Weber & Coulomb's LyteFyba BMS

How do you store and manage your electricity?
User avatar
weber
Site Admin
Posts: 2936
Joined: Fri, 23 Jan 2009, 17:27
Real Name: Dave Keenan
Location: Brisbane
Contact:

Weber & Coulomb's LyteFyba BMS

Post by weber »

It's about time we started a separate thread for our open-source BMS development. And what better way to start than with the debut of the next-generation design of our fibre-optic cell-top management units.

Earlier development history can be found in the build thread for LyteFyba's rolling test-bed, the electric MX-5 called Mexy.

For an overview of the features, see http://dkeenan7.github.io/LyteFyba/

I'd appreciate any suggestions or questions. I'm only posting the schematic at present, since we're still working on the PCB artwork.

Click on the preview image below, to open a zoomable PDF.

Image

The new features:

* They are designed to bolt on top of all the common sizes of prismatic cell from 40 Ah to 200 Ah in the brands CALB, Sinopoly and Winston (and the older Sky Energy and Thundersky). They are adaptable to other sizes by soldering wires to their "snap-off ears". [This mechanical feature doesn't show on the schematic.]

* We now have terminal-to-strap voltage-drop measurement. The terminal bolt is treated as high-impedance access to the cell-terminal so we can pick up voltage drops under load, between the copper strap and the cell-terminal, due to loose bolts or corrosion. No third wire is needed between CMUs as in the previous attempt. We simply use non-plate-thru holes for the terminal bolts and we measure the difference between the top and the bottom of the PCB at each terminal. The top surface connects to the cell terminal via the bolt. The bottom is against the strap.

* We now drive two of the LED colours (red for error and blue for activity) from a single processor pin. We plan to send it AC when we want magenta (red and blue at the same time). This was done to free up the extra analog input required for bolt voltage sensing.

* We've doubled the bypass capability from 400 mA to 800 mA to cater for the larger capacity cells. But half of the bypass resistors can be snapped off, along with the unneeded terminal "ears", so the CMUs can still fit on the smaller cells.

And we thought this BMS needed a short snappy name, hence "LyteFyba".

[Edit: Added link to MX-5 thread for earlier development history]
Last edited by weber on Tue, 12 Jan 2016, 08:39, edited 1 time in total.
One of the fathers of MeXy the electric MX-5, along with Coulomb and Newton (Jeff Owen).
Astroboy
Groupie
Posts: 53
Joined: Mon, 13 Jan 2014, 06:33
Real Name: Jase
Location: Melbourne, Au

Weber & Coulomb's LyteFyba BMS

Post by Astroboy »

The snap off sections sound like a neat idea, especially with the number of load circuits reducing as each snap off section is removed. Less load for smaller capacity cells and higher load for larger capacity cells. Sounds good to me.

Will be interesting to see the PCB.
When i look at schematics they always look like they have a high component count but then when it is layed out neatly on a PCB it doesn't look so bad.
I am curious about the three series resistors on each of the sensing inputs (R3a,b,c & R4a,b,c). Is it physically 3 separate resistors?

I can only think of one suggestion for a possible improvement and that is the tx + and -. Depending on the connector chosen it can be easy to mistakenly connect the tx of one module to the tx of another module. If tx+ connects to tx- of the next module would the current cause some damage? The rx ports are fine but maybe the tx ports could do with a current limiting resistor in there. Maybe one on each leg of the port.
If using the optional optic modules then there is no issue.

The optic modules did seem a bit pricey.
A very quick search showed them to be around $4 (in small quantities). So $8 per cell top module for the optic modules. Are they available cheaper?

Do you have an estimated component cost per cell top module?
User avatar
coulomb
Site Admin
Posts: 6357
Joined: Thu, 22 Jan 2009, 20:32
Real Name: Mike Van Emmerik
Location: Brisbane
Contact:

Weber & Coulomb's LyteFyba BMS

Post by coulomb »

Astroboy wrote: Will be interesting to see the PCB.

The artwork isn't quite finalised yet, but this may give you an idea.

Image

Pretty snappy, huh? Image

The above really a view from underneath; I guess DesignSpark puts this side on top by default since it's the main component side. Also, its guesses about the sizes for some components is way off, largely because we don't care enough to tell it reasonable sizes.

One of the most difficult to place components has been a 100 uF capacitor needed only for the IFO senders (opto LEDs for the fiber connections, when used). It was in the way of a track, so I moved it around ninety degrees, only to be told by Weber "Bzzzt!" (not his exact words). It has to clear the vent and I think also the "rim" of the cells. In fact, it would be great to get it on the other side of the board, even perhaps as a through-hole component, but it seemed hopeless. Nevertheless, Weber the magician has put it there. Still as a surface mount component. Here is a view from the other side, with the capacitor seemingly intersecting with the piezo beeper:

Image

You can see it's at a 45 degree angle, jammed between the piezo beeper and two bypass resistors. The outlines of these components are usually a little larger than they really are, and Weber assures me he's tested the spacing in real life.

