Analog Tape Can Never Be HD: Here’s Why
First, let me state that I’m not against analog recording. I have been using analog tape machines for over 30 years and still have my NAGRA IV-S machine and QGB 10 1/2″ reel adapter in my studio (I just noticed a QGB went for over $5K on eBay…but I cannot part with my machine). However, I continue to read on other forums that analog recording is the “brass ring” or ultimate standard which all other recordings should be compared… including HD PCM digital.
Nothing could be further from the truth. Even some very well respected audio engineers insist that analog tape machines deliver a superior reproduction of music. I do agree that it produces its own sonic signature that is preferred by some engineers/producers (T. Bone Burnett among them) and consumers, but their preference couldn’t possibly be because the playback fidelity of the recording matches the acoustic reality of the original musical instruments/voices. Analog recording just cannot produce the same level of fidelity (dynamic range and frequency response) that a well-done HD PCM recording can.
Take a look at the printed specifications as listed in the NAGRA IV-S INSTRUCTION MANUAL: Note the frequency response curve. It clearly shows that at 15 ips (inches per second), the highest frequency that can be captured on a master tape begins to drop quickly after 20 kHz. This is for the master tape, the one used at the time of the original recording. Each analog copy to copy degrades the SNR by an additional 3 dB. We consumers never get to hear the original source tape. The normal production process was/is to record to an analog multichannel machine (which BTW has less than half the track width of my NAGRA and therefore 3 db less SNR, mix down to a 2-channel stereo machine and then finally to a master. This is then cloned using analog to analog methods (another 3 dB of SNR loss) and the vinyl parts cut. The point is that these numbers are best case examples.
NAGRA IV-S Frequency Response Graph at 15 ips
The next image is also from the manual and shows the dynamic range of the NAGRA IV-S. The very best that it can do isNAB 72 dB at 15 ips using NAGRAMASTER EQ. I’m written on this site that 12-bits of word length (using PCM coding) is enough to fully capture the dynamic range of a well done analog recording. Note that another 1.5 dB can be achieved by having wider tracks on the heads. If you use Dolby SR or another noise reduction system, the number can jump up substantially but it will never equal plain old 16-bit PCM digital and the amount of signal processing applied during Dolby SR encoding and decoding is ridiculously high. I used to do these alignments at the studio where I first started engineering records. Another interesting specification on this list is the crosstalk and phase fluctuations that occur between tracks on an analog tape machine. These are not issues with HD PCM digital encoding (not that PCM doesn’t have unique issues of its own…I will write about those as well).
NAGRA SNR Specifications
Part I of this article is largely taken from the NAGRA IV-S INSTRUCTION MANUAL. It is full of specifications and charts…some of which I will quote and reproduce here. I know many will dismiss the information, specifications and charts out of hand. They want nothing to do with specifications. Their usual defense goes something like “just listen and use your ears” or “analog is just warmer and more emotional”. However, the facts don’t lie. Of course, music recordings, music production and music reproduction are more than just specifications just as they are more than pure emotions and feeling. However, there are certain baseline facts that cannot be refuted. Imagine if someone was to say that a video image at 720 x 480 pixels running 15 frames per second was HD? There is value is establishing technical standards/specifications and I believe that the world of music recording and delivery is way overdue for some of the rigor that is applied to video.
Here’s the discussion of noise from the manual.
Section 5 of the Nagra instruction manual is entitled, “Noise added to a Signal by a Tape Recorder”. It describes the types of noise that are invariably as part of audio recording using analog tape. Below is the section in the manual:
Like all other elements of an electroacoustic chain, a tape recorder adds noise to the useful signal. Manufacturers of both equipment and tape are constantly preoccupied with the problem of reducing noise to the lowest audible level.
