11
\$\begingroup\$

Goal

I'd like to design a potentiometer, with a range from \$0\:\Omega\$ to about \$0.3\:\Omega\$, for calibrating the zero-reading of a COTS digital voltmeter. I'd like it to be easily fabricated by any hobbyist and easy to use in practice, while also consistent with the exacto knife construction techniques shown in this video by "Leo's Bag of Tricks":

enter image description here

The potentiometer context is discussed next and shown in the schematic there.

Context

I'm proceeding forward on a simple instrument design to quickly characterize small signal bipolar transistors, both NPN and PNP, for three key parameters: saturation current (\$I_{\small{SAT}}\$), emission co-efficient or non-ideality factor (\$\eta\$), and DC current gain (\$h_{\small{FE}}\$ or \$\beta_{\small{F}}\$):

editable schematic

schematic

simulate this circuit – Schematic created using CircuitLab

or an easier-to-access image

enter image description here

I want to use a common \$9\:\text{V}\$ battery for the project. The JFET inputs are helpful here, and the TL082 device is cheap. And TI's equally cheap H version sports a very low \$1\:\text{pA}\$ bias. There's a lot that can be done to improve the design. And I may need to, if testing experience demands it. But a week into it, and so far, the results are promising.

The COTS panel meters are nice. I've tested all five of those I've received, and they consistently exhibit almost exactly \$1.23\:\text{M}\Omega\$ input resistance, regardless of the supply rail voltage. I tested with a variety of supply voltages and used them to measure a variety of voltages, and all five of them always come out at \$1.23\:\text{M}\Omega\$. The impedance doesn't vary with the signal being measured.

(The panel meters also all burn between \$16\:\text{mA}\$ and \$17\:\text{mA}\$, also independent of the battery source voltage I use. Since they are spec'd to operate from anything between \$4\:\text{V}\$ (though they run more dimly on \$3\:\text{V}\$) and \$28\:\text{V}\$, it would substantially aid the \$9\:\text{V}\$ battery if I added a switcher to produce \$4\:\text{V}\$. But that can be deferred as an improved version, should I decide to care enough about it. I'd have to spend time that I cannot afford right now, digging through the weeds to find something that is designed to work off a \$9\:\text{V}\$ alkaline battery and for a \$20\:\text{mA}\$ compliance. (A quick scan on Digikey and I saw most targeting much higher currents.) Or I could just run the display off of a separate Infinity X1 rechargeable, for example. But I'm joking about that -- being a \$9\:\text{V}\$ alkaline battery project is a requirement, and it is working fine, and the draw requirements are squarely within the targeted devices for such alkaline batteries.)

What I'm doing now for that calibration potentiometer and why I want to change

For now, I've been using about two feet of #30 wire-wrap wire from OK industries. I just make it too long and then cut bits of it off until the meter zeros out perfectly. Then I'm done.

But I expect drift. And I'd like to prepare something that others can consider doing and using. (No one may care. And that's okay, if so. But I'd like to at least smooth the road a bit, just in case.) They should be able to adjust the potentiometer from time to time, when calibrating the meter for zero. (There's a switch position to gain access to the LM4040 reference for gain calibration purposes.) One could just write it down and subtract it every time. But I'm finding it a bit annoying, and I'm sure others may, as well.

The offset I've found in the meters is consistently positive and shows up as between \$3\:\text{mV}\$ to about \$4\:\text{mV}\$. It's not much. But it is several counts of the second least significant digit. And it just looks better to get a nice stable zero out.

For the #30 gauge wire, the "resolution" needed is about one inch. Anything finer than that probably isn't needed. This means a resolution of about \$5\:\text{m}\Omega\$ would be more than enough. Probably \$10\:\text{m}\Omega\$ would be fine. But no worse than that. And it should not feel as though you just can't get it down to the pixel so to speak. It should be easy to zero the meter without dealing with physical hand-jitter, making it too painful a process to get it accidentally "just right."

