The blackbox project

[Westbranch]

Hi Greg, yup, that was what got me thinking of a chart to keep all these bits organized for us old farts   that think a bit slower  than you bright sparks .
My mind started to implode after I had read all the overnight postings from you 3 leaders...

[Gregy]

Not as young as I'd like to be !
Time diff gives the advantage .... You get to read them "all" when you awake I'm guessing
I'm working on some more for tonight - to fill out suggestions for HW options for each
Of the modules Under my proposed structure

[RossW]

*ding*  Blinding flash of light....  use the "top" of the previous cell! Ie, differential amps. Each one references its output on the negative of the cell under test!!  0-2V (nominal) out. I will still need a small bias or divider to be able to measure over 2.048V (really need to get to about 2.5V). Simple 4:3 divider with 1% resistors still leaves me around 3mV/bit resolution. If we're aiming for 10mV or better, we're comfortably in!

Just knocked it up, quickly. You will need a java-enabled browser to view this circuit simulation:

Even the forum software is breaking it!

http://support.rossw.net/batteryzero.cgi

DoubleClick on the batteries to change their voltage and see the effect on the output.
I'm thinking that I'd use an 8-pin RESNET for each opamp to ensure they all track together, which means the output will be 1:1 for the inputs and I'll need a divider or level-shifter.

Got some tricks up my sleeve to optimise the ADC range, will see if I can pull it together some time in the next week or so. (Will be interstate for a few days from tomorrow)

[RossW]

OK, so I thought I'd knock up a schematic while I was having my coffee.
The picaxe 08M2 will (just) do what we want, but only bare-bones. Having two input-only pins cramps my style a lot.
So I looked at the price - and the Australian distributors want about $4.10 for the 08M2 and $4.70 for the 14M2. For the extra functionality I think I'll just skip the 08M2 and go straight to the slightly bigger one. For a 24-cell (48V) bank, that means 8 boards, so about $5 difference in component cost. "no brainer"!

So, I've dumped my original schematic and will go back to scratch. This chip has enough outputs to put useful indicators on it too.

[Westbranch]

Ross, did you delete the contents ?  (link in #92)

[RossW]

Ross, did you delete the contents ?  (link in #92)

No, but it looks like the process of making it a tinyURL did. I'll try to fix, thanks for the heads-up.

[RossW]

Ross, did you delete the contents ?  (link in #92)

Sheesh.  Tinyurl broke it. The forum software itself breaks the raw URL. Another URL shortening scheme didn't work. Tried to paste the code directly into the forum without the URL tag and it wouldn't copy-and-paste properly either. Just goes to show, when you want something done right, you have to do it yourself!

http://support.rossw.net/batteryzero.cgi works here at least...

[dgd]

Ross,
Your link redirects to falstad.com circuit simulator site that seems to be overrun with ads for Asian women profiles...
dgd

[RossW]

Ross,
Your link redirects to falstad.com circuit simulator site

Yes, it's the sim - the parameters (the "circuit itself") is on the rest of the (long) command line.
If you don't have a JRE, it won't spawn the applet though. (It should start Java in new window)

[RossW]

OK, here's the *START* of what I was talking about - but no circuit simulation.
Those of you familiar with opamps will see what I'm doing

Before you complain about the zener - if this has to work and allowing for one cell to be dead short (0V), we have less than 4V we can be sure of. And the device needs a minimum of 2.3V to work at all, and more likely at least 2.6V to be able to monitor a cell thats  around 2.5V, it doesn't leave much for a regulator! Add to that, most regulators take at least 3mA (even a 78L05) which is more than we wanted to blow.

With 6V (nominal), we'll be dropping 0.9V across R1. Some of that will be from the circuit itself, the rest may be from the zener - if I need 1K, then that's 0.9mA quiescent. Probably acceptable. There's scope to improve on that by adjusting R1 in production once we know the actual current draw.

FAR FROM COMPLETE, just a discussion point.

[Gregy]

Ross
Great progress - I think the current drain is acceptable - after all - any battery has a "self discharge" rate
Even when at rest/unloaded ... I'm guessing we are playing in that range

I realize you are still in design phase ... But thinking ahead to "implementation"
1) can we put a " disconnect" switch on each module - that would enable it to be switched off or /into a Very low power draw mode - that way if for whatever reason batteries were to be left in a situation
With little or no charging - the module could be put into a very low draw (zero?) mode
2) we need to think thru the fuse aspects - eg each sensor line to battery should have a fuse inline
I'm guessing for best practise - probably as close to terminal (which means no impact to your design)
3) isolation
- your idea of "potted" sounded interesting, I guess that would mean a set of flying leads emerging
From the pot (may be best idea to preserve isolation integrity in lieu of using a screw down connector block )

And a design question
Can the LED proposed be a tri/color or similar? I know you are thinking of using a series of flashes
For flashing status ( brilliant)
..... But if the LED could be "biased" (not sure the correct term) to also
Have red or green (eg a very visual way to show "all good" or "Houston we have a problem")
I think I see a spare output pin? Perhaps a second led in red? (Note we don't
Want to add to current draw if we can avoid)

My 2c worth

[Gregy]

And at Xmas could we have a different flashing sequence ?
Only kidding!!

