The µCurrent Mark II

[Pilot3514]

I am interested in hearing what you thought about a new and improved version of the µCurrent.  (Is that being redundant?  Could it be improved without being new?  But I digress.)

I have been trying to follow along with some of the resent projects and both of the resent power supplies have microcontrollers.  What do you think of doing a µCurrent with a microcontroller?

Features (if practical):
- new slimmer case, as used on the USB µSupply.
- rechargeable battery, same as the USB µSupply.
- LCD display, same as the USB µSupply.
- USB interface, to both charge and output readings.
- auto-ranging.
- no switches (except on/off)

I am thinking of a current meter that does not need another meter to work.  When you need to read current, you don't need to drag out both your µCurrent and your multi-meter.  If the current you are going to measure is more than an amp, reach for your Fluke.  If less than an amp, grab you µCurrent.

I was also thinking that you may be able to offset some of the cost increases from the need for a micro and display by using 1% resistors instead of 0.1%.  The error in the resistors can be calibrated out in the micro.  Also by using some of the same components as the USB µSupply you might get more economies of scale.

In the µCurrent you have a 75K and 24K in series to get 99K.  Could that be replace with a 100K and calibrate out the difference?

Could you use some kind of FET to switch in the shut resistors?  I don't know if this will work or what the difference is for the parts costs and assembly costs.

I saw the video about how the µCurrent is tested to see if it is within spec.  My thought is that the Mark II should be design to be easier to test.  With the input lugs on one end, same as the USB µSupply, and the USB port on the other end.  It should be easy to plug them up to a test jig.  The tricky part is the current source.  If you have a calibrated current source that can be computer controlled, then software can do everything else.  This is where auto-ranging would be so useful because the calibration and testing software could sweep through all ranges.

If we assume that you don't have a calibrated current source that can computer controlled then we build one.  If you manually calibrate one µCurrent, then that becomes your calibration standard.  Put it in series with the one to be calibrated, you can then calibrate one to the other.  (Correct me if I'm wrong here.  I watch the videos on calibration but I am no expert)  All you need is a variable current source, even if it is not a calibrated source.  That can be your µSupply, USB µSupply, or Atten PPS3205T-3S and a limiting resistor.  How long would it take to calibrate one µCurrent, 10 seconds?  If it takes too long, put several of them in series and calibrate them all at the same time.

Let me know if my rose colored glasses and wishful thinking are getting in the way of reality.
(Reality is for those poor soles that can not handle fantasy)

[alm]

Features (if practical):
[...]

Sounds like this would seriously increase your BOM costs.

I am thinking of a current meter that does not need another meter to work.  When you need to read current, you don't need to drag out both your µCurrent and your multi-meter.  If the current you are going to measure is more than an amp, reach for your Fluke.  If less than an amp, grab you µCurrent.

Measuring the voltage with even 3.5 digit accuracy will require a fairly decent ADC. At least 12 bits of performance, which is not something you generally get from the internal ADC.

I was also thinking that you may be able to offset some of the cost increases from the need for a micro and display by using 1% resistors instead of 0.1%.  The error in the resistors can be calibrated out in the micro.

Assuming they are constant. You still need resistors with a good tempco. The resistor will heat up as you increase the current. Standard 1% resistors might have a 100 ppm tempco. Heating them up by 10°C will already change the value by up to 0.1 %, which was the accuracy you were trying to get. I guess you could attempt to compensate for this by doing a non-linear correction.

Could you use some kind of FET to switch in the shut resistors?  I don't know if this will work or what the difference is for the parts costs and assembly costs.

The lowest resistance shunt resistor is 10 mOhm. What's the RDS(on) of your FET?

[tom66]

In the µCurrent you have a 75K and 24K in series to get 99K.  Could that be replace with a 100K and calibrate out the difference?

No, because then the gain would be 101%.

99K and 1K = gain of 100

[Pilot3514]

In the µCurrent you have a 75K and 24K in series to get 99K.  Could that be replace with a 100K and calibrate out the difference?

No, because then the gain would be 101%.

99K and 1K = gain of 100

I understand that it would alter the gain of the amplifier but why could that not be taken into account by the software.

I am also proposing that tolerance of the resistors be lowered from 0.1% to 1% so the gain might be anywhere from 100 to 102.

Provided the value is consistent and does not drift with time or temperature.

[Pilot3514]

Sounds like this would seriously increase your BOM costs.