Classification
To distinguish between the different sources the noise are classified into the following groups:
Group B
Group C
Group A: Background Noise
In a high quality recorder, background noise is essentially due to the tape and the recorder manufacturer cannot do much about it, except by making it less annoying by raising the recording level (distortion compensating circuits of NAGRA IV-S) or by using recording standards that attenuate the noise (NAGRAMASTER standard).
Group B: Modulation Noise
Modulation noise is a specific shortcoming of the magnetic recording process and it can be caused in several ways, especially by tape vibration noise, amplitude modulation noise and head magnetization noise.
Tape Vibration Noise
The tape is not transported in a perfectly regular manner; its speed varies (wow, flutter and vibration) and the frequency of the variations may be quite high. The frequency of a sine signal of constant level is therefore modulated by speed variations. Wow and flutter are well known subjective experiences, whereas a high frequency vibration of the tape simply make the recording sound dirty. As the noise disappears with the signal, it is obviously a modulation noise.
The problem is very serious with equipment of simple design, having no mechanical filters or – worse – having pressure pads on the heads. The tape vibration noise in the NAGRA IV-S has been rendered negligible.
Amplitude Modulation Noise
Record a continuous sine signal and reproduce it. Its amplitude ought to be perfectly constant, but in reality it varies by a few percent, thus producing an amplitude modulation that is heard as a noise. The amplitude fluctuations are due to several reasons, as the irregularities of the tape, the tape edges and an imperfect magnetic layer:
Irregularities of the Tape
The magnetic layer of the tape is not perfectly homogeneous; its structure and thickness are not quite regular and this is an important source of the amplitude modulation noise.
Tape Edges
The poor condition of the tape edges can also cause amplitude modulation. It is for this reason that the width of the playback head is slightly less than that of the tape.
Asperity or Drop-out Noise
When some foreign substance, such as a dust particle or any heterogeneity of the magnetic layer of the tape moves across the recording head, the particle lifts the tape off the head. The amplitude of the useful signal drops and this is audible as a noise. That noise can be notably minimized by cleaning the tape, by increasing the specific pressure of the tape on the head and by over-biasing:
Cleaning the Tape
Experience has shown that drop-out noise is worse with certain European tapes with a matt back and it appears that particles from the back loosen themselves and contaminate the magnetic layer. Sound engineers have adopted the habit to scrape the tape in a very simple an efficient way using the QRAC (a tape cleaning blade) accessory before the tape reaches the heads when doing high quality recordings with new tape, this phenomenon occurring principally only when new tape is used. With certain tapes, this cleaning operation has been very pronounced effect in reducing the drop-out noise.
Increasing the Specific Pressure of the Tape on the Heads
While a tangible increase of the tape pressure increases head wear, it also presses foreign substances into the tape – due to its elasticity – and this reduces drop-out noise.
Over-biasing
When over-biasing the tape, we operate on that part of the bias/efficiency curve where a reduction of the bias causes the efficiency to increase. Lifting off the tape reduces the bias, and the increase efficiency compensates partially for the drop-out of the recording field. This method has been used for many years with NAGRA recorders operating in Europe, where the use of matt back tape is of long standing. With 3M tape No. 206, the improvement is not sufficient to justify the drawbacks in over-biasing, but nevertheless, the operation point is still on the falling slope of the bias/efficiency curve. Those recorders adjusted for 3M tape no. 138, and with which tape No. 206 is used, operate with under-bias, which means the drop-out noise becomes very audible.
Group C: Head Magnetization Noise
A recording head through which DC flows, will record a DC signal on the tape, and this signal cannot be played back, since conventional playback heads do not permit DC reproduction. This signal should be inaudible in any case. The DC signal is, however, a source of modulation noise, and the latter is perfectly audible.
DC is not even necessary for being the cause of this problem: an asymmetrical magnetic bias, a magnetic interference field, or permanent magnetization of the heads or of the shielding have the same effect. The earth’s magnetic field in particular is sufficient, unless the head is well-shielded.
