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

[GGoodrum]

After at least two years of frustrating, and expensive, R&D efforts, the three of us are finally happy with the "new" BMS design. So much so that we gave it a real name, instead of just BMS-v4.3.4.5.3.2.6.3k. For continuity's sake, and to continue to keep track of updates, we will keep the version numbers around. This initial "production" version will be v4.4.

This first post will be used as a placeholder to the latest instructions, BOMs, etc., and to list the most current/up-to-date set of features. This initial offering will come in two "flavors", a 16-channel version, and a full 24-channel variant. The board, shown below, is designed to fit in either one standard Hammond extruded aluminum case, or in two cases stuck side-by-side. The 16-channel version will fit in a single 4.06" x 8.63" x 1.20" box, and the full 24-channel version will fit in two side-by-side 4.06" x 6.30" x 1.20" cases. The 1A+ shunt resistors make thermal contact with the case(s), which then acts as a heatsink.



All in all, this is not functionally a lot different than the original v2.x series. There are still shunt circuits on each channel, and a charge controller that throttles the charge current if a cell tries to go over the level the shunts can handle. The similarities pretty much end there, however. There is a big change in the philosophy of how the shunts operate with this new BMS design. With the old versions, the idea was you would set the charge voltage a half-volt, or so, above the sum of the cell voltage that the shunts come on at, and then the HVC signal was used to throttle back the charge current to keep the cell voltage right at the point the shunt would start to go into overload. At the end of the charge, you'd simply wait until all the shunts were fully cooking away, and that would mean none of the cells were taking any more current in, so the pack would be full and balanced. That was okay with the old system, which had max shunt currents of about 400-500mA. The new shunt circuits can have max shunt currents of over 1A, so trying to use the same sort of philosophy was not possible without active cooling.

What we are doing now is to set the charge to voltage equal to the sum of the desired charge voltage for each cell. The shunts are designed to come on at just above this point. The net effect is that the shunts only come on "by exception". If the cells are perfectly balanced, the shunts don't come on at all. If a cell gets full sooner than the rest, its shunt will come on, and keep it there. The HVC signal is used more like a failsafe, tripping only if the shunt gets swamped. With 1A+ shunt currents, however, the cells have to be pretty far out-of-whack in order to overload the shunts enough to trip the HVC signal.

What this "by exception" scheme allows now is for the current to drop all the way down to 0A, if we want. Before, the current would drop to the level of the shunts, but no lower. That made doing an effective end-of-charge detection pretty much impossible. With the new charge controller we now have an adjustable current sensing circuit that will shut down the charging when current drops below a preset value, which can range from 0 to about 2A.

In most of the v4 variants we've been testing the last couple years, we needed a "Start" or "Reset" button to start or restart the charge process. I was actually okay with this, and I think Andy was as well, but not Richard. finally, he broke his brain for a couple weeks, and came up with a very clever way to eliminate the need for the Start/Reset button. Actually, it would have been much easier if I would've let him get rid of the red-green LED. :lol: I held firm, though, so he had a splitting headache for awhile, until he came up with the fix. Now, when the charger/supply is connected, the LED comes on orange during the normal CC and CV phases. If the HVC starts tripping, the LED will blink. Once the end-of-charge shutdown happens, the LED will change to green. It will stay green until the charger/supply is disconnected, and then it goes off again.

Initially, we will have the 16 and 24-channel boards, shown above, but multiple cell circuit sections can be combined, to support larger setups. For instance, two 16-channel boards can be used to support a 32-channel setup. In this case, a single charge controller section can be used. As before, both LiPo and LiFePO4 setups are supported.

The PCBs, along with a printed set of detailed assembly and test instructions, will be available at my TPpacks.com website. I'll add the link here when they are available. In addition, Andy, and maybe Richard, will eventually offer pre-built/tested versions.

The first boards have been ordered yesterday, so I should have them back by Monday. In the meantime, Andy, Richard and I will be working on the instructions, and the BOMs. Links to these will also appear below, and will be kept updated. I also add a bit more detail later on.

-- Gary


Latest Info:

Latest revision: v4.4.3 (01-Jun-11)

Channel Variants: 16, 24

Order site: http://www.zenid.com/goodrumfechter.htm
(updated 01/21/14)

Bill of Materials (BOM) file:

View attachment 1

Assembly/Test Instructions:
View attachment Zephyr BMS Assembly-Test & Operating Instructions-v4.4.2 Zenid2.doc

 

[amberwolf]

Zis is nifty, but is it not usually "Zephyr" instead of "Zepher"?