That is a good point.  That is why I was looking to use part already included in other projects.  I understand that there are many new parts required.  Operating off a rechargeable battery means you have the cost of the battery and the cost of a battery charger.  The micro, which leads to a voltage regulator and crystal.  A USB interface leads to a mini or micro B connector.  Pull up resistors, decoupling caps, voltage reference and so on.  How much does this increase the cost of the BOM.  Not only that but what does this do to assembly costs?  When things are all added up, are the new features worth the additional cost?  I don't know the answer.

Measuring the voltage with even 3.5 digit accuracy will require a fairly decent ADC. At least 12 bits of performance, which is not something you generally get from the internal ADC.

That is very true.  If you need to know how many milliamps you are using down to the microamp level.  Or how my microamps, down to the picoamp level.  Then a 10 bit ADC will not give you that information.  That is why you look at the specs up front. 
- What are the best specs you can get with a 10 bit ADC?  Is that good enough?
- What is the cost of adding a 12 bit ADC?  Is the additional accuracy worth the cost?
These answers are a judgment call.  Two people can have different answers and both be right.

Assuming they are constant. You still need resistors with a good tempco. The resistor will heat up as you increase the current. Standard 1% resistors might have a 100 ppm tempco. Heating them up by 10°C will already change the value by up to 0.1 %, which was the accuracy you were trying to get. I guess you could attempt to compensate for this by doing a non-linear correction.

I may be worn but I don't see the shunt resistors dissipating very much energy.  The whole idea of the product is to reduce the "burden voltage".  So the voltage drop across the shunt should be very low.  And the product is intended to measure low currents, hence the name.  So the power dissipated is very low voltage times very low current.  What does that equal "very low squared" power?  I think more of a problem would be changes in ambient temperature.  Are you in a climate controlled work area or out in the field on a summer or winter day?
I had assumed that the temperature coefficient would be more a function of what the resistor was made of and how the resistor was made.  I did not know it was somehow related to it's tolerance.  I guess you could try to calibrate the unit across a wide temperature range but that would be a real pain and probably not worth the effort.  If the drift where known and predictable you might be able to add that to the software but now you have added a temperature measuring device to the BOM.  Maybe you could sell that as a feature - "Not only can you measure current down to picoamps, you can also tell what the temperature is on your work bench"

The lowest resistance shunt resistor is 10 mOhm. What's the RDS(on) of your FET?

If it where always that value and consent across a wide range of temperature and currents then, of course, you could use the FET instead of the resistor.  However, that is too much to ask for.  The simplest solution is to do what most multimeters do in that situation, you plug your lead into a different jack.  I also considered suggesting a switch or reed relay for that range but I do not know the contact resistance of the relay, it would also add to the cost (but so would another binding post).

All of your points are very good.  They raise many more questions than I have answers to.  I'm not an engineer.  Instead, I'm like an adolescent, I see all the possibilities and none of the problems.

[tom66]

I understand that it would alter the gain of the amplifier but why could that not be taken into account by the software.

I am also proposing that tolerance of the resistors be lowered from 0.1% to 1% so the gain might be anywhere from 100 to 102.

Provided the value is consistent and does not drift with time or temperature.

Because Dave probably values his time. I understand he currently does not calibrate these but simply validates their performance. In which case, your proposed method would save ~$0.01, and cost Dave maybe 30 seconds of his time per unit. If he works for any less than $1.20/hr then it may be worth it to him.

[Pilot3514]

Because Dave probably values his time. I understand he currently does not calibrate these but simply validates their performance. In which case, your proposed method would save ~$0.01, and cost Dave maybe 30 seconds of his time per unit. If he works for any less than $1.20/hr then it may be worth it to him.

As I understand, he validates the performance to see if it falls within spec for one value in each range.  That my be good enough and does take some time.  Since with the aid of a computer, the units could probably be validated and calibrated in about the same amount of time, I don't see where any of Dave's valued time is lost.  By calibrating the units, they might even have better specs than they currently do.
Since the calibration and validation is being performed by a computer, it would also be possible to profile the unit and provide a calibration certificate and a printed graph/chart with each unit.  That my be gilding the lily, but it is an option that would cost little more that paper.

[c4757p]

I think the whole idea is gilding the lily. The µCurrent is an adapter. Once you add this many features to it, you've taken a nice, well designed adapter and made a half-assed one function "multi"meter. It's like taking a perfectly functional bike and adding a lawnmower motor. Either keep it a bicycle or get a motorcycle.

