Film-Tech Forum ARCHIVE: Perspecta decoding with a pc

This is topic Perspecta decoding with a pc in forum Film Handlers' Forum at Film-Tech Forum ARCHIVE.

To visit this topic, use this URL:
https://ft-forum.com/ft/cgi-bin/ubb/ultimatebb.cgi?ubb=get_topic;f=1;t=012372

Posted by Antonio Bizzarri (Member # 9890) on 11-16-2018, 03:50 AM:

Hey, guys, this is a really great forum!
I lurked for a long time, as I had nothing to contribute.
However, as I think (I hope!) I have devised a simple and cheap method to decode Perspecta sound in real time, I decided to break my silence. However, this is all in theory, as I don't have access to a 35mm projector or a perspecta encoded print. Anyway, as the system is simple enough, I would like to share it with you, hoping that someone tests it and reports back, and, ideally, share a few audio clips of the decoded audio (but I would be very happy even if someone had any kind of success with it).

However, this is the basic idea: a computer to do the real time decoding. Given, of course, that I did not completely misunderstood how Perspecta works!!!

Needed hardware:
1) A relatively modern computer (a 5 yeard old decent laptop will do fine)
2) An audio interface with at least 6 inputs and 3 outputs
3) A splitter/distribution amplifier (for example, the Behringer MX882 which costs less than 100$ - but any distribution amplifier will do, as long as it can split the signal in 6 identical copies)

Needed software:
1) Any Digital Audio Workstation program that supports VSTs (I use Cockos Reaper)
2) Two free Vst plugins: A parametric equalizer (the one built in Reaper is fine) and a Sidechain expander (I use ATK sidechain expander). Both are free.

So here is the idea:
1) Feed the mono signal out of the optical sound head to the splitter, so that 6 identical soundtracks are obtained;
2) Connect each of the six outputs to the sound card;
3) Assing each of them in Reaper to a different channel, and label them as follows: 1)Left 2)Center 3)Right 4)30hz tone Left 5)35hz tone Center 6)40hz tone Right
4) (This explains only the left channel, but the center and the right are identical)
4a) Insert the parametric eq vst on channel 4 (30hz tone left) and set it as a bandpass filter, so that only 30 hz can pass
4b) Insert the sidechain expander on channel 1 (left)
4c) Route the output of channel 1 to inputs 3 and 4 of channel 1 (after you insert che sidechain expander the inputs will become 4 instead of 2)
4d) Set the desired attack, release, threshold and make up level (a little experimentation will be necessary), and set the dry/wet knob all to wet. The expander will rise the level of the channel as soon as the control tone enters the sidechain input, and will lower it when no signal is present on the sidechain.
5) Repeat the same steps for channels 2-5 and 3-6
6) Connect outputs 1,2,3 to the cinema processor so that 1 goes to left, 2 to center and 3 to right

I tested this method using video files and artificial control tones, and it works.
However, of course, testing it on actual equipment is a different story.
Anyway, wanted to share this with you. Thank you for reading!

Posted by Steve Guttag (Member # 268) on 11-16-2018, 07:52 AM:

Part of the Perspecta process is not merely switching the channels on/off but also controlling the LEVEL of the channel (using VCAs or the equivalent). Thus they could pan dialog across the screen (had to be careful about foley effects on this) without it sounding like someone was merely flipping a toggle switch.

Furthermore, the "Q" of the perspecta filters have to be pretty precise. You need to be 20dB down if you are 5Hz off in either direction and should be within .5dB if you are +/- 2Hz (to allow for motor speed variances). The filter is described as a "shelf" centered at the desired frequency with very steep skirts on either side. I don't think your EQ plan will give it the right characteristic. Even forgetting the shelf effect, can you set the Q of the EQ filter high enough (narrow bandwidth)?

I believe I have a viable Perspecta solution for someone with QSYS though that will have the levels ride with the Perspecta tones. I have not put this through testing yet so, like you, I don't know if I have a working solution or not but I think it will be something that can be made to work if not working on the onset.

Posted by Antonio Bizzarri (Member # 9890) on 11-16-2018, 08:40 AM:

This is not a switcher, as the ATK is an expander, that increases or decreased the level of the LCR channels depending on the levels of the respective pilot tones. The proportion of the increase or the decrease can be varied based on the perspecta specs. The variation can be equal (1db increase per 1db increase of the pilot tone) or proportional (2db - or 3, or whatever - for 1db). The attack and release parameters of the expander set the responsiveness to variations in level of the pilot tones.

