Howard Winter wrote:
> Phil,
> 
> On Tue, 27 Jun 2006 22:30:59 +0100, Philip Pemberton wrote:
> 
>> ...
>> What's bugging me is that the manufacturers are saying one thing (wait for dV 
>> to hit -5mV per cell)
> 
> I'd be careful with this - I have a feeling that a worn-out cell doesn't go into negative dV, and you end ep 
> cooking it.  I can't say for sure, but I've had very sophisticated chargers end up frying cells that I 
> believed to be a bit dodgy, but wanted to see if they'd charge...

That's why I'm planning to use 0dV or Inflection Point sensing.
What really bugs me is that all the cell manufacturers are harping on about 
negative-delta ("has the voltage dropped by more than N millivolts below the 
peak?") charge termination, yet most of the reference material says that the 
negative delta occurs when a cell goes into overcharge.

I've pretty much figured out how to implement a "homebrew hack" of Galaxy 
Power's Quicksaver algorithm, which seems to consist of:
   - Make sure there's a battery connected to the charge output
   - Pre-charge. Ramp the charge rate up from C/10 to 1C over the course of 
two minutes
   - Fast charge. Charge the battery at 1C until a termination threshold is 
detected (e.g. inflection point)
   - Top-off. Two hours at C/10 to finish charging the battery

It seems that the reason the batteries run cooler when they've been charged 
with an ICS170x chip is because it's designed to cut off early, and top the 
battery up. Inflection-point sensing trips at the point where the delta-V 
starts to drop - at that point, the battery is 90% charged. Follow that up 
with a C/10 top-off charge which still overcharges the battery, but does it at 
a rate that the cell can handle without venting.

Seems GPI have disappeared, which is a shame. Their website has been gone 
since March last year (but it's still in Wayback) but all the data I've seen 
on the street mapping sites and the stuff on PA state's website suggests 
they're still in business... Anyone know for sure?

> And I think you need to be even more careful if you're charging a battery rather than individual cells - one 
> cell's +dV may mask another's -dV !

Another reason to use inflection-point or zero-delta sensing.

I notice the Maxim MAX712 datasheet suggests that zero-delta sensing is only 
suitable for NiMH packs, but negative-delta is suitable for all battery packs. 
Most peculiar.

I'm just trying to solve what appears to be the last of the Nasty Design 
Issues (tm) now. I want a decent amount of resolution out of the A/D 
converter. I've got ten bits (1024 steps) and a reference of 2.048V. That 
gives 2mV per step.
That's fair enough for a single cell, and leaves a good bit of overflow. 
Problem is, if I want two cells, I have to halve the input voltage. I still 
keep the same per-cell resolution, but then if I go back to a single cell I 
can only get 4mV per step resolution.
It gets worse when I design for four cells - with a single cell, I only get 
8mV resolution, which is nowhere near sufficient to pick up the negative 
delta-V...

So far the best idea I've come up with is to have swappable battery holders 
and a voltage divider. The pulldown resistor is in the charger, the input 
resistor is in the battery holder. A five-pin connector connects the pack to 
the charger (2 pins for +/-, one pin for sense, two pins for telling the 
charger how many cells the holder is designed for.

Or I could redesign the voltage amp using a digital potentiometer or 
something. Divide it down as if there was a four-cell pack connected, then use 
the digipot to vary the gain of an op-amp stage... Decisions, decisions.

But at least the other 80% of the hardware is finished. Rather typically it's 
the last 20% of the design that takes 80% of the time. Then there's the 
software...

Thanks.
-- 
Phil.                         | Kitsune: Acorn RiscPC SA202 64M+6G ViewFinder
philpem@dsl.pipex.com         | Cheetah: Athlon64 3200+ A8VDeluxeV2 512M+100G
http://www.philpem.me.uk/     | Tiger: Toshiba SatPro4600 Celeron700 256M+40G
-- 
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist