My friend Nick and I were talking about a tool called the Tinkerer's Tricorder, which, if it existed, would tell you everything interesting about an electrical component as soon as you connected it across its terminals. Inductance, capacitance, resistance, voltage, frequency, diode characteristics, resonant frequency, Q, and so on — all with no danger of damaging the device!
Some of Nick's notes are at http://dangerousprototypes.com/forum/viewtopic.php?f=19&t=5366&p=52238.
I had wasted some time earlier that same day trying to debug a circuit without any test equipment. I ended up programming an Arduino to be a crude volt/ohmmeter. Despite its imprecision and the need to use a laptop to see its readout, it had three big advantages over a standard multimeter:
It, or rather the laptop, recorded a series of measurements instead of a single one, so I could see not just the current state but recent states.
It measured both voltage and resistance without any need to switch modes.
Most of all, it could be simply downloaded and installed on existing hardware, without requiring any additional hardware — not so much as a resistor soldered across a pair of pins!
So I'd like a Downloadable Tinkerer's Tricorder — the closest possible approach to the Tinkerer's Tricorder on common hardware like an AVR microcontroller or an Arduino, without needing any nonstandard hardware — not even so much as a resistor or capacitor soldered across a couple of pins. You just reprogram the flash, connect the device under test between two pins, and the AVR tells you what it is. Maybe you go through some kind of calibration step if you want accurate results.
Nick's project is far more capable, but I think its parts cost is comparable to the cost of a prebuilt LCR meter.
How capable could a Downloadable Tinkerer's Tricorder be, struggling under the limits of common AVR hardware?
Basically an AVR digital GPIO pin can be in one of four states: output high, output low, input high, and input low. Although some exotic Arduinos have DACs, common Arduinos don't have anything but digital GPIO pins plus an ADC. (This is deeply unfortunate, since the core of the AVRs' successive-approximation ADC is a DAC — but its output isn't routable to any pins!)
Output high ties the pin to the 5V rail with relatively low impedance; it can source up to 40 milliamps, but if you connect the pin directly to ground, the chip doesn't blow up; it's internally limited to about 40 milliamps. This means that, at least at high currents, the internal output impedance is about 25Ω. This is convenient for connecting AVRs to LEDs and speakers; it pretty much guarantees you won't blow up the AVR, the LED, or the speaker no matter how ignorant you are, unless you hook the LED up to your power supply instead of the AVR.
Output low is similar, but ties the pin to the ground rail through the ≈25Ω impedance.
Input high ties the pin to the 5V rail via a pullup resistor, which is nominally 20-50kΩ on the ATMega328, which is not what you'd call a really fucking precise specification.
Input low tristates the pin, i.e. sets it at the very high input impedance characteristic of CMOS inputs, around a megohm; the maximum rating from the datasheet is that these pins draw up to 10 μA.
The pins are also diode-clamped to the power rails, which should protect them from voltage spikes if you're testing an inductor and slam a pin to input state. The clamping diodes aren't well-specified in the datasheets, but various AVRFreaks threads claim that these diodes are equipped to handle only up to about 1.5mA before possibly going into latch-up and killing your chip. The ATMega328P datasheet instead gives a spec for the voltage range: -0.5 V to VCC + 0.5V.
The datasheet max specified I/O pin capacitance is 10 pF, although I suspect 1 pF is probably more typical. I haven't measured.
The usual AVRs, including the ATMega328, have a 10-bit ADC with a sample-and-hold frontend, which measures relative to either the power-supply voltage (typically 5 volts), and an internal 1.0-1.2 volt analog bandgap reference voltage; the ADC is claimed to have a 38.5kHz input bandwidth, but people have reported success clocking it up to 4MHz and still getting an Effective Number Of Bits around 5.
Furthermore, the AVR can use its ADC to measure an internal temperature sensor, supposedly accurate to ±10°C. This is important because it means that you can, in theory, compensate for thermal variation in things like pull-up resistor resistance. (The two major factors of variation in the behavior of a given electronic device are supply voltage and temperature. You can apparently measure the internal bandgap relative to the supply voltage, getting a value somewhere around 200, and thus measuring the reciprocal of the supply voltage, by setting ADMUX to 01x01110.)
I'm assuming the internal voltage reference is usable without any external hardware, even a capacitor attached to the AREF pin, but I'm not sure.
