CellLog 8 hacking

[Otmar]

For those of you adding jumpers to run equally off all cells, I have some concerns about it. I've posted on DIY about this but seeing there is a lot of interest here I figure I'll repost. I wrote it before this hacking thread taught me about the LM358 supply voltage tap issue.

I would suggest that people do not do only this mod and run 8 cells. I've been wondering why they did not connect this from the factory? So I reverse engineered the power supply circuit to find out.

The 3V loads on my non-logging model are 8.8mA peak and 4.4mA with the backlight off. This power is controlled by a B817-40W NPN transistor whose base is tied to the top of a 431 Zener. The Zener takes it's control voltage from a resistor divider on the emitter of the NPN. There is a 2K pullup resistor on the top of the 431 that provides base current to the NPN and keeps the 431 running. The trouble arises with the power dissipation of that resistor and or the transistor.

If you assume your max worst case cell voltage is 4.2V then I figure that resistor dissipation is .233W for 6 cells at 25.2V and .45W for 8 cells at 33.6V. This is a 1206 resistor which is commonly rated 1/4W and can also be found in 1/2W for a lot more money. Those ratings are on a cold day so normal practice is to halve the rating. I tested mine at 25V input and the resistor got hot, but not too hot. Putting in that jumper can double the heat produced and I think it would fail before long. It could also start a fire and that would not be good.

Overall I think the listed spec of 43V is quite optimistic!

I've been wondering how to work around this and I think that if we replace R37 with a 1/2W 4.7K ohm resistor it may survive and allow running off 8 cells. In that case I calculate that at a 8 cell max input voltage of 33.6V the resistor would dissipate .191W and the transistor would take .08W which should be safe. I will add 10 of the 541-4.70KUCT-ND on my next digikey order and report back when I've tested it.

Followed by:
Quote/
Originally Posted by swoozle
Then they mention changing another resistor (looks like the one you refer to) to reduce power usage:
"The red circle in attached pics-- 202 resistor is 2k, if you measure 7s pack, you can replace it to 10k, which would decrease the power consumption."

http://www.rcgroups.com/forums/showpost.php?p=13452879&postcount=143
Attachment 15088

Otmar, does this address the overheating danger sufficiently?
/Quote

Thanks for those links! Now I'll get nothing done today due to reading all the forums!

I'm concerned about the 10K for R37 and here is why:
At 33.6V input, R37 drops 30V for 3mA and .09W heat. That's fine.
But then the Q6 the B817-40W has to carry the rest of the heat load.
Max current draw is 8.8 mA, R37 provides 3 of those leaving 5.8 mA for Q6 to dissipate at 30V of drop.
30V * 5.8mA = .174 W. The internal silicon temperature is rated to rise 625 deg C above ambient per watt. So we multiply .174W * 625 and get 109 deg C of temp rise.
Theoretically if the air in the unit is under 40 degrees C (104F) then the transistor should not blow up right away, but in my book that's way too hot to be reliable!

My suggestion of using a 4.7K 1/2 Watt rated resistor moves more of the unavoidable heat into the resistor. 30V at 4.7K = .191W. and less into Q6 30V at 2.4 mA = .072W for a temp rise of 45 degrees C which is probably ok. There still is the possible problem that the heat from the resistor will heat up the transistor due to proximity and that's why I intend to test it on the bench. A good rule of thumb is that if you can hold your finger on it for ten seconds without it being so hot that you have to pull away then the part is not too hot. I'm not sure the resistor will pass that test, but we'll see.

These are coming from this thread: http://www.diyelectriccar.com/forums/showthread.php/needed-celllog8s-relay-circuit-56509p3.html

I hope this helps,

 

[Gregory]

Thanks and welcome to ES Otmar.

 

[Stevil_Knevil]

Wow ..Otmar Ebenhoech joins Endless Sphere?!

Roll out the red carpet, boys - we have been graced by EV Greatness!!

 

[Otmar]

Thanks for the warm welcomes!

I found that this thread had some pretty sharp folks discussing how the CellLog8 works and I figured that makes this the best place to share knowledge with people working on it. Other than this I had never heard of ES before. Is it is maybe mostly for bike sized EVs? I haven't looked around much yet.

