Early analogue recordings & an evaluation of analogue technology ...

[A.S.]

... A question: I recall that a small number of recordings available on vinyl were recorded "direct to disc" - i.e. straight to the cutting head, bypassing tape entirely. As an analog medium, does this potentially offer greater fidelity to the original than magnetic tape with all of its inherent issues, including print-through?

Thanks for the feedback. I frequently suspect that I'm talking to myself. Not a problem in itself because the value of this material can only appreciate with time and long after I'm pushing up the daisies!

Yes I do recall the direct-to-disc (or more accurately, direct to disc stamping master) and the (then) fantastic fidelity of Thelma Houston's Pressure Cooker from Sheffield lab which I still have. It's hard to appreciate just how much care and attention was needed from the musicians to the mastering engineer to squeeze just a fraction more performance from the analogue recording chain. But even Sheffied Lab embraced digital when it arrived, which tells its own story I guess.

The answer to your question (about tape print through) is tantalisingly within your own grasp. Can you ascribe a name to the mechanical equivalent of tape print through in the medium of the gramophone record? It's fundamentally the same problem: information from one rotation potentially bleeding through and corrupting adjacent ones.

Im not sure if I made it absolutely clear in my demonstration in post #18 why this print-through is such a crisis (amongst a list of others were going to look at) for analogue tape. You may have the impression that the print-through only effects the gaps between the words (or music) as I demonstrated. That's not at all true. The print-through is audibly revealed when the wanted signal drops in loudness - i.e. the wanted signal ceases to mask the pre-echo (in my example). In fact, the amount of pre or post echo is constant throughout and 'underneath' the wanted audio, We can't hear it because it is masked, but it's definitely there, as a fidelity-robbing continuous drone. It, like other technical characteristics of the analogue system 'softens' the wanted audio just as viewing an oil painting under subdued rather than natural light hides belmishes. But it is takes us away from a faithful representation of the sound waves that hit the microphones.

We've hardly started on the long list of Analogue Anomolies which were sidestepped at a stroke when digital arrived. There are so many techical oddities of analogue to demonstrate I just don't know which to turn to next.

[Euler]

Thanks for the feedback. I frequently suspect that I'm talking to myself. Not a problem in itself because the value of this material can only appreciate with time and long after I'm pushing up the daisies!

Alan, you should know that many of us read your posts with deep appreciation for the time you put into them and the manifold insights they contain. At the same time, we may not reply, because we don't know enough to make an intelligent contribution. Every morning I check for new posts in the HUG, and I'm especially delighted when I see your initials A.S. So never think that you are talking to yourself, but rather that you have a very full class of fascinated students, only a few of whom are raising their hands to speak.

This is, in any case, how I choose to interpret any lack of interaction in my own classes!

Bruce

[A.S.]

Kind of you to say so. I'm not seeking parise and glory but it would be nice to know occasionally whether what I'm saying is appropriate, logical an above all else delivered at a pace which leaves a lasting impression with the reader/viewer. In particular, by far the quickest way for me to 'knock-up' a "tutorial" (big quotes) is to use the new video screen-cam software. It has the one big advantage to me that recording time is strictly limited to five minutes with a very small countdown timer in my screen that's easy to miss. I've been down to the last few seconds of recording time and have had to rush a bit to avoid a lock-up and need to re-record the lot. Five minutes duration means I have to get the point across at a pace which leaves no time to go into much detail - that may or may not be a plus. I aim to do a half-take just to check recording levels etc. and then to proceed with an unscripted one-take. In fact, even if it were rehearsed, with only five minutes available you can't take your eye off the screen to read a script, and there is no editing facility: what you see is in effect delayed live.

Ok, you prompted me to plan the next video and again that's on analogue tape's limitations. I've spent an hour experimenting with how best to convey this and the technical results are worse than I feared. At a stroke it explains why some listeners prefer the so-called 'easy' sound of analogue. But as we'll see, it's just another illusion. A pleasant one some say, but I don't think that you can argue with the audio analyser's display. What's interesting is that if you listen to the off-tape sound for a while, and then switch over to the perfect source, it is a bit of a shock to the ears. The digitally perfect source signal sounds rather cold, hard and with cleanly etched 'edges'. The off-tape analogue interpretation of the source has the edges sandpapered smooth and demands less attention.

