EEVblog Electronics Community Forum
Products => Test Equipment => Topic started by: SharpEars on October 03, 2015, 09:41:47 pm
-
I have a nice Fluke 5 1/2 digit ammeter that is well within its 1 year calibration spec. Unfortunately, it only goes to 2 A max. I am trying to calibrate a 10V/5A power supply. Is there any way I can take a reasonably accurate (i.e., 4 1/2 digit is fine) current measurement with the 2 A max ammeter of the max output of the supply (i.e., 5A)?
Any advice, even if it involves other parts would be great. Unfortunately I don't have anything that can reliably act as a current shunt (for a voltage measurement). Everything I have is very sensitive to changes in temperature and it is very difficult for me to make <100 milliohm accurate measurements of resistance to divide the voltage by in order to do a calculation of current. I would definitely like to use the ammeter to perform the measurement, but somehow very accurately reduce the 5A being put out by the supply in constant current mode into 2A.
I guess I am looking for a way to divide the current by 2.5 that is insensitive to changes in temperature.
-
I'm thinking of a current mirror with three transistors.
That would split the current into three separate paths but it will need three very well-matched transistors to get 4.5 digits.
-
I'm thinking of a current mirror with three transistors.
That would split the current into three separate paths but it will need three very well-matched transistors to get 4.5 digits.
And therein lies the problem, I have no way I can think of to match three transistors. And, in any case I have no transistors that can take on amps...
I do have wire of various gauges and I've tried building a current shunt out of it. But, even if I can calculate the resistance of a small/large run (which in itself is error prone at a 4 1/2 digit accuracy level), changes in heat to the wire cause voltage measurements across the wire to drift up as its temperature changes due to the fact that 5 A are flowing through it (i.e., more and more voltage is needed to cause 5A to flow through the length of wire as its resistance increases due to a rise in temperature).
-
4.5 digits, that's like 20ppm which is pretty good resolution to be calibrating a 5A power supply to.
I would think that most applications that need to regulate the amperage of a current so precisely would use a current shunt and a closed loop feedback system against a precision voltage reference to ensure the stability and accuracy of the current at those levels. Unless the load has very little dynamic change then usually just the capacitance and inductance and loop transfer function of the output of the PSU would give larger variations in current than that even in a "constant current" mode since usually the loop feedback isn't all that fast.
Anyway to a limited accuracy you could take three "identical" somewhat long (for the reason explained hereafter) wires and split them between the source and the identical load terminal. Pass one through the ammeter. If the shunt and leads to the ammeter are insignificant in resistance compared to the wires used then roughly 1/3rd the DC current should flow in the sample wire compared to its two identical mates. I sort of doubt you're going to get 20ppm accuracy this way but it would be something better than nothing.
If you need this kind of test equipment percision for your application you should just get the right test fixtures to measure / calibrate it. Current shunt, or, if the load is resistive and well known then just the voltage drop across your actual load.
If I had three of these ammeters, I could just wire them in parallel. Unfortunately I have two, which means there would be 2.5 A going through each if I were to wire them in parallel, which is in excess of each one's 2 A limit. Having a third would have made this exercise trivial, unfortunately I only have two. Using wire as a shunt or voltage divider for any of this is error prone due to the fact that it has very low resistance, unless small gauges and long lengths are used and the resistance of the wires would change with temperature.
-
Exactly. It sounds to me like SharpEars is trying to get calibration lab accuracy, but with inadequate equipment. It's not going to happen.
What application needs that kind of accuracy from a consumer-grade bench power supply, anyway? Voltage drop across a suitable resistive load is going to work fine for ordinary mortals.
-
Exactly. It sounds to me like SharpEars is trying to get calibration lab accuracy, but with inadequate equipment. It's not going to happen.
What application needs that kind of accuracy from a consumer-grade bench power supply, anyway? Voltage drop across a suitable resistive load is going to work fine for ordinary mortals.
At what point did I say this is a consumer-grade bench power supply? Your assumption is incorrect.
-
That's 80 resistors in parallel. Double that if you only have 1/4 Watt resistors. :-)
This might not be too bad. Get some really thick wire, strip it and tape two strands down in parallel, about half an inch apart. Lay a line of resistors along it and attack with soldering iron...
If the resistors are still in the paper strip then it's a five minute job. :popcorn:
-
I have no way I can think of to match three transistors. And, in any case I have no transistors that can take on amps...
The only other way I can think of would be to put lots of resistors in parallel to make your shunt. This spreads the heat out and minimizes thermal effects.
You'll need a lot of resistors for 5A though. eg. If you aim for, say, a 2V drop across your resistors to get a nice accurate voltage reading then that's 10W of power. With 1/2 Watt resistors you should maybe go for 1/8W per resistor to avoid thermal effects. That's 80 resistors in parallel. Double that if you only have 1/4 Watt resistors. :-)
Hey, nobody ever said 4.5 digit current measurements using McGyver parts would be easy...
Yeah, I have 1% 1/2 watt resistors, but not nearly enough of one value for this sort of gadgetry. Also, 1% is probably too much error per resistor in any case...
-
The burden voltage of the ammeter will screw you if you attempt to split the current.
Your best bet would be to make a four terminal current shunt to use with a millivoltmeter and calibrate it in series with the ammeter at up to 2A then extrapolate to 5A. Keep the duty cycle really low and the thermal mass high or the significant positive temperature coefficient of resistance of most metals will also be a problem. A serpentine PCB trace, with solder mask in direct track side contact with a large hard anodised heatsink with thermal grease in the interface could be a good place to start.
Calibration would be to determine the mV/A coefficient, not to adust the actual shunt resistance.
-
I have a nice Fluke 5 1/2 digit ammeter that is well within its 1 year calibration spec. Unfortunately, it only goes to 2 A max. I am trying to calibrate a 10V/5A power supply. Is there any way I can take a reasonably accurate (i.e., 4 1/2 digit is fine) current measurement with the 2 A max ammeter of the max output of the supply (i.e., 5A)?
Any advice, even if it involves other parts would be great. Unfortunately I don't have anything that can reliably act as a current shunt (for a voltage measurement). Everything I have is very sensitive to changes in temperature and it is very difficult for me to make <100 milliohm accurate measurements of resistance to divide the voltage by in order to do a calculation of current. I would definitely like to use the ammeter to perform the measurement, but somehow very accurately reduce the 5A being put out by the supply in constant current mode into 2A.
I guess I am looking for a way to divide the current by 2.5 that is insensitive to changes in temperature.
Build yourself a 2A current source you can calibrate with your ammeter (or use one you have laying around). You can decide how fancy you need. I'd perhaps consider doing some sort of rudimentary oven - it just has to be stable enough to be able to transfer a 'reference' during calibration. Calibrate this source with your ammeter as is.
Then, paralllel a shunt (even if it's homemade (aka length of copper or aluminum foil) with a horrible tempco) of about 1/2 the internal shunt resistance on the ammeter. This should divide the reading by roughly 3. Read the 2A reference with the ammeter, and use the resulting difference in reading as a multiplier.
-forrest
-
As far as temperature effects are concerned, it is fairly standard practice to use short pulsed measurements so that the energy dissipated in the sensor doesn't change the temperature of the sensor more than suitable for the time needed to complete the measurement. So a 2 second or less current pulse might be used.
I don't know, it is like a case of having only a hammer and every problem in the world starts to look like a nail, even the ill suited ones.
You could use a current shunt and measure the voltage. You could use calorimetry. You could apply the current to an inductor acting against a balanced reference current and null the saturation. You could use coulometry. You could measure the magnetic field generated by the current. You could calibrate up to 2A and extrapolate.
As current shunts aren't that expensive, I'd probably look into that option.
You can buy them ready made or get some foil and make something custom. You could even use a hunk of random wire if you characterize the current at the 1A and 2A settings to sufficient accuracy, then measure the shunt voltage at each current, then raise the current to 5A, measure the shunt voltage, and it'll be pretty close, especially if you can arrange the measurements to occur in short pulsed operational modes.
All good ideas, especially the pulsing of current. But sadly, I am limited to how my power supply allows itself to be calibrated. To be specific, like most supplies it gives you a low ball current and a high ball current near the max. You measure these and set the supply to whatever your ammeter (or voltage measurement across shunt) tells you. Sadly, this process takes time and the time is in seconds. Over these seconds, there is a change in temperature in whatever is sensitive to such things (e.g., wire, resistors and everything else I have tried).
-
I have a nice Fluke 5 1/2 digit ammeter that is well within its 1 year calibration spec. Unfortunately, it only goes to 2 A max. I am trying to calibrate a 10V/5A power supply. Is there any way I can take a reasonably accurate (i.e., 4 1/2 digit is fine) current measurement with the 2 A max ammeter of the max output of the supply (i.e., 5A)?
Any advice, even if it involves other parts would be great. Unfortunately I don't have anything that can reliably act as a current shunt (for a voltage measurement). Everything I have is very sensitive to changes in temperature and it is very difficult for me to make <100 milliohm accurate measurements of resistance to divide the voltage by in order to do a calculation of current. I would definitely like to use the ammeter to perform the measurement, but somehow very accurately reduce the 5A being put out by the supply in constant current mode into 2A.
I guess I am looking for a way to divide the current by 2.5 that is insensitive to changes in temperature.
Build yourself a 2A current source you can calibrate with your ammeter (or use one you have laying around). You can decide how fancy you need. I'd perhaps consider doing some sort of rudimentary oven - it just has to be stable enough to be able to transfer a 'reference' during calibration. Calibrate this source with your ammeter as is.
Then, paralllel a shunt (even if it's homemade (aka length of copper or aluminum foil) with a horrible tempco) of about 1/2 the internal shunt resistance on the ammeter. This should divide the reading by roughly 3. Read the 2A reference with the ammeter, and use the resulting difference in reading as a multiplier.
-forrest
Sounds like a great idea, sadly the building of the shunt that is 1/2 the internal shunt resistance of the ammeter is a problem. I can't build something whose resistance stays stable with temperature changes and anything that takes on 3 A that I have tried building will change in temperature rapidly messing up the reading. Also, I think my ammeter's resistance is somewhere on the order of 0.9 ohms. For example somewhere in the neighborhood of 1.9 V fed directly into the ammeter produces a reading of 2 A. It would take a lot of wire to make a say 0.5 ohm resistor.
-
Yeah, I have 1% 1/2 watt resistors, but not nearly enough of one value for this sort of gadgetry. Also, 1% is probably too much error per resistor in any case...
That doesn't matter. You're not building a 0.01% resistor, you're building one that can take a lot of heat. Just combine lots of resistors until you get to about 0.5 Ohms (or whatever value you choose for the shunt). Measure the final resistance to 4.5 digits, use that number in your calculations.
The more resistors you use, the better thermal coefficient it will have.
-
Yeah, I have 1% 1/2 watt resistors, but not nearly enough of one value for this sort of gadgetry. Also, 1% is probably too much error per resistor in any case...
That doesn't matter. You're not building a 0.01% resistor, you're building one that can take a lot of heat. Just combine lots of resistors until you get to about 0.5 Ohms (or whatever value you choose for the shunt). Measure the final resistance to 4.5 digits, use that number in your calculations.
The more resistors you use, the better thermal coefficient it will have.
I can't make a 4.5 digit resistance measurement on 0.5 ohms. I have no device that can do this accurately and repeatably. I think the best idea so far is to place wire in parallel with the ammeter, but making a 0.5 ohm wire is not easy - that's a lot of wire (and I don't want to use too high of a gauge).
-
Hey guys, it just dawned on me that i have 5 W 1% 0.1 ohm resistors. Maybe put five of these in series and put the result in parallel with the ammeter, then use scaling to approximate? The problem is that the 3 A going through the resistors will skew their resistance by as much as 50% I think, screwing up the test.
I guess I need to mock something up and see how stable of a reading I can get when the power supply maintains a constant high current. The issue is going to be the fact that the temperature of the resistors when presented with a 1 A load will be different than when presented with a 3 A load.
