"Zephyr" - Finally, the "v4" Fechter/Goodrum/Hecker BMS...

[Zenid]

Incidentally, here's my own pretty picture of where the diode goes, from my blog:
http://zenid10.files.wordpress.com/2011/09/gfmod04b1600x12001.jpg

[fechter]

Now I've put the extra main MOSFET on, should I be okay with a 15A or 20A charger? I built it solidly with nice thick gauge wire everywhere. Are there any issues I should be aware of before I try it with anything more powerful (apart, of course, from making sure the EOC is stable and works okay now)?

Should be no problem. A single FET can handle up to about 20A. The current shunt will get hot, but it is made to take it. Just avoid having any wires in contact with it as the insulation could possilby melt.

[Gregb]

Still not terribly impressed with the instructions. In the Charge Control section Test It says use a current limited PS set to 100mA but no mention of initial voltage. It then says turn voltage up to 20V... From what ???
There is still no instruction for 8 cell units that after cutting the board then term 9 or B8 should be strapped to +.
and it is as though the 47uF (c101) capacitor across the leads doesn't exist....

[Zenid]

Still not terribly impressed with the instructions. In the Charge Control section Test It says use a current limited PS set to 100mA but no mention of initial voltage. It then says turn voltage up to 20V... From what ???.

From zero would be a safe guess...

I didn't have any fancy equipment, so I used his alternative approach of simply running a power supply to it in series through a 100W bulb. That protects the board in case there are any shorts. You can test both the shunt circuits and the main control circuit that way. If you look earlier on in the thread, you can see where I'm running through the tests and double-checking with Richard (Fechter) if I'm doing it right, what to expect and what to do next.

I will get round to writing this up at some point, but I've been busy with other things. If you don't have bench power supply equipment, just run a charger through a 100W light-bulb. Use a little screw-in light-bulb holder, and run the positive through the bulb and use the wire that comes out of it as your "protected" power source to attach to the positive of the board, to check that the main LED lights and that shorting the EOC pins turns the LED red. If something is badly wrong with the board then the bulb will light up brightly.

[Gregb]

Yes I have been collating all your chats about testing. It is just that the tests and instructions seem to be written for advanced users and not people like yourself, new to the game. The charging or discharging of the 47uF could be confusing to people trying to measure/compare the chan leads. And the other things I have listed should have been included by now. There is serious money involved in building this. Also many bench supplies don't like being started at zero. I have written quite a few instructions for techs over the years and these are not among the best I have read. There doesn't seem to be a consolidated list of corrections yet or the differences of the various marks of boards. Especially the addition of the switching diode. And Gary's web page is still down. I have been caught up with a lot of other things recently including making a bench PS so have put it aside for the time being and just read the thread when I can. Don't get me wrong I think they have done a good job, but when you charge money for things, you have to keep it up to date.

[GGoodrum]

I agree, the test instructions could be better, but I must say, this was never intended to be a project for "first-timers", so to speak. Nonetheless, the whole point of this thread is to provide help for those who are putting one of these together. The big problem with this is that although this board will work with a very wide variety of configurations and charging setups, each one of these might operate a bit differently, in terms of how the circuits operate. It makes it hard to have a "complete" set of testing instructions, covering all possible cases.

In the next version of this, which Richard and I are working on right now, we are using a new control scheme that should work the same way with virtually every setup. The end-of-charge detection logic will work the same way, for all setups, and there's no "adjustments" required. The circuit is simpler, with fewer parts, because we were able to eliminate the current measuring portion of the circuit. Anyway, this should help simplify the testing instructions for these new variants.

when you charge money for things, you have to keep it up to date

You make it sound like we are making money off of this stuff. The "profit" from these boards is about $5. Even if the margin was $100 a board, I'd have to sell literally hundreds of these in order to come close to making up what I've spent on test boards and unsold older versions. Also, the website is still down because I have been concentrating on these new versions, plus the related "Lite" variants, and haven't had time to build a new website.

Anyway, if you have a problem with a setup, post as much detail as possible here, and we'll all try and figure out what is going on.

-- Gary

[fechter]

Quite right, I agree the instructions are not very good.
If you have any problems figuring out the testing procedure, you can ask here or pm me.

