Voltage divider with diode acting strange ? 2

[Chancy99]

I'm missing something fundamental here ... I have a typical voltage divider circuit, consisting of two resistors, a 39k and a 22k, plus a 10k pot to adjust things. The 39k is the top leg for input, the 22k and 10k pot are the bottom leg to GND. When adjusted for about 25k, it will divide 12V in at the top of the 39k to about 5V at the junction of the 39k/22k.

Works fine. Stick in 12V, get 5V out.

Now I add a 1N4625 5.1V zener, to make sure that the input voltage can't surge up and burn things out. As I understand it, the zener shouldn't conduct at all until 5.1V.

But it seems to be acting as another resistor to GND. WIthout the zener, I adjust it to show a clean exact 5.00V. Add the zener in, and the output drops to 4.61V. I can increase the input voltage to get 5.00V out of the divider - I have to go to 16.4V.

With 16.V input, if I remove the zener, the output jumps from 5.00V up to 6.2V.

I know this is going to be a DOH type slap the forehead deal ...

D.

 

[peebee]

You've got some big resistances in there, so a rather small change in current will translate to a rather large change in voltage.

You're measuring a change of about .5 volts across a 39k resistance. As V/R=I, .5/39k = .012mA. That's tiny. I'm not that familiar with the performance of zeners, but I'm surprised that the leakage current would actually be so small.

There's not much you could do about the leakage current through the zener (unless you could find a zener that performed better, which would probably not be worth the effort or time or cost -- maybe I'm wrong there. . . .).

I'd think your best bet would be to increase the size of the resistors.

Just curious -- what are you using this circuit for? It wouldn't seem to be a particularly effective power supply or voltage reference. . . .

 

[IRstuff]

Seems to match the datasheet specs for the zener.

It has 10 uA nominal reverse leakage.

(10uA+180uA)*39K ~ 7.4V.

TTFN

 

[buzzp]

The zener will conduct, some, before it reaches the zener voltage. The tolerance of this voltage is likely +-5%. At the zener voltage, the zener will be fully conducting. Your best bet is to increase the zener voltage to something in the 6V area otherwise the 5.1V zener will be conducting some current most of time, consuming more power. As the zener conducts more, the impedance decreases, causing the current to go up thereby changing the output voltage you desire.

The 6.X volt zener should work fine for protecting your circuit (at least it better). Otherwise, you would be better off using like a 7805 regulator since they will not hog so much power. A zener is a power hungry animal especially for regulation. If power is not a concern, get rid of all the resistors for the voltage divider and just use a larger (higher wattage) zener for regulating the voltage to 5.1V. Although, the 7805 will regulate much better than a zener. Also look at the adjustable regulators if the 5V might need to be changed to something else in the future. (Oh BTW, I am assuming all this discussion is for DC voltage).

 

[Chancy99]

Yup - that's it. I reread the datasheets, and picked up on the reverse leakage. Being so small, I've never really taken much notice of it.

This is the front-end to an automotive measuring project. The divided voltage goes into an opamp voltage follower to buffer the signal, then into the ADC inputs of an AVR.

Given that there can be some major voltage swings in a car, the zeners are kind of a last-ditch defense. If something huge develops across the 39k, the zener should clamp it down tight.

I'll change them out for 6.2V zeners - 1N753A. The internal diode clamps in the AVR should be able to handle the extra 1.2V ...

I was thinking of using a Littelfuse SP723 TVS clamp on the signal before it hits the divider. I don't know how much that will contribute though, since if there's a load dump, the 12V supply will leap up as well - where would the TVS clamp to ?

Given the horror stories you always hear of the automotive environment, I want to make this as bulletproof as possible.

peebee - you said to *increase* the size of the resistors ?

D.

 

[peebee]

By the way, unless you have a super-accurate meter, you're not really getting 5V at your output anyway.

Seems to me most multi-meters are about 10k-ohms per volt (can anyone confirm?). That's 50k in parallel with your 29k (22+pot). That means once you've adjusted to precisely 5v output, the instant you remove the meter it will rise by about another volt.

Any load above about 0.005 mA you add to the output will similarly drop the voltage. Hence my statement about this not being a particularly effective voltage reference. Not much you can do with a couple microamps.