Obviously, two feet of #30 bare copper with a sliding clip works. Been there, done that. But it would be a lot bigger than the board, and it is hard to tuck it into the project box, later, without all that bare copper touching something awkward.

Finer wire starts to become too easily broken, I think. But I am currently working through ideas like that, which can be simply mounted on the board. A slider bar style mechanism on stand-offs would be okay.

Perhaps consider a way of using the copper-clad surface of the board itself? Or some other idea that can be mounted on said board using the above construction technique and fabricated by a hobbyist.

A serpentine trace going around and around the board came to mind. But even if I could figure out a clip for it, I can't imagine anyone wanting to use their exacto knife for that mission.

I'm looking for construction ideas to explore that would fit well into this specific project, be easy to make and use, and serve the purpose.

The device is working very nicely. But this meter zeroing is my current problem in front of me, and I'm open to ideas to try.

Motes

It is a requirement that this be a \$9\:\text{V}\$ alkaline battery project. That's immutable.

For the current version-0 project, it is also a requirement that parts count is kept as low as possible, consistent with maintaining functionality, low cost, and part availability.

This means that I don't use a switcher IC to produce a lower voltage for the COTS voltmeter, despite the benefits. Or negative voltage rails.

In the next version, the door is open to a switcher IC in order to gain reduced draw from the battery by using a lower voltage for the COTs meter. (Or, alternatively, find a COTS voltmeter that is a transflective LCD display with 5 digits and can replace the one I am using now.) If it turns out that the next version goes in the direction of a switcher IC, it may then also be in the cards to consider generating a negative rail.

The panel meter I'm using really is quite good. One count in the least digit is \$100\:\mu\text{V}\$. But it's not a noisy digit. Given a precision voltage, it will sit there quite precisely. And if I place my hand onto the circuit being measured to alter the voltage (touch a transistor and stuff moves pretty fast!) it will see it. When I remove my hand, it settles right back to the exact same place. So I'm developing some serious respect for the COTS device that I hadn't been sure of earlier.

What this means for calibrating it down to zero is that as one gets closer, then it will show 5 digits of 0 and then either a - or + with one digit changing for a moment. This repeats at some rate. But when I nail the zero correctly, it does not move around. It also does not move around when I change the power supply voltage from \$4\:\text{V}\$ to \$19\:\text{V}\$, for example. Despite the excess heat that must be going on. And I have let this set on the table for an hour. No motion.

I'm just liking this particular unit, so far.

When I above write about "cut in one inch" pieces, then test, then another inch, etc., note that this reflects a change of about \$7\:\text{m}\Omega\$. That seemed to work okay for the five times I did this calibration step. But keep in mind the \$17\:\text{mA}\$ draw. With \$100\:\mu\text{V}\$ per count in the least digit, I get \$\frac{100\:\mu\text{V}}{17\:\text{mA}}\approx 5.9\:\text{m}\Omega\$. If the least digit is toggling back and forth around zero, but not settling on it, that means I'm getting very close. Then, fixing it means maybe not going by more than a \$50\:\mu\text{V}\$ shift to avoid completely skipping over the null point to the other side.

The solution I've been using works very well. It's almost free and requires very little board change. And I've had no problems with it. It's just awkward.

A 3" long stretch of #32 platinum wire, together with a grabber clip, would seem viable:

enter image description here

But it's not a solution I'm looking for here.

The platinum wire would be USD12, right now. (And the Pomona 72902 is double that much!)

(I'm still going to buy some and test it just because I'm curious.)

