Is it possible to determine whether a phone call is local or long distance by analyzing the audio?

Did they have anything to do with Tandberg ? They used that scheme and I worked on one recently, they seem to allude to owning quite a bit of credt for the invention.

Or did they license it like Dolby ?

I had an Akai reel to reel from the late '60s, that I bought used in '71 or so from a friend who brought it back from Japan. It was a decent deck but it couldn't be considered "audiophile" class. It had the cross-field heads.

I don't recall who owned it and don't want to burn the time doing a patent search. Well, maybe a quick search. This looks like the original patent, files in 1961 and published in 1967: The owner "Iit Res Inst" is the IIT Research Institute, Chicago Ill. The references at the bottom do not include either Akai or Tandberg, which suggests that they did not do further research or obtain patents on the technology.

I also found this clip from 1965: which proclaims "This is the Crossfield head - only Akai has it". My guess(tm) is that it's not quite true.

This claims that Akai licensed it from Tandberg: Akai debuted the M-7 model tape recorder in 1962 and its successor the M-8 in 1964. Equipped with the then new X'Field magnetic recording system (licensed to Akai by Tandberg), these two models gave superior frequency response which resulted in a vast improvement in tape economy. "Tape economy"? Huh?

The TB 3 Stereo from 1957 was Tandberg's first stereo system. In the 1960s Tandberg introduced the cross-field recording technique in the TB-6X model, allowing their recorders to handle higher frequencies than competing models. Tandberg licensed the concept to Akai, who used it widely in the 1970s and 80s in their Akai and Roberts recorders.

So, I guess Tandberg was the first. So much for a "quick" search....

Yes, probably. It might also be that cross field heads allows the use of 0.5 mil tape instead of the usual 1.0 or 1.5 mil tape. Or maybe guaranteed to break acetate tapes instead of mylar. Cassette tapes, dictating machines and logging recorders were about the only machines I recall that ran 1-7/8 ips and 15/16 ips most of which came later.

On a sunny day (Sat, 14 Mar 2015 18:50:48 -0700) it happened Jeff Liebermann wrote in :

They probably mean they could use lower tape speed for the same bandwidth, resulting is less tape to buy. But I doubt that claim. IIRC the head gap is the main issue in playback, in recording it does not even count so much. ???

BTW Akai was the name the thing sold under in Europe.

Roberts was the importing company that imported the first Akai tape machines. They had an exclusive contract with them, and they were private branded for them. The head of the company saw them on a buying trip to Japan, and liked what he saw.

I was just looking to see if you (or anyone) knew off the top of their head, it is not that important. I know your time is valuable.

But I do appreciate the info. Thanks.

You know some weird shit.

even >count so much. "

There is a limit. The optimum record head gap might be 1.5 or so than the o ptimum playback head gap. Actually I think it depends on tape speed somewha t but that does not easily compute. But it does because the optimum tape he ad gaps would be different for different speeds actually if you bring scien ce into this. Unless I missed something there is what cold be considred a w

were so critical in just about everything.

I did some reel to reel ripping recently an dI was surprised at this mid en d Akai having so much dynamic range. And I mena on tapes made back in like

But they did not even leave residue on the heads !

Nice f***ed up ZERO post attributes there, idiot.

Will you EVER learn how to post to Usenet properly?

I see two in a row here just as stupid.

IME, print-through on .5 mil tapes makes them unusable. Even 1.0 mil tapes printed through after a couple of years. Every one.

snip

I have a BIC cassette player/recorder (in mint condition, sans belts, etc. it all still currently works though. That will do double tape speed, and gets 18k response at that.

No appreciable hiss increase with chrome tapes.

The one thing you're good at, AlwaysWrong, is irrelevance.

The issue at hand is reel to reel recorders and the Akai with its cross field heads, in particular (in this branch of the thread, anyway). Neither your Bic Click nor chrome tapes have any relevance at all.

Interesting stuff!

I assume that the digitized versions of these recordings (which is what I assume everyone is talking about analyzing here) were done at sample rates well below the 200-300 kHz needed to capture the bias signal directly. Then you'd be looking for the alias of the bias signal, at some lower in-band frequency... true?

So if the bias is normally -60 dB on the tape itself, how much of it gets through the anti-alias filter when converting to digital? Or do they digitize with special setups that use *no* (or minimal) anti-aliasing specifically to allow these sorts of analyses?

Best regards,

Bob Masta DAQARTA v7.60 Data AcQuisition And Real-Time Analysis

Yes, that's one method that barely works. The 2nd harmonic of the common 44 KHz audio sampling frequency would bring the bias signal into the audible region at 12 KHz. 48 KHz would be better with a 4 KHz mix. The problem is that the sampling frequency is usually quite symmetrical, and therefore lacks sufficient 2nd harmonic energy to do much mixing. 4 and 12 KHz are also in the audio region, making it difficult to see under that recorded audio. Originating from a computah clock, which might be dithered (spread spectrum clock) to reduce emissions, the result is a wide, dirty, low level, and useless mix. The 3rd harmonic of 44 KHz would be at a higher level, but it mixes to 32 KHz, which might be better, but is out of the range of my equipment. Might be worth a try.

In the distant past, I've detected the tape bias signal by slowing down the original 7 ips tape to 15/16 ips. That kills most of the audio, and shifts the 100 KHz bias signal down to 12.5 Khz, which can be detected. Lots of problems with this method and more than a few tricks involved, but it can be made to work. The big problem is that it has to be done with the original tape, which is often unavailable.

