Author |
Message |
willi
Joined: Oct 20, 2008 Posts: 13 Location: Portland, Oregon
|
Posted: Thu Nov 10, 2011 7:27 am Post subject:
CCD based delay? Subject description: thoughts... |
|
|
Hello,
I know many "analog" delays use BBD chips and they are generally scare or costly, or both. Has anyone ever designed a CCD delay? This might not be as silly as it sounds at first...
According to Wikipedia, the BBD and CCD were both developed in the late 1960's, with the BBD coming from Philips and the CCD coming from AT&T. "The concept of the bucket-brigade device led to the charge-coupled device (CCD) developed by Bell Labs." Yes, the most common use of the CCD these days is for digital imaging. But the device itself is more general purpose than that. The first CCD was basically an 8-bit shift register and one of the initial intended purposes of the design was as a delay line. The first device was addressed through wires (not optically). While many modern CCDs have 2d XY optical sensors, there have been many 1d sensors made for fax machines. These sensors (along with CMOS sensors) are also used in barcode scanners.
http://www.datasheetcatalog.org/datasheets2/57/57697_1.pdf
1728 bit shift register. This data sheet seems to indicate the 'bits' are analog charge values, clocked at 1MHz.
Are these devices available in non-optical forms? Is there an effective way to address the CCD optically, but conveying audio information, perhaps with an LED output of the signal feeding the CCD input? Can the CCD hold a variable charge or is the bit somehow digital? Wouldn't it be possible to clock the CCD fast enough that a binary delay line would work? Sony has some audio gear that uses 1-bit encoding and very high sample rates. Would there be any advantages over BBD's (aside from IC availability, which is of course a big consideration as the situation is today) or compared to 'traditional' digital delays? Perhaps kludging an LED input would be sloppy, in as much that it would be very difficult to isolate the LED input to a single CCD cell -- but perhaps that would be a pleasing 'smear' of the quantized capture?
I wouldn't be surprised if BBDs are still commonly used just because that is what was originally used long ago. Of course, I also wouldn't be terribly surprised to find that there is some reason why a CCD would not work in this application. But so far, it seems like an obvious source of lengthy quantized-analog delay lines...
On a slightly related note, perhaps it is possible to create BBD circuits using an FPGA or FPAA.
Thanks for your time and insights. |
|
Back to top
|
|
|
DrJustice
Joined: Sep 13, 2004 Posts: 2114 Location: Morokulien
Audio files: 4
|
Posted: Thu Nov 10, 2011 8:12 am Post subject:
Re: CCD based delay? Subject description: thoughts... |
|
|
willi wrote: | Can the CCD hold a variable charge or is the bit somehow digital? |
Yes, each charge is proportional to the amount of photons that has hit it during exposure. A CCD is really a specialized BBD - I don't really see any advantages (or even possibility) of using it over BBDs, especially as they're not designed to shift a signal through them, only out of them.
willi wrote: | On a slightly related note, perhaps it is possible to create BBD circuits using an FPGA or FPAA. |
With FPGAs being purely digital, the only way would be to digitize the charge on entry and convert to analogue on exit, which constitutes a digital delay line. The last time I checked out FPAAs, they weren't really suited to this kind of application, but that was a few years ago so it could be worth reinvestigating.
Sorry, I don't mean to be a spoilsport...
DJ
-- Last edited by DrJustice on Thu Nov 10, 2011 10:50 am; edited 1 time in total |
|
Back to top
|
|
|
willi
Joined: Oct 20, 2008 Posts: 13 Location: Portland, Oregon
|
Posted: Thu Nov 10, 2011 8:38 am Post subject:
|
|
|
I don't think you are a spoilsport. I just think it is an interesting concept, regardless of whether it proves feasible.
How about stimulating one pixel at the end of a 1d CCD with a highly focused laser? That might be a way to switch the optical representation of the signal fast enough. If I understand correctly, each pixel is then shifted as the clock increments. Adding a laser might not be worth the hassle or expense. Perhaps it is possible to directly stimulate the initial pixel electrically with a very low voltage?