I now have learned that my non-logging CellLog8M version will spew serial data (just saw it on the scope, have not tried to read it yet) out the serial pin on the Atmel if I initiate logging with the center button. This leads me to believe that if I use an ADUM5241 to isolate the serial out, and maybe use the inbound channel to depress the center switch for startup and use the 10mA isolated 5V supply of the 5241 to avoid having to run the CellLog power from the main pack we may have a good isolated interface to a Chassis 12V referenced Arduino central unit. This would allow the remotes to be turned off remotely. Of course this requires some very fine pitch solder patching, some custom boards and mods but it may prove to be worthwhile.

Anyone know of a good i2c driven UART? 4 Serial ports will not be enough for packs over 100V. If not I have some other ideas that may work to combine the data streams with micros in the serial data stream just adding packets together.

Also I'm a bit concerned by the reports of alarms that stopping working. I am going to very curious to see if the CellLog keeps working in a terribly noisy environment like the one that happens anytime a Zilla or MZM PFC charger starts running.

 

[amberwolf]

I found that this thread had some pretty sharp folks discussing how the CellLog8 works and I figured that makes this the best place to share knowledge with people working on it. Other than this I had never heard of ES before. Is it is maybe mostly for bike sized EVs? I haven't looked around much yet.

Welcome!
Never thought I'd see you post here on ES. (although I've often wished you'd suddenly appear with a new 3-phase brushless controller design. :lol: )


ES is here for any size EV, rather like DIYEC, but also like DIYEC has lots more people with car-sized EVs than anything else, with ES there's a lot more people with bike-sized EVs and smaller. A few members have both kinds. Some of us are members on both forums, as well.

Anyone know of a good i2c driven UART?

Not sure how good it is for your purpose, but googling your phrase found this one:
http://www.nxp.com/documents/data_sheet/SC16IS740_750_760.pdf

 

[Otmar]

Not sure how good it is for your purpose, but googling your phrase found this one:
http://www.nxp.com/documents/data_sheet/SC16IS740_750_760.pdf

Thanks for the welcome and part reference. It had been a few years since I googled SPI UARTS and you know how it is; when you talk to someone you think of options so after I asked you all I looked some more and found this same part, as well as Arduino drivers for it! It looks pretty good for a up and running quickly result. Still, for $2 a channel I'm guessing the long term solution is to make the same effect out of a small micro for $1 and some code which can run without a xtal.

I have now soldered a micro USB receptacle off of an old Vonage phone to the CellLog8M board and temporarily run the TTL serial data through it (otherwise I'm sure I'll break something with those light wires). I discovered my TTL serial converter is blown so I found a backup and ordered extra chips, maybe soon I'll be reading the data it spits out (except it's getting late again, darn!). Anyone have the format for that data stream documented yet?

Fun Stuff!

 

[fechter]

Wow ..Otmar Ebenhoech joins Endless Sphere?!

Roll out the red carpet, boys - we have been graced by EV Greatness!!

Yes, welcome! Glad that somebody else can help reverse engineer these babies.

Their design leaves a lot to be desired, but it's one of the best out there in the super-cheap class.

Hey, here's a picture I took when we were all together. Stevil was there and I got to ride his I-lean. Otmar's car is in the background.

 

[fechter]

BTW: check out this related thread. Heathyoung's buffer amp mod to reduce and equalize standby drain.
http://endless-sphere.com/forums/viewtopic.php?f=14&t=45254

 

[Otmar]

BTW: check out this related thread. Heathyoung's buffer amp mod to reduce and equalize standby drain.
http://endless-sphere.com/forums/viewtopic.php?f=14&t=45254

That one is certainly interesting. I like it.
There is always the tradeoff of complexity versus performance and I'm not sure I'm ready to put anything on the input side of the CellLogs yet. I think I'm willing to put up with the small loading imbalance so long as the main processor power comes from outside.
At the moment I am tending toward powering and communicating using a ADUM5241 while chaining the data streams of any number of units with a small Atmel at each CellLog node (since I do want the cell sense wires short for installation neatness) but all that can change since I'm just playing with this and not intending to make a product.

 

[Otmar]

For those of you adding jumpers to run equally off all cells, I have some concerns about it. I've posted on DIY about this but seeing there is a lot of interest here I figure I'll repost. I wrote it before this hacking thread taught me about the LM358 supply voltage tap issue.

I would suggest that people do not do only this mod and run 8 cells. I've been wondering why they did not connect this from the factory? So I reverse engineered the power supply circuit to find out.
...........
Overall I think the listed spec of 43V is quite optimistic!

