Camera Removal?

Re: Cable termination
Before going with pullup/down terminations, I would try a series resistor termination at the signal driver. For best results the value of the resistor needs to be tuned to the cable/driver combination using an oscilloscope. This is actually an interesting point. What is the signal driver in this case? I suggested 100 Ohm pullup assuming it is an open collector type output (ie a current source). The main problem is then to avoid the echo from the driver back to the camera so the load seen by the signal `on the way back' is (infinity (the driver) || the pullup) = 100 = Zc. I guess 100 Ohm is pretty close to the standard ribbon cable Zc. In this case a series resistor is not going to do much for the echo back to the camera and the echos from the camera would have to be handled by terminating at the camera end. What other output stages could there be? A standard push/pull (a NOT gate kinda thing?)? Then one would have to go with a series R. Has someone checked this already? Tuning it with the scope is necessary in either case but there is only one `fast' clock line so this should not be a problem. Anyone had any luck finding a connector that fits?

Output drivers
The output drivers of the chip are almost certainly standard complementary pair push-pull drivers. Series termination is the way to go. It takes a rather beefy driver to handle a 100 ohm pullup or pulldown. Series termination does not add any additional load to the driversin the chip.

Re: Output drivers
Another good point, texaspyro, The drivers are definitely not open collector (the schematic shows no pullups) so one must use serial R to terminate. On the other hand, while the 100 or so Ohm termination does not add to the load of the driver, the combined load of 100 (Zc of the cable) and 100 termination is still too much for the driver to handle. I do not have the specs for the Zoran chip, of course, but the outputs of the camera itself would have to be terminated, as well; now, at 5V that little chip would have to drive a 200 Ohm load which means it would pump 25mA. The specs for OVT7660 state that the driver can output about 8 mA in the HIGH state (about 15mA in LOW) so even if they factored in a safety factor of 2, it would still overload the driver. If all this stuff about Zc is not convincing, think about driving a large capacitive load (which a long cable definitely is) at 30MHz. Anyone willing to try what length cable we can get away with (terminating at the sensor will be a ... )? The output drivers of the chip are almost certainly standard complementary pair push-pull drivers. Series termination is the way to go. It takes a rather beefy driver to handle a 100 ohm pullup or pulldown. Series termination does not add any additional load to the driversin the chip.

Series termination
Nope, series termination does not add any extra load to the driver... it actually cuts it down. Without any series termination (what you have now with a directly connected sensor) the effective series termination would be zero ohms... which would be infinite current... which is not the case.

Re: Series termination
Nope, series termination does not add any extra load to the driver... it actually cuts it down. Without any series termination (what you have now with a directly connected sensor) the effective series termination would be zero ohms... which would be infinite current... which is not the case. Ah, but this is without a mile of cable in between. What we have now is the driver seeing the sensor chips' inputs directly, so the impedance of the load now is effectively infinite. When there is a cable involved, the driver will see first the series resistor (you are absolutely right, it will actually make the impedance higher) and then the cable which looks like a large capacitive load (if not terminated on the sensor side) or even a 100 Ohm resistive load (if we decide to terminate at the sensor's end). Now when the driver switches, say, to HIGH it will have to start charging that `cable cap', and at the speeds involved this would make it output more current than it could handle. If the cable is short enough, though, this cap is not that large (so, in our case, without the cable, there is no cap at all). As you said, a series termination would help ... but it would not solve the problem (of course, a pullup that I suggested now does not make any sense). Making the series termination higher is not going to help since the impedance of the cable has to be matched. The argument that the cap does not `consume' any power does not help either, for the cap will surely discharge back into the driver, `returning' power but the current flowing through the driver is still excessive at the beginning. So the only way out I can see is to keep the cable short. A quarter wavelength is probably ok (~7ft at 30MHz). This 30MHz figure would actually have to be revised since this is the frequency of the clock. To keep those rising edges clean, we are probably looking at double the bandwidth so 3--3.5 ft is probably safe but beyond that---only experiment would tell. May be we can push it to three times, who knows? Any luck with finding a matching connector?

Whew.. you guys are redefining "over my head". Thanks for all the input though! You have me confident that my 5" extension project has some chance of success :D Regarding the connectors, I struck out with a $30 assortment order from DigiKey. I was trying to get FPC connectors and flatflex cable (just like the LCD).. but unfortunately they only come in 1-row and 2-row varieties. We need the 2-row, but that means the pins are staggered, not symmetric on either side of the connector. None of them were wide enough either. I drafted up a two-sided PCB that would put a CVScam connector on one side and a 22-pin FPC on the other. Pad2pad.com quoted me $98 for a minimum order of 107 of them. Problem is, to complete this approach one would need: 1x sacraficial CVS cam =$30 2x OR667CT-ND 22pos 0.5 SMT =$2.10 1x HFF-22U-05-ND 5" flat flex cable =$4.95 1x Digikey shipping & handling = $8.35 ... making this about a $47 "5 inch extension cable" assuming I could sell off those other 105 PCBs... It's a $150 extension cable if I can't. Plus you still have to get a clean desolder of two CVScam connectors and 4 successful SMT solders on the new PCBs. Blah. The hunt continues!!!

Whew.. you guys are redefining "over my head". Thanks for all the input though! You have me confident that my 5" extension project has some chance of success :D Oops. Yeah, it did get a bit too academic, the next step would have been to whip out a Laplace transform :wink:. So none of those connectors worked? Has anyone tried to use a PCB millling machine to make a connector like this? I will ask our lab wizard next week about this (and maybe try it out as well). We cannot reproduce the exact connector but maybe a series of grooves on a copper clad board would make a good enough connector? Then one could solder the ff cable to it. IF this were reliable enough, we would not even have to desolder anything.

In the interest of keeping this thread up to date, "Microfunguy" posted here the two connectors' part numbers which could be a 1-for-1 replacement of the CCD connectors... but you'd still probably need to unsolder the existing and solder these in their place: Molex part numbers: 52465-2291 53307-2291 I've actually found these before, but never in lots smaller than 3000 so I wrote them off. I never had the exact Molex part number, however, so maybe someone else has a suppler I could buy ~8 from? They're board-to-board connectors, so making a cable is still going to be a hassle.. but for folks interested in trying it, they might do the job. For me, I'd either want to replace the connectors with something that readily accepts a ribbon cable of some sort, or I want to make an adapter that plugs into the CVScam connector, and then has a cable connector on the other side.

I have managed to remove the 22-way connector from the camera board with the help of very fine desoldering braid and a stereo microscope. It wasn't easy. there is a large contact area between the receptacle legs and the PCB so you have to make sure as much solder as possible is removed. For my application (compact stereo camcorder), I am going to solder wires directly to the receptacle pads and the other end of the wires to the removed receptacle. A blob of epoxy on each to protect them. I am only aiming for 50mm extension. Apart from timing considerations, longer lengths are going to be prone to noise pickup, even with shielding. It will be interesting to see if anyone can make it work at longer lengths.

View Full Version: Camera Removal?