The most capable FPAAs that I am aware of are actually integrated into SoC designs... but those are getting cheaper every day.
Any of these concepts would almost certainly require a custom circuit to implement; I don't envision directly replacing BBDs with CCDs in established circuits. But existing circuits could provide a starting point for experimentation perhaps. |
|
Back to top
|
|
|
DrJustice
Joined: Sep 13, 2004 Posts: 2114 Location: Morokulien
Audio files: 4
|
Posted: Thu Nov 10, 2011 11:01 am Post subject:
|
|
|
willi wrote: | I just think it is an interesting concept, regardless of whether it proves feasible. |
Absolutely!
Quote: | ...How about stimulating one pixel at the end of a 1d CCD with a highly focused laser? That might be a way to switch the optical representation of the signal fast enough. |
I like that idea! Perhaps there would be some 'optical compressor' style goodness to the sound
Quote: | ...Perhaps it is possible to directly stimulate the initial pixel electrically with a very low voltage? |
The only input to these devices (that I know about) is by light.... Do you mean to stimulate it capacitively with an electrode very close to the first capacitive element on the CCD die?
DJ
-- |
|
Back to top
|
|
|
willi
Joined: Oct 20, 2008 Posts: 13 Location: Portland, Oregon
|
Posted: Thu Nov 10, 2011 5:08 pm Post subject:
|
|
|
Yeah, I think CCDs are susceptible to damage from laser light due to the beam collimation, but perhaps a suitable filter would prevent this. Laser diodes are extremely cheap these days, so maybe that would be a possible path. A dead old fax machine or scanner, and CD player might be a source of free parts to experiment with. But considering that the photons are somehow converted to electrons when striking the CCD perhaps directly injecting electrons would be a better or cheaper solution, if possible. |
|
Back to top
|
|
|
willi
Joined: Oct 20, 2008 Posts: 13 Location: Portland, Oregon
|
Posted: Thu Nov 10, 2011 5:59 pm Post subject:
|
|
|
Actually, looking at the datasheet further, pin 4 is for Serial Input. I understand this to mean a voltage/current can be applied directly to this pin and then carried down the delay line. This would be for putting multiple 1d CCDs in a row and extracting all of the final output serially. Interesting, no? |
|
Back to top
|
|
|
DrJustice
Joined: Sep 13, 2004 Posts: 2114 Location: Morokulien
Audio files: 4
|
|
Back to top
|
|
|
willi
Joined: Oct 20, 2008 Posts: 13 Location: Portland, Oregon
|
Posted: Thu Nov 10, 2011 10:50 pm Post subject:
|
|
|
Yes I agree, the red arrow seems to be the point that light enters the circuit. It seems a little confusing because I am under the impression they are more like photodiode, but whatever. It's quite possible this is a CMOS sensor; I probably just searched for 'fax sensor' or something. But, what about Ao? That is listed as an analog output on Pin 1, and the Timing Chart seems to indicate the pixels can have varying values, and those (greyscale) variables can't be encoded digitally with only 1-bit per pixel... I don't really understand why you would want to inject bit patterns into the shift register, but maybe that is for some kind of electronic shutter function. Perhaps it is not the best datasheet to reference for this concept, anyway.
http://www.thorlabs.com/Thorcat/18100/18143-M01.pdf
Here is a Toshiba CCD with 3k+ pixels. Unfortunately, the diagram of Pin Names does not seem to describe the function of Pin 4 from the pinout diagram on the first page. This chip includes a S+H and 'Electronic Shutter', and if you look at the timing chart, the actual dump from the CCD is surrounded by some other output (the shutter function for facilitating data capture somehow, I guess). So this chip is also probably a bad choice for this concept.
