Constant Current Source; Matching Resistors

[liquids]

I've found some info online about 'constant current sources' for the sake of testing.... 

I am trying to 'confirm' the measurements I am getting from my DMM in regards to 'matching' some 100 ohm 1% resistors. 

Absolute resistance is less important than 'matched' resistance...

My DMMs are clearly not up to the task of stable, accurate, reliable measurement down to 1/100th of an ohm; however I seem to be able get repeatable measurements enough to 'bucket' sort resistors them withing +/- .1 ohm....maybe.

Anyhow, I was told that if I had a constant current source (say, 1-10mA), I could measure voltage drop across the resistors to 'confirm' or refute my 'bucket' measurements/matching, via measuring voltage drop across the resistors under constant current load.

I've never tried this before, and a little concerned I'll sacrifice a lot of op amps and resistors trying this and that out without further assistance....so I'm asking for help with arranging a constant current source, in detail, specifically for this purpose....and with some understanding so that if I should encounter this need again with other values I'd be able to tweak it accordingly.

From what I read, a rail/rail (CMOS?) op amp or one that the output can safely swing to ground (LM358??)...a transistor with the base-emitter junction in the loop...some resistors....etc.

Ideally I'd get a 1mA or 10mA current source....and one that will be able to handle very small resistance loads (10R-100R)....from there I can measure voltage drop across the resistors and match resistors?

I'm confused about how ideally to power the op amp, how and what voltage to 'feed' it's inputs, and how to avoid currents like 10mA where the heat produced might affect the resistor and then the voltage drop, ad infinitum..

http://www.ecircuitcenter.com/circuits/curr_src1/curr_src1.htm   

[Gurner]

Whatever voltage you present at the opamp's +ve input will end up across the Rsense resistor (by virtue of the fact an opamp will always equalize it's inputs)

So just choose a value for Rsense...the it's just V=IR to decide on what voltage you should put at the opamp's +ve input.

Example....

Rsense of 10R
constant current required = 10mA (0.01A)
voltage drop across Rsense = 0.01A X 10(R) = 100mV

so with a Rsense of 10R, you'd apply 100mV to the opamp +ve input to get a constant current of 10mA through Rsense & your load resistor.....i.e. whatever current flows through Rsense also flows through your Load resistor - & it's the load resistor that you would be swapping in & out to suit - but whatever value you put in, it's the comibination of R sense & the voltage presented at the opamp +ve input that sets the constant current.

Note: 100mV is getting perilously close to the ground (rail) for a single suply circuit, so I'd probably go with a larger R sense resistor...

Rsense of 100R
constant current required = 10mA (0.01A)
voltage drop across Rsense = 0.01A X 100(R) = 1V at the opamp input    .....much easier for an opamp amp to handle

[liquids]

Seems like this is difficult to do since I am trying to measure 100ohm resistors...and lower currents are is better (~1mA) so they don't get warm.

[defaced]

Is it worth just buying 0.1% MF resistors for this task?  

[Gurner]

Seems like this is difficult to do since I am trying to measure 100ohm resistors...and lower currents are is better (~1mA) so they don't get warm.

You're welcome.

You're not grasping this - using the info I gave you (& info in the article you linked to), if you want 1mA then just change the Rsense value.

Rsense of 1K
constant current required = 1mA (0.01A)
voltage drop across Rsense = 0.001A X 1000(R) = 1V at the opamp input  

Now you'll get a constant 1mA runniing through your load resistor  

[liquids]

I'm trying to match a set of 100ohm resistors.  Changing that defeats the goal.

[liquids]

Is it worth just buying 0.1% MF resistors for this task?  

Probably!  But it's too late.  (=   I have 60 resistors that are probably all withing ~.02% of each other, consistently, according to my DMM via repeated measurements over time - about 20 of which are 100.X ohm resistors with a larger value resistor in parallel to achieve such measurements.  I'd have gotten 0.1% and paid the premium if I had realized....live and learn!

[JDoyle]

I'm a bit confused as to what you are trying to do here: are you trying to match resistors, build a CCS, or are you trying to build a current mirror?

