Precision Very Low (down to 5mV ) Voltage Power Supply

[theo92]

I need a rather unusual power supply for a very different purpose. I need an output of 5mV to 50mV to power a incandescent lamp to measure it's cold resistance. You know, for precision cold resistance measurement it must be power up by very low voltages so that the filament heats up very very little, so that it becomes negligible. I'm using a 6V lamp (or a 12V automotive lamp), which starts to glow faintly at 400mV, so 5-50mV might be good for cold resistance. I'll use MAX4372 as a CSA.

In next step, I'll need a not-so-precision power supply (may be separate from the previous one) to power up the lamp upto 12V. I used PWM+Power transistor for this purpose. But I realised that there was a problem. As the current flows through the filament and it heats up, the resistances decreases, and the voltage increases slowly. So, the voltage is never stable, increases about 40mV per half minute!

So, I need to replace the whole concept. A LM317 + digital potentiometer may be used, I think. But LM317 Vref is 1.25V, so getting down to 0 volt is not easily possible. Any alternative option for LM317, Which can deliver down to 0V. Any other suggestion for the purpose?

 

[Ian Rogers]

If you don't need much current... Create a -3volt rail and sit your 317 on that or shift your ground up to 1 volt (I would use two diodes 1.25)

 

[JimB]

Why not just put a fixed resistor in series with the lamp?

If the cold resistance of the lamp is about 10ohm, then use a 90ohm resistor in series with the lamp.
Then when you put 5v across the lamp/resistor pair, you will have 500mv across the lamp.

Similarly with a 990ohm resistor, 5v will give 50mv across the lamp.

JimB

 

[unclejed613]

why not just use a DMM? most DMMs have a 5 to 20mA current source, and cause little or no heating of the bulb element.

if you need an incremental amount of current for recording a resistance curve, use a variable current source rather than a variable voltage source.

 

[Reloadron]

Why not just run a low current through the lamp and note the subsequent voltage drop? Along the lines of how a DMM would measure the resistance.

Ron

 

[theo92]

Thanks for all the suggestions from different perspective.

If you don't need much current... Create a -3volt rail and sit your 317 on that or shift your ground up to 1 volt (I would use two diodes 1.25)

I don't need much current to measure the cold resistance(CR). You mentioned the three possible ways to get output down to 0V from 317. But the problem is, for the negative rail I'd need a symmetrical power supply, which complicates the thing. Shihting of the GND may be done, but It'd need a 1V rail, another hassle! Using two diodes is the simplest process, might be used but it decreases the ripple rejection.

Why not just put a fixed resistor in series with the lamp?

If the cold resistance of the lamp is about 10ohm, then use a 90ohm resistor in series with the lamp.
Then when you put 5v across the lamp/resistor pair, you will have 500mv across the lamp.

Similarly with a 990ohm resistor, 5v will give 50mv across the lamp.

This is a nice idea. Can I use a digital potentiometer instead of the fixed resistor?

Reloadron

Why not just run a low current through the lamp and note the subsequent voltage drop? Along the lines of how a DMM would measure the resistance.

Ron

Today, 04:41 PM
unclejed613

why not just use a DMM? most DMMs have a 5 to 20mA current source, and cause little or no heating of the bulb element.

if you need an incremental amount of current for recording a resistance curve, use a variable current source rather than a variable voltage source.

Are you talking about DMM at resistance measurement mode? As far as I know this is not a good method for CR measurement because most of the DMMs are non linear at very low and very high resistances.

Some people advised me to control the current rather than the voltage. But why? Certainly it has a reason which I don't know. Generally, Voltage is regarded as an independent variable rather than the current.

It is really possible to get a resolution of 5mV out of a LM317 using a digipot? And any replacement for LM317 which outputs down to 0V?

 

[Reloadron]

What I was suggesting as to using a controlled low current amounts to doing a four wire kelvin measurement. Keeping the current low will yield pretty accurate resistance numbers with some basic math applier. For example running 5 mA through a lamp typically run off of 12 V at 200 mA. The heating effect would be negligible. I think you will find it much easier to use current over voltage. Anyway, that is how I would approach it.

Ron

 

[squishy36]

You want to control the current. Make a Kelvin measurement just like Reloadron suggests -- it's simple and foolproof when done correctly (I use it constantly in my work). A voltage reference, an op amp, and a MOSFET can make a nice, simple, cheap constant current source. However, if you want accurate measurements, you'll need to measure the constant current accurately -- and this means you'll need either an accurate shunt or a suitable in-calibration meter. Easily solved if you can toss some money at it; a bit harder if you're on a pauper's budget like I am...

If this is a one-off measurement, just use any suitable DC lab power supply that has a constant current adjustment. Set the current to, say, 1 to 10 mA, then measure the current with an ammeter and measure the voltage drop with a voltmeter. (Put in a shunt resistor if the power supply's current doesn't go low enough.) Simple and fast.

 

[theo92]

Now I've figured out why current is preferred over voltage.

For example running 5 mA through a lamp typically run off of 12 V at 200 mA

Can you please explain it a bit.

I didn't know about the four wire resistance measurements, now I see, this is just perfect for my purpose. Actually, I'll not use DMM or DC lab power supplies, I'll embed everything within a single device. I'll use a shunt+CSA for current measurement and PIC's onboard ADC for voltage measurement through a voltage divider, and a constant current source. And a PIC microcontroller of course.

A voltage reference, an op amp, and a MOSFET can make a nice, simple, cheap constant current source.

It can also be done using a LM317. Isn't it?

 

[Reloadron]

OK Following another fun filled entertaining work day the attached is an operational amplifier low current source (adjustable) I have used variations of for years. If you do a Google of Op Amp Current Source you will get plenty similar to this one.

The 5 ohm resistor is a sense resistor in the circuit. I would try and find a really accurate 5 ohm resistor in a 2 watt family. As drawn the circuit should produce -100 to 100 mA for a voltage in of 0 to 1 volt. Be it this circuit or similar I suggest a 10 turn pot for the Vin. That gives you much better resolution. Using larger transistors like a MOSFET or Darlington Pair will yield higher currents.

I am not saying use this circuit but rather snoop around (Google is your friend) and find a like circuit. You likely don't need the -100 mA to 100 mA so a single transistor and single ended supply could be used. All you want is a stable current source that fits your needs.

The idea here is as I mentioned. For example we want to measure the filament resistance of a standard 24 MB Lamp. We look at the data sheet and see it is a 24 volt lamp with a normal current of 73 mA. We know if we want to measure the resistance we need to keep the current low. So could adjust our current source and measure the voltage drop with maybe 5 mA flowing through the lamp. We measure the voltage drop right at the base of the lamp.

Ron

 

[EternityForest]

yes you can use a lm317 as a current source. i think they lose accuracy below 10ma.
an op amp current source or a lm317 with a shunt may be better.

you can control the current with any voltage so you can use a potetiometer or a dac to calibrate it.
build in a precision resistor as a calibration standard.

or you can use a DMM