Still, there are a lot of chips out there. Although the laser trick is neat, it seems like it would be ideal to inject electrons directly. |
|
Back to top
|
|
|
willi
Joined: Oct 20, 2008 Posts: 13 Location: Portland, Oregon
|
Posted: Thu Nov 10, 2011 11:14 pm Post subject:
|
|
|
Ok, here are a couple of Fairchild sensors that definitely include analog delay lines. In fact, they each include 2! That might be annoying and perhaps these are still not the right sensors to target, but it could conceivably be useful for stereo processing or perhaps for linking the output of one line into the input of the other. However, these chips do include some 'extras' like anti-bloom circuitry, and the timing output diagram seems to indicate some sort of 'prescan output' that precedes the actual pixel data. Perhaps these are higher end sensors and it is better to find a very low cost CCD designed specifically for barcode reading. At any rate, these datasheets are much more informative than the last few.
http://www.fairchildimaging.com/documents/CCD143ADataSheetRevA.pdf
2048 photosite units
http://www.fairchildimaging.com/documents/CCD143ADataSheetRevA.pdf
6000 photosite units! |
|
Back to top
|
|
|
DrJustice
Joined: Sep 13, 2004 Posts: 2114 Location: Morokulien
Audio files: 4
|
Posted: Fri Nov 11, 2011 5:24 am Post subject:
|
|
|
willi wrote: | Yes I agree, the red arrow seems to be the point that light enters the circuit. It seems a little confusing because I am under the impression they are more like photodiode, but whatever. |
Photodiode or phototransistor, for the principle it amount to almost the same; as light enters the devices, both of them starts conducting in the forward direction. A phototransistor basically have more gain.
Quote: | It's quite possible this is a CMOS sensor; I probably just searched for 'fax sensor' or something. |
Yes it's an old'ish fax sensor - a CMOS PPS.
Quote: | But, what about Ao? That is listed as an analog output on Pin 1, and the Timing Chart seems to indicate the pixels can have varying values, and those (greyscale) variables can't be encoded digitally with only 1-bit per pixel... I don't really understand why you would want to inject bit patterns into the shift register, but maybe that is for some kind of electronic shutter function. Perhaps it is not the best datasheet to reference for this concept, anyway. |
The datasheet does indeed lack a good functional description. Therefore we have to follow the outlined circuit. And yes, it has an analogue greyscale output.
The bit(s) that are clocked through the shift register (BIT 1, BIT 2,...BIT n) will in turn switch on the Field Effect Transistor below each phototransistor (blue conductor), allowing them to conduct a current proportional to the light (red) that currently shines on the phototransistor. The opened path (green) allow the current to reach the analogue output pin (Ao). No charge is stored or shifted along, the light intensity at each pixel is read out instantaneously as they are addressed by the digital shift register.
The reason why we 'inject' our addressing bit(s) into the shift register (at the SI pin) is because we need to be in absolute control of the timing, so that we know exactly which pixel is being output at each clock pulse.
It's a very simple circuit, with a leisurely read rate of 1MHz, which leads me to think it's very old - used in early fax machines. In conclusion, this does not fit our application, but at least we learned how it works
Quote: | http://www.thorlabs.com/Thorcat/18100/18143-M01.pdf
Here is a Toshiba CCD with 3k+ pixels. Unfortunately, the diagram of Pin Names does not seem to describe the function of Pin 4 from the pinout diagram on the first page. This chip includes a S+H and 'Electronic Shutter', and if you look at the timing chart, the actual dump from the CCD is surrounded by some other output (the shutter function for facilitating data capture somehow, I guess). So this chip is also probably a bad choice for this concept. |
A CCD or CMOS APS (Active Pixel Sensor) will typical store the accumulated charge from the phototransistor in a capacitor. There may be control circuitry to switch on and off the flow into the capacitor (S&H) acting as a 'shutter', and a switch to discharge the capacitor in order to clear the image.
In a CCD the pixel charges are shifted from pixel-bucket to pixel-bucket, and row by row into an output BBD for a 2D sensor. In CMOS APS and PPS sensors the pixels are read out by direct addressing instead, and they can have the same charge capacitor and switches (or not, depending on the intended application of the device) to control current flow and to clear the image.
