Audio Distortion Help Needed (from 16mm film)

johnmeyer wrote on 10/13/2011, 2:01 PM
I have two questions about the audio from a 1964 film I just transferred. Here's a ten second clip:

Distorted Audio Clip

First question: Could this be caused by an amplifier issue? I captured this from the projector's external speaker output. However, I simply plugged this output jack into the Line Input of my Sony FX1 camcorder. I used a 10dB pad in between the projector and camcorder, but I did not directly load the output of the amplifier.

The audio from the projector speaker definitely has some of the same distorted quality, but because the built-in speaker is so poor, I can't tell for sure. The amp is a transistor amp (it is an Eiki SSL projector).

Second question: Is there any technique for "smoothing" this harsh sound? I tried the "smooth" function in Sound Forge but, as always, it does serious violence to the highs. The Vinyl Restoration noise reduction has a side effect that it can remove a little of this distortion, but that's the best I've been able to do so far.

Thanks in advance for any ideas or advice!


JJKizak wrote on 10/13/2011, 2:25 PM
That could be caused by mis-alignment of the sound tube pick-up or some dirt therin, or the sound tube is bad if it has a sound tube pick-up. The alignment is critical for distortion purposeses. (double critical) The normal response is from 50 to 6khz. and not one drop more for 6 khz. I would try it on another projector if available.
Your hookup is a bit unorthadox as that output is designed to drive a high current speaker load. (maybe 8 ohms.), So you might just have a huge impedance mismatch. As far as fixing that distortion? I don't think "Abby" on "NCIS" could fix it.

musicvid10 wrote on 10/13/2011, 2:33 PM
Here's my take on it.
All of the usable audio information lies between 200-800 Hz.
Everything else is noise (static + thd) and drops off pretty rapidly (just trusting my ears at this point).
Very typical of those optical stripe projectors.
The peaks are not hard-clipped so your recording setup was OK.

So if you put a fairly steep low-pass filter at 1200 Hz, you should almost get rid of the harmonic distortion spikes at 1750, 2625, and 3500 Hz. That alone will help the clarity.

If there is something (perhaps in Izotope?) that will emulate / restore the upper overtone series to the narrow audio band , that would make it sound more natural, rather than like a newsreel.

Some would suggest a noise gate, but a more natural approach would be a little compressed negative feedback to reduce background noise in the spaces between words. Less is better so it doesn't "pump." I'm sure you know the technique.

Let us know how it comes out. I may have time Saturday to put my suggestions into a test and try to isolate some of those frequencies and see if they are distortion or actual overtones, if you haven't already.

Opampman wrote on 10/13/2011, 3:28 PM
John - This is pretty typical of 16mm optical tracks, but the THD averages around 46% which IS excessive. I'll play with it tonight and see what I can do. A quick test with a paragraphic EQ at the harmonics made only minimal improvement but as I said, it was a quick test. I'll try to spend some time on it tonight and get back to you.

farss wrote on 10/13/2011, 3:45 PM
Mightn't be a bad move to replace the audio electronics in the projector from the pickup head itself. They were far from top shelf designs in their day.

paul_w wrote on 10/13/2011, 4:19 PM
Tapping into the signal from the head preamp rather than the speaker outlet could help too. It sounds a lot like cross over distortion from the power amp section driving the speaker. Looking at an audio scope would confirm this.
A guess would say the signal going into the power amp section may be a lot cleaner, and closer to the level you want. But does mean diving into the projector and fitting a switched jack socket between head preamp and the power amp sections.

[edit] hmm, i dont see any cross over distortion from the sample on a audio scope. But could still be a problem in the power stage.
Looks more like distortion at the peaks, as if recorded hot. But the optical detector itself may need checking /cleaning too as mentioned earlier. Sounds like a critical alignment is needed to reduce distortion.

paul_w wrote on 10/13/2011, 4:54 PM
Found this, its wonderful.