I need to order some simple parts to be able to adjust the levels going into the tape recorder for a level-perfect comparison with what comes out, off tape. I'll do that over the next few days.

[EricW]

I agree with Bruce (Euler). Not seeking to stroke or flatter, but I genuinely think I've learned more about audio from the HUG, particularly Alan's posts, than I have from any other source. I truly appreciate the time and effort Alan puts in.

Above all else, I think I've developed a much better appreciation of the difference between facts and marketing.

[Pluto]

...the long list of Analogue Anomalies which were sidestepped at a stroke when digital arrived.

Print through....the fault that dare not speak its name. The final ten years or so of analogue dominance was a game of roundabouts and swings. When recording, it was always level and noise versus distortion, but the unspoken enemy that always showed itself when a big dynamic change came along was print-through.

Paradoxically, print-through became more of a problem with the arrival of a new generation of lower noise tapes in the seventies because the reduced noise floor showed up print-through that would previously have been masked by tape noise. The transfer function of print through was linear with changing level, that is to say a 10dB increase in recorded level would result in a 10dB increase in print-through although not all frequencies 'print' equally. Sadly, the worst frequency for print-through was typically smack in the middle of the human voice range.

This is the principal reason that tapes in long term storage are kept tail out. Print-through is worse in outer layers than inner layers because, to be audible, the printing source signal must reach the top of the oxide on the adjacent layer of tape. To reach the top of the oxide in an outer layer, the printing signal need only pass through the backing material; to reach the top of the oxide in an inner layer, it must pass through both the backing material and the complete oxide thickness. Because pre-print is subjectively worse than post-print, tapes are stored tail-out to ensure that pre-print is minimized as much as possible. This practice is also beneficial because a tape can be played and then stored without further fast spooling - most machines stack the tape far less tidily when spooling at high speed than when playing at 15ips.

It is sometimes found that spooling a tape through a slightly magnetized head can reduce print through (as print through is a phenomenon that resides at the top of the oxide layer) but this is obviously an extremely risky process.

Such is the vulnerability of analogue tape that you need to assume that every pass will be the very last opportunity to capture the content of that tape - you may well feed a roll of decent-looking tape into the head block, only to find a pile of iron-dust emerging from the other side.

[kittykat]

I have a few direct to disc & direct cut LPs & they mostly sound more dynamic, transparent & detailed than ordinary LPs.

The musicians would be have to be mentally dynamic as well, as any error on their part cannot be corrected. It would mean a retake and tossing out the metal. :-)

[A.S.]

Well, I have nothing else to do today but take it easy - the perfect excuse to make Analogue Anomalies Part 2. I realised after a couple of 'takes' that I just couldn't setup the test, explain it and get to the analysis in one rigidly fixed (by the program) 5 mins. so please watch this and then holding the final scene in your mind, watch Part 3 (not recorded yet).

I sincerely hope that you can follow my line of attack: you can appreciate that the objective was to generate a perfect* digital multi-tone signal that represents the first five notes of an octave and then feed it into what was considered to be a very good analogue reel-to-reel tape recorder, even now. What we are interested in is the off tape signal in comparison with the perfect digital signal we delivered to the tape recorder's input sockets. For clarity I just used one channel. The record level was marked on the Revox's VU meters as 50%.

The tape is 'BBC Type 200' standard play. I aligned the Revox some years ago - in fact, this is the very machine that I wrote up in my HiFi News article many years ago (to follow).

*What is a perfect digital signal? I'd say that any signal where the random noise floor is 100dB or more below the maximum signal possible is for all intents and purposes, better than our own ears hence perfect.

Here is >>> Analogue Anomalies Part 2 <<<

Please confirm if I have presented this in an easy to follow way. The attached chart shows the relationship between the five tones I generated.

>

P.S. In experimenting with Part 3 I've devised a cunning way of showing you how much (or how little) dynamic range performance a good audio system requires. And it is a shock to me and puts the whole pursuit of sonic (digital) excellence into perspective.

[Don Leman]

Thanks for taking the time and effort to do this Alan. I too have a free day. I was listening to the Messiah and reading the new posts. I paused Mr Handel to watch Part 2 and enjoyed seeing your B77 in action. These old decks still are fascinating to watch. I have a couple of A77's but they sit gathering dust as most of the pre-recorded open reel tapes purchased during the 70's suffer "sticky shed" problems.

Now back to "his yoke is easy and his burden is light".