[Gregb]

Sorry Gary was not inferring you were making money or or being duplicitous. apologies if it came out that way. Some time ago I sent fechter a reference to a SM technology lifepo charger from Microchip. I have been busy lately on volunteer things for handicapped people and haven't been able to give it any time, but have you considered changing your philosophy slightly? Give away the shunt method, but keeping your control panel use it to switch on individual channel chargers when one channel hits full on bulk charging which would then be turned off. If they built a version on a more useful package size I would have tried it myself. You have hinted that SM technology is available to you. I bought one of their test boards and it performed well with accurate cut off voltages etc but I don't have the tools to mount them. The chips are very very cheap and the specs are pretty impressive.

regards
and apologies if necessary

Greg

[fechter]

Yes, I've looked at dozens of chips and so far the best ones I've seen are the Seiko ones. It's still on the drawing board, but we want to get a solid through hole version first. I'm not so good at hand building with those microscopically small smd components. The capacitor coupled balancing scheme still has potential too, but will take a lot of time/money to get built and tested. This would avoid the need for shunts and all that wasted heat.

[Zenid]

Yeah. Goodrum & Fechter are spending too much time on luxury cruises, and lounging around in gold-plated swimming pools off of all the money they made...

Come on guys! We want one of these right away:



Seriously though I will eventually get round to doing a test section on my site, complete with the usual pretty illustrations and photos. I'm just a little swamped right now...

[GGoodrum]

Yes, the surface mount stuff is an attractive future option, mainly because of the reduced footprint and the possibility of being machine-assembled. With these two new (16 and 24-channel...) Zephyr variants, we have addressed both issues, however, by borrowing from the "Lite" variants I'm still working on for LiPo setups. What we did was switch to using resistor arrays and use quad versions of the opto chips. These two changes alone have significantly reduced the build time and the parts count, for a typical board. For example, it used to take me literally hours, just to bend the leads, install, solder and trim the hundred, or so, resistors used on a typical 12-channel board. With the use of the arrays, the time to install the equivalent number has been reduced to about 15 minutes.

The use of these arrays also allows us to really reduce the footprint of both the 16 and 24-channel versions, as you can see below:

24-Channel Zephyr BMS-v4.4.8m.png

(202.94 KiB) Downloaded 3 times

Both are now sized to fit in a single Hammond extruded aluminum box, the 4.06" x 6.3" version for the 16-channel, and the slightly larger 4.74" x 6.3" variant for the 24-channel board.

There are now three opto busses, LVC, Any HVC and All HVC. The reason for the two HVC lines is that the Any HVC optos are connected, like always, in parallel. The All HVC set of optos are connected in series, so it doesn't trip until all the shunts are finally on, which says all the cells are full and balanced. The new charge control logic uses this All HVC signal to trip the end-of-charge (EOC) shutoff logic. The any HVC signal, which trips at a slightly higher point than the All HVC signals, is used to cycle/throttle the charge current, just like on the current Zephyr variants. This provides for HVC protection for errant, or grossly imbalanced cells.

We just finished these layouts, and they are being turned in for a test run on Monday. Richard will be testing a 16-channel version with his 16s5p a123-based LiFePO4 setup, and I'll do a 24-channel version, to try with one of my LiPo setups. I've already been testing the new control logic, with the Lite variants, and the basic cell circuits are pretty close to what we've been using for some time, so we really don't expect much in the way of surprises. Anyway, I will post our progress here.

-- Gary

[Gregb]

Please, please, please, get away from this through hole plated technology. Just had a faulty FAN and getting the components out for fault finding is impossible. By all means use double sided boards (you have to) but solder both sides of any through connection..... and some of the track- hole spacing is a bit tight and allows solder across....

[fechter]

We don't really have an option on the through hole plating unless we go to another board house. If the holes were slightly larger, removing the parts would be easier. I use a desoldering station and that works great for getting parts out in one piece. You can always cut the legs off a part and remove the legs separately, then clear the holes.
I try to avoid close spacing wherever possible to make soldering easier, but there are going to be some tight spots no matter what.

Always hold the board up to a light after soldering to inspect for bridges. When the board is back lit, any bridges will be pretty obvious.

[GGoodrum]

I've done quite a bit of parts replacement, with all the test versions we've done, and I use an inexpensive solder sucker/vacuum device, from Radio Shack, to clear the holes. Works great.