 

[AndyH]

We interrupt this message to bring you an editorial comment...please stand by...

When Gary tried to explain his new-fangled 'shunt by exception' hooey, I thought he'd slipped a gear or two. I didn't expect it to make much difference in performance - and to prove it I threw some really ugly imbalanced packs at the channels. Things like a 4S PSI pack with one cell fully charged and three cells 1Ah down. Or a 4S PSI pack with three cells full and one 2Ah down. And my 21S scooter pack with the PakTrakr feeding off 9 of the 21 cells and it's 'balanced once a quarter' condition.

And the damn thing worked. The high cells hit the shunt LED, then the shunt as voltage rose and the channel started making heat. And then the HVC tripped and the circuit started to cycle away while the low cells caught up. Once the cells got somewhat close, the HVC stopped tripping and the 1A shunts kept the high cells under control. Once things got late into the CV stage of charging, the shunts cooled down and only the LEDs were lit.

But just because the shunts have two layers of protection doesn't mean they need it. I've been running the V4.1.5 channels since last summer some time (after stealing the harness from...er...retiring my faithful V2.3 BMS) and it's had rough duty. The channels were working well early on but the charge controllers? Not so much...and it shows...




I've had to resolder some of the shunt resistors (and have had to protect my motorcycle with a piece of fibreglass...) but haven't had to swap a single part on the channels.

Chargers... The V2 units were designed for lead-acid and other 'dumb' chargers and power supplies - and they worked really well with those power sources. They would also work with some 'smart' or 'too-smart' chargers, but not all, and not without some tweaking. Two of the reasons I retired my V2.3 unit is that it wouldn't throttle my Thunder Sky charger - and my ElCon really gave it fits.

But I saw something Wednesday night I'd not yet seen - the V4.3.5 mod c (seriously!) auto-starts when the Thunder Sky charger starts, does its job as the TS charger works through it's computerized charge profile, waits for the charger to finish and shut down, and the automatically turns off.

Folks in a number of threads ask how to have 'plug in and walk away' charging. Here it is.

 

[adrian_sm]

Fantastic. Congratulations on the release.

Bub after all this time, did you have to release it on April Fools Day!

Better be true.

 

[GGoodrum]

Zis is nifty, but is it not usually "Zephyr" instead of "Zepher"?

You are right, of course. I edited the title.

 

[GGoodrum]

Fantastic. Congratulations on the release.

Bub after all this time, did you have to release it on April Fools Day!

Better be true.

Well, technically the first post was still on March 31st, here in California. Yes, it is definitely real. We'll spend the weekend doing BOMs and the instructions, and then I'll enable the boards for sale on Sunday night.

-- Gary

 

[Gordo]

Fantastic. Congratulations on the release.

Bub after all this time, did you have to release it on April Fools Day!

Better be true.

Well, technically the first post was still on March 31st, here in California. Yes, it is definitely real. We'll spend the weekend doing BOMs and the instructions, and then I'll enable the boards for sale on Sunday night.

-- Gary

How sad for me. What totally shitty timing on my part. I quit reading the "pie in the sky, bye and bye" a 1000 post ago, just took a look at the last post every once in awhile. YESTERDAY, after the sun came out and the temp hit 15*C, I ordered 20 mini-BMS with a head! $353.60 Delivered!

Don't get me wrong. I'm not knocking the V4 in any way. Just crying about my very unlucky decision to wait a year, for a much needed BMS and give up hours before the announcement.

 

[joe tomten]

Gary, Richard and Andy,

Thank you for your persistance and perserverence. Serious respect is in order.

I imagine that I am one of many, many people in this forum who have learned more than they ever imagined about this topic having followed the design and implementation of your bms systems.

Eyes are peeled to the tppacks.com site.

Don't want to miss out .

Joe

 

[GGoodrum]

We interrupt this message to bring you an editorial comment...please stand by...

When Gary tried to explain his new-fangled 'shunt by exception' hooey, I thought he'd slipped a gear or two. I didn't expect it to make much difference in performance - and to prove it I threw some it some really ugly imbalanced packs. Things like a 4S PSI pack with one cell fully charged and three cells 1Ah down. Or a 4S PSI pack with three cells full and one 2Ah down. And my 21S scooter pack with the PakTrakr feeding off 9 of the 21 cells and it's 'balanced once a quarter' condition.

And the damn thing worked. The high cells hit the shunt LED, then the shunt as voltage rose and the channel started making heat. And then the HVC tripped and the circuit started to cycle away while the low cells caught up. Once the cells got somewhat close, the HVC stopped tripping and the 1A shunts kept the high cells under control. Once things got late into the CV stage of charging, the shunts cooled down and only the LEDs were lit.

