Interesting observations

[frackers]

Nearly 4 weeks with the new Dallas one wire battery monitor chip on the battery bank and I'm seeing some useful data already.

The main thing is the efficiency of the deep cycle lead acid flooded cells I'm using. The charge controller is the NE555 based voltage comparator so I run the 1kw load when I hit 27.5 volts and turn it off when I go just below 24 volts on my nominal 24 volt system. The DS2438 has registers that accumulate the total charge in and out of the battery over the battery lifetime and I'm already seeing 90% come out compared to what went in (currently 1281 amp/hr in, 1187 amp/hrs out). This has implications on tracking the charge when I implement the charge controller based on coulomb counting rather that on volts.

It also has implications on the display I have that shows the residual battery charge - the graph keeps heading skywards because so far I'm not accounting for the inefficiency of the charge cycle - what goes in does NOT come out!!

As soon as I have the GPIO one wire chip that can operate a relay sorted out then I'll be experimenting with the coulomb counting controller - I think I have the algorithms right, time will tell.

Another project in progress...

[Colaman]

Don't forget to implement Peukert's number in your coulomb counting controller.

Here's a page on it. Or does the DS2438 take that into account already?

I'm not sure on the size of your batteries, but if your typical loads are markedly different the 1kW dump load will be misleading you a bit when you use it to calculate the efficiency of the system.

High-end battery gauges can either automatically calculate Peukert's number (after a number of charge/discharge cycles they can then self-adjust as the batteries age) or have provision for the user to enter it. Lower-end gauges merely reset the gauge to 100% once they figure out that they've reached float.

[frackers]

Interesting - I didn't realise that the effect actually had a name although I was aware of it, hence only discharging by 20% maximum (ie. leaving 80% in the battery). I certainly see its effect from the non-linear impedance of the battery when charging and discharging at high rates (greater than C/10) and the effect it has on terminal voltage.

Due to the difficulties of charging to 100%, I'm hoping to run my application in the 80 to 95% charge range with the occational (perhaps once a month) equalisation charge to rebalance the cells. With luck I'll still be using the same batteries in 5 years time :-)

[Colaman]

I'm hoping to run my application in the 80 to 95% charge range with the occational (perhaps once a month) equalisation charge to rebalance the cells.

Hmmm. Might need to implement Peukert's number (or similar fudge-factor) into your system then, otherwise it will probably drift fairly quickly without the known starting point a full charge gives. Once you've got that factor in there, it should be relatively easy (hah! famous last words) to auto correct it when you full-charge it monthly. (Start from current battery level, estimate with present factor how much needs to go in to get 100%, measure how much actually goes in to get to float voltage/current, tweak factor according to difference. A few charge cycles should get a factor that's pretty close.)

Finally had enough cash to order a Victron battery monitor for my setup yesterday - It's been a bit hard to judge battery charge with just a voltmeter and an ammeter on the charge side of things.... I keep getting nasty surprises when I go and check the SG. Besides, the other half is not too sure on the whole concept so a simple "% full" gauge is the way to go for her.

Robin - Down Under (or are you Up Over)

I too am Down Under.... in fact, if I went much further Down Under, I'd be bobbing about in the Southern Ocean :-p I believe I'm a little further west than you though.

[commanda]

I'm currently back of an envelope designing a battery monitor using a picaxe.

If you use 2 current sensors, one in the charge line and one in the discharge line, and a timed interrupt (1/2 second), and accumulate AHrs into 2 variables, ChargeAccu and DischargeAccu, then;

remaining capacity = TotalCap minus DischargeAccu plus ChargeAccu times efficiency

if Vbatt = float then

     efficiency = DischargeAccu / ChargeAccu

     ChargeAccu = 0

     DischargeAccu = 0

if Vbatt = flat then

    TotalCap = DischargeAccu plus ChargeAccu times efficiency

My application is a 200 AHr SLA pack in a caravan, running a radio, tv, and some lights, so I don't expect the current drain to ever get high enough to worry about Peukert.

Amanda

[frackers]

This sounds very close to what the DS2438 chip does internally. It samples about 30 times a second and accumulates the current state of charge into a register that goes up and down to match the total charge in the battery and also into two other registers - one for the total charge and one for the total discharge.

Its this second pair of registers that I'm using to get the overall charge efficiency of the system and applying the value derived from here to modify the tracking that first register does.

I'll be publishing all the code and the hardware circuits under the GPL in due course so that others can follow what I've done and hopefully improve it...

[Rover]

Never looked hard at DS2348, looks decent... a few questions

1. Are you measuring current? If so what are you using for the shunt?
   
2. Any idea how well the onboard temp measurement works
   
3.  Since the ADC is 10 bit (1024 full scale) what range did you use for the voltage divider , I'm guessing you may have used 30V as a max bringing it down 3 to 1 to 10V so you needed a 3X multiplier .reducing accuracy.
   
   

I use a similar method but with a straight 12 bit ADC (4096) , course doesn't have the nice eeprom and accumulator.

[frackers]

Its a very handy chip - very easy to interface electrically and using the OWFS (One Wire File System) package under Linux on the Linksys router its also easy to talk to it.

The current shunt is shared with the digital panel meter and is a 50amp 0.001 ohm from a local store.

The onboard temperature sensor is only as good as how close the chip is to the thing you're measuring the temperature of. In the case of the small lithium battery pack that this chip is aimed at then it right on top of the battery but in my case its about 1m away to it only indicates the ambient temperature.

The maximum input voltage as you have discovered is 10 volts so I'm using a 3 to 1 stepdown to get into range. That gives a resolution of 30mV with an accuracy of probably nearer +/- 60mV which I think is plenty for a great hunk of lead-acid cells!! I'm powering the chip at 5volts and using the auxillary ADC input to measure battery volts.

I think the best thing going for it is the charge counting it does internally which makes accounting for what goes into and out of the battery bank easy to accomplish.

I'm not using the elapsed time registers at all since I can get sub second accuracy from the Linux system driving it using NTP (Network Time Protocol).

Cheers

[Rover]

Cool...I'm going to use 4 high side current monitors (max4173H)s with 4 50 amp shunts to my MAX186 8 channel 12bit ADC to my stamp (this is a 12 volt system)... the max186 has a 4.096v top end . My divider is 4:1 using 20v as a top end. Battery voltage is a straight divider to one of he ADC pins.

God I hate soldering SOIC chips

[frackers]

I've etched up a few PCBs that convert SOIC to DIL. I made half a dozen 16 pin with standard 0.1" pitch between pins and 0.6" (old EPROM width) between rows and crop off the ends for smaller ICs. Means I can prototype on stripboard and heatgun the chip onto the converter at the last minute!!

[Rover]

similar to this.. because I'm thinking of buying some, at 50 cents each...?

http://www.futurlec.com/SMD_Adapters.shtml

[frackers]

Yup - thats the ones. Not available in New Zealand as far as I could see so I made some...

[Rover]

Naw they they make down under.. just reversed

Rover

 Canadian/American