But it did make me think what a 48V. battery room might look like at night

[RossW]

The LED as drawn is purely indicative of function. It wouldn't be like that on the final design because it would (probably) draw more current than the series current limiting resistor could manage. For the minimal cost of it, I was thinking an open-collector transistor which could drive LEDs (plural) to full brightness off the entire "nominal" 6V supply.

One of the LEDs (probably only fitted to the last device in the chain) would be a higher-intensity IR LED for communications to the "outside world".

red/green LEDs probably. Not sure about running a RGB if we may only have a little under 3V supply (this thing may even have to run off around 3V, someone could have NiFe, NiCad or NiMH cells...

I was thinking of board layout and perhaps having a large through-hold plated hole, large enough for an M10 or thereabouts size bolt. One either end of the board, so it could be bolted directly to either + or - of a cell.

Disconnect might be a little tricky if you want to disconnect the sense lines too, or do you think that the miniscule load imposed by a 10K resistor is ok? (I could up the resistors to 100K, should still work fine)

I already plan to put the device to sleep after its done its thing. That will reduce its drain, but the zener is still going to be chewing up the lions share of it. I considered using an opamp as the voltage regulator, but the rail-to-rail devices I've got have a Vcc limit of 6V, so no good there.

Almost all the pins on the PIC can be used as inputs or outputs, so there are plenty of spare pins.

[Gregy]

Direct connect to battery terminal could work - but it would then need an onboard fuse
For that connection - and then still needs a flying lead to the other batteries I assume ... We are dealing with gigantic arc welders here! any fault is going to turn the module into molten goop
(If anything left at all) ... Fuses are our savior

I looked at the current draw from just having the resistors and opamp load ... Perhaps I missed
Something, because I looked at it as being very high impedance ...
Either way I was thinking of the power draw from having the chips powered and of course Zener/voltage ref cct powered up - so I was thinking a spst toggle in the postive rail from highest
Battery eg prior to R1
I was thinking of disconnect on the supply rail only, and use fuses for complete manual disconnect from
Battery if required

Running of "3V " .... Yeah I guess your right eg fully discharged NiCd @1Vpc x3

Wow you are really designing for all cases

this raises a question - how are you planning to "program" the device for different battery
Chemistries? (Eg if your planning some basic on board fault /status indicators .... The voltage
Thresholds will need to change accordingly (DIP switch? Pin headers ?)

what's the high end limit?) I'm guessing the more
Likely future scenario will be increased use of LIXX cells (aka need to measure up to 4.2Vpc)
.. Ok ... I think I hear you say lets crawl first..

No mention of the Xmas season mode - very dissapointing

[RossW]

Direct connect to battery terminal could work - but it would then need an onboard fuse
For that connection - and then still needs a flying lead to the other batteries I assume ... We are dealing with gigantic arc welders here! any fault is going to turn the module into molten goop
(If anything left at all) ... Fuses are our savior

Fuses protect wires. (think about it!)

I looked at the current draw from just having the resistors and opamp load ... Perhaps I missed
Something, because I looked at it as being very high impedance ...
Either way I was thinking of the power draw from having the chips powered and of course Zener/voltage ref cct powered up - so I was thinking a spst toggle in the postive rail from highest

The devices should draw milliwatts. The regulator is almost certainly the most hungry element by far.
A switch or jumper is fine though.

Running of "3V " .... Yeah I guess your right eg fully discharged NiCd @1Vpc x3
Wow you are really designing for all cases

No guarantee we'll get all bases covered, but to knowingly design without considering them is foolish.

this raises a question - how are you planning to "program" the device for different battery
Chemistries? (Eg if your planning some basic on board fault /status indicators .... The voltage
Thresholds will need to change accordingly (DIP switch? Pin headers ?)

what's the high end limit?) I'm guessing the more
Likely future scenario will be increased use of LIXX cells (aka need to measure up to 4.2Vpc)
.. Ok ... I think I hear you say lets crawl first..

My intentions (initially at least) were:
* Let each board collect and report the voltages of its own directly-connected cells
* Let each board receive the status of every other board
* Have a threshold which applies across all boards, which each learns and passes on.

Initially, I was only intending to display an alarm on imbalance, not on high or low absolute values - that can be done by the monitoring/logging device.

To program - a simple RS232 input. (They'll already be communicating via serial, it's just a matter of levels)
Alternatively, the "first" board could have a 20-turn pot on it that is used to set the maximum delta. Lets say it read a proportional input as 8 bits. I can't imagine where more than 1V difference would be acceptable, so lets say the pot would trim from 0-1V in 256 steps - 4mV/step. To set, wind the difference down until lights start to pop up across your bank, then back off a bit.