But just because the shunts have two layers of protection doesn't mean they need it. I've been running the V4.1.5 channels since last summer some time (after stealing the harness from...er...retiring my faithful V2.3 BMS) and it's had rough duty. The channels were working well early on but the charge controllers? Not so much...and it shows...

View attachment 2
View attachment 1

I've had to resolder some of the shunt resistors (and have had to protect my motorcycle with a piece of fibreglass...) but haven't had to swap a single part on the channels.

Chargers... The V2 units were designed for lead-acid and other 'dumb' chargers and power supplies - and they worked really well with those power sources. They would also work with some 'smart' or 'too-smart' chargers, but not all, and not without some tweaking. Two of the reasons I retired my V2.3 unit is that it wouldn't throttle my Thunder Sky charger - and my ElCon really gave it fits.

But I saw something Wednesday night I'd not yet seen - the V4.3.5 mod c (seriously!) auto-starts when the Thunder Sky charger starts, does its job as the TS charger works through it's computerized charge profile, waits for the charger to finish and shut down, and the automatically turns off.

Folks in a number of threads ask how to have 'plug in and walk away' charging. Here it is.

Yes, Andy's poor v4.1.5 boards have taken quite a lot of abuse, while we sorted out all the different charge control schemes.

Once we switched to 1A shunts, that's when all the fun started. We had two major problem areas heat dissipation and unwanted oscillations from excessive noise. For the former, we had all sorts of schemes for controlling fans, but I was never happy with this approach. Once I got my head around the "by exception" idea, the heat problems more or less went away. Now, a proper heatsink is all that is needed. What seems to work best for me is to mount the shunt resistors on the bottom, and then silicon glue a 3/16" bar across them and then slide the board into the Hammond box with the bar making thermal contact with the case.



This lets the case itself act as a heatsink. Since we don't ever have a case anymore, where all the shunts are cooking away full blast, this has proven to be more than enough to get the heat out. In my worst-case tests, where I had all but one cell full, the case was warm, but not so hot that you couldn't hold on to it.

Richard finally found a rock-solid way to stop all the weird oscillations that we've seen over the last two years, and all it took was a single capacitor in just the right spot.

I've struggled with layout options, mainly worrying about making sure the cell LEDs were visible. In most previous versions, I included a custom lid for the Hammond boxes that had holes for the LEDs. Unfortunately, this would just about double the price of the PCBs. At one point I even argued to get rid of the LEDs altogether, using the logic that if these boxes are buried inside somewhere, you wouldn't be able to see the LEDs anyway. Then Andy came up with a solution. If the LEDs are to be used, we can just substitute a piece of clear polycarbonate in place of the lid. Problem solved.

Anyway, back to work...

-- Gary

 

[AndyH]

This is how the board looks of one wants to use one of Gary's earlier heatsink techniques. I used it very successfully with a 'scooter BMS' product that's still in the field. It works well with the hotter-running 'old style' shunts so is more than up to the task for this new 'exception' shunt method.

The resistor spacer is a repurposed slice of V2.6c BMS. (Yes Amberwolf, you're rubbing off!)







It's very easy to install the resistors with a spacer, and there's enough case contact for positive heat transfer.

And it's much easier installing all the LEDs flat on the board than trying to space them to fit thru holes in the lid.

 

[velias]

I wonder if we can substitute a single 15 ohm shunt resistor on the cells instead of the paralleled 6.2 one?
I've been using a 15 ohm one on the very original BMS without any problems, my pack has always fully balanced within 5 minutes max.
It makes a big difference in less heat generated.

 

[GGoodrum]

I wonder if we can substitute a single 15 ohm shunt resistor on the cells instead of the paralleled 6.2 one?
I've been using a 15 ohm one on the very original BMS without any problems, my pack has always fully balanced within 5 minutes max.
It makes a big difference in less heat generated.

You could, but why? This version will still run cooler than your's running full blast with 15 ohm shunts.

-- Gary

 

[amberwolf]

Well, if you want it to balance at the higher current rate, you'd need a 3.1ohm single resistor. If you use your higher-resistance shunt, it'd be 1/5 the balance current. I'd guess that's fine if you don't need the higher charge/balance rate?

 

[Alan B]

Great progress!