-
Serpentine PCB trace or copper foil shunt, lacquer to protect it and an ice-water bath to keep it at constant temperature.
You have a 5 1/2 digit ammeter so as long as you have a millivoltmeter of comparable accuracy and a reasonably stable current source that can provide approximately one amp, you can certainly measure the shunt resistnce accurately *IF* you can keep it at a constant temperature (e.g in an ice-water bath).
-
Hey guys, it just dawned on me that i have 5 W 1% 0.1 ohm resistors. Maybe put five of these in series and put the result in parallel with the ammeter, then use scaling to approximate? The problem is that the 3 A going through the resistors will skew their resistance by as much as 50% I think, screwing up the test.
How do you figure that 3A will go through the resistors?
-
Hey guys, it just dawned on me that i have 5 W 1% 0.1 ohm resistors. Maybe put five of these in series and put the result in parallel with the ammeter, then use scaling to approximate? The problem is that the 3 A going through the resistors will skew their resistance by as much as 50% I think, screwing up the test.
How do you figure that 3A will go through the resistors?
Well, if I make them equal 2/3s of the resistance of the ammeter's shunt, then 2 amps goes through the ammeter and 3 amps will go through the resistors. That may not be 5 0.1 ohms resistors (probably around 6), I have to actually calculate this more accurately.
In the end, assuming I have a reliable 2 A current source, which I can measure with my ammeter. I then make the "concoction" and measure the 2 A current source again. Now I know my scaling factor. Then I measure the 5 A power supply as part of its calibration, apply the scaling factor and enter the result when the PS asks me what it's high current output should read.
The biggest issue is keeping the scaling factor constant as the load on the resistor group goes from 1.2 A (60% of 2 A) to 3 A (60% of 5 A), because there will surely be a temperature rise in the resistor group.
-
Why don't you divide the current equally between three parallel and identical shunts? Then the temperature coefficient won't matter because it will be the same in each current path. For example, you could make the shunts out of long straight lengths of steel wire suspended in free air.
Then put the ammeter in series with one of the shunts and null out the meter burden voltage with a microammeter. With care and precise measurement you should get much better than 1% that way.
-
So I mocked this up and I don't know, resistors going from 144 mW to 900 mW will probably screw up my multiplier pretty badly due to them heating up over the course of several seconds. The resistors are rated for 5 W, but 900 mW is enough to change their resistance pretty rapidly due to a fairly high ppm/C:
(http://i.imgur.com/oDqZ9mY.png)
-
I use 1 mOhm thru 10mOhm precision shunts. At that resistance, heat is not a huge factor. You can, as other said, pulse it. The higher the resistance, the more it dissipates, the more its Tc will come into play.
This one is $2 US .5% 5mOhm 2W
http://www.digikey.com/product-detail/en/LVK25R005DER/LVK25R005DERCT-ND/5125272 (http://www.digikey.com/product-detail/en/LVK25R005DER/LVK25R005DERCT-ND/5125272)
5A will dissipate 125mW and give you 25mV which ain't that bad. You can use mineral oil to prolong the temp change but keeping it close to ambient. Your meter should also have enough resolution with a 1mOhm. Use one resistor and keep it simple.
Hope that is helpful.
-
I use 1 mOhm thru 10mOhm precision shunts. At that resistance, heat is not a huge factor. You can, as other said, pulse it. The higher the resistance, the more it dissipates, the more its Tc will come into play.
This one is $2 US .5% 5mOhm 2W
http://www.digikey.com/product-detail/en/LVK25R005DER/LVK25R005DERCT-ND/5125272 (http://www.digikey.com/product-detail/en/LVK25R005DER/LVK25R005DERCT-ND/5125272)
5A will dissipate 125mW and give you 25mV which ain't that bad. You can use mineral oil to prolong the temp change but keeping it close to ambient. Your meter should also have enough resolution with a 1mOhm. Use one resistor and keep it simple.
Hope that is helpful.
With a 1 mOhm shunt and 5 A current source, I would have to measure 5 mV across the shunt with 4 1/2 digit accuracy/precision. I don't think I have a device that can do this accurately.
-
I think I have "stumbled" upon a solution. I have four 2 ohm 100 watt power resistors. If I connect them in parallel and taking their tolerance into account, I get a measured resistance of something like 0.575 ohms. The exact value is unimportant as long as it is under 0.6 ohms.
Also, because these are heatsinked 100 W power resistors, they don't change their resistance too much when 3 A is applied to the bunch in parallel (0.75 A per resistor at a fairly low voltage of 1.725 V or so = 1.29375 W per resistor).
So, if I take the four in parallel and connect my 2 A ammeter in parallel as well, I should be able to measure 5 A (as just under 2 A on the ammeter) and then scale the result based on a prior measurement of a known 2 A current source to get an accurate reading.
Any thoughts or flaws with this idea that I have overlooked?
This is the power supply I am trying to calibrate with this Frankenstein approach (don't laugh!!):
http://www.keysight.com/en/pd-838347-pn-6611C/special-order-power-supply-10v-5a?cc=US&lc=eng (http://www.keysight.com/en/pd-838347-pn-6611C/special-order-power-supply-10v-5a?cc=US&lc=eng)
-
With a 1 mOhm shunt and 5 A current source, I would have to measure 5 mV across the shunt with 4 1/2 digit accuracy/precision. I don't think I have a device that can do this accurately.
The OP has a 5-1/2 digit. My Fluke 87 3.5 digit in mV mode can do .01mV resolution which is not hugely special. 1mOhm should be fine.
-
Y not
use 4 same value resistors in parallel?
Through each resistor you'll have 1/4 of the current flowing through it
-
With a 1 mOhm shunt and 5 A current source, I would have to measure 5 mV across the shunt with 4 1/2 digit accuracy/precision. I don't think I have a device that can do this accurately.
The OP has a 5-1/2 digit. My Fluke 87 3.5 digit in mV mode can do .01mV resolution which is not hugely special. 1mOhm should be fine.
