44
« on: February 15, 2013, 05:44:57 PM »
I don't hear it as much today as I used to, but every once in a while someone brings up the question of self erasure of magnetic tapes.
Usually, these questions arise in general audio forums and folks who say that this degradation happens don't know anything about tape and use it as a reason why digital recordings are superior. (typical internet BS)
I usually defend tape by pointing out that any "remastered" releases of old recording are dubs of master tapes and of course our beloved Tape Project tapes couldn't be better evidence of just how good these 50 year+ master tapes sound in spite of repeated wear and tear over the years.
Well, Stuart Rohre from Applied Research Labs (associated with University of Texas Austin) has begun a long term experiment to determine what, if any, sound degradation occurs over long periods of time and also the results of repeated baking.
There have been many discussions of whether repeated baking and the way that sticky tapes are baked (length of time, temp etc) have a negative effect on tapes.
What follows are quotes from Mr. Rohre explaining the methods under which this experiment is being conducted. I think you can all agree that great pains have been taken and that this research will be a reference once the results are published.
"We have commenced after baking the test tape for two months. (It had sticky shed).
We recorded a series of test sine waves across and beyond band edge.
Tones were recorded at 7.5 IPS.
We also did a sweep over the band of the recorder to beyond band edge.
Each tone and each sweep was recorded for a minute, after ascertaining that we would have several rotations of the 10 1/2 inch reel at the speed used. The concern was to average out any once around effect from a bulk degaussing signature, if any. The tape was a new in 1991, Ampex type 797 instrumentation wideband tape, but used on an intermediate band Racal recorder, which duplicates our archive of data tapes. (They are overbiased 10 dB to give a robust recording with intermediate electronics on wideband tape.)
However, our test tape has Racal autocal 2 dB over bias, which might be easier to self erase. Thus our test is worst case, compared to our archive tapes re bias level.
We are most interested in any self erasure due to repeated baking or storage for varying periods of time.
The tape was recorded and then digitized. After 24 hours, it was digitized again today, and will be again tomorrow. The next test
will be to bake the tape for a period of time, and again digitize the tone tests. Another set of longer baking times is planned, and after each, a digitization. We will use computer statistical analysis to determine the degradation of any tone, if any. Particularly of interest to this group will be any losses at near band edge, representing the highest audio frequencies you might record.
To date, we had experienced no observable losses in our acoustic recordings but those do not extend to the extreme of band edge at the original speed of our data tapes. This test tape was designed to cover signals you could record at an original speed of 15/70 inch per second.
(Our recordings in the archive range from one week to one month per tape, or in the case of programmed recorders, one year in the Arctic.)
We used a spectrum analyzer to do a quick survey of the recorded tests, and individual tones were yielding the expected IRIG 3rd harmonic distortion of 40 dB down from the 1 volt RMS fundamental. No second harmonic was observed, indicating the heads were well degaussed.
The instrument used for this observation was a Hewlett Packard 3561A Real Time Spectrum Analyzer.
The signal source was a Rigol Arbitrary Waveform Generator, and monitoring was by a Rigol digital oscilloscope as well as bar graph metering of the Racal Storehorse tape recorder, where each of the channels has a peak level bar graph during recording and playback.
A passive multi-channel box of ARL construction, fed the separate channels from the one 50 ohm generator output.
For those unfamiliar, Racal manufactured the first fully auto calibrating recorder, the Storehorse, in England, but sold to IRIG labs at the research labs here, and White Sands and other missile ranges.
Rigol is a fairly new player in Instrumentation, but builds the entry level scopes for Tektronix in China. Their product quality seems high; as we have had our Rigols for 3 years and they are self calibrating, and very versatile. Results of set ups can be recorded on Thumb drives for repeat use. The scope is capable of computing a FFT as well as being a conventional oscilloscope. Digital tags are applied to signal levels and timing.
A pattern of even and odd channels was recorded with an unrecorded chanenl as a guard band around each active channel. (Racal has selective channel recording).
The HP RTA was also in the playback path for single channel monitoring,
and a selector switch on the Racal determined which channel at any one time was directed to the HP analyzer."
When asked if there are changes immediately after recording Mr. Rohre responded;
"We are only a Matlab program analysis away from knowing the answer to any changes from moment of recording until a short period of time later.
However, I should say that one could expect SOME change, merely from the fact that once you hit stop and rewind, the tape is going to move from the path it had forward and have an ever so slight path different in the reverse rewind and then again when it is put into playback. You always hope for perfect tracking but wear eventually causes slight path changes.
The other thing that you will see is a once around change due to a degaussing signature. Most bulk degaussers used to produce 'blank' tape
have either a noticeable or slight once around effect on amplitude, depending on the type of degausser used and its control of the declining magnetic field.
For the immediate playback, my long experience in producing in house test tapes has been that you would not notice any decline by usual observation which includes a spectrum analyzer and certainly a true RMS/ dB meter. This is the first time we have had something recorded and then digitized, and digitized again in a short period of time, with intent to process to the most detail of amplitude levels that we can.