\$\endgroup\$
12
  • \$\begingroup\$ Could you use an active circuit? A simple one using an opamp and power FET may work, but requires a negative power supply. The offset is a constant with an active circuit. \$\endgroup\$ Commented 2 days ago
  • \$\begingroup\$ @qrk No negative supplies!! Just one 9 V battery that is readily available everywhere. I talked about the extra battery for the display only as a joke. I won't be doing that! Few parts, low cost, good value, for hobbyists who may want to bin and/or match bipolars to some reasonable degree to get better results in some other project. The fewer the parts, the better, consistent with retaining functionality. \$\endgroup\$ Commented 2 days ago
  • \$\begingroup\$ If I understand correctly, the Vbe and Vce measurements you get from this device require you to subtract the Vref in any case. So what does it matter that the GND does not read 0, when you are not really measuring against GND? \$\endgroup\$ Commented 2 days ago
  • \$\begingroup\$ @jpa I've not included the equations used here. But you are right that I can just leave things alone and handle the offset as a manual subtraction. That's what I've been doing this last week. Works fine. I wrote all about this already. But I just want to make this a little less tedious for others. This is more about making it "easier" to do and relaxing the math a little. There's no actual need for all the trouble, except that I've found it a bit more tedious because of the offset issue. I was hoping to make it slightly easier for others. \$\endgroup\$ Commented 2 days ago
  • 2
    \$\begingroup\$ How about an adjustable constant current source parallel to the panel meter and a fixed resistor in the GND path? \$\endgroup\$ Commented 19 hours ago

7 Answers 7

Reset to default
5
\$\begingroup\$

You've kind of constrained yourself into a corner, IMO.

One solution that might work would be to purchase 4x 0603 resistors of 0.02/0.04/0.08/0.16Ω and wire them in series with solder jumpers across each. Digikey's price for 0603 parts (all in stock) is around 0.25 USD each.

That would give a 0 to 0.3Ω resistance with a resolution of 20mΩ (or +/-10mΩ in marketing-speak). You could add a series trace with 20mΩ resistance and a moveable solder jumper tap to get 'infinite' resolution.

\$\endgroup\$
5
  • \$\begingroup\$ When I started this, I assumed I would just subtract offsets. And that worked. But two things have changed. One is that I slowly grew to realize this project may be useful to others and I started considering how I might address that. Then, once that thought hit I started thinking about respecting the time of others. The repeated subtractions, looking backward, can be avoided and if I'm going to put this "out there" then I should put some thought into this issue. So I didn't constrain myself so much as that's just where the project dragged me towards. \$\endgroup\$ Commented 2 days ago
  • \$\begingroup\$ Your suggestion about 0603 resistors poses multiple problems. One is the skill set of someone using an exacto knife to cut lines that an 0603 can work with, and then solder it, pushes my sense of reasonableness. Also, that would mean a lot more exacto knife cuts in the copper. I'm trying to make this as easy as possible. I am picking resistor values that are sold by Taiwan's Viking Tech for almost nothing -- most values are only sold by them in 5000 lot, Digikey. But certain values are very cheap and available. Plus, the cost of each of these resistors being about as much as the TL082 itself? \$\endgroup\$ Commented 2 days ago
  • \$\begingroup\$ And finally, the idea just is not convenient as drift sets in. (Or temperature changes, requiring an updated zero-set exercise.) It needs to be a slider of some kind or a potentiometer. If someone hands me a good idea, I'm running with it. If I come up with one, I'll post it and select it. Either way, though, I will find a good answer. (The switches I picked are about 50 cents each, sadly. But their case is designed to be soldered down, as well as the three leads, too. Only one I could find that was reasonable price, available, and rigid once placed.) \$\endgroup\$ Commented 2 days ago
  • \$\begingroup\$ Hmm. The switches were 40 cents in 1s and 30 cents in 100's. \$\endgroup\$ Commented 2 days ago
  • \$\begingroup\$ I said "cut in one inch" pieces, then test, then another inch, etc. So it's shaving off about 7-8 milliOhms per step. I didn't do your calcs yet. But does that fit my need? Or is your solution not meeting that requirement? (I'm just still thinking about your idea to see if I can somehow get around to liking it.) Keep in mind 17 mA draw. And 100 uV for the least digit. 100 uV / 17 mA = 5.9 milliOhm. If the least digit is toggling back and forth around zero, but not settling on it, then fixing it means maybe 50 uV shift? \$\endgroup\$ Commented 2 days ago
5
\$\begingroup\$

You have ruled out a -ve power supply for biassing the voltmeter input down.