[A.S.]

Part 3 displays the spectral content of the signal going into the tape recorder compared with the signal recorded onto tape and played-out by the recorder. In addition, I show that the noise floor of the Revox even using good quality tape gives it a dynamic range between the loudest signal it can record and the all pervasive tape hiss (when the recording VU needle is registering 50%) is only about 40dB in the proximity of the tones we are recording.

Hiss or noise is the enemy of all and every hi-fi system. Hiss or noise of any type obliterates quiet signals that are at or below the threshold of that noise. That is why a truly high fidelity system must have an adequately silent noise floor. This is an uncomfortable truth. We'll show exactly why that is in Part 4.

Here is >>> Analogue Anomalies Part 3 <<<

[A.S.]

... Hiss or noise is the enemy of all and every hi-fi system. Hiss or noise of any type obliterates quiet signals that are at or below the threshold of that noise. That is why a truly high fidelity system must have an adequately silent noise floor. This is an uncomfortable truth. We'll show exactly why that is in Part 4.

Hopefully you have been following my demonstrations thus far. In this final part, I have generated two tones (500Hz and 650Hz, midrange tones very typical of the spectrum of western music) but not as close in frequency as the previous five tones which represented the first five notes of an octave. I've injected those two tones into the the tape recorder. We compare the signal sent to the recorder with the off-tape signal.

Again, this shows conclusively that any and every microtone that has a loudness below that of the background hiss just cannot be reproduced - ever. Once a musical element is covered by hiss anywhere along the reproduction chain it is impossible to recover it. Random noise cannot ever be converted into musical tones. Once low-level transient musical tones have been swamped by hiss in the microphones, mixing desk, master tape recorder, DAC, sound card, LP, cartridge, CD transport, amplifier or even in the speaker cones themselves, they're lost forever.

Here is >>> Analogue Anomalies Part 4 <<<

Aaaaghhh! I thought I'd finished this demo but out of curiosity I thought I'd play some music for myself ..... more later!

[A.S.]

What about music indeed.

Just before I packed away the Revox for another few years I though I'd play some music and compare on and off tape. The very first music file I had to hand had a nice strong high frequency component, so that would do very nicely.

You'll recall that I demonstrated the absolute curse of unwanted background noise. It covers-over all the fine tonal detail with a random spluttering and once masked by the noise, the fine tonal detail is irrecoverable. Because of the quirks of analogue (tape) and our hearing, the effect of the very poor frequency resolution is often masked in the middle frequencies, but as frequency increases, numerous additional limitations of analogue (tape) reveal themselves.

By now you should appreciate that the vertical scale in the frequency spectral plot is signal loudness, in (arbitrary) dBs. We can slide the signal up and down that vertical scale just to give us a nice presentation. All that matters is the relative strength of frequency bands relative to others.

If you listen carefully to this excerpt you can hear and see that the high-hat is not covered by the music and can sing out very naturally. The spectral display reveals that it has much energy above 10kHz, specifically above about 15kHz and extending right to the edge of the horizontal display - over 22kHz. We then record that high quality source onto tape and compare the on/off results. Whilst the frequency response at the top is quite similar in extension, note how the analogue off-tape signal does not ever drop down in level below about the arbitrary -84dB above about 15kHz. But the digital source drops to around -96dB. That means there is about 12dB more dynamic range in the high-hat reproduction than the analogue tape can reproduce. The problem was reveals in the previous episodes: tape hiss masking the quietest cymbal microtones. And this is revealed a slight softening of sound, which as we all know, some audio enthusiasts like.

Here is >>> Analogue Anomalies Part 5 <<<

Complete digital source and off-tape music tracks as follows both as moderate-rate MP3s. You decide which is which after considering the points covered in this thread about dynamic range, tape hiss, frequency response and resolution.

Loading the player ...

Loading the player ...

Conclusion thus far: if you want to achieve more resolution from your high-fidelity replay system, the simplest, cheapest and easiest way to hear more of the microtones in music is to chose a replay system than has the lowest background noise be it noise in the medium (tapes, discs etc.) or in the driving equipment. But there is a practical limit of perfection. A continuation of my experiments here show that providing the dynamic range between the loudest sound reproducible and the background hiss is about 80dB or so no further improvement is audible. So chasing equipment with a noise floor at -100 or -120dB is pointless. So the real issue then is pushing down the noise floor between what we have seen is around -50dB in analogue tape (proximity of close tones) and the edge of audibility at around -80dB. That's really very difficult indeed for any analogue system to achieve.