I have several 6-1/2 digit DVMs for voltage measurement. That doesn't mean I trust them at 0.01mV resolution, let alone the 0.0001 mV that would be needed to measure a 1 mV shunt with 4 1/2 digit accuracy. As I said previously, I am not calibrating a $50 Chinese Hung Low supply here, I am calibrating this:
http://www.keysight.com/en/pd-838347-pn-6611C/special-order-power-supply-10v-5a?cc=US&lc=eng (http://www.keysight.com/en/pd-838347-pn-6611C/special-order-power-supply-10v-5a?cc=US&lc=eng)
Note in particular the fact that this power supply measures down to 0.0001 A (i.e., 100 microamp resolution). I want to retain a reasonable level of accuracy here.
-
With a 1 mOhm shunt and 5 A current source, I would have to measure 5 mV across the shunt with 4 1/2 digit accuracy/precision. I don't think I have a device that can do this accurately.
The OP has a 5-1/2 digit. My Fluke 87 3.5 digit in mV mode can do .01mV resolution which is not hugely special. 1mOhm should be fine.
I have several 6-1/2 digit DVMs for voltage measurement. That doesn't mean I trust them at 0.01mV resolution, let alone the 0.0001 mV that would be needed to measure a 1 mV shunt with 4 1/2 digit accuracy. As I said previously, I am not calibrating a $50 Chinese Hung Low supply here, I am calibrating this:
http://www.keysight.com/en/pd-838347-pn-6611C/special-order-power-supply-10v-5a?cc=US&lc=eng (http://www.keysight.com/en/pd-838347-pn-6611C/special-order-power-supply-10v-5a?cc=US&lc=eng)
Note in particular the fact that this power supply measures down to 0.0001 A (i.e., 100 microamp resolution). I want to retain a reasonable level of accuracy here.
Well...why didn't you say so in the first place ?
You should have shown us this in the beginning and saved everyone a lot of typing.
You need to get your hands on a more capable measuring device.
-
You need to get your hands on a more capable measuring device.
+1
Calibrating to 100uA needs something that is calibrated and accurate to 10uA. Take it to a lab?
-
You need to get your hands on a more capable measuring device.
+1
Calibrating to 100uA needs something that is calibrated and accurate to 10uA. Take it to a lab?
It also appears measuring current is just part of the tip of the iceberg if "relative accuracy" is to be maintained.
-
You need to get your hands on a more capable measuring device.
+1
Calibrating to 100uA needs something that is calibrated and accurate to 10uA. Take it to a lab?
Just because it displays 100 uA, doesn't mean it is really accurate to 100 uA. I would be happy to take 5 A +/- 500 uA.
Also, a lab would probably charge me >$250 to calibrate this thing...
-
You need to get your hands on a more capable measuring device.
+1
Calibrating to 100uA needs something that is calibrated and accurate to 10uA. Take it to a lab?
It also appears measuring current is just part of the tip of the iceberg if "relative accuracy" is to be maintained.
You see that Guildline 9230/15 current shunt they recommend to calibrate the 5A range? Good luck finding one!
The rest of the equipment they list is unnecessary to calibrate the thing. Specifically, I was able to calibrate the low current range with my Fluke 5 1/2 and a high precision 500 ohm resistor, the voltages with my handful of 6 1/2 DVMs and the high current range are all that's left - that's the hard part...
Getting an accurate ammeter that can do 5 A is a b**ch.
-
There are some better ones at digi key for about $30-40 I think.
Sent from mobile device.... Keeping it short and mis-spelled
-
There are some better ones at digi key for about $30-40 I think.
Sent from mobile device.... Keeping it short and mis-spelled
If you know of one that can produce max voltage (>= 0.1 volts) and load 10 volts with 5 A (i.e. <=2 ohms) and doesn't suffer from temperature related drift and is accurate to four places in terms of its specified resistance and costs $30-$40, please let me know because I can find no such beast.
-
Physics cannot be trifled with. I was not suggesting a miracle.
Sent from my SAMSUNG-SM-G890A using Tapatalk
-
So I made my own current shunt out of:
1 foot of 8 AWG tinned copper wire
2 Nakamichi banana plugs (the wire was pre-tinned and soldered into the plugs to make a damn good connection)
1 16 AWG pig tail soldered to each banana plug for the voltage sense
Even with this contraption, that barely changes temperature when presented with a 5 A load, there is a decent amount of temperature related voltage drift even at 1 A, as the PS tries to maintain the 1 A load. It seems to me that short of a long thick copper bar, I don't know how you can get a shunt to not change its resistance as the load increases into the amps. Pulsing the load is still out due to the fact that I am not in control of the power supply's calibration procedure.
I have somewhat lost faith in current shunts though...
-
Have you looked through this to see if there is a meter that will do
https://www.eevblog.com/forum/testgear/multimeter-spreadsheet/ (https://www.eevblog.com/forum/testgear/multimeter-spreadsheet/)
-
Don't lose faith, just use a lower TC shunt material or apply temperature compensation.
https://en.wikipedia.org/wiki/Manganin
You can buy manganin foil to make shunts with, for instance, or premade shunts.
That will help the TCR issue.
Also for many ampere currents it is not uncommon to use significantly heavier conductors such as thick foil or thin sheet.
You could also air cool or water cool the shunt. Obviously heatsinking and forced air cooling is commonly used for power electronics applications because as you said the temperature does tend to rise without it.
If your wire is insulated, try coiling it into a bucket of room temperature water.
You could also just measure the voltage drop across the wire and look for changes proportional to time to try and estimate a correction factor.
Wow, so many informative ideas in one post.
Thank you!!! :-+
-
solder one thick cable with two small wire at both end, its will make it as big shunt resistor. put 2A current (tested using your amp meter) and measure the voltage so you get the resistance...
and you get a fairly accurate shunt resistor
-
solder one thick cable with two small wire at both end, its will make it as big shunt resistor. put 2A current (tested using your amp meter) and measure the voltage so you get the resistance...
and you get a fairly accurate shunt resistor
If you look at Reply #36 to this thread, I have already done this. The problem is that the resistance of the thick (8 AWG) wire changes with time/temperature. It does this enough to make this a fairly inaccurate shunt resistor for what I am trying to do.
