Tuesday, December 22, 2009

OK. HERE IT IS. SAG. VOLTAGE. MORE

I've heard a lot the last couple of weeks regarding our top balance/bottom balance situation. Most of the online forum engineers did NOT skip a beat. They went off of current shunt balancers pretty much cold turkey, but immediately lunged to monitoring individual cell voltages for a "low voltage limit" as the holy grail of the NEW BMSpeak. It was awe inspiring to watch. The entire conversion from shunt balance to LV cell limits took what looked like 8 minutes. Several new forum topics were started on the new design and ONE guy is already taking orders for his NEW design. I had a bit of a private e-mail conversation with him to avoid embarrassing him and it became immediately clear he had no clue how a charger worked to charge a battery.

So today, after a weeks work on this opus, I've posted a 102 minute epic adventure on the website at http://evtv.me on how to charge a LiFePo4 battery, and what a constant current/constant voltage charge curve is. If you already know, FEEL FREE TO SKIP THIS ONE. It was boring even for me.

But what is lost in all this is the question I keep asking them, and they seem TOTALLY UNFAZED by it. You are monitoring for WHAT and you will do WHAT with the information if you find it?

The problem of course is that voltage is a function of current based on the internal resistance of the cell. And that internal resistance varies both ACROSS the entire discharge curve, and by temperature. So whatever voltage you set, will cause a different result at a dramatically different temperature. So design and test in December will not have the same results in June. And the REAL problem is the voltage at all. It sags at the beginning of the curve, but it sags MUCH more at the end of it. And at some discharge currents can drop to very low values with NOTHING wrong with the cell, in fact everything very very RIGHT with the cell, since it can deliver that kind of current.

So if I WANTED to set an alarm, and could magically do it at NO effort or expense, such that if ANY cell fell below some voltage, and simply set off an alarm, I would have NO idea what to set it at anyway. Or what to do with the information if it went off.


But I have gotten some new equipment and can do some pretty good current rates out of single cells. I did some discharging of a Thundersky 160 Ah at 300 amps with a duty cycle of 20 seconds at 300 amps discharge, followed by 20 seconds rest and repeated until full discharge. It appears this CAN cause a modest Peukert effect.

Today, I had a bit more productive fun with a THundersky 90Ah and a Sky Energy 100 Ah cell. Topped them both off quite nicely, and then did a discharge curve in an odd combination with the West Mountain automated load and writing down the values. I set it to discharge for 180 seconds and rest for 60. For both cells, that worked out to about 5% increments of total capacity for each 180 second pulse - 4.5AH for the Thundersky and 5.0 AH for the Sky Energy.

What I found was encouraging and discouraging. First, these little batteries are REMARKABLE devices in how much power and current they can spew for hours on end. Discouraging? Well, I had been working on a little Arduino board to drive the ancient fuel gage in the Speedster's original combi gage. Had it working pretty good and just assumed that the voltage decline, while very slight, would be very linear and at least detectable. Wrong.

I've been spending way too much time on the ugly end of the discharge curve. Up at the GOOD end where all of you all want to be, the voltage differences are very small. Several people have told me you cannot monitor pack state of charge by voltage. I pretty much blew them off. The voltage changes with discharge, I've got meters. I can measure it.

Well, I STAND CORRECTED. They were precisely correct. UP in the good part of the curve, there is so negligible a change in voltage that it is just not really very useful. I will have to do AH counting with the Arduino instead of simple voltage measurements. No fuel gage based on voltage will work.

Why did I think it would? Well down on the ugly end, if you can get unloaded for a minute and check your pack voltage, that's a pretty GOOD indicator down there. When it goes below 3.00 v static, you are pretty much done. And I've always said that there wasn't much left after that.

Well how TRUE TRUE TRUE. And with good reason. Your static voltage will dip below 3.00 v at about the 95% discharge level. And there really ISN'T any left. But recall that we get 2000 cycles at 80% DOD and 3000 cycles at 70% DOD. Up on THAT part of the curve, voltage is essentially useless.

See the graphs below. They tell the story.





I guess the only other interesting thing is that the Sky Energy cells, which I had said seemed to have a flatter discharge curve, seemed that way because they have a flatter discharge curve. Actually quite a bit flatter.

So a couple of things to learn from this. 1. Current load causes dramatic changes in voltage, but they are almost impossible to relate to useful information. They are just going to sag under heavy currents. These graphs are at 1C. It is much worse at 3C or 4C.

2. A fuel gage for an electric car using LiFePo4 cells, that is useful in the upper range of the battery and useful for detecting either 70% discharge or 80% depth of discharge, can NOT use voltage as a metric. It MUST count actual Ampere Hour usage. I know of no other way.

Enjoy.

Jack Rickard