esd diodes as diodes
in [Design]
|
Prev: Dumbed down consumer electronics: Adding DTV channels
Next: The Payback for Political Correctness...
From: Paul Keinanen on 6 Aug 2010 04:12 On Thu, 05 Aug 2010 18:11:47 -0700, John Larkin <jjlarkin(a)highNOTlandTHIStechnologyPART.com> wrote: >On Thu, 5 Aug 2010 04:39:59 -0500, "Tim Williams" ><tmoranwms(a)charter.net> wrote: > >><miso(a)sushi.com> wrote in message >>news:37c28984-3697-42e3-b1e9-30edbfe39f3b(a)j8g2000yqd.googlegroups.com... >>> But you have to provide two new power rails? This isn't a passive >>> solution. At some point you are just better off buying the protection >>> diodes, probably from both a cost and reliability standpoint. >> >>What's hard about this? I've been doing this for years, I hope it's not >>novel? >>http://myweb.msoe.edu/williamstm/Images/Input_Protection.png >>Clamp diodes unmarked, substitute as desired. > >What's hard is that I'd need 256 low-leakage diodes. That's a lot of >parts. Remember to use diodes in non-transparent packages. A protection diode in a transparent package will make a nice light meter :-).
From: Tim Shoppa on 6 Aug 2010 06:57 On Aug 3, 6:09 pm, John Larkin <jjlar...(a)highNOTlandTHIStechnologyPART.com> wrote: > I want to clamp some signals before applying them to an analog mux, so > that customer overloads don't blow through the mux and trash other > channels. One obvious way is series resistors and clamp diodes. > > It occurred to me that the cheapest way to get pairs of low-leakage > clamp diodes is to use the esd diodes on some really cheap IC, like a > cmos AND gate or something. Has anybody done this? I'm going to state the problem a little more generally. Signal processing power per square inch of PCB space is simply ginormous these days. A chip the size of a fingernail can (like John's analog mux) marshall a hundred analog inputs. The issue is, (correct me if I'm wrong John), is that the input signal conditioning to protect against even mild overloads often takes many more parts and (this is the kick in the teeth for me) PCB space. In the multi-hundred to multi-thousand channel systems I deal with, we have been dealing with the problem at the cabling level. Signals come in (for example) on 60-conductor twist-and-flat differential pairs. We have a standard PCB that terminates the twist-and-flat, applies signal conditioning, and then comes into the marshalling stage. We just copied this from other companies in the 80's. But even the above doesn't scale sufficiently. It was kind-of OK for the 80's but even the wiring systems have gotten much smaller since then. IMHO we should be moving the signal conditioning to the place where the analog voltages are being made, do the A/D there, and haul back everything over some high-speed optical or differential copper bus rather than do a "home run" of every analog voltage in the laboratory or car back to some common rack. I have succeeded in some cases with this but it'll take a broad systemic redesign, not some quick and dirty thing I can do at the data acquisition end. Am I too far off John?
From: John Larkin on 6 Aug 2010 08:13 On Fri, 6 Aug 2010 03:57:03 -0700 (PDT), Tim Shoppa <shoppa(a)trailing-edge.com> wrote: >On Aug 3, 6:09�pm, John Larkin ><jjlar...(a)highNOTlandTHIStechnologyPART.com> wrote: >> I want to clamp some signals before applying them to an analog mux, so >> that customer overloads don't blow through the mux and trash other >> channels. One obvious way is series resistors and clamp diodes. >> >> It occurred to me that the cheapest way to get pairs of low-leakage >> clamp diodes is to use the esd diodes on some really cheap IC, like a >> cmos AND gate or something. Has anybody done this? > >I'm going to state the problem a little more generally. Signal >processing power per square inch of PCB space is simply ginormous >these days. A chip the size of a fingernail can (like John's analog >mux) marshall a hundred analog inputs. > >The issue is, (correct me if I'm wrong John), is that the input signal >conditioning to protect against even mild overloads often takes many >more parts and (this is the kick in the teeth for me) PCB space. > >In the multi-hundred to multi-thousand channel systems I deal with, we >have been dealing with the problem at the cabling level. Signals come >in (for example) on 60-conductor twist-and-flat differential pairs. We >have a standard PCB that terminates the twist-and-flat, applies signal >conditioning, and then comes into the marshalling stage. We just >copied this from other companies in the 80's. > >But even the above doesn't scale sufficiently. It was kind-of OK for >the 80's but even the wiring systems have gotten