Battery capacitance tester...

[Silvio Klaic]

Its looks like I need to start new project.
I came across desulfator schematic for lead-acid batteries while ago:
http://www3.telus.net/sail/sj23/e_electrical_tips/e04.html
http://leadacidbatterydesulfation.yuku.com/forums/3/t/Pulse-Charging-and-Desulfation.html

And build one to see if I can restore some of few batteries from my back yard.

I figured that desulfation process should work something like on logarithmic scale, definitely not linear. Cristal structure will star slowly to erode, then as it becoming thinner and thinner process should go faster and faster until hits the end – clean plate, and become slower and slower.
In other words, initially there will be almost no change in periods of charging and discharging thru desulfator.
As desulfator is used in next periods, charging and discharging time should be significantly longer and longer until you hit near clean state of the lead plate.
After that time all next desulfation periods should be at the same time or insignificantly longer.

So far I tested desulfator on two batteries:
First is some sort of gel-lead based battery. After running desulfator for month or so there was no difference that I can see. Battery is charging and discharging at the same rate as before.
So I start to wonder if desulfator works at all.

Next battery which I used was real lead-acid, so I clean it up, refill with distilled water and try with desulfator again. There was something really interesting started to happen.
After couple of weeks, periods of charging and desulfation become shorter and shorter.
I end using this after I was charged battery under an hour and desulfator is empty it in 5-6 hours.
Basically it looks like battery is totally dead, however I noticed that something is strange whit it. When charging, initially voltage goes at 20V and after minute or so it drops to about 13V and charging is continued normally up to 14,4V.
I didn’t see this on any other battery, but this looks like effect of negative current- radiant energy which John Bedini (http://johnbedini.net/) is described for charging with back EM pulses.

But if you look design of desulfator it’s at functional level almost the same as Bedini monopole energizer. Difference is that Bedini uses magnet on wheel to match battery frequency.
To solve this problem of fast charge/discharge I need to fully discharge battery (probably short circuit it) and then recharge it in few times.
However if this is not negative current/radiant energy problem, then this process will destroy totally battery.
So I need to build battery capacitance tester to see if discharging/charging cycle increases or decrease capacitance. If capacitance is decreased, then desulfator is burn my battery.

[Silvio Klaic]

Commercial battery capacitance testers don’t promising. Basically they cost too much for something simple as measure capacitance of battery.

Those testers who test capacitance in 10-30 minutes are most definitely not for something what I need.
They basic operation is to make some constant load (i.e. 1A) and measure voltage drop. After that microprocessor calculates capacitance, something like this:
Battery is under constant load of 1A (~12W), after 6 minutes voltage is changed from 12,6V to 12,5V, and then capacitance is calculated:
Note: battery discharging is not linear; therefore next calculation is inaccurate because it uses linear method and its here only to show how can be done.
12,6V = full battery (100%)
12,5V = battery at 87,5% (or 12,5% less in 6 min)
Time to reach empty state-11,8V=(100%/12,5%)*6min=8*6min=48min=0.8hour
Battery capacitance=1A*0.8hour=0.8Ah

So as I mention, battery discharge is not linear, so you need include that into calculating remaining time to be fully empty. However, some used batteries may in some states (plate damage, high sulfation, etc.) show different behavior at different charge levels. Because of that, testers with calculation in these cases newer tell correct capacitance.
I need to test just those batteries, so this type of testers is no use for me.

Other testers which test capacitance by totally discharging battery are big, mainly because it uses passive cooling for test load. And each time you discharge/charge battery you lose small bit of capacitance.
After thinking about this I reckon that this is not exactly what I need, because if I use this tester I need to charge battery after and repeat process, not to mention writing measured results and time...
There is no way that I will spend that much time on it, I’m not some stupid scientist...