Optical Sound Tracks on Film[/link]

johnmeyer wrote on 10/13/2011, 5:15 PM
That could be caused by mis-alignment of the sound tube pick-up or some dirt therin,When I first got this projector, I mistakenly loosened the set screw on the sound lens. I didn't transfer any 16mm sound film for a long time, so I didn't deal with it right away, although I did keep looking for the SMPTE buzz test film. I found the film on eBay, but it was too expensive. When I finally got some sound film to transfer, I just loosened the set screw and then moved the sound barrel up and down and also rotated it until I got the clearest sound. I'm pretty sure I got it calibrated pretty close to what I'd get using the SMPTE tape and a VOM (which is what the service manual suggests).

However, getting back to the main part of the story, I have some important new information ...

I went back to my various manuals (yes, I know, RTFM) and found the following page. You don't need to read the tech explanation, unless you are an engineer. Instead, just look at the phono plug:

Well ... this wasn't what I had expected. When I connected to my camcorder, I used a mono phono to RCA adapter. This only has a tip and sleeve, and no ring. I pulled a stereo phono plug out of my parts box, and it is clear that by using the RCA mono adapter instead of the stereo adapter, I was not only using the speaker output instead of the 600 ohm output, but I was also shorting out the 600 ohm output.

So, it's time to go back and re-capture, but this time with the correct adapter.

BTW, I also use a pair of RCA to CAT5 converters, back to back, in order to break any ground loops. I use these all the time when running long distances (from sound boards to my camera, for instance) and they do a wonderful job of providing balanced line shielding for long runs. However, they also provide isolation as well, and is why I was using them here.

I'll post back after I've recaptured the audio. Capturing from the 600 ohm unbalanced output should eliminate any issues of properly loading the speaker output, and also make it much easier to more or less match impedance.

P.S. Paul, you posted while I was writing. Nice explanation of how the audio is picked up by the focused slit lamp.

paul_w wrote on 10/13/2011, 6:08 PM
Sounds good John, hopefully the AUX 600 ohm out pin is a signal before the power amp and not just a resistor network on the speaker connection.
How fortunate you have the manual. That is a very stange plug wiring system! Would never have guessed that. Tell us how it goes.

johnmeyer wrote on 10/13/2011, 7:15 PM
Well, I just put another two hours of learning under my belt, but with no results.

I re-did the audio transfer from the Eiki SSL projector, but this time using the "correct" plug. I also loosened the focus adjustment for the optical sound and listened on good headphones as I re-adjusted it, moving it up/down & rotating. The adjustment is quite sensitive, but it has little to do with distortion. Instead, as you wander away from the sweet spot, the audio gets dull and then quiet. I'm quite certain I have it set correctly (and it was set correctly before).

I spent quite a bit of time with levels (the 600 ohm output is still affected by the main volume and tone controls), both on the projector and on the camera, trying to minimize distortion. I'd upload the result of my efforts, but you won't be able to tell any difference. After all this work, no real difference.


So, either the amplifier is creaky, as Bob suggested, or perhaps the audio on the Kinescope film just wasn't that great to begin with.

Based on advice above, I did open the back cover on the projector with the aim of pulling the amplifier and taking a look, but the darn thing is buried in the bottom of the projector and looks like more than a casual job to get it out. It is possible that a coupling or filter electrolytic has gone bad, although having re-capped all my late 1950s and 1960s tape recorders last year, those old caps actually hold up pretty darned well.

So, unless anyone has ideas for post-capture restoration, I'm going to finish up this job tomorrow and ship it off to Chicago.

FWIW, since there are few old farts on this board, the voice you hear is a radio broadcaster named Cliff Johnson. He had a Chicago radio show in the 1950s that was broadcast directly from his house. His son, of the same name, who I went to school with, was the wise-cracking little-kid star of the show who later went on to become a fairly famous (in the Midwest) singer in a band called Off Broadway.

TMI, I know.

musicvid10 wrote on 10/13/2011, 8:07 PM
John, that audio sounds just like it did 45 years ago. It is what it is. A mechanical "fix" isn't worth pursuing because one doesn't exist. Analog optical pickups are sloppy, noisy, and narrow band. And the original audio recording probably wasn't any bertter.
Did I mention I was the school projectionist in junior high ;?)

I should have some time on Saturday to run some tests and see if the harmonic distortion can be filtered out somewhat. I don't have Izotope so maybe someone else can do better.
johnmeyer wrote on 10/13/2011, 8:45 PM
John, that audio sounds just like it did 45 years ago. It is what it is. A mechanical "fix" isn't worth pursuing because one doesn't exist. ... I should have some time on Saturday to run some tests and see if the harmonic distortion can be filtered out somewhat. That's pretty much what I too have concluded.