10 bits isn't much, but the device can measure time much more precisely. The internal RC oscillator is nominally accurate to 10% (3 bits), unless calibrated by the user, in which case it's accurate to within 1% (7 bits). However, in the common case that the AVR uses an external quartz crystal, its timing accuracy depends on that crystal's accuracy, which is normally about 0.005% (15 bits). Current normal Arduinos (Pro, Pro Mini, Uno) use a 16MHz Murata ceramic resonator that's accurate to 0.5% (8 bits), and older Arduinos (everything before Duemilanove, including Mini) use a quartz crystal. The AVR microcontrollers currently used can run at 20MHz, but the Arduinos run them at 16MHz for compatibility with the 16MHz ATMega8 used in the first Arduino.
(It appears that temperature and supply voltage are the major factors in the speed variation of these resonators, says this guy who was tracking his against atomic clock broadcasts from WWVB and this guy who was measuring against GPS. So maybe with the temperature and supply voltage sensors, you could reduce the errors. Temperature-controlled crystal oscillators can reach three orders of magnitude better than raw crystals, i.e. 0.05 ppm temperature stability (24 bits).)
Nick tells me ESR meters usually work at very low voltage to enable in-circuit component testing without activating any nearby transistors — under 0.1V. I don't think that's practical for the Downloadable Tinkerer's Tricorder because the chips I want it to run on don't have controllable-voltage output pins.
A tricky thing is that, when you hook up some unknown device to your AVR, the AVR doesn't have the opportunity to consult its datasheet before applying power to it in order to figure out what it is. So it seems like a good idea to design the tricorder's strategy to minimize the risk of this.
The first and most obvious kind of damage is overheating. Too much voltage across almost any electrical device will cause it to generate more heat than it can dissipate until it gets really hot, hot enough to damage or destroy it. Since we're talking about a max of 5V at 40mA here, or 200mW, this is certainly a possible danger here; there are plenty of tiny resistors and things out there that can't dissipate 200mW at any kind of reasonable temperature. However, for macroscopic components, we can rely on their thermal mass to allow us to dissipate 200mW for a short period of time — if we can manage to use a duty cycle of 5%, we should be able to use 5V with near impunity, at least on discrete components, since we'd dissipate an average of 10mW.
Even the lowest-current-rated discrete diodes I could find on Digi-key were rated for 10mA average, so I think briefly passing 40mA through an unknown diode should be safe.
Integrated circuits are more delicate when it comes to brief overheating; a presentation I dug up from Cypress on "electrical over-stress" mentions, among other things, "melted or vaporized bond wires", with horrifying photos. Still, I don't think we're talking about 40mA there.
MOS circuits in general can be destroyed by overvoltage, even very briefly: it punches holes in the "O" layer in between the "M" and "S" by means of avalanche breakdown, and those holes are permanent. However, I don't think there are any MOS devices out there that will be destroyed in this way by 5 volts, so I don't think this is a real danger.
I seem to recall that semiconductor junctions can be damaged by high current density, even without high temperatures, but I can't find a citation for that now.
Electrolytic capacitors in general can be damaged by reverse-biasing, sometimes exploding spectacularly. The super-teensy tantalum capacitors are particularly famous for this. I dug up a NASA paper on tantalum capacitor damage by reverse bias, and it seems like the damage takes place by means of a microamp (!!) or so of current. But it doesn't seem to happen below about 3 volts, and at 3 volts, it takes several hundred seconds. Even at 5 volts, it takes several tens of seconds. I suspect that short exposures to reverse bias can be corrected by a similar and rapid exposure to forward bias, reforming the dielectric layer that may have been disrupted.
Electrolytic capacitors, however, have the great advantage in this case that they have very high capacitance, 0.1μF or greater (typically 100μF or greater), even when back-biased (until they start conducting). If you're charging an 0.1μF capacitor through a 20kΩ pull-up resistor, your RC constant is 2ms. This should give you plenty of time to determine that you have a lagging voltage waveform, and therefore a capacitor, and turn the current around before the voltage reaches 3V.
Diodes in general will conduct if you put enough reverse voltage across them, either by avalanche breakdown or zener breakdown; diodes that are not designed for this will be destroyed. LEDs tend to have fairly low reverse voltage limits, but Nick tells me they're still over 5V, so not to worry. (I did find a Skyworks small-signal Schottky diode with a 1V maximum reverse voltage. But I could only find three diodes that fragile in the entire Digi-Key catalog.)
So, the simplest thing to measure would seem to be a DC voltage applied to a pin, or across two pins. The ADC can measure it directly, as long as it's positive.
For 1.1 to 5V, ADC values of 205 or so to 1023 measure the voltage with 0.25% to 0.05% quantization error: 0.005 volts. Of course, this is relative to the supply voltage, which probably isn't regulated to within even 0.25%; Joris van Rantwijk found that a single GPIO pin dumping 30mA introduced a 1% error.
