keithe@tekgvs.LABS.TEK.COM (Keith Ericson) (04/20/89)
In article <7970@pyr.gatech.EDU> is813cs@pyr.gatech.edu (Cris Simpson) writes: >I am developing an interrupt driven data collection system running >on a Compaq 386-25. It works fine, except that I am getting >tripped up by an unknown interrupt, which makes my latency high (1.8ms). >It occurs at about 230Hz. Can anyone suggest what this might be? (I'm responding to offer some help, and not to generate flames. So back off if you feel so inclined, eh?) We have been told by sources we hold to be reliable that the AMD version of the interrupt controller (the 8259 as I recall) can suffer from a problem with the internal pull-up resistors. In the process of doing a die-shrink these resistors became so high in value as to be inneffective in preventing noise from being interpreted as valid interrupt inputs. We have developed a work-around with a SIP resistor package to add 330-ohm (I think it is) resistor to +5 to stiffen up the pull-up resistors on these interrupt controllers. You can probably come up with something as well. If you know enough to figure it out from this description you're likely competent enough to do it. If you _can't_ figure it out then don't try it and don't ask for more details - you risk really screwing up your computer. kEITHe
geoffs@gssc.UUCP (Geoff Shapiro) (04/22/89)
In article <7970@pyr.gatech.EDU> is813cs@pyr.gatech.edu (Cris Simpson) writes: >I am developing an interrupt driven data collection system running >on a Compaq 386-25. It works fine, except that I am getting >tripped up by an unknown interrupt, which makes my latency high (1.8ms). >It occurs at about 230Hz. Can anyone suggest what this might be? >I'm getting 6000 ints/sec except when I'm in the mystery int? > What kind of hardware are you using to do the data collection? Perhaps an A-to-D card which is generating interrupts at a frequent rate? The reason that I ask is because I have seen similar behaviors here when running cards in the bus which are generating too much noise on the interrupt request lines of the bus. To understand this you must understand the 8259 and how it handles noisy interrupt input lines. A little known feature of the 8259 is that when noise generates a pulse on one or more of any of the interrupt input lines and this pulse is not of the length or level required for recognition as an unqualified interrupt source, the 8259 will internally trigger an IR7. Normally, in the DOS system IR7 is vectored to a null interrupt service handler and therefore is blown off harmlessly except for the degradation of system throughput when the lines are real noisy. I would check this out as a possible source of your problems... ...!{tektronix!verdix}!sequent!gssc!geoffs Geoff Shapiro (503) 641-2200 Graphic Software Systems * This space intentionally * 9590 S.W. Gemini Dr. * left blank. * Beaverton, OR 97005