Designing a Battery Cycling Circuit for 200µA Charge/Discharge Using LM334

Hello,
I test batteries for a living. I'm looking to build a simple circuit that will charge, and discharge a battery at constant current for hundreds of cycles.
I was thinking of using a LM334 to supply constant current. My problem is I don't know how to tell the current to 'switch' once certain voltages have been reached.
I typically cycle my batteries between 0 and 1 V, at currents of approximately 200 micro amps (they're small).
Essentially I'd like to circuit to pass 200 micro amps until the battery reaches 1 V, then pass -200 micro Amps until the battery reaches 0 V, then switch back to 200 micro amps; and repeat indefinitely.
Does this sound possible?
Thanks

Shane

my first question - why would you ever discharge the battery down to zero volts???

By this currents you can use a LM334 directly. But there is no way to go to 0 Volt, (not to switch or measure)
About what type off battery's are you talking that allow 100% discharge to zero.

I work on a daily base with all types of battery's but what you want to do sounds like destroying the battery.
And in case you want to test them you need to use a micro controller, then you can measure the time it takes to charge discharge, current, voltage and a cycle counter for your statistics.

Or are you working on the new silicon anode Li-ION battery.
If my mind is correct it has 10x the capacity and cheaper to make but i did not know it can go down to 0 Volt.
That will be great when i want to replace my batterys at home yes i am on solar power.

If you have a supply that goes below "zero" you can much more easily test for "zero". I'd do this by raising the battery ground above the logic ground, say to one volt. Then you would test for 1 volt and 3 volts at the battery.

The problem with discharging exactly to zero is that all circuits have some inaccuracy.. Say for example that the voltage comparison accuracy is +- 10 mv. If the comparison is on the low side of the spec then the battery would have to be 10 mv below completely discharged to quit. I'd cycle the batteries at enough less than fully discharged or fully charged to be sure you get there.

The big problem I see with this is that the LM334 current source won't output -100 mA.

If you used a sense resistor and a variable voltage source you could mesaure the voltage through the resistor and continuously adjust the voltage to produce the current you need. The same controller that reads a ADC (analog to digital converter) (probably built into the uP) could drive a DAC (digital to analog converter).

With one side of the batteries at one volt a DAC output below 1v would be a negative current. I think you could easily find a DAC that would have sufficient output..

Then the basic components would be a uP with an ADC, a DAC and,a 1v (or so) source. I'd bring the 1v into the ADC so it didn't have to be that accurate. The more you can do in the processor the less you have to concern yourself with absolutely accurate analog components. You can also modify the operation by changing the program rather than changing the circuitry.

more comment ...

If the battery "ground" is 1 v or so and you are reading it with a microprocessor you could even use a couple of resistors for a voltage divider to make 1v. As the current in or out changed the divider voltage would change some but the processor could adjust for that. This could be easily be done with just a standard 5v or 3.3v supply.

Yeah, I was going to ask the same thing.

You can go to zero if you use a split supply (i.e. positive and negative rails).

It's definitely possible. Do you also want some indication of the discharge/charge time or is this just something like a stress test or life expectancy (or masochistic torture .

I'm studying new and novel materials for Li-ion batteries. Although in a 'normal' battery you would never get to 0 V, in our test cells we do.

Yes, Si based materials for Li-ion batteries. We test vs. metallic Li so the voltage does reach 0 V on 'discharge'

Thanks for your comments. We test at voltages very close to 0 V, so 'errors' of +/- 10 mV would be too large. I'm using an analogue to digital converter on a Raspberry Pi to measure and log the data. I don't want to get into anything too complicated for controlling the current. I just want a circuit that will switch current direction depending on the voltage of the battery.

I'll be logging the time and current via a Analogue to Digital converter. We use the time and current to get capacity. I then record the capacity for each cycle to see how quickly the battery is fading. This is a measure of how 'good' the new battery material is.

The reason I suggest a microprocessor current control is both that I don't know any other easy way to get negative current and you probably could use it to recognize the high and low voltages.

The problem recognizing a zero voltage exists no matter what the resolution is. I just used 10 mV as an example. Most analog gets a little hinky at either rail. If you have even 1mv of offset you might not recognize exactly zero. Working from 1v allows an input to be off by a half bit of your resolution and still be seen. Of course if you have +- supplies this is not so much of an issue, so long as everything agrees exactly where zero is..

The processor has to have an analog to digital converter so it seems silly to add another. if you want to log the performance it could be easily sent in a serial format from the microprocessor in whatever format the programmer would like to send, as a comma delimited file for example. Most simple microprocessors have serial ports so this is real easy. Other interfaces are possible, especially if you use a fancier processor. Many of these also contain non-volatile memory so they could accumulate data and generate complete reports.