Much better is a ferrite tape head that offers expanded frequency response, typically to 1 Mhz. These are now commonly available but were previously rather specialized devices. They're used in systems that phase(?) lock onto the bias signal to provide an AFC (automagic frequency control) to eliminate flutter and wow from the tape: (See links at bottom of page)

I don't have any examples available, so I'm guessing from memory. I would guess(tm) a 16 bit digitized bias level would be about -100 dB below the peak audio level. That's fairly horrid when the audio band noise floor is probably about -60dB. Trying to extract a signal 40 dB into the noise is not my idea of fun. Yet, given time, it can be done. Plenty of articles on detecting signals below the noise floor:

The easiest is a sliding narrow band filter, that slowly and repeatedly scans across the 100 KHz area of interest, collecting signal level data and bin counting. I built one of these which produced its output on an x-y plotter. It was a crude autocorrelator, that looked for coherent signals. After about 1,000 passes, I could see a bump at the expected frequency, if the pen didn't rip the paper: It was also mechanically sensitive enough that I could see doors closing and changes in air pressure on the plot. The catch is that it could easily take a day or two to see anything meaningful.

Today, there are certainly better methods of audio forensics. I'm mostly familiar with older analog techniques and am somewhat lost in todays digital DSP world. The stuff I do today is far less sophisticated. Mostly it's signature analysis by looking at waterfall plots, spectrograms, sonograms, etc of radio transmissions, looking for residual tones and junk that can help identify the source. I've used your program (Daqarta) for this, but prefer Spectrum Lab for signal analysis.

The big problem with using a computah to look under the noise is that the FFT requires a huge number of samples. For example, with a common

I don't believe there's a single established method for extracting such signals. There are too many different types of recorders, media, and encoding schemes for one solution to work well with all of them. The wide band tape head is probably the most useful, but only with original media.

I don't know of any arrangement that does not use some form of anti-aliasing filtering to keep the audio and the bias signals from mixing. The better recorders place their cutoff frequencies quite high (about 50 KHz) to prevent group delay problems at the high end of the audio spectrum. Some people claim that they can hear phase shifts, so preserving the original waveforms has become a requirement.

How exactly would you be expecting to get the output from a tape player to analyse for 100kHz signals?

If the player isn't right there in front of you playing the tape nothing is likely to have a frequency response up anywhere near as far as that.

Tell me about it! I am so freakin' envious (in THE most positive sense!) of you guys. I have a degree in EE, but one of my frustrations in life is that I seldom/never get to use it. Imagine the coolest car, locked in a garage, with 0 miles/km. in the odometer. :-(

I do get to use my other car... err, engineering degree (CS), a lot though.

Bob:

I can fill some of the blanks. Marie Fonzi is the widow of Gaeton Fonzi, the guy who was friend of Bill O'Reilly. He helped Bill a lot when he was a nobody. Gaeton was an investigator (it pisses me off when people say that he was a reporter, or writing a book -- He was doing the business of *We The People*, since the House is even more of that than the Senate) who recorded all his job-related calls.

But I digress. The familiarity of Marie with technology is nothing short of tragic. It is a tragedy. (Pardon my hot-blooded hyperbole). She can post in forums, and e-mail, though. She told me that she was worried that after all these years of playing them (for book authors, investigators, etc.), the tapes may break, and some actually have. At this point, I was pulling the few hairs I have left.

She e-mailed me: "Ramon, the CNN cameraman is here, he is asking what kind of copy you want".

The cameraman spent 3 hours, in 2 phases: first, he plugged a cable between her cassette player and his huge, tripod-mounted camera. Next, he repeated the exercise through the air, into a "huge microphone" (her words).

When she said: "I gave him 2 [virgin] DVDs, they are going to make 2 copies in CNN-Miami, one for me, the other for you" my cynical reaction was "Boy! They are really in dire financial straits!! Fox is beating the crap of CNN! They can't even afford a couple of optical disks!!". Then I realized that all she wanted was to put a little pressure on the guy:

"You are not going to steal disks from a poor window, are you'?"

Back to techie-land: Let's say this was needed to save the earth: Would you analyze the info in the tapes or in the optical disk? I am betting that at the highest sample rate, you get everything, correct?

-Ramon

Brian: That was my concern (minus the specific numeric frequency, of which I am clueless) as well.

See below Jeff's post. Maybe the residuals that he mentions contain some of the info that I need?

Regards,

-Ramon

===============================

Yes, but I'm not going to get involved in a conspiracy theory resurrection.

In the 1970's all long lines telephony was analog. There were channel bank filters and mixers that upconverted base band audio to higher frequencies for transmission, and back down at the destination. Also known as FDM (frequency division multiplex:

The conversion process is not perfect and it is possible to see mixes and intermodulation products of the carrier and local oscillator frequencies of this up/down conversion process on the resultant audio[1]. There was also about a 1 to 5 Hertz Bode frequency shift introduced to prevent feedback and oscillation. You can't hear the beatnotes and frequency shift, but you can see them with PC based spectrum analysis. I suggest:

or better yet Spectrum Lab:

You will need to look at the "blank" spaces between the words, where there is no voices or background noises to muddle the display. You're looking for continuous carriers, buried well under the voices.

Autocorrelation:

is a big help for seeing these tones by removing the audio and background rubbish.

Interpreting the residual tones, and separating them from recording artifacts, is going to be difficult. You'll need to find someone with experience in 1970's telco muxes as well as some clue as to where these recording have been, whether they were converted from previous recordings, and possibly what equipment was used. You'll also need to know which CO handled the call, which will point to which carrier handled the call (AT&T, GT&E, ITT, etc), and then what brand and model of carrier equipment might have been used. That's not going to be easy and will probably be a huge time burn for little benefit.

Hopefully, I've given you enough hints to get started. You're on your own. Don't bother sending me email as I won't help.

Even at 16 bit 44.1 kHz sampling ( which is pretty mild these days ) it'll be fine. -96dB digital noise floor, 20-21KHz bandwidth.