-
solder one thick cable with two small wire at both end, its will make it as big shunt resistor. put 2A current (tested using your amp meter) and measure the voltage so you get the resistance...
and you get a fairly accurate shunt resistor
If you look at Reply #36 to this thread, I have already done this. The problem is that the resistance of the thick (8 AWG) wire changes with time/temperature. It does this enough to make this a fairly inaccurate shunt resistor for what I am trying to do.
Can you dump 5A through it for long enough to reach equilibrium then disconnect the PSU and take a very fast measurement of the resistance. You should be able to do it inside two seconds if you have the resistance probe in one hand and the other hand on the PSU switch.
-
For simple ammeters that don't have active compensation for the shunt temperature, to get 3 1/2 digit accuracy with a copper shunt, the shunt temperature *MUST* be controlled to within 0.1 deg C. This is an unavoidable consequence of copper's temperature coefficient of resistivity of 4.29E-3 per deg C. To get 4 1/2 digit accuracy you need a Constantin shunt, which is two orders of magnitude less temperature sensitive (3E-5 per deg C), and you'll still need to control its temperature to within 1 deg C. Manganin is better than Constantin, but even so the temperature will need to be controlled to within a few deg C
There are no workarounds - if you are dissipating significant power in the shunt, temperature control in anything other than a stirred ice-water bath will be a major problem, and using water baths for electrical work even at low voltage is a royal P.I.T.A
-
solder one thick cable with two small wire at both end, its will make it as big shunt resistor. put 2A current (tested using your amp meter) and measure the voltage so you get the resistance...
and you get a fairly accurate shunt resistor
If you look at Reply #36 to this thread, I have already done this. The problem is that the resistance of the thick (8 AWG) wire changes with time/temperature. It does this enough to make this a fairly inaccurate shunt resistor for what I am trying to do.
Can you dump 5A through it for long enough to reach equilibrium then disconnect the PSU and take a very fast measurement of the resistance. You should be able to do it inside two seconds if you have the resistance probe in one hand and the other hand on the PSU switch.
Wire up a stable 4 wire resistance measurement of something like 7 milliohms and measure it with accuracy and repeatability in a total 2 seconds? I am not the Flash :scared: and neither is any of my 5 1/2 or 6 1/2 digit multimeters.
In all seriousness, even if I was superman I cannot disconnect the load from the supply without aborting that phase of the calibration procedure. Remember that the PS is sourcing 5 A into the load and the load needs to stay relatively constant throughout the procedure until its current draw is entered into the PS (and not disappear for a quick 4-wire resistance measurement mid calibration).
-
Remember that the PS is sourcing 5 A into the load and the load needs to stay relatively constant throughout the procedure until its current draw is entered into the PS
Have you said how long you have to run it at 5 A ? A lot of responses say 2 seconds but my guess is you need more than 2 sec.
-
Remember that the PS is sourcing 5 A into the load and the load needs to stay relatively constant throughout the procedure until its current draw is entered into the PS
Have you said how long you have to run it at 5 A ? A lot of responses say 2 seconds but my guess is you need more than 2 sec.
I need a heck of a lot longer than 2 seconds. The power supply will supply 4.x A of current and I need to measure it and enter the value into the power supply accurate to 4 decimal digits (e.g., 4.7824 A ) without disconnecting the load from the PS. So, I need 100-200 uA level accuracy at 5 amps (i.e., enter a value that is +/- 200 uA of current sourced).
Perhaps this explanation will elucidate the difficulty of my task.
-
Are you planning to sell this once you've calibrated it with bits of wire and buckets of water or is this for your own use?
If for your own use and you have nothing accurate enough to calibrate it, then what's the point of the accuracy you're chasing? i.e. if the rest of your equipment was up to the standards you're trying to calibrate the PS to and you NEED that accuracy I can see the point, otherwise....
Or is this some kind of volt-NUT thing going on?
cheers,
george.
-
Physics cannot be trifled with. I was not suggesting a miracle.
Sent from my SAMSUNG-SM-G890A using Tapatalk
You can solve problem with finesse or the brute force way. I recall the story of the SR71 spyplane.
There was a problem of the fuel tank at high altitude, they had to take in account of the rate of expansion of metal at high temperatures flying at mach3. So the designer made it in such a way that while plane was on the ground it was leaking fuel and when up at high altitude it has a perfect seal, no leakage. :o
So coming back to your problem, can you calibrate in a cold room/ freezer or at least use dry ice in a container
( there will be condensation ) ? The question is will the instrumentation work at low temperatures?
-
Forget the ammeter completely, build (or buy) a shunt and measure the voltage drop across it instead, you still have your accuracy, and is actually likely to be more precise than a mishmash of ammeter and parallel shunt.
-
Forget the ammeter completely, build (or buy) a shunt and measure the voltage drop across it instead, you still have your accuracy, and is actually likely to be more precise than a mishmash of ammeter and parallel shunt.
I am leaning in this direction and have just purchased 4 different shunts from 5 A to 20 A, to see which of them will help solve this problem.
-
Are you planning to sell this once you've calibrated it with bits of wire and buckets of water or is this for your own use?
If for your own use and you have nothing accurate enough to calibrate it, then what's the point of the accuracy you're chasing? i.e. if the rest of your equipment was up to the standards you're trying to calibrate the PS to and you NEED that accuracy I can see the point, otherwise....
Or is this some kind of volt-NUT thing going on?
cheers,
george.
To try to answer all of your questions:
Personal use, there is some volt-nuttedness going on and the rest of my equipment is way beyond that level of accuracy.
-
Well, if it is the volt-NUT stuff, then good luck to you and hopefully you can sleep well at night knowing that all your equipment tolerances are drifting as they age :)
Hope your shunts allow 4 wire connection/measurement or you'll have to factor in the inaccuracies of your probing/measuring and wiring....
cheers,
george.
