Go Down

Topic: ESR meter with Arduino (Read 79320 times) previous topic - next topic

Sreg

Hi all,

Since I do not have an Arduino, I have built szmeu's original circuit around an Atmega8.
The only difference is that I'm using 1N5821 instead of 1N5822.
The firmware has been adapted from Totoro's code.
When connecting a cap or small resistor across pins 1+2 and GND, I'm getting nothing, the ADC value is always 0. (2.56V internal reference)
However, when connecting a potentiometer accross +5V and GND with the wiper going to the ADC (ADC0/PC0 in my case), I'm getting values, so the ADC seems to be working ok.
Charging pulses are generated, measuring about 30mV pp at pins 1+2 (maybe too low?)
Any pointers as to what could be wrong?





dhenry

When you use a resistor to charge-up a capacitor, the current goes down as the capacitor is charging up, exponentially. This is less of a problem initially; but as the capacitor is charged up, the error (from a linear charging up process produced by a ccs) gets bigger (small resistor or small capacitor, or long charge-up time / cycle).

One way to solve this is to taylor-expand the integral of current over time. You only need to expand to its 2nd order derivative to maintain fairly good accuracy.

Totoro

Sreg,
First, just to be sure we're on the same wavelength, short pins 1 and 2 on CN1 (on szmeu's schematic) together and put the  device-under-test (DUT) between these shorted pins and ground. For now, just ignore the possibility for a four-wire setup (actually three wire with the circuit from szmeu).
OK, you've done that. Now, with a 30mV pulse across your DUT, you can do the math and you should have approx. 600milliOhms +/- for the DUT. This assumes that at top of the 100 Ohm resistor, R8, (where it connects to the emitter of Q1) you see a 5V pulse.
If that's the case, then you may want to make sure that the oversampling is set up correctly for your device. You may need to change some of the register set stuff in the original code. This stuff...
  sbi(ADCSRA,ADPS2);
  cbi(ADCSRA,ADPS1);
  sbi(ADCSRA,ADPS0);
Refer to the datasheet for your device and the Atmel document that szmeu mentions.
Then again, it could be something simple like the discharge transistor, Q2, is always on. Or you didn't change the value of Vref in the code. You mention the internal ref on your device is 2.56V. In the code it's 1.1.
This is all just guesses on my part. Maybe you have checked all this stuff. My 2 (euro) cents worth, anyway.
Good luck!
- Totoro

bobale

Totoro, have you designed PCB by any chance?

Totoro

bobale,
Sorry, I am just a breadboard/point-to-point/proto-board type. My main need for an ESR meter is to troubleshoot old audio gear and test equipment.


bobale

I also need it ocasionally, but I would rather build standalone device, but I'm really bad at PCB design, so I'll wait for someone else to do it (someone who needs it more than me) :smiley-mr-green:

Sreg

#51
Jan 25, 2013, 12:33 pm Last Edit: Jan 25, 2013, 12:38 pm by Sreg Reason: 1
Totoro,

thanks for your reply!

Sreg,
First, just to be sure we're on the same wavelength, short pins 1 and 2 on CN1 (on szmeu's schematic) together and put the  device-under-test (DUT) between these shorted pins and ground. For now, just ignore the possibility for a four-wire setup (actually three wire with the circuit from szmeu).

Yes 1+2 are shorted together amd DUT is placed between those and GND.
Quote

OK, you've done that. Now, with a 30mV pulse across your DUT, you can do the math and you should have approx. 600milliOhms +/- for the DUT. This assumes that at top of the 100 Ohm resistor, R8, (where it connects to the emitter of Q1) you see a 5V pulse.
If that's the case, then you may want to make sure that the oversampling is set up correctly for your device. You may need to change some of the register set stuff in the original code. This stuff...
 sbi(ADCSRA,ADPS2);
 cbi(ADCSRA,ADPS1);
 sbi(ADCSRA,ADPS0);
Refer to the datasheet for your device and the Atmel document that szmeu mentions.

I'm fairly sure my code is ok, but I'd be glad if you could take look at it - please see attachment.
I'm also getting about 3V pulses measured at the emitter of Q1.

Quote

Then again, it could be something simple like the discharge transistor, Q2, is always on. Or you didn't change the value of Vref in the code. You mention the internal ref on your device is 2.56V. In the code it's 1.1.

Unfortunatley the Atmega8 has only 2.56V internal reference voltage, but I think I can live with any decrease in resolution this will introduce.
My problem is that no values whatsoever are displayed when connecting the DUT as described above. (ok not entirely true, -1 is displayed since I return -1 from the calculation if millivolts are 0 - please refer to source code).
Q2 should not always be on, I get pulses on its base according to the discharge pulses.