I've been wondering how to work around this and I think that if we replace R37 with a 1/2W 4.7K ohm resistor it may survive and allow running off 8 cells. In that case I calculate that at a 8 cell max input voltage of 33.6V the resistor would dissipate .191W and the transistor would take .08W which should be safe. I will add 10 of the 541-4.70KUCT-ND on my next digikey order and report back when I've tested it.

Ok, time for an update.
I've spent most of the evening wearing out ohms law and learning a bit in the process.
I got my 4.7K resistors and built an Arduino based IR thermometer so I could compare the temperatures of different setups.
The first thing I noticed when I actually started measuring things, as opposed to calculating them while staring at paper, is that I was not properly understanding the emitter follower supply. Funny how measuring teaches us things. I was planning that all the extra bias current from R37 would go through the emitter of Q6 to the load. No, that's not how it works. Once the transistor gets enough base current to cover the load (a factor of the transistor gain) the base stops drawing more current so really the collector is dropping 99% of the output power. This makes the transistor problem worse than I expected.
A note: My initial thermal calcs were made with backlight at the factory default, later when testing modifications I realized this and turned it up three clicks to full, about 1 mA more draw. Also all temperature measurements were made with with a Melexis IR sensor and so are subject to emissivity and field of view. Ambient temp was stable at 28 C. I tried to make them consistent by finding the highest reading possible, but the actual temps are likely a bit higher than indicated due to sensing error.

I have pages of notes, but I'll try to keep it to what may be interesting. Max pack voltage was set at 33.6V (4.2V * 8 cells)
The stock cellog drew 22.8 mA and resistor (R37) and transistor (Q6) temps were 94 and 88 deg C respectively. Way too hot!
With my 4.7K R37 swap it dropped a bit. 14.1 mA draw and temps of 61 and 68 deg C.
I then added a 200 Ohm resistor instead of the bridge between D4 and D5 which actually changes little, it does drop 2.7V worst case, but that drop can't be too high or else cell 7 will start to feed it.
WIth the 4.7K and 200 ohm mods I calculated power dissipation of the 4.7K R37 at 0.19W and the transistor at .275W, still way too high.
So I finally found a mod that I'm happy with and I decided to test it on a CellLog 8S that someone kindly donated for research.

This mod runs off 7 or 8 cells and also USB, but will no longer run off 6 or fewer cells:
Start with a stock CellLog8M or CellLog8S
Remove R45 which is actually a diode, also remove D2.
Get two 750 ohm through hole resistors (1/2W each ideally, but mine may have been 1/4W) and put them in series for a double power 1.5K ohm resistor.
Wire the 1.5K dropping resistor between the junction of R29 and D5 cathode (the old jumper spot) on one end to the junction of R45 and D2 cathodes on the other end.
For those who are more visual like me, here are the pictures:

At 33.6V input this mod puts 20V of power supply drop at 23.2 mA on the added resistors, putting that heat in a much safer place.
The dropping resistors came in at 62 deg C but best of all Q6 now only gets to 50 degrees C compared to the original 88 degrees and it is easy to hold your finger on it to a count of ten.
The modified unit will run at full backlight down to just over 15V and so should be fine for any 7 or 8 cell packs.

While on the bench, I found my 8 cell faker (eight 500 ohm resistors in series with taps at every cell point) was showing 0.4V high on the upper two cells. I'm thinking this is the uneven load from the LM324 on cell 6. I measured current across D4 of 0.85 mA higher load on cells 1-6 for an imbalance of 1 Ahr every 49 days. Initially I don't think breaking the ground connection solves this either, but I may be wrong there. On cars without active balancing I'm thinking that it may be worth it in the end to just use stock units on 6 cells (but always 6 or the ones with fewer will have less load) and buy more of them.

The usual disclaimer: I've only bench tested this and any use is at your own risk. I very likely may have done something wrong.

 

[fechter]

Great work! that looks easy enough.
I think you could leave D2 in there as they would normally be reverse biased and not be doing anything.

I always kind of worried about the power for the LM358 too, as the maximum supply voltage is rated at 32v. Not much that can be done about that unless you limit the unit to 6 cells. On the other hand, I haven't seen it blow up yet, so I'm sure the rated voltage has some headroom.

 

[Otmar]

Great work! that looks easy enough.
I think you could leave D2 in there as they would normally be reverse biased and not be doing anything.

I always kind of worried about the power for the LM358 too, as the maximum supply voltage is rated at 32v. Not much that can be done about that unless you limit the unit to 6 cells. On the other hand, I haven't seen it blow up yet, so I'm sure the rated voltage has some headroom.