[Pluto]

Well, I'll wager that the first of the two clips (A) was the off-tape sample - there was no significant loss of "sparkle", but the ride cymbal had slightly less attack than was evident in the second clip (B). There was also, surprisingly, a subtle difference in the sound of the brush work on the snare.

A point of significance is how good a job the old lady (B77 analogue tape recorder) made of this recording after several years on ice and the obvious problems that showed up with the earlier multi-tone test.

Having watched the Part 5 video, it is interesting how the direct recording shows an obvious steep low pass filter at about 15kHz (probably an artefact of the MP3 conversion) which was not so much in evidence with the off-tape sample. It's as though the MP3 engine has been "fooled" into thinking that it was all done by about 12kHz, only to "find" that there really was more up top that had to be acknowledged!

Now if I had been asked to choose which of these two I might prefer from looking at the spectra alone, I'd probably have opted for the off-tape version without the cliff-edge dip at 15kHz.

[Woodsail]

One of my favourite pieces of music is the Beecham recording of Peer Gynt made in stereo by EMI in 1957 - the year I was born. I had an old reel-to-reel tape back in the 60s, and now I have the CD. It still sounds fresh - and better than any other version I have heard.

Back on an earlier comment about MP3 taking us backward - I personally am not too concerned - lossey music is just a short term by-product of storage constraints. We have already seen the default compression rate rise from 128 to 256 on iTunes and to 320kb/s on Amazon. Vendors are bringing more sophisticated rip and store devices to market that capture CD quality. And we are starting to see more High Definition music become available.

[A.S.]

Well, I'll wager that the first of the two clips (A) was the off-tape sample - there was no significant loss of "sparkle", but the ride cymbal had slightly less attack than was evident in the second clip (B). There was also, surprisingly, a subtle difference in the sound of the brush work on the snare.....

Actually Pluto, I just realised that I made a mistake presenting these examples and need to re-do them.

You will recall that we were using the 'Monitor' switch on the front of the B77 to listen to the signal going into the recorder or in the other position, the signal that had just been recorded onto tape. The mistake I made was this: when the switch is in the 'input' position, we are not monitoring the signal directly at the input sockets on the recorder, we are monitoring the signal after it has passed through some amplification inside the Revox. I did know this because if you look carefully on any of the videos you will see that there are second/third/fourth etc. harmonic tones present in the 'input' position which are not there on the source but I forgot to consider this when presenting the music examples. Harmonic distortion is evidence that some amplification (and not very good amplification) is in the signal path. For the purpose of all the demonstrations I made, this is irrelevant because the dominant problems are from the tape itself, not the analogue electronics inside the Revox.

But when we come to compare the first generation 'source' recording with the off tape, we must directly compare the source with the off tape and not rely on the Monitor switch on the Revox.

So here we are again, this time with the first generation digital sound and the off tape. This is a true comparison of the effect of passing a digital recording through an analogue tape recorder. I have cut-off the lineup tone used to equalise the levels.

Loading the player ...

Version C

Loading the player ...

Version D

[A.S.]

It's interesting how one idea leads to another.

My guess would be that those clever folks who designed the MP3 coding system thoroughly understood the characteristic of analogue media. They would, for sure, have known how quite tones are masked by louder ones. My guess is that their starting point would be to devise a data-reduction process which mimics the noise masking process that we have seen applies to analogue tape.

If that is so, it should be easy to see noise profiling in the MP3 encoding system. More information to follow after some experiments (in hand).

[Pluto]

Actually Pluto, I just realised that I made a mistake presenting these examples and need to re-do them.

OK, this time item C is off-tape - the difference is rather more obvious this time (which makes it easier, assuming I was correct last time!!).

OTOH it doesn't show the Revox electronics in a very good light; either the Old Lady really does need a full electronic overhaul (probably duff capacitors after all this time) or the sound quality of cheap electronics really has progressed after thirty years. Maybe both are true.

I get rather annoyed with those who allege that IC op-amps, in general, 'degrade' the sound* - the concept of demonstrable transparency doesn't mesh easily with the audiophile creed.

* the point being that the Revox A/B77 electronic design precedes the widespread use of op-amps because, at that time, they were NOT good enough. These days, entirely the reverse is true.

