| ??? 08/27/06 15:03 Modified: 08/27/06 15:10 Read: times |
#123138 - Cables... Responding to: ???'s previous message |
Jan said:
I thought about it more during the night and I concluded, that for the "bomben- und idioten-fest" version we should ban the "in the DB25 shell" version. Need not to be mounted WITHIN the DB25 shell. You can still solder the DB25 plug to the printed circuit board very easily. If you allow a cable here, then insist on a "printer" cable, which is short, is a flat ribbon cable and has shielding ground wires between the signal lines!! If we finally have to struggle with capacitive cross coupling within an unsuited and superflous printer cable, we should forget the idea of letting a newbie fabricate his own "ISP cable"! Jan said:
In fact, we should strongly discourage using long cables from "the circuit" to the target. Your posts show that this IS a concern. As I recommended series termination, this is not so much of concern compared to circuits omitting them. Nevertheless, it's always good to keep cables as short as possible. More than recommending the newbie to keep the cable short, we should focuse on the importance of having "grounded" shielding wires between the signal lines!!! So, changing the cable or even only the header pinning is a huge NO-NO! Jan said:
However, I still believe it might be a good way. The AT89Sxx inputs, although not exactly Schmitt-trigger, certainly exhibit enough hysteresis to accomodate a slow ramp - it's still the SAME input as used for normal operation, and no slew rate is specified for those in the datasheets, anyway. It's not so much the slew rates of these comparators I concern about, but the propagation delay time! Remember, Atmel's ISP software let the clock line toggle and recieves the MISO signal afterwards. As long as we do not exactly know the delay between the clock signal toggling and the MISO signal recieving we should not add additional delays (neither for the clock signal nor the MISO signal) which are much higher than the original circuit. Some tens of nanoseconds is ok, but hundreds of nanoseconds or even microseconds? Jan said:
Today, I'd expect CMOS inputs with heavy protection and a reasonable amount of hysteresis. I assume the same, but can you rely on it?? Kai |
