to be made. My project partner has other
motivations also -- but all together the motivations are only positive.

I understand that the problem with poor technical quality reaching the public is NOT the mastering alone -- sometimes it is a set of constraints on the mastering. However, there is no excuse for the poor
decoding quality (or non-existant decoding) reaching the customers after the DHNRDS becomes fully available.

Of course, there is always the matter of choice, and choosing not to decode... That has actually happened.
For myself, I really enjoy hearing the material in the way it was meant to sound (unless it was originally done with NR distortions already considered -- some material DOES seem like that.)

Even for my own listening enjoyment -- I would NOT tolerate anything but near perfection (ask my project partner -- he has apparently been a little 'frustrated' by my excessive pickiness.)

John