For 0 to 1.1V, you can set the internal bandgap reference as the reference voltage, so as to measure with millivolt precision. However, the internal bandgap reference is specified to have ±10% error; presumably it's highly stable over time on a given chip, but varies from chip to chip.
If you can connect the voltage source between two pins, instead of to a single pin (implying that the voltage doesn't share a common ground with the AVR), then you get a couple of additional benefits:
If you set pin A as low input and pin B as low output, then you'll measure a positive voltage on pin A if the positive end of the source is connected to pin A, and a zero voltage (and some heat dissipation in the clamping diode) if the negative end of the source is. I think you can also measure a positive voltage on pin B in that case, since whatever current is being sourced by the clamping diode to ground on pin A is being sunk through the nonzero output impedance on pin B. But in that case, you can set pin A as low output and pin B as low input, and you'll measure the positive voltage on pin B.
This also means you can measure AC voltages, although only either the upper or lower half of the waveform at any one time, by switching back and forth.
For 3.9 to 5.0V, you can get higher precision measurement. Suppose WOLOG that you've measured that pin A is the more positive end of the source. Then you set pin B as low input and pin A as high output. Now pin B sees a voltage of (5.0 - DUT) volts, which it can then measure with millivolt precision by using the internal bandgap reference. This trick is limited by the stability and precision of your 5V reference, which, unfortunately, is probably worse than 0.25%. (You can re-measure it to 0.25% against the internal bandgap frequently, but you can't get the 0.02% precision this seems to promise without a more precisely controlled reference voltage.)
This range of tricks gives you variable voltage precision as follows:
This gives you at least 0.25% precision for -5 to -2.2 volts, -1.1 to -0.4 volts, 0.4 to 1.1 volts, and 2.2 to 5 volts; at least 0.45% precision from -5 to -0.22 volts and 0.22 to 5 volts; and at least 10% precision from -5 to -0.01 volts and 0.01 to 5 volts.
(As usual, linear analog-to-digital conversion is a bad deal here. If we wanted 0.45% accuracy for anything up to 5 volts and had 10 bits to work with, a smooth exponential distribution of levels 5/1.0045**n for n up to 1024 would give us 0.45% accuracy down to 51 millivolts, rather than 220 millivolts; that is, 10 bits would cover an entire order of magnitude.)
The four different ways to configure the two pins to measure a voltage between them (LI LO, LO LI, LI HO, HO LI) provide two different ways to measure any given voltage, to different precisions, while only allowing microamps of current to flow. If these measurements disagree, then you know the device isn't actually connected across the terminals, or isn't a voltage source. Since none of these configurations allow more than microamps of current to flow, none of them should be dangerous to electronic components that aren't generating voltage.
If the voltage source is just connected to pin A, then pin A might read the same voltage in the LI LO and LI HO configurations (that is, changing the state of pin B might not affect pin A much). But if pin A is floating, it's not a great idea to depend on it to do much of anything; it might follow pin B through internal leakage or capacitive coupling. So you might try the HI LO and HI HO configurations instead, just in case, but perhaps only briefly — the pull-up resistor could source enough current to damage, say, back-biased tantalum capacitors.
If you know that you have a voltage source between your terminals, the next thing to do is presumably to evaluate how much current it can deliver, which you should maybe be able to do by setting both terminals to low output (LO LO). In this voluptuous configuration, the negative terminal of the voltage source is forced below ground and fed current from the clamping diode, while the positive terminal feeds up to about 40mA of current to ground through its impedance. You should see 0V on one terminal and a small voltage on the other, which voltage should increase (probably nonlinearly!) with the current being sucked from the voltage source. If it can provide more than about 40mA (AAA batteries can provide 500mA) you won't learn anything useful.
LO LO should be safe for any component that isn't generating voltage itself, since it's just trying to put both of its terminals at the same voltage.
The trouble with LO LO or HO HO with a voltage source is that you're almost certain to be overtaxing the poor little clamping diodes. It might make more sense to use LO HO or HO LO, with the appropriate polarity to keep either of the terminals from going outside the power rails.
A resistor between the two terminals will tend to bring their voltage to equality in any of the five configurations suggested for measuring voltage sources in the previous section (LI LO, LO LI, LI HO, HO LI, LO LO). In the first four configurations, the input pin will follow the voltage on the output pin, and quite quickly too — if you have a 1pF pin capacitance and a 1MΩ resistor, your RC time constant is 1μs, 16 clock cycles on the Arduino, far too fast to digitize the charging curve with a mere 35kHz ADC. You might be able to beat that by doing a lot of transitions and digitizing one sample after each transition, at a variable delay, but you're going to be out of luck when you get to 10kΩ resistors, let alone 100Ω resistors.