Instead of putting a 0-300 mΩ rheostat on the meter's -ve supply (let's call it what it is, it's a rheostat, not a potentiometer), how about put a fixed 1 Ω nice low tempco SMD resistor in the supply? This will have your voltmeter reading solidly below zero. Or use a lower value, 0.5 Ω or 0.33 Ω.

Now you can tune the voltmeter input up, at the (nice consistent input resistance) input, with a sensible value for a commercial board-mounting twiddle-pot.

One problem with the OP's idea, and indeed this extension of it makes it worse (by using the small difference between two larger values), is that it relies on the DVM current being constant, to generate a constant voltage offset in PZERO.

A better solution is to generate an effective -ve rail, by putting a silicon diode in series with the -ve supply pin, lifting it a whole diode drop. If the present two-amplifier circuit cannot be reconfigured to drive the offset DVM, then switch to a TL084, the quad version, and dedicate one of them to handling the offset.

\$\endgroup\$
20
  • \$\begingroup\$ Potentiometers are often wired up as "rheostats". So making a show of word-use doesn't impress. That said, please show me what you mean by "in the supply" and provide more detail about how you see it helping out. I'm listening. But I do not yet follow. \$\endgroup\$ Commented 2 days ago
  • 1
    \$\begingroup\$ @periblepsis my distiction is that a potentiometer is a 3 terminal device that measures the potential (hence the name) along a wire, so the wiper current can be / should be zero, and a sloppy contact does not change the value. A rheostat is a 2-terminal device where the wiper carries the full current, and sloppy contact changes the value. I might have been bitten more than most in my career by changing wiper contact resistance. I'm surprised that a 'big beast' like you should post an almost-illegible schematic. It appeared to me your pot was in the DVM -ve supply, where my resistor goes. \$\endgroup\$ Commented 2 days ago
  • 1
    \$\begingroup\$ @periblepsis It seems that the zero circuit works by raising the -ve supply pin with your up to 300 mohm. Raising it 'too far' with a fixed resistor means you can now inject a positive offset into your high resistance measurement input, a much easier proposition with normal pot values. \$\endgroup\$ Commented 2 days ago
  • \$\begingroup\$ If I understand correctly, the idea is: RZERO = 1 ohm normal resistor to provide negative offset. Then have potentiometer connected to voltmeter VIN, probably through a large e.g. 10 Mohm resistor. To get low enough adjustment voltages, the potentiometer could be between GND and VCC, with a series resistor on the VCC side. Extra benefit is that the adjustment can also compensate for negative offset in the voltmeter. \$\endgroup\$ Commented 2 days ago
  • \$\begingroup\$ @Neil_UK No. The voltmeter references everything to its (-) power rail. I can alter the (+) rail anywhere from +4 V to +28 V without breaking it and it changes nothing in its readings from the sense line, which is always referenced vs its (-). Only makes a difference when the (-) side is altered with respect to the circuit ground. The voltmeter ground needs to be moved. \$\endgroup\$ Commented 2 days ago
3
\$\begingroup\$

Your serpentine trace idea has merit, if you can spare the board area. A thicker wire shorting bar can be soldered across to reduce resistance, provided the serpentine trace is left with no solder mask.
Be aware that copper has a temperature coefficient.copper trace with wire shorting bar