 

[buzzp]

Change the voltage divider circuit and use a real voltage regulator such as a 7805. You will be much happier (the input voltage can go up to 30-35V (cant recall)). You could then put a TVS on the input to the regulator so the regulator is not sacrificed when a large transient comes along. Depending on whats on the 5V bus(any reactive loads), you may want one on here too. It will also run much, much cooler.

 

[Chancy99]

buzzp - I'm not looking for a 5V voltage source, I'm looking to measure several signals from the car ECU. Mostly they range from 0-12V, which needs to be translated down to 0-5V for the ADC in the AVR. There's actually 5 channels of this setup ...

We want to keep the resistances high so as not to load the signals we're measuring. The opamp voltage follower after the divider is to provide a nice clean low-impedance source to the ADC.

The power supply section is rock-solid clean. +12V and GND come in through inductors, then to an RBO40 TVS clamp, then through a couple of ferrite beads into an AN7705 automotive regulator.

The ADC reference uses an LM336-5.0 fed by another 10uH inductor with 0.1uF cap to GND. Clean and quiet.

Argh - now I have to desolder the 1N4625 5.1V zeners to put in the 6.2V ones. I *hate* desoldering.

D.

 

[peebee]

Re: peebee - you said to *increase* the size of the resistors ?

Oops, typo. As a power engineer, I tend to think of resistors as rated in kW of load rather than ohms. Smaller resistance = bigger load, hence "bigger resistor". Physically, that would be bigger too. But I realize my statement was confusing given this context.

Back to the subject, now what I realize you're trying to accomplish, maybe disregard some of my recommendations above.

I'm not quite sure you want to jump right into replacing the 5.1's with 6.2's either, though.

Please realize I'm way out of my area of expertise here -- others would have better ideas -- but I'm thinking you might want to put (3) 5.1's in series across the whole 12v input, get them out of your voltage divider circuit altogether, they will only add non-linearity.

Decreasing the size of the resistors would do the trick too -- but make sure you don't overload the output that you're tracking. Careful you check the burden imposed by the input to your ADC, too, although I suspect it's low enough that it would not overly impact the accuracy of your circuit. Accurately adjusting your trimpots could be tricky using a voltmeter, for the reasons mentioned above -- you'd do better to adjust to 5v output by looking at the output of the ADC.

But you'd do way better to get advice from somebody else than me!!! When I first read read your question, sounded like you were just tinkering around for fun -- if I'd realized this was rather more serious I probably wouldn't even have responded!!

 

[IRstuff]

I'm not sure that will actually help that much. I couldn't find a datasheet for the 6.2V version, but I suspect that the leakage rate will still be in the same range.

The only way out of that is to buffer the signal first and then run a lower resistance divider.

Another approach might be to measure the error and simply back out the leakage current in the calculation. Since you can treat the zener as a 10 uA current sink, the Thevenin equivalent circuit can be used to back out any measured voltage into the correct input voltage.

TTFN

 

[OperaHouse]

Instead of a zener on the input, put a diode from the resistor tap to 5V zener with its own pull up.

 

[Chancy99]

IRstuff -

That's what I'm doing right now with the first board as built. Works fine. Kind of a hassle since all the calculations are off, so the calibration constants in the code are off, so ....

Another option is to just do away with the zeners completely. With the 39k, the amount of current coming in is very small even with a large transient, and the opamps should be able to handle it fine I should think. The opamps are fed from the main 5V supply, so their output can't go above 5V, so the AVR inputs should be safe.

There is NO room to add another opamp buffer up front for each of the 5 channels

So the input range is 0-16.5V now instead of the original 0-12.75V, mapping to 0-5V for the ADC input. The original conversion factor of 500 to convert from an ADC max count of 255 to 12.85V now changes to 647. That's with keeping the 39k/22k/10kpot setup. I can adjust those to handle the original 12.75 input max if necessary.

Always has to be a gotcha in there somewhere. I breadboarded it all without the zener, since I was using a nice clean lab powersupply. No big transients coming in here, so no need to protect against them on the breadboard. Never dawned on me that it would affect it like that - I was expecting the knee of the curve to be above the 5V max being measured.

Dean.

 

[IRstuff]

But it is. The spec zener current was something like 250 uA. That's a 25:1 margin to the leakage current.

TTFN

 

[Skogsgurra]

Chancy,

The salient words here are "horror stories" used in one of your first postings. These horror storiess are often told to people that do their first automotive designs. And they do not help very much - they just steer focus away from the work at hand and makes you think a lot about protection instead of function.