In a recorder like the NAGRA IV-S – unless the oscillator has a breakdown – the bias symmetry is one order better than is necessary whilst the shielding of the recording head is “just good enough”, allowing for accessibility. The earth’s magnetic field, being essentially vertical in our latitudes, can cause an increase of the phenomenon when the NAGRA is turned into a vertical position.
Moreover, when operating in the RECORD, NO LIMITER position, transient signal may magnetize the recording head sufficiently to produce an audible noise. A sound of sufficient amplitude (saturation), decaying progressively, will fortunately demagnetize the recording head. The shielding, too, can be magnetized; a phenomenon particularly observed on recorders that have been shipped by air. It is therefore necessary to demagnetize them from time to time. Another effect of a magnetized head is noteworthy: It records a DC signal. As long as the tape is normally transported, only the modulation noise during playback is heard, but, when the tape is accelerated from stop to normal running speed, a “pop” can be heard. It marks the differential of the recording of the DC signal with respect to the accelerated time. This effect can be used to check whether the recording head is magnetized or not.
I find this information very informative and enlightening. I hope that readers will insist that digital download sites provide accurate information regarding the provenance of any track being listed for sale. If it came from an analog source, it will be limited in its dynamic range AND frequency response. An HD container will provide the best possible playback of the source tape (even safety copies etc), but it is NOT THE SAME AS A NEW HD RECORDING and doesn’t deserve to be identified as an HD track.
[NOTE: Since I wrote this piece, I’ve learned that I was incorrect about the 6 dB of signal degradation when transferring an analog tape to another analog tape…I have updated this article to reflect this new information]
These are some great facts…I’m sure there’s a lot of engineers that swear analog tape is HD that have NO IDEA about some of these specs.
Pingback: Mark Waldrep is claiming that PCM 24/96 is superior to DSD - Page 21
Pingback: Angry Metal-Fi: A Higher Resolution | Angry Metal Guy
I read the article and I think he pretty much got it. He mentions John Siau from Benchmark and his post on keeping everything in the chain at high-resolution levels.
Using the specifications for a battery powered portable tape recorder is suspect (at best). Not dogging the Nagra, I have one myself, but it does NOT represent the highest capabilities of analog tape machines.
At the zenith of analog tape machines frequency response could be had out to 40Khz or more (Sony APR-24)
Dynamic range of the last produced machines could reach 80dB unaided by noise reduction.
A late machine running with the mentioned Dolby SR has a published dynamic range of 110dB which is between 17 and 18 equivalent bit depth.
The article is misleading.
I think you would find very few engineers that have used Nagra recorders (the portables or the T machines) not include them in the “highest” level of analog tape machines. Certainly, the guys that I know who specialize in analog reel to reel do. And using DC power is a plus not a negative as far as I’m concerned.
Of course, there are lots of wonderful machines that have been developed and noise reduction systems that bring them to near digital specs. But the vast majority of masters made during the golden age of classical rock ‘n roll were done with Ampex machines without noise reduction…and those master don’t approach the level of high-resolution audio. That was the point of the article. Either you believe that old analog two-track masters are high-resolution or not. I don’t.
An 80dB unaided dynamic range can be achieved by a bit depth of 14 bits. If you add Dolby noise reduction to analog, you have to add in-band perceptual shaped dither to digital to keep it fair, in which case a bit depth of 16 bits can exceed 110dB of dynamic range. Also, if we agree that the upper limit of human hearing is around 20kHz, then CD resolution can match or exceed master tape performance for all intents and purposes.
The chart of frequency responses is the electronic response of the direct amplifier. It is not the response of the recorder including the tape. The high-frequency cutoff is intentional, to remove unrecordable frequencies that might cause distortion or beat against the bias frequency. The low-frequency cutoff is similarly intentional; indeed, a purist might think that the best recording might be obtained with the filter in “flat.” That is not true, as flat allows in frequencies that cannot be recorded on the tape, increasing distortion. The best wide-band recording is obtained in the “music” position, which attenuates unrecordable rumble.