The reason I ask is bc you really don't need matched resistors in a current source - there are a number of articles on the web about building quality CCSs and most use only a couple of transistors and the (nearly) constant Vbe of one of them to set the current (when dropped across a single resistor); take a look at audio power amp designs - but in a current MIRROR using discrete transistors, you will need matched resistors to balance the (possibly) different Vbes of both transistors in the mirror.

We may be able to help you more if we know what you are trying to do with the matched resistors.

I always have more confidence in a DMM's voltage measurements than it's current readings - it's easy to make a consistantly high input impedence, not so easy to make it low, consistantly and repeatedly.

Regards,

Jay Doyle

[Gurner]

I'm trying to match a set of 100ohm resistors.  Changing that defeats the goal.

Again, you're clearly not getting it ...the Rsense resistor that I'm saying make 1k (to get 1mA of constant current) simply sets the constant current (in combination with the DC voltage presented at the opamp +ve pin, which should be 1V for a constant current of 1mA & a 1K Rsense)... once you've all that set up, your 100R resistors (that you wish to test) would actually go where the RL is in this diagram...

http://www.ecircuitcenter.com/circuits/curr_src1/curr_src1.htm

then, no matter what value resistor you put in the RL position, you'll get the same constant current flowing through RL each time you swap it out (due to the voltage across Rsense feeding back to the opamp -ve pin - and since Rsense would always be the same, you'll always get the same amount of current flowing through it). therefore you can then measure your voltage drop across the RL (your 100R resistors)  & check for variations knowing that each time you put in a new one, the current through it remains the same as the last one you checked. I'm not convinced your DVM will have the accuracy...1mA through a 100R resistor is a 100mV drop - how many decimal points in voltage does your DVM measure?

[cthulhudarren]

You could always use a small linear taper pot in series with a smaller value to match exactly.

[liquids]

I'm a bit confused as to what you are trying to do here: are you trying to match resistors, build a CCS, or are you trying to build a current mirror?

I'm trying to match resistors.  I've got convincing matching with a lot of effort via a cheap dmm, isolating variables, getting repeatable results, and paralleling resistors to get as close to a slew of resistors within 0.1% tolerance (realistically that's probably the absolutel closest I'm actually getting via DMM resistance readings) from a batch of 1% MF resistors.  Think analog synth keyboard controller resistor string.  

However, I thought I might get some confirmation - or indication of less precision than I believe I have - via using a constant current source, to measure voltage drop across each resistor, to a degree that will confirm that I have resistors in that matching tolerance range, or slightly more variation than I currently think I have.  

I started this by using a TO-220 9v regulator, and simply putting the voltage regulator output to ground via the 100 ohm resistors.
Well, a 1A regulator can handle 90mA, but it gets a resistor rather hot, and that affects the measurement.  Hence, probably not the best method, I quickly realized.  I'm told that a constant current source (instead of a voltage source) is a better way to go...and here I am.

I always have more confidence in a DMM's voltage measurements than it's current readings - it's easy to make a consistantly high input impedence, not so easy to make it low, consistantly and repeatedly.

Regards,

Jay Doyle

Exactly, I have more confidence in my DMMs ability to measure voltage drop across a resistor than resistance (or current, as you suggest) given that I'm working with 100 ohms of resistance.
I may only get one digit or 1/10th of a percent more accuracy than I currently have, but I'm trying to measure two ways, solder once.

[cthulhudarren]

for a constant current source, try a LM334.

See page 6:

www.datasheetarchive.com/dl/Datasheets-22/DSA-423659.pdf

[PRR]

If the resistors are close, a constant-current source adds little.

The DVM works best near 199mV (that's the internal basic ADC). 199mV in 100 ohms is 0.3 milliWatts, quite low. To get 199mV in 100 ohms needs 2mA. Given 9V, you want 4K5 total. You don't want the meter auto-ranging past 199mV due to range-error, so aim low. A common 4K7 resistor in series with 100, all across a 9V source, gives a nominal reading of 187.5mV.

[liquids]

Thanks, PRR.

I'm replying quickly so I imagine I could do the math...but I should expect to see significant voltage drop difference with variances in resistance?  I think last time, I had my DMM set to current...and changing from 100ohm to a 10ohm resistor showed little difference in current measurement...I wasn't clued in enough at the time to measure voltage drop.   

