| ??? 08/23/06 04:53 Modified: 08/23/06 04:55 Read: times |
#122822 - Driving cables is nasty!! Responding to: ???'s previous message |
Richard said:
They won't tolerate multi-meter-long cables, but they do seem to work. I never had any trouble with the ATMEL's original "ISP cable". It's true that the used 74VHC244 buffer isn't TTL compatible, but there's no cable at all to the printer port, because the circuit is soldered directly on a DB25 plug. So, this incompatibility will not hurt in the very very most cases. But, imagine what happens, if someone adds a cable here... Driving cables is a very very nasty thing. Many people think that a cable is something which is many meters long, so will only cause trouble, when I try to connect a 4m long flat ribbon cable, for instance. But that's wrong! A cable is something, where a simple connection turns to behave like a transmission line, means where an arrangement no longer can be explained by electrostatics, but where electrodynamics takes place, means where a connection turns into a transmission line. In a transmission line wave packets are travelling, producing echoes and ringing at all kinds of inhomogenities. So, not using the proper series termination and not using shielding wires between the signal wires will make the signal become totally distorted, shoing stair case profiles, heavy ringing, undershots, etc. This the more, the faster the driving buffer is. One unsane effect of electrodynamics is, that the load which originally was only a bit stray capacitance and stary inductance with a transmission line becomes nearly a short circuit! A 100Ohm transission line looks like a load of 200Ohm with proper series termination! This is the reason, why you cannot drive cables by a standard output of a CMOS chip, why buffers with high current outputs are fabricated, why in certain situations even a 74STTL-buffer should be used, etc. So, it can be one simple resistor being omitted somewhere, which makes a functioning circuit to fail. It can be the choose of the wrong cable or even only the wrong header pinning to make a proper functioning circuit to totally fail. I recommend to everyone, to sit down and look with scope what at the end of a flat ribbon cable arrives from the original nice looking signal. And you will no longer be surprised, why so many "one chip solutions" don't work reliably!!! By the way, for 74HCMOS each connection being longer than 20cm is a transmission line! With the todays fast chips like 74VHCMOS and 74LVCMOS already a 10cm connection turns into a transmission line... Kai |