[amspire]

I have been trying to follow along with some of the resent projects and both of the resent power supplies have microcontrollers.  What do you think of doing a µCurrent with a microcontroller?

They type of integrating ADC in a multimeter does a much better job at getting accurate and stable measurements in the presence of a noisy DC then an microcontroller type sample and hold ADC.

Down at the sub-microamp  level, there is often a lot of noise.

The integrating ADCs are totally monotonic by design (a small increase in current will never give a small decrease in the reading), and usually have extremely good linearity. Microcontroller ADCs may not be monotonic, and their linearity may have issues.

Basically the uCurrent/Multimeter combination is a very good solution.

[EEVblog]

I am interested in hearing what you thought about a new and improved version of the µCurrent.  (Is that being redundant?  Could it be improved without being new?  But I digress.)

I have been trying to follow along with some of the resent projects and both of the resent power supplies have microcontrollers.  What do you think of doing a µCurrent with a microcontroller?

Features (if practical):
- new slimmer case, as used on the USB µSupply.
- rechargeable battery, same as the USB µSupply.
- LCD display, same as the USB µSupply.
- USB interface, to both charge and output readings.
- auto-ranging.
- no switches (except on/off)

I am thinking of a current meter that does not need another meter to work.  When you need to read current, you don't need to drag out both your µCurrent and your multi-meter.  If the current you are going to measure is more than an amp, reach for your Fluke.  If less than an amp, grab you µCurrent.

It is now not a uCurrent Mk2, but a standalone low current meter. An entirely different product, and one which would not get much use. The uCurrent is designed to be simple for that occasional use when you need it.
The BOM cost has gone up a lot, plus all the usual stuff about how it looks and works etc.

Dave.

[EEVblog]

As I understand, he validates the performance to see if it falls within spec for one value in each range.  That my be good enough and does take some time.  Since with the aid of a computer, the units could probably be validated and calibrated in about the same amount of time, I don't see where any of Dave's valued time is lost.

Most of the time in testing is taken up by physical handling.
I've optimised this on the uCurrent by having a press jig and spot checking each range. No cables need to be plugged or unplugged, no buttons pushed.
Having a unit with a USB interface and software to contend with would be messy and time consuming.
And I do the reference current generation in batches. So I take 50 units and test all of them on the mA, then switch my current source and test all 50 again for the uA range, and then again for the nA range. So 3 test runs of 50 units.
It's pretty quick and efficient.
If I had to deal with a computer and software and USB, and data recording etc, you can multiply that time five fold, easy.

Since the calibration and validation is being performed by a computer, it would also be possible to profile the unit and provide a calibration certificate and a printed graph/chart with each unit.  That my be gilding the lily, but it is an option that would cost little more that paper.

How long does profiling take?
I'd need all new PC controlled low current sources.
It gets ugly.

[george graves]

It is now not a uCurrent Mk2, but a standalone low current meter

Dave is being nice.  So I hope you don't miss his subtle point.  If you want to take someone's design, and make a better version if it, it's common curiosity not to take the name.  So calling it a "uCurrent Mk2" is a big boo-boo.  And there are a lot of people that would turn their nose up at you taking then name.  I'm sure you were just using that name as a shorthand, but it's worth pointing it out just in case you missed that.

[EEVblog]

Dave is being nice.  So I hope you don't miss his subtle point.  If you want to take someone's design, and make a better version if it, it's common curiosity not to take the name.  So calling it a "uCurrent Mk2" is a big boo-boo.  And there are a lot of people that would turn their nose up at you taking then name.  I'm sure you were just using that name as a shorthand, but it's worth pointing it out just in case you missed that.

Given that it's in the Suggestions board, I don't think he has any intention of making one himself.

[Pilot3514]

I think the whole idea is gilding the lily. The µCurrent is an adapter. Once you add this many features to it, you've taken a nice, well designed adapter and made a half-assed one function "multi"meter. It's like taking a perfectly functional bike and adding a lawnmower motor. Either keep it a bicycle or get a motorcycle.

Your right, it's very much like taking a bike and adding a motor, you then have a motorcycle.  It is not the same thing, it does not completely replace the other or take away from it's original utility.  You just end up with something "better".

I think we can agree to disagree.  I feel that a single purpose meter that does one thing well was value.  I would not go so far as to say it does something "half-assed" since it does not exist and can not be tested.

Maybe I can not leave well enough alone.  Maybe I see everything as an opportunity to add a computer to it.  I just saw a need for the µCurrent to go from an adapter to a standalone functioning device.