Thank you very much for all your help, but please don't bother to work on it Saturday, because, given the feedback here, I'm finishing the project tonight and it will go out FedEx tomorrow.

Did I mention I was the school projectionist in junior high ;?)Et tu? I had that job in grade school because the teachers, apparently, didn't know how to thread a projector. In fact, when people ask if I remember where I was when I heard Kennedy (JFK) had been shot, I remember quite exactly: I was threading the projector in the back of the auditorium at my grade school during lunch period.

PixelStuff wrote on 10/13/2011, 9:54 PM
Wouldn't it be cool if someone made a plugin where we could capture the video image of the sound track and then digitally convert the image into an audio file.

No I'm not the first one to think of it. In fact I think I remember a fictional TV show where a scientist was trying to scan vinyl records in 3D so he could play back a perfect replica of the sound from that.

And here is this...

johnmeyer wrote on 10/13/2011, 11:09 PM
Wouldn't it be cool if someone made a plugin where we could capture the video image of the sound track and then digitally convert the image into an audio file.That reminds me of the old idea of using a laser beam to read the grooves in records, thus eliminating the wear.

Well, what do you know ... a quick Google search, and apparently someone has actually built one:

ELP turntable

I have no idea if it is real, but it is amazing, given the market, that someone would have done it.

farss wrote on 10/14/2011, 2:40 AM
I'm very much inclined to agree with Musicvid, that does sound a lot like newsreel audio from 50 years ago. Things improved a lot in the following decade or so though and todays optical sound tracks on 35mm prints are very good, most cannot tell the difference if the DD track fails and the projection system switches to the optical track.

Optical reading of vinyl has been tried by many over the years. Someone sent me a link to a project someone undertook that enable you to play vinyl by scanning it on a conventional optical scanner.

JJKizak wrote on 10/14/2011, 6:41 AM
That Laser groove reading system did work but the beam picked up every little nuance in the groove---noise and clicks and pops and required on the fly software to filter it out which kind of negated the advantage of pristene audio playback. Also playing back your original sound sample was scads better than the first time I played it back. There is basically no problem with it now.

RalphM wrote on 10/14/2011, 11:02 AM

The naarator's voice quality sounds very much like those of the 40's and 50's. There may be nothing at all wrong with your projector.

If you want to ship the reel to me, I can run it through my EIKI NT-0 which is also an optical pickup. I've transferred very little 16mm sound, but one of those I did was from my neighbor who was an actress in a training film. The sound quality was surprisingly good.

johnmeyer wrote on 10/14/2011, 11:48 AM

Thanks for the offer, but I FedEx'd the package back to the client this morning.

Hopefully I'll get some more 16mm sound film to transfer soon and I'll be able to tell whether the problem was specific to that reel. I did transfer some 1-hour sound films from a TV station (with commercials embedded) for another client about a year ago, and don't remember having this problem, so I think in the end it may just have been bad sound on the original film.
musicvid10 wrote on 10/14/2011, 12:18 PM
This looks like the start of another interesting set of tests which "may" lead to an actual restoration technique or workflow template for audio on 16mm reels. There are tons of them lying around, and I think broadcasters who use clips of this footage don't actually pay much attention to the audio.

If my theory is correct (and it hasn't been tested) it "may" involve limiting the frequency range to the usable program audio, applying some noise reduction, then emulating the natural overtone series to a small degree to enhance the frequency range. Expanding the dynamic range may help a bit too.

Although no help to the topic at hand, I may still play with it a bit this weekend just to see if it can be improved a bit.

johnmeyer wrote on 10/14/2011, 1:02 PM
I may still play with it a bit this weekend just to see if it can be improved a bit.If you do, I certainly will be very interested in your technique and results.

I'm waiting for another assignment to be emailed to me, so while I'm waiting, I played with a snippet of the audio. I've dealt with this sort of distortion before, a long, long time ago, and remember using some of the de-clicking features in Sound Forge. This is a somewhat unorthodox "off label" usage of this feature, since the audio doesn't contain any clicks, per se. So, I just tried doing something similar using the "decrackle" feature in iZotope RX2 and what do you know, I did indeed get some improvement. I also tried the iZotope "declip" feature, even though the peaks are not clipped. You can force RX to reconstruct the waveforms, even though they are not clipped. I thought I got a little improvement, but not as much as using decrackle.