much smaller since >then. IMHO we should be moving the signal conditioning to the place >where the analog voltages are being made, do the A/D there, and haul >back everything over some high-speed optical or differential copper >bus rather than do a "home run" of every analog voltage in the >laboratory or car back to some common rack. I have succeeded in some >cases with this but it'll take a broad systemic redesign, not some >quick and dirty thing I can do at the data acquisition end. > >Am I too far off John? Well, my narow problem is to prevent customer over-voltage inputs from blowing through my analog multiplexers, and do that without putting a couple of hundred parts on my board. But the issue you bring up is real. Running fire-hose equivalent cables to a central location is expensive. We've been considering scattering i/o out in the field, as small boxes with one ethernet or maybe PoE cable running into the system. That has its problems too, like if the boxes would wind up in a temperature/altitude chamber near the DUT, sixty feet off the ground. But that little ethernet box still has ICs that need overload protection! Most of our VME modules have a calibration connector and a relay per channel, so our customer can switch every channel to a traceable dvm/source and verify calibration before and after every test run, without disconnecting field wiring. That works well with all the gear in one rack, but would be very messy to attempt with distributed i/o. John
From: Didi on 6 Aug 2010 09:57 On Aug 6, 3:13 pm, John Larkin <jjlar...(a)highNOTlandTHIStechnologyPART.com> wrote: > ... > Most of our VME modules have a calibration connector and a relay per > channel, so our customer can switch every channel to a traceable > dvm/source and verify calibration before and after every test run, > without disconnecting field wiring. That works well with all the gear > in one rack, but would be very messy to attempt with distributed i/o. Why do you say it would be messy? Replacing say 32 analog cables with a 10/100 Ethernet link should only make it easier from where I look at it (clearly not from the same point as you). I ask because I was asked recently about a tiny (50x100mm) ADC board with 16 inputs, to etherner, the motivation being cabling. [Nothing came out of it but then the inquiry was from Pakistan, either the floods got them or it was one of the so many "first ask then think" inquiries coming from these parts of the world :-) ]. Dimiter ------------------------------------------------------ Dimiter Popoff Transgalactic Instruments http://www.tgi-sci.com ------------------------------------------------------ http://www.flickr.com/photos/didi_tgi/sets/72157600228621276/
From: Joerg on 6 Aug 2010 10:01
Jim Thompson wrote: > On Thu, 05 Aug 2010 16:56:36 -0700, Joerg <invalid(a)invalid.invalid> > wrote: > >> Jim Thompson wrote: >>> On Thu, 05 Aug 2010 14:46:44 -0700, Joerg <invalid(a)invalid.invalid> >>> wrote: >>> >>>> Grant wrote: >>>>> On Thu, 05 Aug 2010 09:36:05 -0700, John Larkin <jjlarkin(a)highNOTlandTHIStechnologyPART.com> wrote: >>>>> >>>>>> On Wed, 04 Aug 2010 16:10:24 +1000, Grant <omg(a)grrr.id.au> wrote: >>>>>> >>>> [...] >>>> >>>>>>> You can run the mux chip at 6V? Clamp to 5V at the front for mux, >>>>>>> look after the ADC separately on other side of mux? >>>>>> I can run DG408-type mux chips at +12. One recent board, I ran them at >>>>>> +-17, by bootstrapping 5-volt DC-DC converters on top of my +-12 >>>>>> rails. A whole nother story. >>>>> Used to be +/- 10V signals on +/- 15V rails was plenty for everyone ;) >>>> Same with cars. Nowadays everyone "needs" 150-200 horses to commute to >>>> work. Back when I was a kids dad pulled a 1.2-ton travel trailer with a >>>> 55HP car, family of five in the car, lots of luggage, food, a tent, plus >>>> an inflatable boat inside the trailer. >>>> >>>> >>> [snip] >>> >>> On level ground, with a tail-wind ?:-) >>> >>> I prefer 345hp myself... 25+MPG at a steady 80MPH :-) >>> >>> As for MUX protection, how about simple-minded... >>> >>> http://www.analog-innovations.com/SED/MUX_Protection.pdf >>> >>> Diodes, transistors, resistors, capacitors: users-choice. >>> >> The right one looks quite busy WRT to SMT placement costs. > > Which is why I asked about using arrays early on. > >> The left one >> is nice but you might want to up the cap to 10uF or so in case the >> mother of all spikes comes waltzing in. > > I don't believe spikes were the issue... simple over-range protection > for a MUX. > Depends on far "range" ranges :-) -- Regards, Joerg http://www.analogconsultants.com/ "gmail" domain blocked because of excessive spam. Use another domain or send PM. |