Thanks Fechter, I agree that D2 could be left in, I had not looked at it that way. The D2 pad did give me more mounting strength for the resistor.
From what I read somewhere else, I heard the GND of the LM358 is connected to the 3V supply, which makes it just barely legal for 33.6V (8 cells at 4.2V) but I agree they may be counting on using some headroom. I just did a quick check on a LM358 National part (date code M47AJ probably from the mid 1990 since it was in my stuff and in a dip package). It drew less than 1 mA at 52V with only a gradual current increase with voltage. I'm guessing the rating is pretty conservative!

Since doing the mod last night, I was thinking about how much the current draw varies with voltage (and therefore number of cells) and how that will unbalance a pack that is not set up to use the same number of cells on every CellLog. One elegant solution would be to put a depletion mode fet as a preregulator above the original regulator with its' gate tied to 3V instead of the 1.5 resistor that I have now. This would make current draw mostly independent of supply voltage.

See figure 10 of http://www.ixys.com/Documents/AppNotes/IXAN0063.pdf for a diagram and read the whole document for those not familiar with D-FETS.
At first I thought that they are all much to large for the package, years ago these were hard to find in less than a TO-220 IIRC, but I see you can now get them in SOT223 which should take the power and is thin enough to fit, CPC5602CTR Just may work! This would also drop the total current draw back toward the ideal of 8mA that the base circuits use. I suppose I should do the thermals: 30V drop, 8mA, .24W, easy to dissipate in that package. I've added two to my next Digikey order. It may be a some time before that ships, but it's fun to have another improvement in the pipeline to test.

 

[fechter]

I just did a quick check on a LM358 National part (date code M47AJ probably from the mid 1990 since it was in my stuff and in a dip package). It drew less than 1 mA at 52V with only a gradual current increase with voltage. I'm guessing the rating is pretty conservative!

Like I always say, one test is better than a stack of datasheets. That's comforting to know. Yes, I forgot the ground of the 358 is at 3v, so still within its rating with 8 cells.

I couldn't get the link to work for the FET. I'm interested in that one. I'll find it eventually.
The transistor will need some kind of heat sink I think, or at least some minimum surface area of copper pad to dissipate the heat. Somewhere around here I have some super thin PCB material that was intended for multi-layer boards. Something like that might work.

 

[Otmar]

Here's a DK link: http://www.digikey.com/product-detail/en/CPC5602CTR/CLA298CT-ND/1739360
Almost all parts I find are at DK since I like their search but I try to quote actual part numbers instead of -ND since they are more universal.
For a heatsink I was thinking that a package that big is usually good for 1/2W by itself, but as an extra precaution it would be mounted on a "heavy" resistor lead. Of course a small PCB would be excellent. We'll need to verify my 8mA assumption on the completed circuit since that plays heavily into the power dissipation.

 

[fechter]

It would be easy enough to just build one and try it. You might be right about the dissipation rating. I find a TO220 with no heat sink can get pretty warm dissipating 0.5W though.

I've made linear regulators using enhancement mode FETs but never tried a depletion mode one. I just don't have any in my junk box to play with. One advantage of the depletion mode is it will have a lower dropout voltage compared to an enhancement mode, which is generally around 2v with the ones I've tried.

I've found many times when such a regulator is desireable, usually as a pre-regulator for a 78xx that needs to operate over 35v.

 

[Otmar]

The greatest benefit I see to the depletion mode fet is that it doesn't need a resistor biased gate drive from the input voltage and therefore can maintain a stable current draw over very large input voltage swings. I use one in the Zilla as a precharge sensor to run from 4V to 450V while holding a 1mA constant current.

 

[fechter]

I'll have to get my hands on one to play with.

You may need to drive the gate off the cathode of the 431, otherwise the Vbe drop of the Q6 regulator transistor might be enough to turn it off. I can't really tell from the datasheets to predict.

 

[dnmun]

The greatest benefit I see to the depletion mode fet is that it doesn't need a resistor biased gate drive from the input voltage and therefore can maintain a stable current draw over very large input voltage swings. I use one in the Zilla as a precharge sensor to run from 4V to 450V while holding a 1mA constant current.

so this functioned like a common collector pnp transistor? and the current flowing from the fully charged battery to the controller lead through the resistor pulls down the gate voltage which is connected to the controller lead, as would the base for the common collector?

this was a very educational read, thanks a lot.

if D2 remains, it would still be capable of measuring the 2S if that was all?