[Pluto]

Another thing springs to mind in the light of your revelation about the switching: are you sure about the Revox response plot shown briefly in the video? While I didn't notice the graph scaling, I remember thinking that it was either extraordinarily good or not actually off tape! What tape speed was running?

[A.S.]

It's interesting how one idea leads to another. My guess would be that those clever folks who designed the MP3 coding system thoroughly understood the characteristic of analogue media. They would, for sure, have known how quite tones are masked by louder ones. My guess is that their starting point would be to devise a data-reduction process which mimics the noise masking process that we have seen applies to analogue tape...

And indeed, if we make a few simple experiments we can see that is exactly the line of thinking that led from analogue to pure digital and then to analogue-like (MP3) data compressed audio. It's so blindingly obvious that I don't know why it hadn't occurred to me before.

Fortunately I have already covered the basics of how the noise floor of analogue tape (unlike digital) is elevated in the proximity of recorded tones in the previous Analogue Anomalies videos a few posts back. This noise floor variability has something to do with the fact that in analogue tape, the sound is actually converted to magnetism, and stored in tiny bar magnets glued to the plastic tape. Those particle magnets do not exist in isolation (they are heaped up on one another) and hence, unlike a digital 'bit' which is uniquely addressable and turned off or on, they will be influenced by the general magnetic field in their proximity. And if there is a loud signal (therefore a strong magnetic field) that field will bleed into adjacent magnetic particles of the tape and will certainly swamp (or even actually erase) weaker signals. We've seen how the noise floor ramps-up in the vicinity of tones. We have also looked at how there tiny magnets have sufficient stray field that they can print-through onto adjacent layers (see a few videos back) so they are quite strong.

Well, guess what. Armed with what we've discovered about analogue tape we can now use the same method to evidence that MP3 uses fundamentally the same 'encoding' process as analogue tape. In the absence of any tones or music, that is, recording silence, MP3 as is typical of any good digital system, and has a very low background noise level. In fact, when recording silence we'll see that the MP3 system has the exceptionally low background hiss level of the best pure-digital systems. But the really cunning thing about MP3 we will see is that the MP3 coder actually generates than cleverly manipulates a background shaped noise (i.e. frequency specific hiss) just as I did to show how the profile of analogue tapes noise could be added to the pure digital signal (a few videos back). The MP3 codec ingeniously actually introduces noise where there was no noise before. And we've seen in the previous video that no signal can be retrieved from under the hiss. Turning that around, we can say that if there is hiss of a certain quantity (loudness, level) there is absolutely no need for the MP3 encoder to even attempt to encode those tones that would be inaudible under the hiss. So, unlike the analogue tape recorder where all signals regardless of how loud or quiet, regardless of what frequency band they are in, all are presented to the tape record head (and only some are detectable off-tape as we've seen), the MP3 encoder actually makes second-by-second decisions based on the audio content which signals are louder than the hiss and worth encoding, and which ones to discard as swamped by the hiss. Very cunning indeed. Greatly reduces the data rate.

I'll make a video of this later.

UPDATE 1: After some experimentation, it's clear that the MP3 coder is far smarter than just a crude emulation of the characteristics of analogue recording tape. It has evident intelligence. It certainly does analyse the signal to be encoded and it seems to have, as I suggested, defined code/don't code loudness thresholds. The way it behaves if a signal is just a little above the threshold is very different to the behaviour just below the threshold, in the 'don't bother attempting to encode' area. But surprisingly it does introduce significant amounts of harmonic distortion - that is, the coder generates (2nd) harmonics that are completely absent in the source audio signal as indeed analogue tape does. Is this a deliberate strategy to 'warm-up' the sound?

[A.S.]

I specially modded this B77 years ago to be damned nearly flat. 15 ips. I'll find the HFN article I wrote on it.

Note: This article is not on our server. That means it has not been scanned. I will find the original HFN and scan it asap.

[A.S.]

Another thing springs to mind in the light of your revelation about the switching: are you sure about the Revox response plot shown briefly in the video? While I didn't notice the graph scaling, I remember thinking that it was either extraordinarily good or not actually off tape! What tape speed was running?

Actually Pluto - and I'm a bit surprised your golden-olden ears didn't detect this - the burbling in the lineup tone gave the game away as to which recording was, beyond a shadow of doubt, off tape!