Yes, it is true that a load dump can produce 80 - 100 V in a 12 V automotive system. But this overvoltage seldom hurts your signal paths. If you have a design where the resistance is somwhere between 10 and 100 kohms, then your maximum current will be between 1 and 10 mA, worst case.

Most internal clamps can handle that with no problems at all. And if there is no clamp you can simply add a diode (1N4148 type) from your mux or A/D input to the V+ rail. The reverse leakage of such a diode is very tiny and won't disturb you at all. You need no clamp for negative excursions for the load dump never produces negative voltage.

The "damage path" for load dumps is more likely to be through the power supply unit. A standard linear regulator usually cannot take more than 40 or 50 V and is the first victim. A switcher can be designed to take a lot more, but is seldom used in low-power interference-sensitive applications.

There is also the temperature range to take care of. A typical "under the hood" environment is among the most hostile places you can put electronics. Extreme cold and extreme hot are words that describe the problem.

I suggest that you take a step back and do a realistic evaluation of the project and the challenges involved. But don't let horror stories derail you. Good luck!

 

[Chancy99]

skogsgurra -

Thanks for the words I hear you about getting all worked up with protection circuitry there.

I'm going to try it with the 6.2V zeners, and if it still has a leakage problem, dump the diodes altogether. My attitude in design here has been to add things in to the layout if they'll fit. Don't have to actually stuff the parts, but at least the footprints are there if they turn out to be necessary.

D.

 

[logbook]

You have overlooked Operahouse’s answer, which was spot-on but evidently a bit too brief to catch your attention

If you take a zener diode and run it at say 75% of its rated breakdown voltage, the current drawn is uncertain to a very high degree. You should not use a zener diode in this condition.

Maybe the data sheet gives a reverse current at 2V and room temperature. This is an almost completely unusable fact. What does it do with temperature? You could guess that it follows a "doubles every 10 degree C" law, but it would be a guess. You could look it up in a text book and you would get the wrong answer because the correct data is not given in text books.

Diodes, including zeners, don’t follow "diode equations" under leakage and reverse bias conditions. Manufacturers fiddle with the processing to get good data sheet specs and anything which isn’t specified isn’t guaranteed. Reverse leakage on ordinary diodes is more like a resistive effect rather than the exponential effect that text books would have you believe.

You can get diodes specified for reverse leakage down to the picoamp level if you want. Use a specified low leakage diode to a suitably biased 4.7V zener and the diode will clip nicely at around 5.3V.

Note: for your multi-input system you only need one zener. All the diodes feed to that one zener so it is not a complicated solution. Bias the zener at 5mA to define its voltage according to the data sheet values.

 

[buzzp]

Now that I am clear on your design. I vote for axing the zeners or any other clamping device and just using high impedance as your savior. I have done this many times in industrial designs and one automotive design. I believe someone else mentioned this as well.

I have seen diodes placed from the input to the A/D to the 5v bus in order to bleed off any excess voltage above ~5.6Volts. I never have put this into practice but seems to be fairly common. Microcontroller manufacturers like doing this within the IC although they do not encourage designers to use these internal diodes as a limited voltage device.

 

[Chancy99]

I'm even wondering how necessary those diodes are, since the inputs are feeding a rail-rail opamp. Also, do we really want to feed that excess voltage to the internal 5V rail ?

Say a 40V spike comes along. The diodes will suddenly push the 5V rail up to 40V or so, which won't do the microcontroller any good. I'm thinking that they should feed into the 12V input, just in front of the regulator.

The trouble with that is that an 11V surge will still be fed right into the opamps. That *should* be OK, since their output can only swing to the rails, 0V and 5V. Might burn out the opamp chip package though.

If anything, perhaps a 16V or 20V zener in front of the voltage divider, rather than the 5V or 6.2V zener between the divider and opamp input.

D.

 

[buzzp]

Chancy,
Your comments are exactly why I have never liked the diode solution for controlling over voltage situations. Just does not seem like a good idea. Although, the power supply regulator should be able to handle this. A bypass cap on the micros supply should take care of any "transients" as far as the micro is concerned. A higher voltage zener is not a bad idea.

 

[melone]

A zener clamp can blow during a reverse battery condition. Don't laugh, you will be surprised how often this will occur. (Especially when the zener is place on the front end of the circuit, i.e. ~20V zener before the resistance.) That is why in automotive applications, a diode clamp is often used / preferred.