I need standalone device which will measure capacitance by fully discharge battery, charge it to full, wait battery to stabilize after charging and in meanwhile write constantly and precisely everything...
So basically I need device which I can hook to battery and forget about it for few moths. After that came to see detailed repost with charts what is done.
That’s what I called good science. I’m here to do thinking not working, that’s machine job.

[Silvio Klaic]

To make tester I decide to use computer as platform to control, store and manage test data.
It’s much simpler than to use microprocessor, especially when I have old 486 black/white laptop collecting dust.
For measure I need A/D converter. After searching I found that ADC0804 is ideal:
http://www.sahincnc.com/omersahin/project/recorder/adc/ppadc.html
http://www.quasarelectronics.com/3112-8-bit-pc-data-logger-mark-ii.htm
http://electricly.com/analog-to-digital-adc-0804/
It’s cheap, easy to interface with PC, minimal number of components and to measure up to 15V I need only set resistors to drop voltage 3 times to be in range of 0-5V.
With this and 2 relays I can have full control of starting/ending charger and test load.

For charging I decide to use my regular charger, not to build my own. So I place relay in parallel with AC line voltage for device. With this I can disconnect/connect charger from mains power of electric network without doing any modification to it.
However I find that charger have lamp on it which is lit if charger is connected to battery, newer chargers have voltmeter. To completely remove charger from battery I need relay to disconnect both AC mains and DC charging lines.

The real problem is with constant load for testing.
It’s known that battery is self-discharging after some time. So I need to decide how much current I need to apply to test.
If I apply high current, i.e. 5A (~60W) on battery of say 50Ah, theoretically light bulb will stay lit for 10 hours. In practice usually reaches much less because of internal resistance and heating. For info about all that you can see this page:
http://www.windsun.com/Batteries/Battery_FAQ.htm
I can’t use very light load to do testing because self discharge comes in play and result can’t be accurate.
Standard test uses load for 24 hour discharge, because most of time batteries is used under high loads in short time. To calculate what load is needed for 24 hour test of 70Ah battery, divide 70Ah with 24h and get 2,9167A (~36W light bulb or 4,2 ohm resistor).
For smaller batteries of 45Ah you need 1,875A (~23W).

Now, the big question is do I need higher load for 24 hour testing or lower for 100+ hours? After additional research I conclude that higher load actually make some minor damage to battery and any possible repair done by sulfator may be nullified with testing.
Destructive testing isn’t something what I have in mind, so I decide to use load in range from 0,5% to 1% of battery Ah value. My battery is lead-acid 12V 70Ah, so load for it will be from 0,7A to 0,35A.

Next problem is how to construct constant current load, most definitely with transistors or/and with other IC for current regulation. For load I chose to use halogen light bulb of 12V 5W, because load resistors are big and radiated heat is the same.
Load form this halogen bulb at 13V is 0,44A and all schematic I found for loads higher than 0,3A need heat sink for transistor. That’s starting to become big device and I want something tiny...
After conducting experiments with current regulation I concluded that is easier to use halogen bulb directly without regulator.
Testing different voltage and reading currents, I got this results:

Voltage	Current (A)
13	0,440
12,9	0,439
12,8	0,437
12,4	0,430
12,35	0,429
11,9	0,420
11,8	0,419
11,5	0,415

These aren’t linear results, because light bulb doesn’t work linear as resistor. However in testing segment from 12,6V to 11,8V, current can be recalculated from voltage using constant. From my measured readings I recalculate that constant for calculation is 0,01/6 (0,00166...).
Formula for calculating current is now simple: I=I₁₃ – (13V - Ur)*10*C
Where I₁₃ is Initial current measured at 13V, Ur is voltage for which current is calculating and C is constant for used halogen light bulb – 0,00166667.

Now I can construct schematic for my device and start experimental work.
Here is schematic:

[Silvio Klaic]

After testing I found that ADC chip works excellent on bidirectional LPT port. Problem is in my old 486 laptop, there’s only unidirectional LPT port.
So I need now to change schematic, add multiplexer and reroute it to status pins of port, and other reroute to data pins. At least now it’ll work on any LPT port.