-
I think I have "stumbled" upon a solution. I have four 2 ohm 100 watt power resistors. If I connect them in parallel and taking their tolerance into account, I get a measured resistance of something like 0.575 ohms. The exact value is unimportant as long as it is under 0.6 ohms.
Also, because these are heatsinked 100 W power resistors, they don't change their resistance too much when 3 A is applied to the bunch in parallel (0.75 A per resistor at a fairly low voltage of 1.725 V or so = 1.29375 W per resistor).
So, if I take the four in parallel and connect my 2 A ammeter in parallel as well, I should be able to measure 5 A (as just under 2 A on the ammeter) and then scale the result based on a prior measurement of a known 2 A current source to get an accurate reading.
Any thoughts or flaws with this idea that I have overlooked?
This is the power supply I am trying to calibrate with this Frankenstein approach (don't laugh!!):
http://www.keysight.com/en/pd-838347-pn-6611C/special-order-power-supply-10v-5a?cc=US&lc=eng (http://www.keysight.com/en/pd-838347-pn-6611C/special-order-power-supply-10v-5a?cc=US&lc=eng)
You are stuck on using the DMM in current measuring mode. This is unnecessary if you have an external shunt. Measure the shunt resistance (your in-cal meter will be more accurate than the tolerance of the resistors themselves) and then measure the voltage drop across the shunt as you pass the current through it, and use good old ohms law to calculate the current. If your meter can do basic math (ax+b) then you can even get it to display the result in amps directly. This is, after all, exactly what any digital ammeter does internally.
The basic error spec for this measurement will be the sum of the resistance and the voltage measurement error specs. I'd also measure the resistance of the resistors hot (after use) to determine if their value has changed significantly due to self heating, as this is another source of error.
-
Just throwing this out if using resistors
Heat them up with 5a, quickly measure at 2 a, then measure at 5a ? If quickly is slow, maybe you could insulate the resistors ?
-
You are stuck on using the DMM in current measuring mode. This is unnecessary if you have an external shunt. Measure the shunt resistance (your in-cal meter will be more accurate than the tolerance of the resistors themselves) and then measure the voltage drop across the shunt as you pass the current through it, and use good old ohms law to calculate the current. If your meter can do basic math (ax+b) then you can even get it to display the result in amps directly. This is, after all, exactly what any digital ammeter does internally.
He is looking for a way to "relatively"* accurately measure current, while at the same time he doesn't have anything available that will measure current , voltage or resistance accurately*
This is what I call a lost cause thread
*information from previous posts
-
He is looking for a way to "relatively"* accurately measure current, while at the same time he doesn't have anything available that will measure current , voltage or resistance accurately*
The OP has 6 1/2 digit meters which are way beyond my accuracy requirements but to this forum I think he needs a 8 1/2 meter (its all relative). So I suggest he give out his location and ask to borrow the use of one. You never know who has one in their garage. Another idea - rent one. This is after he gets his shunts and heat controls them. How about a toaster oven, they are cheap.
I have enjoyed this thread as a mind game.
-
Out of curiosity, what are you doing that requires precision regulated current at 5amps?
Some things come to mind, electroplating a precise thicknesses of material? Electrophoresis machine for home paternity test lab? metrology, characterizing shunts? electrohydrodynamic thruster?
In any case, the reason you can't find one is that it isn't really that commonly needed for a single instrument, if you are going to have a very high precision meter, a voltage meter is much more generally useful and external precision shunts can be sized as appropriate for the task. All in one tools make a lot more sense at the picoamp level than the dekaamp level.
Of course, you can always break the budget (and be everyones envy) by getting a high end SMU and solve your power supply and accuracy issues in one go
https://www.youtube.com/watch?v=zT80zDZgOWU (https://www.youtube.com/watch?v=zT80zDZgOWU)
7 amps with uA resolution, pA resolution at lower currents. (watch his review of the 2450 too, great stuff, who else uses liquid nitrogen as part of a review of electronics lab equipment?)
-
He is looking for a way to "relatively"* accurately measure current, while at the same time he doesn't have anything available that will measure current , voltage or resistance accurately*
This is what I call a lost cause thread
*information from previous posts
I don't think that this is a correct summary of the status quo. My understanding is that he can measure resistance and voltage with adequate accuracy (6.5 digits DMM?), but cannot measure current (at 5A) with the desired accuracy. Sharpears, can you please confirm?
If the above is correct, I would also suggest to measure the voltage over a decent quality shunt resistor. Can't you let it warm up and equilibrate by applying 5A (or whatever current setting you want to calibrate) over a long time, then remove the current supply and quickly measure the actual resistance? If you use a power resistor with decent thermal-mass heatsink attached, the temperature should not drift much over the course of a few seconds. And you can of course estimate the drift by continuing to measure the resistance for a few more seconds afterwards.
-
- Borrow a DMM from someone.
- You probably going to need 2A-5A more often since you want it correct in that range, hence buy a new DMM.
-
This is what I call a lost cause thread
This is what I call not reading all of the replies in this thread and coming to an erroneous conclusion.
-
He is looking for a way to "relatively"* accurately measure current, while at the same time he doesn't have anything available that will measure current , voltage or resistance accurately*
This is what I call a lost cause thread
*information from previous posts
I don't think that this is a correct summary of the status quo. My understanding is that he can measure resistance and voltage with adequate accuracy (6.5 digits DMM?), but cannot measure current (at 5A) with the desired accuracy. Sharpears, can you please confirm?
If the above is correct, I would also suggest to measure the voltage over a decent quality shunt resistor. Can't you let it warm up and equilibrate by applying 5A (or whatever current setting you want to calibrate) over a long time, then remove the current supply and quickly measure the actual resistance? If you use a power resistor with decent thermal-mass heatsink attached, the temperature should not drift much over the course of a few seconds. And you can of course estimate the drift by continuing to measure the resistance for a few more seconds afterwards.
Your summary is correct!
The thing people are seemingly ignoring is that at the 5 1/2 digit level, thermal effects are non-negligible. Therefore, pre-heating and other games do not solve the problem. You either have the issue of it heating up or of it losing its heat if pre-heated.