As always, it is important not to make it worse, in some other way, by adding artifacts.

PeterDuke wrote on 10/14/2011, 1:31 PM
My hearing is rather poor these days so I don't know what others hear. However I can still tell an old recording from a more recent one, which has often led me to ponder what it is about old recordings. Could it be non-linear distortion, irregular frequency response, phase distortion, signal processing such as companders (compressor-expander, particularly prior to Ray Dolby) expansion to reduce low level noise but which also truncates reberberant tails of transients, something else, or all of the above?

Early telephone engineers determined that the minimum bandwidth for good intelligibility was from about 300 Hz to about 2800 Hz. Anything less and intelligibility would suffer. For vowel sounds, the first two formants are most important, particularly the second which ranges from about 800 Hz to 2000 Hz. It is movement of the formants that is important, so provided that they stand above background noise and are not masked by adjacent sounds or irregular frequency response, intelligibility will be preserved. I think it probable that the peaks that Musicvid has observed are actually formants. However I also detect significant non-linear distortion, which is as difficult to remove as un-baking a cake.

I think that the best that could be done is to compensate for any broad frequency response errors by using tone controls or a graphic equalizer. If part of the spectrum is missing you can't really regenerate it, although I have seen papers where people have tried. If spurious components have been added by non-linear distortion or phase distortion, once again, you can't easily take them out again.
musicvid10 wrote on 10/14/2011, 2:21 PM
As a science demonstration project a few years back, I helped a student build an optical voice transmitter / receiver with an IR LED and photodiode detector.

It was much cleaner than any 16mm optical sound track I remember ever hearing, and actually worked up to 30 ft. in a semi-darkened hallway. Total cost with miniature enclosures was just over $20.
farss wrote on 10/14/2011, 3:34 PM
"As a science demonstration project a few years back, I helped a student build an optical voice transmitter / receiver with an IR LED and photodiode detector.

Such a setup should be capable of MHz bandwidth. I've seen a video link that used a 1W white LED and photodiode that was good for over 3KM.

Thinking about John's problem some more I'm inclinded to question the providence of the original audio recording. 40 years ago we have Nagras, 24 track audio recorders and crystal locked film cameras. The first time I watched Woodstock it was from a 16mm print. I'll never forget that, it was the only copy in Australia and when I picked up the reel off the flatbed editor the middle fell out of it and we spent hours untangling the mess. Thing is, it sounded pretty darn good.

I've spent a fair amount of time trying to restore old analog audio recordings going right back to acetate. The thing is, as John has mentioned, you can all too easily make matters worse. Those old analog systems most certainly degraded the original sound but in ways that were, shall I say "smooth". You start futzing around in the digital domain and you can lose that smoothness, many times I'd take a break and then go back to the original and realise despite all my best efforts it actually sounded more tolerable than my efforts at fixing it.

In this specific case the audio isn't repugnant to the vision, the tram track down the middle, the soft graphics etc all set the stage and I feel my brain just accepts it as a vintage piece. One could of course simply record a new pristine soundtrack, not that hard a task. I'd wager after doing that the image would suddenly start to look tragic but then that could also be easily redone. Problem then is the original is completely gone!

That's the thing with archival restoration. The intent should be to undo the damage done by time to get back to the original experience. Remastering is a different task where a new work is created. In this case somehow getting rid of the tram track damage done to the print is fine. Even if the soundtrack could be made into a modern prestine master that goes beyond restoration.

Opampman wrote on 10/14/2011, 8:31 PM
John - If this was a kinescope recording, remember, they were shot in a darkened room with a 16mm camera recording with a built-in optical track recorder in the camera. The camera was pointed at a CRT which was fed the live video to be recorded. Having made many many optical recordings when I ran a film lab, the built-in optical valves in (usually) the Auricon 16mm camera were no match for a true optical track recorder which would have been used to make the Woodstock print, for example. So, in actuallity, the recoding is not that bad - if indeed it was a kinescope (which I missed the first time through).