Try this:

https://www.circuitlab.com/circuit/2h5u2d/eeweb_batterytest_04/

or if you can't access it on CircuitLab, then the attached image is the same thing:

You might have to tweak R7 to get it to detect the zero volts point (Peter was right! You can't go to zero volts -- it will have to be a voltage slightly higher than zero)

OR if you use a MCU, you can program it to monitor slope and when the slope goes to zero, then then start charging again. Or better, when slope = 0 AND the voltage is zero. If the slope reaches zero yet the voltage is not at or near zero, then you've detected a fault condition (extra feature

Peter, I stand corrected -- detecting "zero" with a comparator isn't really possible. But with an MCU one can watch for zero slope.

If you have the money the instrument to use is a Keithley Source Measurement Unit (SMU) controlled by a PC.

http://www.keithley.com/products/dcac/currentvoltage

The circuit Steve provided looks good.

Totally agree with Dave on this one. I think you can get a SMU from Keithley for around the 2-3k range. You can program it using labview or using your own custom app using C++ or visual basic, etc.

The benefit here is that you can spend your time writing the test code instead of designing the test hardware, and you don't have to worry about calibration of your test circuit.

Steve that is a nice setup
Yes in deed it is best to measure the zero slop real zero will be difficult.
If this was my test setup i but the test battery and current source in a metal cover to reduce electrical disturbance and the controller even outside.
Below 10 mV any source incl your body is a disturbance.
Also you need to be care full with coding the ADC you can not any more use a simple filter by dropping the last bits of data.

p.s. is there a site that we can follow the news from this battery.

I have lots of K2400 SMU's. Very useful but very expensive. I'm looking for a low price alternative.

Excuse my ignorance but what does this circuit do?

Isn't 'zero' just ground? If I understand correctly comparators aren't good at detecting ground and therefore can't be told to switch when ground is reached?

I'd like to make something like this work but the short answer is I don't know how.

My current project is trying to use a Raspberry Pi with an 8 channel analogue to digital converter to record data as a function of time ans save to a .csv file. If you know of a better way to do this I'm open to suggestions.

It's like this:

One way to alternate a constant current is to have both a source (I1 in the first drawing) and a sink (I2). If the source current is twice the sink current, then half will flow into the sink and half will flow into the battery -- when the switch is closed. When the switch is open, 200uA will flow OUT of the battery into the sink. So, this makes it possible to reverse the current flow in the battery with one switch.

In the full circuit (reproduced, below), OA2:Q2 forms a constant current source set to 400ua and OA1:Q1 is a current sink set to 200uA. OA3 and OA4 from a window comparator that monitors the battery voltage and sends signals to the RS-flipflop formed by NAND1 and NAND2 which store the extreme states of the battery (i.e. 1V charge, and nearly 0V charge).

R7, R8, and R9 form a voltage divider that sets the monitor voltages for the window comparator. They set the voltage on the inverting input of OA4 to around zero (probably should use a variable resistor for this, so you can calibrate it for the lowest possible trip point (somewhere slightly above zero) [something like a 5K pot with the wiper connected to the inverting input of OA4, then reduce the value of R7 to around 8.2K, replace R8 with a pot and tie that pot's wiper to the non-inverting input of OA3, and reduce R9 to around 6.2K -- off the top of my head, so I hope that's right ) So, the - input of OA4 is slightly above 0 and the + input on OA3 is around 1V (might need to be slightly below 1V).

The output of the RS-FF turns Q6 on an off which turns the current source on and off. Q6 is the switch depicted in the other diagram.

Let me know if you have questions :)

Probably good to use 1% resistors throughout and stable pots. Also, use a CMOS 4011 for the NAND gates (e.g. CD4011B) with VDD tied to +5 and VSS tied to -5.

The other (or "first") diagram:

Only if a comparator's negative supply lead is tied to zero volts will the comparator "have trouble" sensing zero volts. If the comparator's negative supply line is tied to some negative voltage (beyond the comparators common mode input range), then zero volts will be easily detectable. In other words, if the comparator is supplied by, say, a + and - 12V supply, then zero volts is as easy to detect as 5V or -3V or any voltage within the input range of the comparator.

So, in the case where the negative supply line of the comparator is tied to ground, then even for a comparator that has a rail-to-rail input range, 0 volts will be on the edge of it's range, and thus will be problematic. BUT, if ground is in the middle of it's input range, because the comparator is supplied by a "split rail" power supply, then no problemo detecting ground!

Seems like that would work -- and if you use my circuit, then just tie the input of the Pi's ADC across the battery.

Perfect, thanks Steve

Wow, thanks for that. I'll start assembling the parts so I can build it.
Thank you very much; I'll let you know how I do.
Cheers