\$\endgroup\$
4
  • \$\begingroup\$ Yeah. That need is one reason I want this to be convenient. If they move it, or if day changes to night, or the air conditioner kicks in, then a new calibration may be needed. But if it isn't hard to do then that's just fine. Your suggestion is what I'd considered above in my question. That problem with it, as I see it right now, is: (1) board space; and, (2) so many knife edge cuts; and,, (3) cutting very close so that the copper traces are thin means asking for skill sets I have little right to expect. I want this to have a good shot even with someone doing this "first time." \$\endgroup\$ Commented 2 days ago
  • \$\begingroup\$ I was handling the copper wire in my hands and/or letting it cool in the breeze while moving around to make other measurements. I saw no change in the meter reading. But I would be holding only perhaps 5 or six inches of the copper. So with a 15 degree shift (?) and 6 inches I figure about 2 to 2.5 milliOhms difference. So that comports with observation. \$\endgroup\$ Commented 2 days ago
  • \$\begingroup\$ A soldered wire will be a problem, though. If the device reading needs to be re-zeroed, it should be very convenient to re-zero. Taking seconds. Not minutes or hours. And contrasting your solution here (which I'd also wondered about) with Spehro's, I'd probably say that Spehro's (if using 0805 or 1216) might be fewer cuts and less skill requirements. But none of these solve the "re-zero quickly" issue, which is pretty important. \$\endgroup\$ Commented 2 days ago
  • \$\begingroup\$ Glen... An idea struck -- about 3" of #32 (most common) gauge of platinum wire would be just about right. If that can be laid in a straight line then a slider across it (some spring tension somehow) would get the job done, I think. I'm going to get some and start some fabrication ideas. The downside is that it will cost me about $12 for 3 inches from a quick check. That's a lot to spend on "a part" just for calibration purposes. But even if it doesn't win out in the end, I will probably enjoy testing the idea, anyway. \$\endgroup\$ Commented 2 days ago
3
\$\begingroup\$

Frame Challenge

  • Do you need to fix the DC offset?
  • Do you need to offer a cal position to measure it?

TL;DR

The 0.1% error due to the DPM offset error is about 0.1%, VR4 has an error of O(1%) due to the TL082 input Voffset error, dwarfing it.

Remove the facility for the zero cal, the user does not need to see a few LSBs flickering around. Remove the 300 mΩ variable resistor, or fix it at 200 mΩ.

Don't over-engineer or unevenly engineer an already excellent value product.

Discussion

Some parameters, VCE, IB, the ratio of emitter currents, are immune to voltmeter offset errors. Note we already have a 0.2% possible error in the emitter current ratio due to component tolerances. The TL082 has typically 5 mV and worst case 20 mV input offset voltage, and that's an error that will be compared to 700 mV VBE.

The only measurement that depends on the absolute voltage is the emitter current. With a 4 or 5 V reference voltage, and a 4 or 5 mV offset error, that equates to a 0.1% accuracy for the current, if left uncorrected. If estimated as a small fixed correction to the current, we could expect to reduce its uncertainty by a small factor, maybe 3, perhaps by putting a fixed 200 mΩ underneath the DPM.

The absolute emitter current enters in two places, β and IS calculations.

You do not need 0.1% accuracy for β. We cannot use a β value to that resolution. How often do we tell noobs to build their bias circuits to cope with a 3:1 variation in β, and not to rely on a predictable value? Not only that, the IB measurement is subject to the proportionally larger amplifier input offset voltage errors.

I don't think you need 0.1% accuracy for IS. We never run a transistor at IS. Its main use is to plug into a model to try to predict VBE (which varies logarithmically) (through temperature and the ideality factor) for some much larger practical IC. A 0.1% error in IC translates into a tiny VBE error, which is going to be strongly temperature dependent anyway. Remember we tell noobs not to design for a specific VBE as well.

And it just looks better to get a nice stable zero out.

Then don't give the user the facility to see there's that small error there!

\$\endgroup\$
3
  • \$\begingroup\$ "And it just looks better to get a nice stable zero out.//>Then don't give the user the facility to see there's that small error there!" Some of us like real noise, because it means a real signal. A dead zero makes me suspicious. \$\endgroup\$ Commented 5 hours ago
  • \$\begingroup\$ @ChrisH this is where we make a distinction between errors we can see, and errors we can't. The Voffset error in the TL082 opamp is typically 5 mV, which is in the context of a 700 mV (VBE) measurement ~ 1%, and we can't see it. The DPM error we can see is a few mV in a few volts, ~ 0.1%. Think carefully about performative specs, and real specs, and where to spend the engineering refinement. \$\endgroup\$ Commented 4 hours ago
  • \$\begingroup\$ Well, I wouldn't try to zero it out perfectly at all, dealing with any offset in the analysis. which I think means I agree with you \$\endgroup\$ Commented 1 hour ago
2
\$\begingroup\$

