The Cookbook CCD camera.
I'm no trained electronics technician, but I've been playing with a soldering iron for more than 20 years. I built my first computer (based on a 16-bit TI 9900 CPU) back in 1978 so I didn't consider that I'd have any problems with the CCD. Besides, Berry et. al have done an excellent job with the diagnostic software.
I made the power supply from bits that I already had in my junk box, but to save time I bought the rest of the bits straight from University Optics. I assembled the electronics over about 2 weeks of spare time. I encountered only 2 problems.
I usually followed the construction plans in the book, although I devised a much easier method of soldering the CCD chip into the camera body. When powered on it worked perfectly - a triumph more for the detailed instructions in the book than my technique. I think that if I can make it,then so can just about anybody else.
The problem with U2 was that every time I connected the pre-amp board, U2 died. U2 feeds some transistors used as level-shifters. I removed the transistors and replaced U2 and ran the diagnostics again - no problems. One transistor added. Output levels OK, but only just. A second transistor added. A dead U2. My electronics isn't great, but I calculated that it was being asked to sink too much current - an LS chip is only rated for 4ma, I believe. A quick check in the diagnostic software says something to the effect that U2 is being asked to do more than it was designed for, but if it fails then just replace it with one of a different brand because it does usually work. Well, I'd just gone through 3 different brand chips and they all died. A 7414 is a pin-for-pin replacement for the 74LS14 and can sink more current, so I tried one of these. It worked, and continues to work to this day. I faxed University Optics who in turn told Richard Berry who faxed me about the problem and said they would look into it. To this day I don't think anybody else has had this problem, but why mine failed and nobody else's has is a mystery to me. (They suggested things like oscillations on the power rails but a check-out with a CRO showed no problems.)
The windscreen-washer pump that UO advertised as being the best pump for the job failed within a few hours use. By the time I had finished testing the CCD it was acting up, and after only a few nights use on the telescope it was dead. Great! I eventually replaced it with a non-submersible water pump that is used to cycle water in evaporative air-conditioning systems. It cost me A$85 but will probably last forever. It consists of a shaded-pole AC (mains) motor in a plastic housing that sits above the water. An impeller sits in the water and pushes the water out through an opening. It produces enough flow for the system, in fact perfect I would say. Unfortunately, the system isn't sealed and so bacteria can (and do) develop if I leave the system with water in for too long. At the moment I add some ethelene glycol (anti-freeze) to the water, but this isn't ideal for a number of reasons. I may replace it one day. Anybody have any suggestions for a perfect replacement?
Knowing the temperature of the CCD is a good place to start when calibrating the performance of the system. I put a temperature sensor IC on the cold finger when I was building the unit. I use an AD590, which produces a current proportional to the absolute temperature. It requires just 2 wires and is quite insensitive to supply voltage and other problems. A single trimming resistor is all that is required to calibrate the device. It is read on a DVM set to mV range and displays in degrees Kelvin. A circuit diagram is here.
In practice, the cooling system is VERY stable. There are certainly no rapid fluctuations, but I find that as the night cools the chip temperature drops a degree or so every few hours. Not quite enough to worry about, but it means that dark frames done at the start of the night aren't quite right for use a few hours later. I'd probably like a complicated temperature regulation circuit, but it sort of goes agains the KISS principal, and as the variation isn't too much I'm not that bothered.
The cooler typically pulls the chip (alright - the temperature sensor) down to about -30°C when the ambient temperature is +15°C. On the hottest summer nights I can't quite make -30°C, while in winter I easily run it at -35°C. This difference means roughly a factor of 2 in dark current which means that summer images aren't quite as good as winter images.
The LDC modification makes an enormous difference to most of the pixels, but there are quite a few hot ones left. Going colder may be one way to solve this. But the next most useful change is in the software - the bias drift control. The fact that the bias level can be relied on to be 100 significantly increases my speed at reducing the data. If you haven't obtained the LDC mods and 245PLUS software then you're missing out on a lot.
I've added a switch that allows me to change the gain of the pre-amp. The gain is set by the ratio of R45 to R43 which are specified in the book as 2K2 and 39K. I changed R43 to be 47K to give a full count (4096) when the chip wells were full. I then added a switch to allow an additional 22K resistor to be inserted into the chain. The original gain setting was 17·73; my new settings are 21·4 & 31·5, almost double the original. In practice, I almost never use the low gain setting.
It should be noted that the book originally specified setting the ref value to about 75 and therefore the reset value was up around 800. But the reference level was only used for the internal double-sample mode and so this was artifically truncating the dynamic range of the system. It is now recommended that the reset value be set to around 100 (which means that ref will be 0) and that R43 be increased to restore the full dynamic range. Rather gratifyingly, I had already been playing around with gain settings and was in the process of composing a message to Richard Berry about the ref/reset values when the LDC mods came out.
After construction, everybody discovered that the reference level wasn't stable. This is due to Q7, the buffer transistor between the CCD and the pre-amp, being very temperature sensitive. There were many methods devised to control this problem (including wrapping old socks around the CCD housing!) but nobody seemed to do the obvious - move Q7 inside the CCD camera body. This requires adding a link on the circuit board and moving one wire inside the camera to the base of Q7, while the other legs of Q7 go to pins 1 and 2 of J2. While this simple change doesn't solve the problem entirely (the op-amp is also temperature sensitive) it does significantly reduce the problem. Combined with the 245PLUS software drift subtract, I don't have a problem any more.
Here is a view of the inside of the camera, showing the location of Q7 and the temperature sensor. I also added some insulation inside of the camera which should slow the influx of heat into the housing from the outside world. Here is a view of the camera with the insulation in place, just before the camera is closed. The inside is flushed with nitrogen immediately before sealing.
One of the nice things about the Cookbook is that the control program will run on any IBM-PC from an XT up - it doesn't need anything fancy. The CCD is controlled from a standard parallel port, the PC driving all clocking lines and controlling timing. I am luck enough to live under dark skies and have my telescope mounted permanently in an observatory so I don't need to use a portable computer. I managed to obtain an old '286 and find that it is more than ample for the task. One of the nice things about progress is that many places upgrade their computer systems frequently and a clunky old '286 is pensioned off for a song. I managed to get a 12MHz Compaq PC, with built-in VGA graphics, 110Mb hard disc, 3Mb memory, and a swag of other bits for A$151. I had to buy a 3·5-inch floppy and a monitor (monochrome is all that is necessary) to make it fully functional, but that's all I use.
I'd really like to get the chip colder. Putting it in a vacuum would help considerably (probably worth ~10°C), and then adding a second Peltier stage for another ~20°C.
I'd move the CCD chip as far forward as possible to help with mounting filter wheels and viewing accessories. This isn't a problem with the average SCT where the focal surface can be pushed back to accommodate these accessories, but for Newtonians every bit helps.
home back to ATM to images
Page last updated 1996/10/28