CCD sensor -> analog delay

[Taylor]

Buried within a thread about how expensive BBDs may or may not be, was this post by amptramp:

Bucket brigade chips may be in short supply, but has anyone considered the video analogue?  Camera chips would permit video entry via a LED and clock the signal through as in a bucket brigade.  These things are being made in the millions for cellphones.  Just black out the video pixels after the first one and read the signal out at the end.  I have a VGA video camera that I ppicked up at a garage sale for a buck last year.  That would give you a 640-cell delay and you could do stereo (in fact, 480 channels, not just two).  You just have to be able to paint or otherwise block the input to the pixels after the first.

This is too amazing an idea for it not to get more attention.

Of course, there will be lots of things to work out, but I think we should discuss the possibility of implementing this. I don't at all think it makes sense as a cost-saving measure. Perhaps you'll save 4 bucks, but the time invested will be worth a lot more than that, even at minimum wage. But as a unique idea with potentially very strange sound  results, I think it's definitely worth pursuing. Anyone have any further thoughts?

[CynicalMan]

Some Malekko Echo delays have CCD in their name, but gut shots on the other forum suggest they are BBD-based.

This has already been done, but you'd have to buy the information:
www.aes.org/e-lib/browse.cfm?elib=4590
www.aes.org/e-lib/browse.cfm?elib=3946

CCD's look like they normally have delays in the microsecond range, and most of them would need a clock in the kHz to get a useful delay. At that frequency, I don't know whether the delay would work, but it's worth a try. If it does work, then we could probably get delays into the low milliseconds, but you couldn't get long delays. The longest ccds you can get seem to be around 2000 stages, like this:
www.datasheetcatalog.org/datasheets/166/200129_DS.pdf

Maybe useful for a flanger?

[slacker]

I approve of this madness

[Taylor]

This has already been done, but you'd have to buy the information:
www.aes.org/e-lib/browse.cfm?elib=4590
www.aes.org/e-lib/browse.cfm?elib=3946

CCD's look like they normally have delays in the microsecond range, and most of them would need a clock in the kHz to get a useful delay. At that frequency, I don't know whether the delay would work, but it's worth a try. If it does work, then we could probably get delays into the low milliseconds, but you couldn't get long delays. The longest ccds you can get seem to be around 2000 stages, like this:
www.datasheetcatalog.org/datasheets/166/200129_DS.pdf

Maybe useful for a flanger?

Hmm. However much I may know or be in the dark regarding analog audio processing, I know even less about video - basically nothing. However, I don't think those are what amptramp is talking about. He's talking about using an actual video sensor as the storage medium. Realistically I've no idea how to hook it up to make a delay, but if a pixel is equivalent to a stage of BBD storage, then a 480x640 sensor is over 300,000 stages... so, I guess that can't be right, because if it was, that would be amazing. That would be an analog 30 second looper with potential for overdub...

[amptramp]

I approve of this madness

I started this madness.  The idea here is that there are large numbers of old VCR's, still cameras and movie cameras with usable IC's in them.  Something that would be really great would be one with a test or cascade input so you could have purely electronic input as well as output, but I doubt much of this sort of thing will show up.  Taylor is correct, I was thinking of using the video sensor as the delay element since the popular delays are out of production.  With input from blue, green and red LED's, you could have a three-channel delay which could be concatenated to provide a single delay of three times the length.  I don't know how well a video sensor would operate in terms of shift frequency range, but I imagine it would be a larger range than it is specified for - if you are generating NTSC video for an old TV, the horizontal scan is about 64 microseconds.  But there is nothing saying the device is limited to that - it just happens to be the design point the device is optimized for.

I am not sure I have ever seen anything that operates an area array as a single line where you would get 640 x 480 = 307200 stages that are serially addressable.  But you could use the array to generate fiunctions of a complex variable which is like running the observation point around a contour map.  Now that would be madness.

[bean]

I seriously have a nerd boner right now. This is fascinating stuff.

[Brossman]

So... what's the deal with this? Can we do it? Am I gonna hafta go raidin' pawn shops all over muh town, n'buy some VCRs'n'stuff ta tear out tha old IC chips?  Idk a whole lot, but I've got a breadboard (and I get the feeling I might need several, haha) And I'm willing to at least put my two cents down...

I'm thinking analog signal processing (CCD), with digital timer... that would probably be the best of both worlds.

I don't know how well a video sensor would operate in terms of shift frequency range, but I imagine it would be a larger range than it is specified for ... But there is nothing saying the device is limited to that ...

I am not sure I have ever seen anything that operates an area array as a single line where you would get 640 x 480 = 307200 stages that are serially addressable.  But you could use the array to generate fiunctions of a complex variable...

If you can control the inputs and outputs (not only what kind, but location, destination, and purpose...), then why wouldn't you be able to use it? Though, going into seriously high ranges like up around 30-sec (a la 555 @15sec) can pose lots of potentioal problems with long-term use. that s*&% gets hot...

[Taylor]

I think if anybody does this they win the DIY award of the decade. I certainly don't have the know-how, or time at the moment, but if anybody thinks they do have the skills or at least time to figure it out, I would in all seriousness pay some bucks for such a thing.

[R.G.]

One of the main uses for the SAD1024 was as a time base corrector for analog/tape video back in the stone age when they were developed. Audio was not the only intended use. It's why the SAD1024 is alone among BBD delays in the ability to be clocked really, really fast.

Or was.

[armstrom]

You could look into some of the linear CCD sensors that are available in the wild. These are commonly used for autofocus systems in cameras (both phase based and contrast based). or as PSDs.

Kodak makes sensors that go from 2098 x 3 up to 8160 x 1.. These are designed for use in optical scanners though, so I'm not sure how slow you can clock them
http://www.kodak.com/global/en/business/ISS/Products/Linear/index.jhtml?pq-path=14426

Just a thought.

[Taylor]

I still think this is such an interesting idea, that I thought I'd bump this old thread. I still don't have the knowledge to do anything with it, and even less free time than back then, but maybe some of our newer members could pursue this line of thinking?

[Rixen]

I suspect the leakage/noise per element would be too great to allow for slow clocking. I would love to see someone try it..

[amptramp]

I can't believe this has been bumped from eight years ago but it should be easier to get old VHS cameras and associated stuff at garage sales or thrift shops.  It is still an idea that some enterprising DIY'er should be able to cobble up if the information about the sensor is still available.  Good luck to anyone who tries it!