Next possible problem is with reading voltage...
ADC can read form 0-5V in scale of 0,02V with error +/- 0,02V. Measuring input voltage range 0-15V, scale is set to 0,06V and error +/- 0,06V.
That is not good, if battery is empty - 11,8V, reading will be 11,82V (2,5% full) or if error in reading is used in count: 11,88V (10% full).

Precision of 0,06V is too low for my needs and after calculating I came to conclusion that precision must be in range of 0,02 – 0,03V to get accurate results.
So ideally will be to measure in range from 10-15V. I only know to do that thru bridge of Zener diodes. First diode make input voltage drop, and second sets output range. Both Zener diodes in series make maximum input voltage.
I can’t use 10V diode, because there is about 1,2V nonlinear current after targeted point so in reality range 10-11,2V will be in nonlinear state and results become inaccurate.
With 9,1V Zener diode I get nonlinear range from 9,1V to 10,3V which is acceptable. For second diode I’ll use 6,2V and that in total produce maximal input voltage of 15,3V with maximal output voltage of 6,2V.
However in here I need to install resistor and trimmer potentiometer to get range in 0-5V. Resolution for this reading from 9,1V to 15,3V is 0,024V and fits required range.

Now after modifying schematic I can start to make prototype version for further testing:

[Silvio Klaic]

Finally after awhile I finished prototype. I did some changes in schematic due incorrect readings under load.
I reduce number of resistors at zener bridge and now measuring consumes about 0.21mA. On top of that I did some corrections in program, because voltage drop from first zener (9.1V) diode is now 8.925V under input of 15.3V. Error in reading is +/- 0.025V, and in worst case scenario can be max 0.05V.


Here is new schematic and device prototype.


When I was charging battery for first time in awhile, it reaches 15.8V and after 30-40 seconds slowly drops to 13.4V. In first (top) graph you can see second charging after discharge. It begins at 13.85V, then drops to 18.3V after 3-4 seconds and continues to charge normally. Still don't know why this happens during first few charging cycles.

On second graph is shown several discharges thru 5W halogen lamp. Here you can see that all are more or less identical. One interesting thing is steep voltage drop after 12.1V to 10.2V. Normal batteries don't have that voltage drop.
On top of all that, as you can see, for charging this battery it takes only 22 seconds, and to discharge it thru 5W halogen about 80 seconds.
So I still can figure it out if battery is really dead or not. This 70Ah battery was before using desulfator usually charged for about 3 hours. I'm going to try few times to drain it with 50W halogen lamp down to 8V and recharge it. Then I'll know for sure if it’s really dead or not...

[Silvio Klaic]

After intense testing I wasn’t happy with some results and decide to do some changes/upgrades to tester.


Here is new improved schematic.

• Voltage reading at and beyond 15V shows some nonlinearity, probably because leakage current through 6.2V Zener diode.
  So I remove it and place regular diode to VCC for input overvoltage protection. Input impedance of ADC0804 is 5k (as is stated in datasheet), so I reduced input voltage divider to 19k.
  Now at 15V consumes about 0.23mA, but reading is more accurate. I also have to change base voltage in program from 8.925 to 8.88V.
• There was some significant difference in voltage at halogen lamp and battery terminals.
  I was hoping this won’t affect calculation, but it does. So I add sense wire which is connected at clamp as +12V wire and all voltage reading are now done at battery terminals.
• I originally placed halogen bulb inside casing, but that will melt casing if left too long working. There is also problem of changing bulb. So I add PTR connector and mount halogen bulb in it.

Here is how prototype now looks like.

However I’m thinking to add resistor + OP amp for current measure up to 2.5A at halogen load.
To switch between voltage and amperage I’m planning to use MOSFET transistors.
This would eliminate calculation of how much halogen light bulb consumes (which only gives estimate value) and add options to easy replace it with more powerful version.