I have ordered some 4-wire shunts from China and will post back on how they fare. I got: 5A, 10A, 15A and 20A in order to see what the level of temperature drift in them will be once used. I would imaging that the 20A has the lowest temp related drift, but also measure 75 mV at 20A, so I will have a reduction in measurement range at 5A.
-
So what does one use a constant current psu for that outputs 5A for personal use?
I can think of only a few main options:
(a) quantitative electrochemistry, coulometry etc.
(b) as a calibration source for some other lesser tightly specified equipment like regulators, loads, whatever.
(c) generating constant magnetic fields for some kind of sensor / field related physics experiment (MRI or such)
I will just not answer your question with links to one of my other threads:
https://www.eevblog.com/forum/testgear/hpagilent-6675a-the-120-v-18-a-2000w-monster-power-supply/ (https://www.eevblog.com/forum/testgear/hpagilent-6675a-the-120-v-18-a-2000w-monster-power-supply/)
https://www.eevblog.com/forum/testgear/hpagilent-6675a-the-120-v-18-a-2000w-monster-power-supply/msg569237/#msg569237 (https://www.eevblog.com/forum/testgear/hpagilent-6675a-the-120-v-18-a-2000w-monster-power-supply/msg569237/#msg569237)
-
He is looking for a way to "relatively"* accurately measure current, while at the same time he doesn't have anything available that will measure current , voltage or resistance accurately*
This is what I call a lost cause thread
*information from previous posts
I don't think that this is a correct summary of the status quo. My understanding is that he can measure resistance and voltage with adequate accuracy (6.5 digits DMM?), but cannot measure current (at 5A) with the desired accuracy. Sharpears, can you please confirm?
If the above is correct, I would also suggest to measure the voltage over a decent quality shunt resistor. Can't you let it warm up and equilibrate by applying 5A (or whatever current setting you want to calibrate) over a long time, then remove the current supply and quickly measure the actual resistance? If you use a power resistor with decent thermal-mass heatsink attached, the temperature should not drift much over the course of a few seconds. And you can of course estimate the drift by continuing to measure the resistance for a few more seconds afterwards.
Your summary is correct!
The thing people are seemingly ignoring is that at the 5 1/2 digit level, thermal effects are non-negligible. Therefore, pre-heating and other games do not solve the problem. You either have the issue of it heating up or of it losing its heat if pre-heated.
I have ordered some 4-wire shunts from China and will post back on how they fare. I got: 5A, 10A, 15A and 20A in order to see what the level of temperature drift in them will be once used. I would imaging that the 20A has the lowest temp related drift, but also measure 75 mV at 20A, so I will have a reduction in measurement range at 5A.
The shunt inside a DMM isn't made of pixie dust, and so it is subject to the same self-heating concerns. It is something that is always there. What about the current measuring shunt in the power supply that you are calibrating/adjusting with the setup in question? You will have the same concerns there too. You might want to have another look at that and decide how far you need to go in the calibration setup in order to be accurate enough.
Using a low resistance shunt will help significantly. The issue that arises then is the low voltage generated across it. You will need to control the thermoelectric effect in the voltage measurement connections, and of course, you will need to null the voltage measurement at zero current. Any heating of the connection could cause a voltage to be generated at the connections, but only if the heating is uneven. If both connections heat up an equal amount, then the voltages cancel and contribute nothing. That is what I meant by controlling the thermoelectric effect - you can't eliminate it.
-
The shunt inside a DMM isn't made of pixie dust, and so it is subject to the same self-heating concerns. It is something that is always there. What about the current measuring shunt in the power supply that you are calibrating/adjusting with the setup in question? You will have the same concerns there too. You might want to have another look at that and decide how far you need to go in the calibration setup in order to be accurate enough.
But it is made of unicorn horns - good DMMs adjust for temperature changes of the shunt and have circuitry to do this. I don't...
How else can a 5 1/2 digit ammeter meter keep a steady reading at 2 A?
-
If the shunt is in an instrument, a sensor to measure the shunt temperature and look up a temperature dependent calibration factor is simple to implement. Its also possible to ovenise it. No pixie dust required.
As I said a lot earlier, for high accuracy bench work, you need accurate temperature control of the shunt. That's difficult without either an actively temperature controlled heatsink, (and its going to be a PITA to hold much better than +/-0.5 deg C), or an ice-water bath (as the melting point of pure ice at 1 atmosphere is well defined and stable, and it takes a LOT of energy to melt ice). A water bath is a PITA because of its conductivity. Unless you use high purity deionised water, you will have to totally seal and insulate the shunt. Fortunately you are working with very low voltages across the shunt, so any water resistant lacquer, varnish or conformal coating should do if you pay particular attention to sharp edges and corners, and make sure the bath itelf is electrically floating. Your next problem is stirring an ice bath without damaging the shunt or its insulating coating. I'd tackle that by fitting a drain 2/3 of the way up the side of an expanded polystyrene cooler or ice bucket, and putting a small submersible pump at the bottom of it with a pipe running to the top, as far from the pump as possible. Fill with cube ice, with the shunt secured in the middle 1/3 of the way from the bottom, and add cold water till it reaches the overflow level. Start the pump, and keep it topped up with ice, and it will hold 0 deg C. The pump supply needs to be isolated, and the drain must drip (rather than a continuous flow through a small bore pipe) to keep the bath electrically floating.
-
If the shunt is in an instrument, a sensor to measure the shunt temperature and look up a temperature dependent calibration factor is simple to implement. Its also possible to ovenise it. No pixie dust required.