Our main work is more than what amplitude are tones on a tape, but in the real world of acoustics what are those tones, what distortion products are apparent from the acoustic path from tone source to the recorder input, and what are the interactions of various tones due to long path vs. short paths, depth of the transducer vs. medium density, and numerous other specifics that make up sonar signal analysis. These days, our recording outputs are processed into 3D pictures, with color changes reflecting sound levels, and you see a lot more than an alteration of a simple sine wave, or even a complex musical combination of tones.
For this research test, we are going to concentrate on what was the level of signal recorded, and then compare that to the spectrum levels
on playbacks.
Thanks to all of this group who commented over the past years on possible loss of high frequencies re music recording. That stimulated our chief scientists to wonder about, and to want to quantify, any effects of self erasure. And to extend the work to study long term effects of baking tapes to recover from sticky shed.
Our own recordings typically don't push the high end of the band for the speed, but we do depend on recovering a timing mark placed near band edge, and
it has been known to vary in quality, but usually due to issues such as care in setting bias by our co workers from the sponsor lab, in the field. They have to deal with recorder check out and final alignment on the rolling deck of a small research ship, or in a plywood hut atop an ice island in the far Arctic. Or while flying at 400 mph in a turbo prop transport plane. Hardly lab quality spaces to do precision alignment of a recorder.
The human ear is an amazing diagnostic and analytical instrument. We use
a monitor speaker in our lab, as it is often the easiest way to identify
shipping signatures mixed into wind noise, or other noise of the recording location. But most science recordists don't have a "Golden Ear", and we have to rely on the mathematical precision of our measuring instruments. We hope to produce useful insights for several recording communities."
For more regarding testing procedures;
Good question about how you would tell what caused the change.
You average the play back signal over the whole time it was stable in the test recording. The tape is not stopped during the playback and digitizing each day of all the test tones and the sweep. Thus all signals have the same tracking error if any. Azimuth was adjusted to RLCO Azimuth standard tape, which is an IRIG test tape used by all the military test ranges.
The recorder used has much more precise guiding than consumer or music recorders of the older types often collected, (Ampex 300 and others). Speeds are servo controlled against a master clock oscillator.
The heads only have 51 hours on them since April of last year, too little time for multiple cut down tape paths to form. 797 is a less abrasive tape than the 795 widely used for NASA ranges and military GSA procured recorder users.
Using mathematical statistical analysis of the digitized signals, (which were digitized at 1,000,000 samples per second, 16 bits), we will come as close as we can to ground truth.
The playback tests will be repeated several times to average out any variations in tape tracking. Short term variations due to tape handling should come out in the averaging processes.
At the end of it all, the several computer scientists involved may only be able to tell me that any change was within the expected range of variation in any repeated process.
We should be able to measure much better than the level changes the ear can hear. That is the key take away, can we get any changes in multiple digital processing of the tape?
BTW, this tape piece was blank and had not been used on the section we recorded. It only had been baked, then rewound to equalize the winding tension and improve the tape packing, which came out factory smooth.
The digital steps after playback reduce the uncertainties to very minute amplitude uncertainties, as the A/D is good to 1/2 bit. BTW, the A/D system is a scientific system made by National Instruments, a company founded by members of the ARL staff many years ago and located across the freeway from our campus. The computer it is attached to is a dual Dell server system, also an Austin product. And if need be, we have access to the UT Supercomputer systems on this same campus, a few buildings away.
One of the many precautions we have to take is to disable bias and recording capability of the Racal for the duration of the playback testing. There are four staff members working this experiment, with a total experience with recording, of 168 years! We are also quite lucky to have access to a consultant, who has worked for four tape recorder companies and an two aerospace companies as well as being a long time recording consultant to the U. S. Navy. He was also an expert witness for one of the premier air crash investigation law firms on the East Coast, specializing in cockpit voice recorder analysis of such cases as TWA 800 and others. We think we have most bases covered for this testing."
"Further comment. ALL extent Ampex tapes of the 1970's to 90's, even new in the wrapper and carton, have been found to have sticky shed; thus you have to initially bake two samples of the same batch of tape, then proceed from there. Our first priority determined by our sponsor, is to evaluate multiple baking cycles. We will of course, also investigate a tape's self erasure due to being recorded and then sitting for various periods of time before being measured for any loss of replay levels compared to first play.
There really are two experiments to do.
Most interest is by force in what happens to tape archives made during the era that faulty tape formulas were sold. Of essential interest, is what happens if you bake and use (play) a tape, but for whatever reason, you want to gather more data from that tape at a later time?
The best example was our tasking some four years ago, to find out if tapes recorded in the 1980's and believed to only contain useful acoustic environmental data down to 100 Hz, could now be baked again, and played at higher speed and digitized with state of the art fast, A/D and inspected for useful signals down to 10 Hz. That effort was fully successful, with the calibration signals on the Master tape yielding the same levels as when the tape was recorded in 1980, as did a wide band noise signal."
Needless to say, I'll keep an eye on this research and post new entries as they become available.