A CCD with the more advanced control circuitry can be used as a delay line (with light entering the first pixel) if the chip gives us manual control of the shutter and clear functions.
I'm sorry, I don't have time to go through the other datasheets now; it takes a bit of time to study them and prepare the text and illustrations. I may get back to that later if needed.
Quote: | Still, there are a lot of chips out there. Although the laser trick is neat, it seems like it would be ideal to inject electrons directly. |
The problem with that is that there are no electrical analogue input. So we're back to the laser then
DJ
--
Description: |
Part of IA2008-MB82A datasheet - read out function illustrated |
|
Filesize: |
53.37 KB |
Viewed: |
307 Time(s) |
This image has been reduced to fit the page. Click on it to enlarge. |
|
|
|
Back to top
|
|
|
The Real MC
Joined: Jun 20, 2008 Posts: 62 Location: Painted Post NY
|
Posted: Fri Nov 18, 2011 10:14 am Post subject:
Re: CCD based delay? Subject description: thoughts... |
|
|
willi wrote: | Has anyone ever designed a CCD delay? |
Yup
Stephen St Croix designed the famous Marshall Time Modulator in the late 70s which is a CCD-based delay.
Unfortunately there is little information on this device, the critical CCD elements are sealed in epoxy, schematics are nowhere to be found, and St Croix passed away some years ago. The only folks who have the know-how behind the device is studioelectronics.biz who have acquired engineering notes and spare parts from the St Croix estate.
Somewhere on the net are scans of a late 70s Device Magazine with an in-depth interview with St Croix who provided some insight of his design. St Croix was so paranoid of industry espionage and theft of his design that he encased the critical parts in epoxy and deceptively hid his patents which have yet to be uncovered. |
|
Back to top
|
|
|
DrJustice
Joined: Sep 13, 2004 Posts: 2114 Location: Morokulien
Audio files: 4
|
Posted: Fri Nov 18, 2011 12:20 pm Post subject:
Re: CCD based delay? Subject description: thoughts... |
|
|
The Real MC wrote: | Stephen St Croix designed the famous Marshall Time Modulator in the late 70s which is a CCD-based delay. |
Fascinating! I even found one for sale while investigating the unit. And here are some pictures of the innards.
Apparently the Boss DM-1 Delay was also CCD based.
Do you know anything about how they did this? It has been suggested that the early spy-camera sourced CCD units in the Marshall Time Modulator had a signal injection feature for testing and diagnostics - sounds plausible. What about the BOSS unit then?
Edit: The Boss DM-1 Delay used a Reticon R5101 CCD (not an image sensor, just a delay). So did some old MXR delays.
Edit 2: Found an interesting file BBDementia.zip. Within this archive you a Reticon.zip archive is embedded. It contains some scans of documents explaining that CTDs (Charge Transfer Device) come in the form of BBDs and CCDs which differs only in the structure of the implementation. In essence a CCD isn't an image sensor before you add photo diodes/transistors especially for that application.
DJ
-- Last edited by DrJustice on Fri Nov 18, 2011 1:50 pm; edited 2 times in total |
|
Back to top
|
|
|
DrJustice
Joined: Sep 13, 2004 Posts: 2114 Location: Morokulien
Audio files: 4
|
Posted: Fri Nov 18, 2011 12:55 pm Post subject:
|
|
|
@ willi : I had a brief look at the three last data sheets you linked to. Unfortunately none of those CCD image sensors has an input pin (which makes sense, really). The only way in is by photons...
DJ
-- |
|
Back to top
|
|
|
sneakthief
Joined: Jul 24, 2006 Posts: 569 Location: Berlin
|
|
Back to top
|
|
|
The Real MC
Joined: Jun 20, 2008 Posts: 62 Location: Painted Post NY
|
|
Back to top
|
|
|
axl
Joined: Oct 28, 2012 Posts: 33 Location: ita
|
|
Back to top
|
|
|
|