As I said a lot earlier, for high accuracy bench work, you need accurate temperature control of the shunt. That's difficult without either an actively temperature controlled heatsink, (and its going to be a PITA to hold much better than +/-0.5 deg C), or an ice-water bath (as the melting point of pure ice at 1 atmosphere is well defined and stable, and it takes a LOT of energy to melt ice). A water bath is a PITA because of its conductivity. Unless you use high purity deionised water, you will have to totally seal and insulate the shunt. Fortunately you are working with very low voltages across the shunt, so any water resistant lacquer, varnish or conformal coating should do if you pay particular attention to sharp edges and corners, and make sure the bath itelf is electrically floating. Your next problem is stirring an ice bath without damaging the shunt or its insulating coating. I'd tackle that by fitting a drain 2/3 of the way up the side of an expanded polystyrene cooler or ice bucket, and putting a small submersible pump at the bottom of it with a pipe running to the top, as far from the pump as possible. Fill with cube ice, with the shunt secured in the middle 1/3 of the way from the bottom, and add cold water till it reaches the overflow level. Start the pump, and keep it topped up with ice, and it will hold 0 deg C. The pump supply needs to be isolated, and the drain must drip (rather than a continuous flow through a small bore pipe) to keep the bath electrically floating.
In lieu of this, what about something simpler like a relatively high current capable shunt (e.g., 20 A), which should not heat up very rapidly. Then I would count on a high accuracy voltage measurement at 2 A as measured on an ammeter (maybe even 4 A, if I put my 5 1/2 digit ammeters in parallel) and extrapolate it to 5 A.
Even better and closer to your idea would be to use 4-8 strands of 16 AWG enameled magnet wire, coiled in a room temp water bath. That way, you get around 1 A going through each magnet wire and the water keeps the temperature relatively constant (i.e., very very slow shift from ambient if a 5 gallon bucket is used).
-
You should be able to take a hacksaw to the 20a shunt to get higher volts
-
You should be able to take a hacksaw to the 20a shunt to get higher volts
Dremel tool more likely, but great idea! I was also thinking about somehow attaching a heat sink to it, with an insulating thermal compound (e.g., Ceramique) for even better heat dissipation. Another idea was to wire 4-8 20 A shunts in series.
By the way, if I submerge the shunt in water will the voltage across it stay constant or will electron flow through the water and electrolysis at 15-20 mV screw up the voltage reading.
-
By the way, if I submerge the shunt in water will the voltage across it stay constant or will electron flow through the water and electrolysis at 15-20 mV screw up the voltage reading.
Yes, propably. I would use oil to cool the shunts. Search for transformer oil.
Greetings
-
Rig a temperature probe in good thermal contact with the shunt. Use a block of polystyrene to add thermal insulation over the probe. Set up to continuously measure the resistance to at least 5 digits using the four wire method. Heat the shunt up with a hot air gun to at least 30 deg above ambient. Let it cool naturally to ambient while logging the temperature and resistance. Run 5A through the shunt and note the temperature rise. You'll soon see if you will be able to hold 4 1/2 digit accuracy over the range 0-5A without temperature control (or if your Chinese supplier has stiffed you by supplying plated brass fake shunts rather than genuine Manganin or Constantan ones).
If you choose to immerse the shunt in a water bath, the lacquer/varnish/conformal coating prevents any significant electrolysis. All connections should be 'gooped' with a setting sealant or preheated and coated with hotglue. Any pin holes in the coating will result in some electrolysis if dissimilar metals are exposed, but the minuscule exposed area and relatively low conductivity of reasonably pure water will limit it to a negligible current, if the coating is applied carefully. Remove the shunt from the bath and dry carefully after use as the coating is likely to break down if stored wet.
The problem with an oil bath is temperature control. You'd need a stirrer in the oil bath and either a heat exchanger to or the whole bath immersed in an ice-water bath. An ice-water bath is the cheapest way to hold a fixed temperature.
-
The shunt inside a DMM isn't made of pixie dust, and so it is subject to the same self-heating concerns. It is something that is always there. What about the current measuring shunt in the power supply that you are calibrating/adjusting with the setup in question? You will have the same concerns there too. You might want to have another look at that and decide how far you need to go in the calibration setup in order to be accurate enough.
But it is made of unicorn horns - good DMMs adjust for temperature changes of the shunt and have circuitry to do this. I don't...
How else can a 5 1/2 digit ammeter meter keep a steady reading at 2 A?
The answer is that they don't.
Check the accuracy spec for the 2+A range of a 5.5 or higher digit DMM. For my Keithley 199 meters, it is 0.1%+15 counts. At 2 A and 5.5 digits, that is an error margin of +-215 counts. For my Keithley 2001 7.5 digit meter, the error on the 2 A range is 500 ppm of reading + 20 ppm of range + an extra 50 ppm of range "above 0.5 A for self heating". So for a 2 A reading at 5.5 digits, that is an error margin of +-114 counts. That's on a $5k meter. The accuracy on the higher current ranges is by far the worst of all the functions and ranges, and you can guess why.
-
This is what I call a lost cause thread
This is what I call not reading all of the replies in this thread and coming to an erroneous conclusion.
I have a 2 page attention span, if you know what I mean ;)
-
The shunt inside a DMM isn't made of pixie dust, and so it is subject to the same self-heating concerns. It is something that is always there. What about the current measuring shunt in the power supply that you are calibrating/adjusting with the setup in question? You will have the same concerns there too. You might want to have another look at that and decide how far you need to go in the calibration setup in order to be accurate enough.
But it is made of unicorn horns - good DMMs adjust for temperature changes of the shunt and have circuitry to do this. I don't...
How else can a 5 1/2 digit ammeter meter keep a steady reading at 2 A?
The answer is that they don't.
Check the accuracy spec for the 2+A range of a 5.5 or higher digit DMM. For my Keithley 199 meters, it is 0.1%+15 counts. At 2 A and 5.5 digits, that is an error margin of +-215 counts. For my Keithley 2001 7.5 digit meter, the error on the 2 A range is 500 ppm of reading + 20 ppm of range + an extra 50 ppm of range "above 0.5 A for self heating". So for a 2 A reading at 5.5 digits, that is an error margin of +-114 counts. That's on a $5k meter. The accuracy on the higher current ranges is by far the worst of all the functions and ranges, and you can guess why.
Actually, on a 7.5 digit meter I can't guess why. One would think that they would have an appropriately thick, heat dissipated and maybe even temperature compensated current shunt that, with 7.5 digit precision on the voltage side, could give a reasonably accurate current reading via voltage across the shunt.