If I get this to work, I'll gladly post my pcb layout.

-- Sreg

Sreg

Finally got it to work, it seems the BC327 was defective  :smiley-red:

Thanks for all the help!
As promised, I attach my layout and source code (built with Atmel Studio 6.0).


-- Sreg

markbee

wow thx for the layout. I was just starting to design one. Saved a lot of work.

markbee
XBee blog: http://lookmanowire.blogspot.com/

Totoro

#54
Jan 25, 2013, 07:59 pm Last Edit: Jan 27, 2013, 10:02 pm by Totoro Reason: 1
Sreg,
Great to hear you got it working!
You mentioned above that you return -1 when millVolts is zero. I wondered why, so I took a look at your code. You have helped me to avoid a bad thing, the potential divide-by-zero bug. Thank you!
- Totoro

bseishen

What Schottky diode is everyone using that has had success? I am currently using a B340A-E3/61T and the pulse peak at the terminal with no DUT is only around 150mv. I think this is causing my reading to drastically be off. I am trying to wrap my head around how the current voltage divider equations every one is using is working. It doesn't take in account of the clamping voltage and power dissipation of the Schottky diode. Any help would be greatly appreciated!


Totoro

#56
Apr 10, 2013, 10:36 pm Last Edit: Apr 11, 2013, 09:30 pm by Totoro Reason: 1
Hi bseishen,

If you read my original text (kept intact, below) then I am sorry. I 'double-thinked' myself into some confusion. I will try for a better take.
First, the Schottky diodes only limit the maximum voltage available to the DUT. Here's what happens. The transistor Q1 turns on and a rising wave edge goes through R8 and is pretty much unaffected by C2/C1. So, R8 is effectively in series with the DUT. This forms a simple voltage divider and the rising wave edge sort of shelves at the voltage determined by the divider formed by R8 and the DUT. For a DUT with, say, 100 milliOhms, that is a pretty small voltage, i.e., a few milliVolts. Since we're dealing with capacitors here, there is some continuing charging that will mess up the measurement if we wait too long to make it. I put together a LTSpice model to play with this. Attached is a closeup of a pulse from a sample run with a DUT of 100uF and .1 Ohm ESR. The yellow trace is the R8 side of C2 and the blue is the DUT side. (I had looked at the same thing on a 'scope and the simulation agrees really well. I don't know why it didn't sink in earlier. It is interesting to do a run with a resistor as DUT. The 'shelving' is more apparent.)
So, pretty much ignore what I said before. Part of my confusion was from using a different power source after changing the diodes. This shows that the value for Vcc in the code is quite critical for consistent and linear results. Also, it is now apparent how two ranges work by adjusting the voltage divider.
I hope this helps. My understanding is better, anyway.

---- here is the original text ----
I went back and really looked at this as well as reading the original document by Dr. Le Hung (link in szmeu's first post). I have to admit that after looking closely, I really don't understand this completely either. Clearly, the pulse can only have an amplitude determined by the schottky forward breakdown. In my circuit I used germanium diodes and that was about 300mV. I just swapped them for 1N5822 (as in the orignal schematic) and now have a pulse of 220mV and my calibration is way off. What seems apparent now, is that the voltage specified in the code as the 'supply' voltage, i.e., 5V, is just a number. I mean that it has to be adjusted to correspond to something that, well, I don't know exactly what it specifies. Dr. Le Hung calls the pulses current pulses. There must be something to that since the two ranges do work (there is a factor of ten, approximately, difference when the pulse 'comes from' the 100 vs. 1000 Ohm resistor).
Try playing with the voltage value and see if you can dial something in. I haven't tried that yet but it will be my next step to try and figure this out.
I am first a software guy and the hardware part of this is now not as clear as I thought. Maybe someone else can add to this.

bseishen

Totoro:
Ended up doing some spice analysis on the schematic. The line is very linear! So with a simple slope formula should be able to get some good results. I currently have some 1% resistors in the mail, when i get them in i'll test this out.

Totoro

#58
Apr 11, 2013, 09:19 pm Last Edit: Apr 11, 2013, 09:25 pm by Totoro Reason: 1
Wow, what timing! You posted while I was composing. Please read my update.
I think your simulation is too simplified. I would be interested in your opinions if you would try mine.

bseishen

Yup, it is super simplified. Just decided to do DC analysis only. I ran your schematic and I'm getting the same voltage per ohm on the ADC pin  :) Have you tried running different Schottkys? It changes the ADC voltage quite a bit! The DUT is incremented in .5 ohms on the attached graphs.

Go Up