Anda TO-220 voltage regulator should work fine here?

[PRR]

> I imagine I could do the math...

I imagine so.

[liquids]

> I imagine I could do the math...

I imagine so.

I should say, I can/will have excel do the math 

[PRR]

It's hardly worth a spreadsheet. You don't actually need exact answers, just the approximate size of the answer to know if it will work, and if your (unspecified-2.5 digit? 3.5 digit?) DVM can resolve small-enough differences.

For "100" +/-1% :

9V*101/(4700+101) = 0.1893V
9V*100/(4700+100) = 0.1875V
9V*99/(4700+99)  =  0.1856V

So 2mV per %. If your meter reads 199mV then it gets real dubious to expect half-percent readings with any direct approach. Some popular-price meters now read to 199.9mV. While I know they can't be trusted to 0.1mV (ANY temperature change overwhelms a tenth-milliVolt) it may be reasonable to expect tenth-percent resolution with repeated careful measurements.

> analog synth keyboard controller resistor string

Then there is a slightly more accurate trick. The plan I proposed relies on the 9V and 4K7 being, not exact, but constant from moment to moment. If you go ahead and build the whole keyboard string, and its driver, then you _know_ you have the _same_ current in _every_ resistor of the string. Bridge your DMM across resistors one at a time. They should all read the same. (This again assumes the string driver is constant from moment to moment.)

Drawback: at 1V/oct each resistor should step-down 83.3333mV. This does not make the most of your available 199mV (or 199.9mV) resolution.

For really-better matching you must turn to bridge techniques. This tends to get fussy.

BTW: you really should be using "4-terminal" technique. Force current across the resistor with two leads. Measure the voltage with two other leads. That way you measure just the resistor, not the leads also.

You know back in 1970 they did not over-think this. Same-batch 1% resistors will come out more accurate than most musical instruments. Do you think a guitar maker sets the frets with 0.1% accuracy? 0.000,1 inch error? That's below micrometer, gets into interferometer accuracy, yet I think most guitar frets are set with ruler and eyeball.

[liquids]

Very interesting points.

As of last night, I have three 'raw' keyboard controlers...and I'm encountering that, the 'best' available keyboards I currently have on hand are the ones that uses bubble switches/rubber&carbon contact or whatever for the key switches...the true 'single buss' keyboard do I have is not cutting it for reliability and consistency...it's old, tempermental, finicky, dare I saw poorly designed.

Anyhow, the switches in the bubble keyboards have their own resistance...in the 50-200ohm range from preliminary measurements...which is sort of frustrating after spending so much time trying to match 1% resistors.

I'll probably work with it for now...and see how much it affects things.  At least to get something to work with until a better keyboard comes along or the 'classic' style single buss can be bough up to spec...

[PRR]

The music keyboards were pretty good when made. Any exposed-contact system is sure to need cleaning 30+ years later.

The contact resistance is not very important. In the classic keyboard chain the contact is not in series with the 100r resistors. It is in series with the voltage sensing buffer, which IIRC is a very high impedance. That's basic philosophy of precision voltage division.

[liquids]

I hadn't bee visualizing that way, that's actually great news.

I had been picturing the switch resistance as a variable resistance to ground, added into the voltage divider network. 
If configured properly, it will just be a resistance from each key's voltage divider 'tap' and the op amp (which has a small value cap, a 10MEG resistor to ground, and 2 series diodes to ground with cathodes facing the oop amp's in put in this case http://musicfromouterspace.com/analogsynth_new/SINGLEBUSSKEYBOARD2007/SINGLEBUSSKEYBOARD2007.php

That said, there might be no harm in 'adding' a relatively small valued resistor between the keyboards bus and the op amp's input anyhow, to minimize the differences in resistance from key to key, press to press...I'd expiriment with that first however.

That said, that makes me think that - while slightly more complicated and expensive, any keyboard might be able to work, in that each key could also simply control a make/break connection between an individual CMOS logic switch with the key control the control pin on the log chip switch, connecting the resistor string to the bus.....and apparently the CONSISTENT resistance (60-200 ohms for a cmos switch) wouldn't be an issue....

Not that I'm going to do that anytime soon, but it makes more sense to me now.  Thanks!