 

[geoff57]

Hi
The board look great glad to see you finally got all the kinks out of the design.
What I want to know is how we connect this to the batteries, what is the battery output connectors, are they the standard JST HX connectors with 22 gauge wires or the microfit 3.0 connector or similar with 18 gauge wires, what will we need on the battery pack, dose this do everything or will some form of LVC or LVC/HVC board be required or can we even get away with just a parallel board if we want to put several packs in parallel say to make anything from a 10Ah pack right up to a 20Ah pack as has always been posable with the boards from TPpacks and me.

I will be waiting for these boards to become available preferably with the option of a kit form or some I'm sure would like ready built versions.

Geoff

 

[gblast123]

Great to hear this, and many thanks for making this available.

The big weakness to making a great bike has been the BMS.. and, assuming that this pans out, everyone
now can get a high powered lithium pack...something that is in short supply out there.

Any idea or guesstimate as to what the cost would be for the BOM for this? ie 16s board?

David

 

[AndyH]

Hi
The board look great glad to see you finally got all the kinks out of the design.
What I want to know is how we connect this to the batteries, what is the battery output connectors, are they the standard JST HX connectors with 22 gauge wires or the microfit 3.0 connector or similar with 18 gauge wires, what will we need on the battery pack, dose this do everything or will some form of LVC or LVC/HVC board be required or can we even get away with just a parallel board if we want to put several packs in parallel say to make anything from a 10Ah pack right up to a 20Ah pack as has always been posable with the boards from TPpacks and me.

I will be waiting for these boards to become available preferably with the option of a kit form or some I'm sure would like ready built versions.

Geoff

The required connections are the same as the earlier BMSs - balance wires from pack negative thru positive - plus a pair of larger gauge wires from the controller to pack pos/neg for the charge current.

I'm personally working with 12-15A chargers at around 77V, so use Anderson's to the charger and pack. I'm using 18AWG for balance wires (often 2-3 feet of run) and 'regular' size Molex MiniFit/Amp Val-U-Lok for the balance taps and throttle connections.

All are free to use what they want, however, as none of the fittings or connectors are specified or are critical as long as they're up to the electrical and physical task. (None are specified off-board. Gary gives details of the board-mounted connector below.)

No additional boards are required for full cell-level protection. This board does LVC, HVC, balancing, and charge control on one board.

This BMS is more flexible than the V2.x units. One can use one 8S section of channels to protect up to 8S of cells, and can use another 8S section for another group. If they charge the sub-packs separately, the only connection they need to swap is the feed from the charger - the balance leads can stay connected.

The only connection between the channels and the charge controller is the isolated HVC signal line - so each 8S section of channels is independent of the others. It means that folks with larger batteries or multiple subpacks can build a section of channels into the pack if they wish, and mount the controller separately on or off the bike.

 

[GGoodrum]

The connector on the cell circuits we are using now is a 9-pin MTA 156 header, with the latch. The reason for this is that the matching connector doesn't require crimped pins, so no expensive crimper is required. The wires don't even have to be stripped, they are simply pushed into place, using a simple $15 T-handled tool.

-- Gary

 

[dnmun]

way way cool, congrats on a job now done better than we ever thought at the start over a year ago. you guys rate!

 

[fechter]

Here are a few features of Ver4:
100% microprocessor-free. No computers are needed to make it work. Software glitches are not possible. One of my main design rules is to make everything as simple as possible, yet still perform the required function (KISS).

Simple two wire connection to the charger. All power for the board comes from the pack and no separate power supply is needed. No third contact is needed to detect the charger. The charger is detected by the voltage differential between the charger and the pack. Anytime the charger voltage is less than the pack voltage, the control circuit automatically shuts off. In the off mode, the control circuit draws about 1uA per volt on the pack, most of which is from the voltage regulator. This is based on the TLV3702 cmos comparator, which draws less than .5uA per amp.

'Standard' build will support up 24s and up to 20A charge current. Higher current/voltage is possible by using higher rated parts. Board sections can be added to increase cell count. 40s or higher should be possible.

Unused cell circuits can be simply left disconnected. Minimum cell count is 4s.

One of the problems with ver.2.x was charger compatibility. The high fequency PWM would drive some chargers crazy or even cause them to blow up. v4 uses a delayed HVC trigger to keep the frequency below 1Hz and still give the proper duty cycle for fastest possible balancing. A single capacitor determines the frequency of operation and can be tweaked for best charger behavior. The stock value has been chosen to be compatible with the widest range of charging supplies. Keeping the PWM frequency low minimizes heating on both the control board and the charger, as well as minimizes the potential for EMI.

