I've also used Testpoint, and it's a great program. It would probably be able to handle such an ISP system , and that's the way I would approach the problem too. It would be a complex thing to produce. Commercial production would probably have to go with the $100,000 solution you mentioned. Got any names of manufacturers? I've goofed around with "assuming" the temperature in the factory and calibrating temperature sensitive equipment, and got burned bad. I finally made a test jig that blows a stream of ambient air at a temperature sensitive component, and simultaneously checks a "known good" component at that same temperature in the same airstream, thus getting around any complex calculations. This was not an ISP system, however, just a go-nogo tester. -----Original Message----- From: Craig Lee To: Lawrence Lile Date: Tuesday, May 04, 1999 9:27 AM Subject: RE: In System Calibration[OT] >Mr. Lile, > >Is this in a temperature controlled factory? If it is, you can simply >assume ambient temperature and read the thermistor circuit. The initial >value you receive will correspond to the ambient temperature. > >Typically the thermistor curve doesn't change very much,(please refute if >your experience has been different), so you can use your measurement to >'get on track' so to speak. > >If you need better calibration than that, you could build yourself a >standard that you place in close proximity to your thermistor circuit. >Then you would need some way to vary the temperature around your unit. >On each degree, or fixed degree change, the standard would request a >measurement from your unit, do a compare, and return to you the correct >factor. > >So the real trick is if you are using OTPs without some non-volatile >storage. You really need something. Perhaps one of those cheap 3-legged >I2C jobs that store a few bytes, etc. > >If you have lots of testpoints, you could do a crude ICT(in circuit test) >of the resistances with a voltmeter, interpolate some constants, and >program the OTP. However, I'd find this even more painful and cost >intensive due to the fixturing. > >As far as commercial equipment, you could purchase an ICT machine for >something like $100,000. Then purchase custom fixtures for each board >at about $10,000 each. Then you need to maintain it at about $50,000 >per year. > >I am doing some test development using a product called Testpoint and >a PCI multi-i/o card. This card measures resistance, current, voltage, >frequency, etc. Test engineering is my primary business, so if you >want me to develop a solution for you, send me a note privately. > >Craig > >> -----Original Message----- >> From: pic microcontroller discussion list >> [mailto:PICLIST@MITVMA.MIT.EDU]On Behalf Of Lawrence Lile >> Sent: May 4, 1999 7:56 AM >> To: PICLIST@MITVMA.MIT.EDU >> Subject: In System Calibration >> >> >> I've been dreading a project that may require in-system >> calibration. Here's >> the scenario - I'm using some cheap, poor tolerance thermistors and >> potentiometers in a controller. I have a number of these built up, and >> found their behavior changed quite a bit from unit to unit, ostensibly >> because of tolerance variations. (I'm not really sure) >> >> I could tighten the tolerance by spending money, but the circuit must be >> absolute minimum cost. So I'm imagining a system that: >> >> 1. Does a final check on resistances in the circuit (and maybe a quality >> control check too?) >> >> 2. Calculates two or three constants based ont he resistances in >> the circuit >> >> 3. Creates the proper code (would it have to COMPILE or just modify the >> right bits in a hex file?) >> >> 4. Squirts the program into a PIC either in- circuit or just before >> insertion >> >> It sounds really complex and capitol-intensive. >> >> Has anybody worked with these before? Is there commercial equipment that >> does this task automatically? >> > >