Which simply reinforces my point that analog tape cannot and should not be considered high-resolution. High-res PCM can and does do a better job without the limitations of tape.
Digital, regardless of sampling rate, always represents a subset of an analog signal.
Can you explain how a digital source file of 96 kHz/24-bit fidelity could be considered a “subset” of an analog signal? Are comparing the real time analog signal or an analog recording? Digital may not sound the way you want but it the best and most accurate recording technology on the planet.
I think Mr Evans is failing to understand digital reconstruction during DAC reproduction and may be assuming that the analogue representation is merely a finite set of amplitude points defined by recorded bit-depth at the recorded sample rate, which it quite clearly isn’t. I can’t think of any other reason for his comment.
As you state so rightly in the main body of your text, measured flatness and width of frequency response along with measured distortion figures DO tell the whole story. There is no room for argument in the face of easily measurable and verifiable facts. If the goal is perceptual transparency, even the best analogue recorders fail.
…and an analog *recording* represents a much, much smaller subset of an incoming analog signal. This *minor* detail always gets completely overlooked by digiphobes.
By the way, over-biasing isn’t an issue if Dolby HX is used during the recording process. More details via email.
i think the graph you have included shows (since you are making it the central point of your argument) that tape performs far far better than standard digital reproduction. only 3dB down at 40 khz … but when we talk about ‘resolution’ it has little to do with either frequency response or dynamic range. if you are going ti argue about resolution and you want to use statistics rather than hearing tests … then you should use an appropriate metric.
I know there are arguments about how to quantify “resolution” and people that want to avoid discussion of frequency response (sample rate) and dynamic range (word length). I believe that these parameters are among the most important for defining fidelity — or potential fidelity. Subjective evaluation by hearing along are meaningless. Personal preferences are just that — personal and have no application in comparing two formats or recordings. I acknowledge that analog tape can capture frequencies above the audio band but in the aggregate, analog tape falls far short of high-resolution PCM audio in both fidelity and accuracy. Resolution is an appropriate metric IMHO and sample rate and word length — among other parameters — help define it.
I would argue PRECISELY the opposite – that ‘subjective evaluation’ is the ultimate standard we can use. I mean – are we not recording in the first place to please PEOPLE – rather than test equipment?? It also seems to me questionable to use S/N ratios and frequency response charts to make such an argument – i don’t think there’s any correspondence between either of these factors and how ‘good’ a recording is .. certainly those are secondary aspects of subjective quality but god knows most performances in the history of recorded music easily fit within a 67dbm or so range – they kind of have to! Although yes – the machines tell us that ‘digital’ reproduction has superior S/N characteristics -so if one were go argue FOR digital and AGAINST analogue – then of COURSE one would trot out the S/N specs … this is the nature of rhetoric.
Thanks for you reply. There’s certainly no absolutes when it comes to personal preference and the subjective approach to listening. It you or others prefer a particular format of recording, then who am I or anyone to argue. But objective realities regarding fidelity (which means accurately reproducing an original) do matter and help define the relative measure of a recorded experience. I much prefer a recording format that requires far fewer sonic compromises regardless of whether most performances fit within an older standard resolution format. It is simply a fact that high quality PCM digital audio has much greater potential fidelity than any analog format ever devised. Analog can deliver great listening but it will never be high-resolution.
