bender@oobleck.Eng.Sun.COM (I want to be eating rich soup in another town) (03/28/91)
I'm working on a serial communications board here and experiencing a crosstalk problem on my receive data lines. The setup is an SBus card plugged into a SparcStation, connected via a 96-pin round, single outer shield cable or 3m length, to a metal patch panel with 9 DB-25 connectors. What happens is that due to the way the wire pairs are twisted in the 3m cable between the card and the patch panel, I get crosstalk on a serial port's receive data line if the other wire in the twisted pair has any serial traffic on it. The line receivers that we are using are the 5180's, octal line receivers, currently sourced to us by Unitrode, but potentially we could also be buying them from Signetics. I'm seeing about 100mV p-p "noise" on the receive data line that's unterminated at the DB-25 end. For those of you familiar with the 5180 part, we have the threshold level set to 160mV on the non-inverting input (signal comes in on the inverting input). We also have the fail-safe pins tied low. I can make the problem go away (really, just mask the problem, because the noise is still there, it's just not causing the receiver to switch) by tying the fail-safe pins to Vcc rather than ground, or by raising the input threshold level to around 210mV. I understand why each of these fixes work; ideally I'd like to tie the fail-safe pin high AND raise the input threshold level to around 800mV..1V, but the hardware engineer on the project (not me) doesn't think raising the input threshold above 160mV is a good idea. The reasons that he gives are that since the ultimate end-user could use something like a laptop computer with 1000-feet or more of cable (in RS-423 mode) to drive our line receivers, if we raise the threshold much above 160mV we stand the chance of not seeing the signal switch. Well, I personally think this is crap. The RS-232 spec shows a +/-3V deadband around 0V, and the RS-423 spec I think lowers that to +/-1.5V around 0V. Given that someone, somewhere will try to drive one of our serial ports with 1000-feet or more of cable from a laptop, what's a reasonable figure for the voltage drop in that much cable, say 22 or 24 gauge stuff that you can get from Belden? The real question I have is that will I have any problems raising the input threshold to 800mV or so? Any help on this would be appreciated... also, I have heard that there are electrical differences between the Unitrode 5180 and the Signetics 5180, and that these differences make the selection of input threshold and important factor to consider when designing a system that could possibly use either the Unitrode or Signetics part. thanks, mike the device driver writer doing double-duty as h/w engineer p.s. I also know that the idea fix would be a combination of what I mentioned and also respecifying the pair twisting of the cable. -- Won't look like rain, Won't look like snow, | DOD #000007 Won't look like fog, That's all we know! | AMA #511250 We just can't tell you anymore, We've never made oobleck before! | MSC #298726