If you're going to use a sliding contact on wire, at least use resistance wire (nichrome). That would allow you to use something more robust then 30AWG, as well as a shorter piece. 20AWG apparently has 0.63Ω/ft (horrible units, but finding a table with AWG gave me feet) so you'd only need around 15cm with decent strength. Or you could go thinner and shorter.

Nichrome is a pain to solder, but if you solder down some all-brass screw terminals to a board, you could run the wire between those just above the surface, with a wiper contact clipped on.

Another interesting suggestion is to use a mechanical pencil lead. I have a 0.5mm HB (~#2) lead here and the resistance is a little too high - 300mΩ needs about 1mm. But going to a 0.9mm lead gives a little over 3× the cross sectional area so 3mm would be needed, and softer leads have lower resistance (higher graphite content). You can even get 3B in a 2mm diameter (16× the area of the one I have here, and lower resistivity, so you'd need about 30mm) or 6b in 3.15mm diameter. You could use screw terminals again but these are brittle so I wouldn't. I'd solder down a brass or copper spacer and glue the lead to it with conductive epoxy.

\$\endgroup\$
2
  • \$\begingroup\$ BTW the lowest value pot I can find from my regular supplier is 4.7Ω, or 10Ω without paying a fortune. They're both single turn so (with the 10Ω, and assuming that the range is a bit under 360° in practice) your full range would be ~10° and your desired resolution ~0.3°. So not impossible, but frustrating to set up - sub-degree turning of knobs is my job, in the sense of aligning optics. Many years ago a firm called Mallory used to make 1Ω pots but they seem to be long gone, and vintage pots probably aren't a solution for precision adjustments \$\endgroup\$ Commented 6 hours ago
  • \$\begingroup\$ I spoke too soon - for production a 20Ω 15-turn trimpot would be interesting, giving ~4mΩ/° resolution at £37 ($/€40-50) \$\endgroup\$ Commented 6 hours ago
1
\$\begingroup\$

Measure against Vref

If I understand your circuit correctly, Q1 passes a constant current through the transistor being tested, and then you measure Vbe and Vce. But doesn't this require you to subtract the Vref voltage from the readings anyway?

How about setting the voltmeter negative supply to equal Vref instead? You can do this with a transistor and opamp, similar to how Q1 and TL082 are currently used.

It does require extra parts, but makes the instrument easier to use. You can even make it selectable whether to measure against E/C for PNP/NPN.

\$\endgroup\$
1
  • \$\begingroup\$ Q1 moves the DUT bipolar to the rough center of the voltage supply. This was earlier done because I was using LT071's and not LT08x devices. And the older ones require their common mode input voltages to be 4 V away from the rails, worst case. Crazy, but true. And with only 8-9 V, this was critical So I wanted to center the DUT. That's the function of Q1 and the opamp driving it. It turns out that there's still value in doing this for several other reasons I hadn't fully appreciated before. So it stays. I'll think more about what you are writing, tomorrow. Thanks for offering your thoughts. :) \$\endgroup\$ Commented 2 days ago
1
\$\begingroup\$

I am with the putting a diode in series with the negative supply to the whole circuit idea.

So you have +8.4 volts and -0.6 ish volts : two rails for free. Use an inverting opamp off the voltage reference with a trimpot to feed a small negative offset current into the meter input via a resistor.

With a careful design a voltage of maybe -0.2 volts from the opamp when fed into a pair of resistors at the input of a meter will give you a small negative offset at the meter.

You can make a small accurate and precise voltage with a potential divider more easily than using a current through a very low value resistor with dubious contacts.

\$\endgroup\$

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.