Great Article. It always worries me when people are so defensive in the face of empirical scientific fact – especially in these troubled times. I love analogue tape too and worked for many years as a location recordist with Nagras and am still professionally involved in audio. I even love the slight click and pops of vinyl and optical sound and grain on film etc. But our senses are a most unreliable metric by which we can measure objectively. This is why professional sound equipment and measuring/metering/monitoring has to be reference accurate – because our senses lie. It is ironic and perhaps inconceivable that the very heroes of music recording, that people cite, from bygone days would not have jumped at the chance to record digitally, if they were still here. For them these ‘qualities’ were often troublesome and unwanted. There are even respected engineers recording classical music on analogue tape and insisting it’s better. This wasn’t the opinion of Decca, Denon and Dr. Stockham, nor the classical artists, in the late 70s when digital recording began. Equally at the other extreme the claims of DSD and high sample rates leave me baffled. Doubled blind testing agrees time and time again too. The single biggest development in recent years IMO has been the often misunderstood 32bit float, for production and post but not release and only because it allows a huge margin of error in capture and processing.
Thanks Steve. I just had my Nagra IV-S checked out by Glen Trew of Trew Audio in Nashville. He commented that it’s one of the best he’s seen…great specs etc. It has not been used a great deal and certainly for the past 20 years. But I cherish it and recognize its place in audio history (along with the QGB!). Cheers!
The point is that you want the best recording you can make – based on your subjective evaluation, “to please people” – to be degraded by the equipment and medium as little as possible. That’s what the measurements are all about.
Well there’s a reason Analog is much preferred and in terms of recorded music, has produced the highest sales, compared to Digital. But some of what you are saying is not really true, either.
I own a Fostex E-2 / E22 (have both 1/4 inch and half inch head blocks / guides) and can easily record up to 70 Khz in 30 IPS mode on Maxell UD tape or ATR. My Otari MTR-10 also does this. The Ampex ATR series was also good at this. The AMPEX 1100 I used when I was 11 years old went up to 25K before rolling off, at 7 1/2 IPS, and also down to 15 HZ on the low end, which was excellent for a 1/4 track machine.
Many analog master recordings are 30 IPS not 15. Almost anything that wasn’t classical, or hobby label, was 30 IPS in the 1950s and 1960’s.
However at 15 IPS, I could get to 40 Khz using a 4 track Tascam 34. Using DBX type 1 model 150 processors, properly set up (transparent), you should be able to get a signal to noise ratio of 120 DB. It works better on lower noise recorders. Using that on a Fostex E-2, on 1/4 inch tape I was at a 127 DB signal to noise ration.
While I admit there is some loss using companding, that loss is audibly less than that of the non-linear staircase of PCM digital At -90 DB, analog doesn’t get fuzzy or grainy.
The main advantages of analog recording are:
1. Dynamic range. With Digital recording, you are always fighting digital zero and digital max. With the exception of my invention (not disclosed) when you record drums and acoustic guitar on digital you are very likely to loose peaks and position points. You spend alot of time adjusting and re-adjusting levels to keep the peaks from passing the highest number. This results in poor sound quality because these levels can’t be increased without clipping. Try to equalize and you get worse problems. To mix in digital, you have to record everything else at a lower level. The result of these problems is the engineers dump everything through compression, resulting in a bad mix, and a loud muddy sound that isn’t pleasing to the ear. And hundreds of thousands of great albums that flop because of bad sound.
With analog recording, there is a noise floor and a point of distortion. However, analog recording, especially using a companding system like DBX allows you to record everything Wide tape or high speed tape (30ips) captures all those peaks. You spend much less time. You can also adjust levels with very little loss, maybe a little noise. You can equalize and you have much more range, where in digital, you clip.
The speed of sound in air is approximately 343 meters per second. When recording stereo, this is where digital looses the most content The 3D image is lost because points in air space are forced to the position of samples. This means that you loose the position of objects in air space, they are rounded to the nearest sample.
A good analogy to describe this is the difference between viewing a movie on film versus a movie on 540 line TV.
Your comments are well taken but your conclusions are incorrect. There can certainly be rare analog systems tweaked to produce results near a high-resolution PCM digital recording but the norm during the era of analog mixing and mastering was 15 ips 2-track without noise reduction. A much better analogy would be a movie on film vs a movie at 4K digital…there’s no comparison.