Post by thread:[PIC] revising analog input protection, 18F4680

I'm reviewing an ugly ADC input circuit which so far has worked very adequately. An op amp's output signal (nom. 0-5v, but could rise to 12v in fault condition) is clamped for safety by a 680R/5V1 zener diode combination. So the voltage into the PIC ADC pin is externally limited to 5v1 and the current from the opamp is externally limited to 10mA if the op amp ever hits 12v. Possibly due to some 5v1 zener diodes going obsolete and being replaced by a close alternative, I find that some PIC ADC inputs are being dragged down by the presence of the zener diodes, as the incoming voltage approaches 5v. This seems to suggest that the current batch of diodes are turning on earlier than those previously specified. Removing the zeners seems to make the problem go away. I'm then left with the worry that merely limiting the over-voltage ADC input current to 10mA may be risky, even though the ADC inputs are very unlikely to rise above 5v during the service life of the equipment. Questions: a) is just a current limiting resistor of 680R (for 10mA) sufficient due diligence, for protecting the PIC ADC input clamping diodes if the source voltage rises to 12v? b) is relying on limiting the internal protection diode current, asking for SCR latchup? Ponderings appreciated, Matt

An extremely good alternative to zener diodes is a clamp regulator such as a TL431(which I tend to recommend to all and sundry). The lower voltage TLV431 has some advantages. This has a much sharper "knee" than a zener and can be obtained in 0.5%, 1% & 2% accuracy versions. "Programming" requires 2 resistors. If you are committed to an existing PCB then a small add in PCB could be made to insert where the zener no longer is. My constant refrain is that any overvoltage risks unexpected PIC operation - but I'll admit that a genuine 5.2V and maybe 5.3V on a genuine 5V supply is statistically not too too likely to cause problems. A TL431 or TLV431 would allow you to clamp very close to 5V. 680r alone probably meets absolute max ratings (see relevant data sheet) BUT certainly violates "normal operating conditions". You can be utterly certain that some PICs will perform bizarrely if you inject 10 mA (or even 1 mA ) via protection diodes. As long as data sheet guarantees non destruction and as long as your system does not mind processor going insance when Vin > ~= 5V then no zener at all is an OK choice. A small Shottky diode from Vin to Vcc = 5V clamps wellish but Shottky leakage may be annoying if you have very high temperatures. The Schottky drop also notionally exceed sthe limits that I preach about but often does OK in practice. Very good is a small Schottky from the middle of a split Rin to Vcc so you clamp to about 5.3V or less with a largish resistor between V and pin. Also violates spec but statistically better. Russell On 6 September 2011 21:07, Matt Rhys-Roberts <spam_OUTmattTakeThisOuTnu-ins.com> wrote: {Quote hidden}> > I'm reviewing an ugly ADC input circuit which so far has worked very > adequately. An op amp's output signal (nom. 0-5v, but could rise to 12v > in fault condition) is clamped for safety by a 680R/5V1 zener diode > combination. So the voltage into the PIC ADC pin is externally limited > to 5v1 and the current from the opamp is externally limited to 10mA if > the op amp ever hits 12v. > > Possibly due to some 5v1 zener diodes going obsolete and being replaced > by a close alternative, I find that some PIC ADC inputs are being > dragged down by the presence of the zener diodes, as the incoming > voltage approaches 5v. This seems to suggest that the current batch of > diodes are turning on earlier than those previously specified. > > Removing the zeners seems to make the problem go away. I'm then left > with the worry that merely limiting the over-voltage ADC input current > to 10mA may be risky, even though the ADC inputs are very unlikely to > rise above 5v during the service life of the equipment. > > Questions: > a) is just a current limiting resistor of 680R (for 10mA) sufficient due > diligence, for protecting the PIC ADC input clamping diodes if the > source voltage rises to 12v? > b) is relying on limiting the internal protection diode current, asking > for SCR latchup? > > Ponderings appreciated, > > Matt. >

I had a similar issue some time back ... 15V source/environment, voltage divider to drop that to <5V at A/D input, 5.1v zener for protection. I too found that the 5.1V zener was altering the signal as it went higher, and it was inconsistent from one diode to another. Up to then I was using a 4.096V reference to the A/D, so I changed that to 2.5V, and altered the voltage divider to divide by 8 (7k and 1k), and removed the zener altogether. With this, I get good resolution still, and the input can safely go to 40V. Cheers, -Neil. Quoting Matt Rhys-Roberts <.....mattKILLspam@spam@nu-ins.com>: {Quote hidden}> I'm reviewing an ugly ADC input circuit which so far has worked very > adequately. An op amp's output signal (nom. 0-5v, but could rise to 12v > in fault condition) is clamped for safety by a 680R/5V1 zener diode > combination. So the voltage into the PIC ADC pin is externally limited > to 5v1 and the current from the opamp is externally limited to 10mA if > the op amp ever hits 12v. > > Possibly due to some 5v1 zener diodes going obsolete and being replaced > by a close alternative, I find that some PIC ADC inputs are being > dragged down by the presence of the zener diodes, as the incoming > voltage approaches 5v. This seems to suggest that the current batch of > diodes are turning on earlier than those previously specified. > > Removing the zeners seems to make the problem go away. I'm then left > with the worry that merely limiting the over-voltage ADC input current > to 10mA may be risky, even though the ADC inputs are very unlikely to > rise above 5v during the service life of the equipment. > > Questions: > a) is just a current limiting resistor of 680R (for 10mA) sufficient due > diligence, for protecting the PIC ADC input clamping diodes if the > source voltage rises to 12v? > b) is relying on limiting the internal protection diode current, asking > for SCR latchup? > > Ponderings appreciated, > > Matt. >