 

[amberwolf]

A depletion mode FET is used on the new v3 CA's regulator, and is vulnerable to overheating and destruction if too much current is drawn from it while the input voltage is above a certain point. Hjns and I have both managed this by accident.

As long as you can be certain of the input voltage maximum and the load on the FET is predictable, it'll probably work fine with just a minimum heatsink (the CA just uses the PCB/traces, which isn't very much at all). If things could be variable, you might wanna put a heatsink on there for the worst-case. (the CA doesn't ahve room for that, and the casing is nearly sealed, anyway)

 

[NeilP]

Time to start hacking mine I think

 

[Otmar]

I've made a little time for playing again. Here is an update on my CellLog Hacks:

I've installed the Depletion mode N channel fet as a preregulator and it is working great. I've also managed to fit an optocoupler inside the package to isolate the alarm output.

Here are the details for those who want to duplicate or modify what I've done:
I removed R45 (the diode) and D2. I did do the high voltage mod jumper as others have described.
I then added the CPC5602 DFet as shown in the diagram and the following pictures. It's hanging out on top of other parts with a stiff resistor lead as a heatsink connected to D5. The Source is on the cathode of the old "R45" and D2, this sits somewhere around 5.5V in use. The Gate is on the emitter of Q6 (which by the way seems to be also the top of R15 unlike what the schematic shows). I cut the small drain lead of the fet for more working room.
Doing this hack both cut the power draw in half and made it more consistent with changing input voltage. At input voltage of 6V it draws 9.82mA, at 15V 10.02mA, at 25.2V 10.17mA, and at 33.6V 10.31mA. This is important for those of you that have some CellLogs with fewer cells than others to avoid imbalancing the pack.
At 33.6V the Fet just gets warm. It's dropping about 28V at 11mA for .31W, it has about a 30degC rise on the bench. No problem there as long as you don't leave these in the hot sun. Of course no black electronics survive the hot sun safely.

As for the alarm output, this needs to be isolated when using more than one CellLog on a pack with a common shutdown. I didn't want to tap the pack for this since again the current draw would vary with voltage. I want my opto outputs to be in series so that on = OK and if any node loses power or alarms the master will indicate an alarm. So I've replaced the alarm output transistor with an optocoupler.

I'm using an ac/dc fet output optocoupler, pn TLP227A so that the polarity of the output wiring doesn't matter.
I removed Q9 that normally drives the ALM output. I shorted the Q9 base and collector pads on the PCB. I removed the ALM socket and put the opto LED side in there. I then added a couple of wires out (note, 20AWG is pretty big, try 24 next time) and used a strategically placed zip tie as a strain relief. The inside of the plastic case could be cut a bit to clear the Opto, or you can be sloppy as I was on this first one and bend the opto leads until it fits.

It works well but it may not when it gets hot and as the opto ages. It may be better to get an opto with a required lower trigger current, or reduce the value of R13 to 604 ohm and take the increased current draw of 2 more mA. This may not be needed, but this particular opto lacks a graph on the data sheet comparing input current to output resistance and instead talks about trigger current which concerns me that it may be different from the common (and larger) H11F1M types which are fine at low input and output current.

Lastly, I also made a master unit using an Arduino nano and a LM293 to drive the outputs. This is for shutting off the charger and limiting the controller in case of a fault. I can go into more detail on this if people want. It took many hours of manual wiring and I'm wondering if I might want to make up a PCB for it and offer it as a bare board or maybe a completed device if there is enough demand. At this time the "schematic" is just a bunch of scribbles on paper so I'm going to avoid posting it.

Have fun!

 

[Stevil_Knevil]

Awww, your CellLog 8 coughed up a cute little hairball :lol:

Nice work!

 

[Stevil_Knevil]

It took many hours of manual wiring and I'm wondering if I might want to make up a PCB for it and offer it as a bare board or maybe a completed device if there is enough demand. At this time the "schematic" is just a bunch of scribbles on paper so I'm going to avoid posting it.

..not sure which machine(s) I would put it on.. but please count me as interested.

Q: Does it hurt to be so awesome, Otmar?

 

[Otmar]

Q: Does it hurt to be so awesome, Otmar?

Thanks, but you know this is all spin! I have a disclaimer in my sig now. :lol:

 

[megacycle]

Would be good if it could sell onto jun-si for standard upgraded device with smd.