The equalizer can narrow the band as much as you want, and you can daisychain more than one of them to make the sides of the band even steeper, so this too should not be a problem.

Anyway,the fun part of this project is that anyone with a pc can do it now at 0 cost. All is needed is a pc, a daw, a couple of free plugins.
I would love to make a demo myself, but I'm missing the most important part, i.e. a file with a perspecta encoded soundtrack. If someone is willing to offer a link to download one (a lenght of 30 seconds or a minute would be fine), I could post the results for everyone to hear and judge. Or someone else could do it.
I'm sure your solution will work too, and I'm interested in hearing the results!

Posted by Steve Guttag (Member # 268) on 11-16-2018, 08:45 AM:

It sounds interesting then. Clearly, I'm not familiar with the equipment/programs you are using.

Personally, i'm hoping to find some home-video stuff where the Perspecta tones survived the transfer (perhaps on Laserdisc). I do know collectors that have Perspecta on 35mm and at the office we could set something up.

Posted by Antonio Bizzarri (Member # 9890) on 11-16-2018, 08:51 AM:

It would certainly be fun to find something to work with!

Posted by Allan Lyman (Member # 9449) on 11-18-2018, 08:25 AM:

I have the demo film,from youtube on a file.can opload it to you
this demo HAS the tones.the current film on youtube,does not,but its the same film

alan

Posted by Antonio Bizzarri (Member # 9890) on 11-18-2018, 08:47 AM:

Yes, please, it would be great!!!

Posted by Mark Gulbrandsen (Member # 72) on 11-18-2018, 08:55 AM:

My German friends have a complete working Perspecta system that is built on some sort of programmable audio processor. Not sure what processor they use, but I may be able to find out. It performs exactly as an original Perspecta decoder would.

Mark

Posted by Sascha F. Roll (Member # 9043) on 11-18-2018, 11:17 AM:

The cinema Mark mentioned is the Schauburg in Karlsruhe, Germany.
Their sound engineer integrated a digital Perspecta-decoder in the BSS Soundweb system. Works perfectly fine.

https://www.youtube.com/watch?v=rM1CekiMtdA

Posted by Marin Zorica (Member # 1716) on 11-18-2018, 12:42 PM:

They have BS BLU80 or simillar......i do have experiance in progamming BSS BIAMP, QSC Qsys.....and some DSP's....

I had similar idea to make NR SR&A decoding, but mainly problem is how to calibrat it at end.....

In this case, when you filter out controll freq, how much level per freq deviatio the output channel level should go +/-dB?

Posted by Steve Guttag (Member # 268) on 11-18-2018, 01:37 PM:

I'd be interested in the test video too.

On the YouTube demo...1st, someone did a REALLY nice job on the UI! Second, I'm surprised that the control tones are bobbing up/down with the signal. One would presume that with the appropriate filters, the audio should have very little effect on the 30-40Hz range. There should be a 3rd order crossover at 63Hz to separate the signal from the tones.

The other odd thing I noted was that I was pretty sure that Perspect limited their tones to about -16dB (referenced to 100% modulation) yet their meters were constantly showing more signal than that. However, since all three tones were playing, I suppose there was crosstalk. That said, the filter skirts are quite steep such that they should be down 20dB 5Hz on either side of the bandpass which should further reduce the influence of the actual audio signal).

Posted by Mark Gulbrandsen (Member # 72) on 11-18-2018, 03:10 PM:

Thanks for posting that Sascha! That's exactly the video I was thinking of.

I would have thought there would have been some variance on the Perspecta tones on just about any of the films. At zero or very low modulation it's gonna drop a little more then the -16 db or what ever it was. And the tones had to still be present.

Mark

Posted by Allan Lyman (Member # 9449) on 11-19-2018, 04:45 AM:

if those who want the video(its ONLY the sound,no picture)but the same you find on youtube today.so send me a pm,with you emailadress,and I will send it to you.

alan

Posted by Marin Zorica (Member # 1716) on 12-25-2018, 05:02 PM:

Antonio idea is ok, as well Steve comment about that BSS solution. Same thing as antonio described for VST's can be done in qsys, bss, biamp or so.......i just have two biamp dsp's as hend so waiting file to give it a try [Big Grin]

Posted by Carsten Kurz (Member # 5396) on 12-25-2018, 05:48 PM:

Software only approach: https://en.m.wikipedia.org/wiki/Pure_Data

- Carsten