part 1 349 bytes content-type:text/plain; charset="us-ascii" (decoded quoted-printable) Hi Matt, Attached is a circuit that I have used many times. This one is for eight inputs and protects against positive swings. It will work for +5 or lower, just attach the diode anode to the same supply as the micro. Cheerful regards, Bob -- http://www.fastmail.fm - Send your email first class part 2 12459 bytes content-type:image/gif; name="input_protect.gif" (decode) part 3 181 bytes content-type:text/plain; name="ATT00001.txt" (decoded base64) -- http://www.piclist.com PIC/SX FAQ & list archive View/change your membership options at mailman.mit.edu/mailman/listinfo/piclist

I'm using a derivitative of bob's circuit to produce a rail to clamp to. That is, two forward diodes and a transistor arranged as in bob's schematic to produce vclamp. The second diode is to compensate for the additional drop... I use BAT54S'es as clamps, so I just use a single 54S ignoring the center pin to get the two diodes.... Then a whole bunch of bat 54S'es used as normal clamps to my vclamp rail and ground. -forrest On 9/6/2011 2:09 PM, Marcel Duchamp wrote: {Quote hidden}> On 9/6/2011 9:38 AM, Bob Blick wrote: >> Hi Matt, >> >> Attached is a circuit that I have used many times. This one is for eight >> inputs and protects against positive swings. It will work for +5 or >> lower, just attach the diode anode to the same supply as the micro. >> >> Cheerful regards, >> >> Bob >> >> >> >> > I like your idea, Bob. > > It might be slightly improved by using 9 transistors instead of 8 plus > the diode in order to better match the Vbe. And if possible, a > transistor array would probably have very close Vbe matching. None of > this usually matters though.

On Tue, Sep 6, 2011 at 2:21 PM, Forrest W Christian <forrestcKILLspamimach.com> wrote: > I'm using a derivitative of bob's circuit to produce a rail to clamp to. > > That is, two forward diodes and a transistor arranged as in bob's > schematic to produce vclamp. The second diode is to compensate for the > additional drop... I use BAT54S'es as clamps, so I just use a single > 54S ignoring the center pin to get the two diodes.... You compensate a PN-junction drop with a Shottky diode? Regards, Mark markrages@gmail -- Mark Rages, Engineer Midwest Telecine LLC .....markragesKILLspam.....midwesttelecine.com

On Tue, Sep 6, 2011 at 5:32 AM, RussellMc <EraseMEapptechnzspam_OUTTakeThisOuTgmail.com> wrote: > An extremely good alternative to zener diodes is a clamp regulator > such as a TL431(which I tend to recommend to all and sundry). The > lower voltage TLV431 has some advantages. > > This has a much sharper "knee" than a zener and can be obtained in > 0.5%, 1% & 2% accuracy versions. "Programming" requires 2 resistors. > If you are committed to an existing PCB then a small add in PCB could > be made to insert where the zener no longer is. > I really like the TL431. It's dirt cheap. The TL431 datasheet says that it needs a minimum of 1mA cathode current. This might mess up A/D accuracy if using a higher value series resistor. The TLV431 is more acceptable, at ~55uA cathode current. These values probably vary slightly depending on manufacturer. -- Martin K

> I really like the TL431. It's dirt cheap. > The TL431 datasheet says that it needs a minimum of 1mA cathode > current. This might mess up A/D accuracy if using a higher value > series resistor. > The TLV431 is more acceptable, at ~55uA cathode current. These values > probably vary slightly depending on manufacturer. Yes. Lower current at regulation dropout is one of the advantages I referred to. The "required" currents are what's needed to get it around the "knee" and into proper regulation. If it's drawing less than that it MAY be in regulation if at about the knee voltage BUT if it's not in regulation it will be drawing less than that. 50 uA at 5V is the equivalent of a 100k load. Current falls semi linearly with volateg for several volts below that so looks roughly like 100k load from say around 2 V up. You can get a much sharper knee by adding one extra transistor (bipolar or FET) and using the TL431 to operate that. Only load then below clamp voltage is the sense divider. Russel