The end-of-charge (EOC) detection is now based on charging current, which allows the shunts to work at lower heat levels near end of charge. V2.x boards required all the shunts to be essentially maxed out to trigger the end of charge. This created lots of heat, all of which is wasted energy. The only adjustment on the v4 board is for the cutoff current. A higher setting will make the charger cutoff earlier. Once the EOC detector triggers, the charging current will be turned off and stay off until the charger power is cycled. When used with "Smart" chargers that have their own end of charge shutoff, the cutoff current adjustment can be turned all the way down, essentially disabling it.

Cell LED indicators are NOT an active circuit element. V2.x boards used the cell LEDs as a zener diode in the HVC trigger circuit. There were numerous issues with consistency of the trigger point and temperature dependency. v4 cell indicators are now just indicators and aren't part of the active circuit. They also come on when the shunts are first active, rather than when they are maxed out like v2.x boards. This allows the indicators to still light up even though the shunts are barely on.

Another feature is 'conduction cooling', which uses the aluminum housing as a heat sink and has a thermal path from the resistors to the case. This allows sealing the case and does not require fans up to a reasonalble power level. Of course you could run the board open like many v.2.x users did or make your own enclosure.

Like earlier versions, this one uses analog linear clamping shunts. As cell voltage tries to rise, the shunt comes on gradually to limit the voltage. HVC trigger is based on the shunt current, not the cell voltage like most other systems. The cell circuit goes from near zero current to maximum + HVC activation over a very narrow voltage range. This ensures the cells are not subjected to any higher than necessary voltage, yet still minimizes balancing time. The oscillation issues were finally solved by using a brute-force filter on the shunt transistor. This puts the 431 clearly outside the conditons that can cause oscillations. Luckily modern technology has brought us tiny little capacitors with huge capacitance at a very low cost, so this solution works out nicely.

 

[gblast123]

So,if I get this right, you can pre-build subpacks of 24v (for lifepo4) and then just pop them in?

If so, it would make finding and replacing any crappy cells a lot easier.

 

[jimw1960]

Sorry to be thickheaded, but I'm having trouble understanding what kinds of arrangements these boards can handle. So let me say specifically what my need is. I have four 4s packs, each with separate 4s balance connector, that I want to put together in series to make a 16s pack. Which board will I need to do that and how would I hook the four separate balance leads to it? I assume I would need to make some kind of 4-to-1 adapter cable?

 

[fechter]

A 16s board could be used with some kind of adapter cable for your existing balance connectors or you could cut the connectors off and wire them directly to the board.
Most of the layouts have cells grouped in 8s sections, each section with a separate connector.

 

[Gordo]

Hi
The board look great glad to see you finally got all the kinks out of the design.
What I want to know is how we connect this to the batteries, what is the battery output connectors, are they the standard JST HX connectors with 22 gauge wires or the microfit 3.0 connector or similar with 18 gauge wires, what will we need on the battery pack, dose this do everything or will some form of LVC or LVC/HVC board be required or can we even get away with just a parallel board if we want to put several packs in parallel say to make anything from a 10Ah pack right up to a 20Ah pack as has always been posable with the boards from TPpacks and me.

I will be waiting for these boards to become available preferably with the option of a kit form or some I'm sure would like ready built versions.

Geoff

The required connections are the same as the earlier BMSs - balance wires from pack negative thru positive - plus a pair of larger gauge wires from the controller to pack pos/neg for the charge current.

I'm personally working with 12-15A chargers at around 77V, so use Anderson's to the charger and pack. I'm using 18AWG for balance wires (often 2-3 feet of run) and 'regular' size Molex MiniFit/Amp Val-U-Lok for the balance taps and throttle connections.

Andy;
I understand you are using 21 TS 40Ah LiFePO4 cells? Are you using 3 BMS boards and just populating the 21 sections? Have you modified the TS charger to get to 77V? I have an original 3.5 M Chen scoot using 20 cells and have 3 spare to swap out for my 3 lazy ones that go low after 4-5 cycles. I want to try the BMS first, before I swap out my lazy cells. And I may add the 21st cell if I can adjust the voltage of the TS charger?
Thanks;

 

[GGoodrum]

Sorry to be thickheaded, but I'm having trouble understanding what kinds of arrangements these boards can handle. So let me say specifically what my need is. I have four 4s packs, each with separate 4s balance connector, that I want to put together in series to make a 16s pack. Which board will I need to do that and how would I hook the four separate balance leads to it? I assume I would need to make some kind of 4-to-1 adapter cable?

Here is graphically what you need:



This just shows the balance connections. It is assumed each 4-cell block is already wired in series into a 16s configuration.

-- Gary