I've been finding solder bridges to be much less of a problem since we started using these resistor arrays, which have legs that don't need trimming. Also, having the main traces on top helps a lot.

Here's our latest variant, which is now a single PCB that can either be 16 or a full 32 channels:

The 16-channel version fits in a 4.06" x 6.3" by 1.2" extruded aluminum box, and the 32-channel version is sized for the taller 2" version of the same box.

-- Gary

[chroot]

@GGoodrum- Have you decide the pre-made BMS?

[ejonesss]

since the link to tppacks does not work do you still have any boards in stock and how much?

[Gregb]

you might be better to send Gary a PM.....

[GGoodrum]

Sorry, I missed that. I'm going to try and make the website a priority this weekend, but in the meantime, you can send me a PM. If I were you, however, I'd wait for another week, or so. Monday I am turning in our latest version of the newer, easier to build Zephyr variant, which makes use of resistor arrays and quad opto chips that drastically reduces the build time.

This latest board will support up to 32-channels, in a single box. For sixteen channels, a 1.2" x 4.06" x 6.3" box is used, and for the full 32 channels, both boards go in the taller 2" version of the same box.

Functionally, this is the same as its been for a the last several versions, as we've been mainly tweaking the layouts. There's three opto lines, one standard one for LVC, and two for HVC, one set a bit higher than the first. The higher one is used as a failsafe, to catch any weak/errant cells that may reach the normal cutoff way before the rest of the cells. This will trigger the HVC cycling mode where the power will cycle between off and on about once a second, until the rest of the cells get closer. The other HVC opto output string is connected in series. This signal goes active when all the cell's bypass circuits are on, which means the cells are all full. This then causes the EOC logic to shut off the charge current and turn the main LED a solid green.

Stay tuned...

-- Gary

[ejonesss]

that's ok for now ggoodrum budget (i am recovering stuff lost or damaged by this year's floods on the east coast caused by a week of off and on rain and i am trying a couple other ideas using a pair of meanwell power supplies and 10 voltphreak chargers.

the temporary house i am in has a bad outlet that caused 2 soneil chargers to blow and take $300.

also will the failsafe mode eliminate the effect that seems to occur on the negative end of the pack where the end 4 cells seem to be slightly lower than the rest?

[Gregb]

"also will the failsafe mode eliminate the effect that seems to occur on the negative end of the pack where the end 4 cells seem to be slightly lower than the rest?"

Could you elaborate on this? Do they measure a lower voltage? and by how much ? and does this effect come in after standing for a while or are they cutting off early ?
None of mine are exactly the same particularly after an hour or so. and are you leaving something like the cell log across them ? It does consume power

[GGoodrum]

"also will the failsafe mode eliminate the effect that seems to occur on the negative end of the pack where the end 4 cells seem to be slightly lower than the rest?"

Could you elaborate on this? Do they measure a lower voltage? and by how much ? and does this effect come in after standing for a while or are they cutting off early ?
None of mine are exactly the same particularly after an hour or so. and are you leaving something like the cell log across them ? It does consume power

This was with the original v2.x boards, which had a tap for the 12V regulator off the top of the 4th cell. All of the v4.x variants have the regulator powered from the whole pack voltage.

-- Gary

[GGoodrum]

Here is the latest layout variant, that we just turned in to run:

Functionally, it is the same, but we've made some accommodations so that this one PCB can support 12s, 16s, 24s and 32s configurations. Also, I think we are a lot closer now to having a final variant. After Richard finishes testing this one, we should be good to go.

-- Gary

[SlyCayer]

Hey,

I bought the board very early around the 11th page of this thread, and I am wondering if the board I have is good enough to trust with A123 20Ah worth of 7000$...

I never had the time to put together after I ordered the board and the parts.

Is there anything that has changed significantly since I bought the parts and board? I will be putting it together soon and if you are telling me it it's no use to put it together due to problems with the original design then I won't lose time with it.

[fechter]

The addition of the diode described on the previous page is about the only major thing. There were a few mismarks on the actual board for the components, so when in doubt, check the schematic. The schematic is correct. I'll send you the updated schematic. The testing instructions suck. Zenid's version will help. Building this board is not a good project for a beginner or someone with limited electronics knowledge.

[gensem]

Gary,

Would I be able to use non standard cell numbers, like 26s, 28s and 30s in the 32s "BMS"?