Re: [ARSCLIST] Bias frequencies and their recovery (was Digital File Specs and StoryCorps)

["Richard L. Hess" <arclists@xxxxxxxxxxxxxxx>]

Hello, Eric,

Here is an interesting reference on 24-bit recording with some 
real-world examples:
http://www.sounddevices.com/notes/recorders/real-world-24-bits/

Hi, Tom,

Thanks for adding those bias frequencies.

If we say that we wish to recover bias up to 250 kHz (for sake of 
argument, let's not go for the extremes of the Sony APR (400 kHz) and 
the Ampex ATR (432 kHz)) we need to look at roughly the half 
wavelengths of these frequencies.

Let's also say that we wish to recover 150 kHz bias off cassettes.

250 kHz at 15 in/s has a wavelength of 60 micro-inches, implying the 
need for a 30-micro-inch gap. Reel playback head gaps are minimum 80 
micro-inches and generally longer.

At 7.5 in/s, we'd need a 15 micro-inch gap and at 3.75, we'd need a 
7.5 micro-inch gap.

For cassettes, using the reduced spec of 150 kHz recovery, we'd need 
a 6.3 micro-inch gap.

My suggestion to Jamie Haworth has been to make a composite head that 
is a normal play head with a hard drive head embedded with the gaps 
aligned. Then phase lock a lower frequency pilot to the recovered 
bias and use that for the flutter signal. It's somewhat like the 
"colour-under" recording on lower-quality VCRs. The bias pilot tone 
should be in the 9-18 kHz region to be safe, IMHO. I think that would 
still provide enough resolution for flutter correction, although not 
perhaps for scrape flutter. The software flywheels that need to be 
put in place to avoid increasing distortion are interesting. If we 
were doing 96 kHz sampling, the flutter signal could be 19-36 kHz or 
so, providing 2x the resolution.

In that way, with relatively simple hardware we can archive another 
track that contains the speed-perturbation-related bias information 
using existing equipment and then Jamie can work his magic on the 
digital files later.

I discussed this over a very pleasant lunch in downtown Toronto 
during JTS with Nadja Wallaszkovits of Vienna who is working on 
archiving the bias using cassette repro heads that are offset. With 
only having a napkin available for calculations, I think she is 
working within the gap length combing region which is why her results 
have been variable. The fact that her heads are offset creates the 
second issue that you are correcting for flutter at a point it's not 
happening. And, the math of using a delay is interesting, but 
complex, since the time between heads varies based on the flutter 
component. It's much like stagger-head stereo images whipping around 
when there is flutter.

Even with 192 ks/s recording, the SD 722 provides decent response out 
to the "Nyquist frequency" (according to their FAQ "for scientific 
purposes") so we can't get better than 96 kHz at 192 ks/s, and I 
would suspect that the real world bandwidth is around 80-85 kHz. Good 
for some bats, I suspect. My Racal Store 4 DS does better at 60 in/s: 
300 kHz, but only plays back 75 kHz at 15 in/s, so it is not really 
useful for bias recovery, either.

The whole bias recovery has been put out there by Jamie Haworth and 
Michael Gerzon, but few other than Jamie and Nadja are doing it, and 
Nadja is only storing the bias, not resolving it as far as I know.

Cheers,

Richard



Richard L. Hess                   email: richard@xxxxxxxxxxxxxxx
Aurora, Ontario, Canada       (905) 713 6733     1-877-TAPE-FIX
Detailed contact information: http://www.richardhess.com/tape/contact.htm
Quality tape transfers -- even from hard-to-play tapes.