But if one pin is following the other like that, you can infer that there is perhaps a resistance connected between them, and try to run some current through it to see what happens.
The first thing to try is presumably HI LO or LO HI, which forms a voltage divider between the pull-up resistor and the unknown resistance. If you somehow knew the resistance of the pull-up resistor — more accurately than "20–50kΩ" — you could use this to calculate the unknown resistance with some precision. Without that calibration, you're just ballparking it, although you can measure the ratios between resistances quite accurately.
(In practice the pull-up isn't actually a linear resistance.)
For the case where the pull-up resistance curve is known accurately — let's say 35kΩ — you have the following scale of measurements on the HI pin:
So far we've only been pushing about 140μA through the resistor. This should be safe for almost anything (except, as mentioned earlier, electrolytic and especially tantalum capacitors over long periods of time.) But if our device is behaving like a resistor — in particular, the measurement we get here is stable over long periods of time such as a millisecond or two, and consistent in both directions — it's almost certainly safe to put more current through it, at least for a while. And we can get a more precise measurement that way if the resistance is lower than around 1-10kΩ.
So if we switch to LO HO and HO LO mode, we're attempting to put a full 5V across the putative resistor. Now we should be able to measure the voltage drop introduced by the pin driver (the sink) on the LO pin, and perhaps less precisely the voltage drop on the HO pin.
These drivers are somewhat nonlinear — their voltage drop isn't purely proportional to the current — but treating one of them as having a resistance around 25Ω worked reasonably well for me in an experiment, over about the 50-500Ω range. Experimentation is needed.
This could involve pushing a full 40mA through the resistor, dissipating 200mW, so you probably want to do it on a low duty cycle. This will involve lots of opportunities to watch its impulse response, which could tell you if it has much inductance or capacitance.
If you're switching from HO LO to LO LO, a capacitor should dump the positive charge it's accumulated back into the formerly-HO pin, keeping its measured voltage above zero for a bit due to the nonzero input impedance, while a resistor will stop carrying current as soon as you stop applying voltage. An inductor, on the other hand, will continue carrying current in the same direction, dragging the formerly-HO pin's voltage below zero (unmeasurably, due to diode clamping) but continuing to dump charge into the still-LO pin, keeping its measured voltage above zero.
That's not the only possible way you could turn the pins off to stop burning energy in the resistor. In fact, of the 16 possible configurations of two pins, only LO HO and HO LO apply 5 volts at low impedance across the load; the other 14 are some variant of "off" or "gently pulled up", so any of them would work. However, of those 14, only LO LO and HO HO provide low-impedance paths for current to continue flowing in the case of an inductor. Suddenly increasing the impedance in a circuit with an inductor is a recipe for potentially damaging or possibly even painful voltage spikes; we can hope that the diode-clamped inputs are up to the stress, but it seems better to minimize the risk.
Even assuming the behavior we observe is memoryless, the different voltages applied by HI LO and HO LO (in the case where current flows) can give us two points on the E-I curve of the device under test; LO HI and LO HO give us two more points, although perhaps less informative ones. If those four points and the origin are collinear, we have a resistor, at least over the voltage range we're testing with. Being able to tell if they're collinear, though, presupposes that we have a good map of the E-I curve of the pin drivers, since that's what we're supposedly inferring the I at high current from, and also a well-calibrated value of the pullup resistance!
Reversing the bias will, among other things, distinguish a diode from a resistor, even if the exponential E-I curve doesn't — unless it's a zener diode with a low breakdown voltage.
Capacitors can be delicate. Inductors aren't, but chips are delicate when connected to inductors. So this requires some care.
When the external voltage drops you observe in HO LO or LO HO and LO LO modes aren't time-invariant, you know you have either a capacitor or an inductor; and when those voltage drops go up rather than down over time, you know you have a capacitor, so you should be careful. But "being careful" might mean avoiding HO LO and LO HO modes altogether XXX
If you reverse the voltage, it should XXX
XXX
XXX
http://xyphro.de/blog/index.php homebrew smart tweezers
http://blog.iteadstudio.com/tag/lc-meter/ the Goliath Arduino-based LC meter.
http://hackaday.com/2011/07/24/using-an-arduino-to-measure-inductance/ http://reibot.org/2011/07/19/measuring-inductance/ another L-measuring circuit.