[ Edit: made it clearer which side was which ]

Edit 2: don't mind the bypass resistors floating out in space. Rest assured that there is a perfectly logical Image reason for this, to do with rotations and how every second CMU is flipped over with 80% of the components moved 180 degrees from its neighbours. (That's so that positive and negative terminals can alternate, the usual case for series strings.)

[ Edit 3: updated 3D component models, re-did one of the images. Twice. ]
Last edited by coulomb on Sat, 11 Oct 2014, 11:09, edited 1 time in total.
MG ZS EV 2021 April 2021. Nissan Leaf 2012 with new battery May 2019.
5650 W solar, 2xPIP-4048MS inverters, 16 kWh battery.
Patching PIP-4048/5048 inverter-chargers.
If you appreciate my work, you can buy me a coffee.
Astroboy
Groupie
Posts: 53
Joined: Mon, 13 Jan 2014, 06:33
Real Name: Jase
Location: Melbourne, Au

Weber & Coulomb's LyteFyba BMS

Post by Astroboy »

Coulomb, that through hole piezo seems massive.
Any reason for not using a much smaller surface mount piezo?
If you could use a smaller piezo then you should be able to get the capacitor further away from the bypass (load?) resistors. Load resistors get quite hot and capacitors usually don't live long when heated.

I like the snapoff load winglets. Maybe add some pads on the main section just in case you ever need to reattach a removed winglet. A couple of pads on each side would give you something to solder some bridging wire to.
Good idea, keeping a lot of the copper intact on the winglets, should help with heat dissipation. Acting as a heatsink.
User avatar
weber
Site Admin
Posts: 2936
Joined: Fri, 23 Jan 2009, 17:27
Real Name: Dave Keenan
Location: Brisbane
Contact:

Weber & Coulomb's LyteFyba BMS

Post by weber »

Awesome 3D images Coulomb. But yes, they should be taken with a large grain of salt.

Thanks for your help on this, Astroboy.
Astroboy wrote:I am curious about the three series resistors on each of the sensing inputs (R3a,b,c & R4a,b,c). Is it physically 3 separate resistors?
Yes they are 3 separate 0805 (imperial) resistors. This is part of the high voltage-withstand of our design. If a cell or bolted terminal connection goes high resistance under load the CMU may see the entire battery voltage across it in reverse. Our CMUs are designed to not arc over and ignite venting electrolyte solvent under such abuse, and even to continue transmitting data. The 1000 V diode, the fuse and the 1F supercap are the other parts that make that possible.
I can only think of one suggestion for a possible improvement and that is the tx + and -. Depending on the connector chosen it can be easy to mistakenly connect the tx of one module to the tx of another module. If tx+ connects to tx- of the next module would the current cause some damage? The rx ports are fine but maybe the tx ports could do with a current limiting resistor in there. Maybe one on each leg of the port.
We used to have such resistors but eliminated them in a drive to reduce parts count, to keep the parts placement costs down. If CMUs were left connected TX+ to TX- for long enough, the MSP430 might well overheat and be destroyed, but they can certainly survive momentary shorts. This includes a short on the output of a single CMU. However our design, using two different PCB isomers in alternation, means that there is no confusion about which side of the board to connect to. They always connect to each other's nearest side.
The optic modules did seem a bit pricey.
A very quick search showed them to be around $4 (in small quantities). So $8 per cell top module for the optic modules. Are they available cheaper?
In fact the red LED device is about $6, so about $10 a pair. They are of course cheaper in quantity. But you would never use them on every cell-top module, only when the signal needs to go outside a battery box or go past an Anderson Connector or an ELV-segmentation contactor. So typically you would only have a pair of the Industrial Fibre Optic (IFO) connectors for every 27 to 36 cells, which works out to less than 40 cents per cell.
Do you have an estimated component cost per cell top module?

It cost us $1280 in parts to build 250 of the previous version. That's just over $5 per CMU. That's excluding the IFO connectors, PCBs and solder stencils but including the solder paste. The new version has the same parts except for an additional resistor.
Coulomb, that through hole piezo seems massive.
Any reason for not using a much smaller surface mount piezo?
Only one reason -- cost. The 17 mm PS1740 was cheaper than smaller ones (from all 3 suppliers, DigiKey, Mouser and Element14). But thanks for prompting another look at this. Element14 now have the 13 mm PKM13 and the 14 mm PS1420 at a lower price than the PS1740.
If you could use a smaller piezo then you should be able to get the capacitor further away from the bypass (load?) resistors. Load resistors get quite hot and capacitors usually don't live long when heated.
What do you consider is the difference between load resistors and bypass resistors? They are used for top-balancing. They are turned on when a cell is fully charged (~3.5 V), to bypass some current past the cell, to allow the remaining cells to catch up.

They are only dissipating about 150 mW each (don't forget the diode drops about 0.8 V when bypassing). We've measured them at max 35 K above ambient after running for hours in a closed battery box, so not very hot. And in normal use they only run for a few minutes at the end of a charge. So they shouldn't have much impact on the life of the capacitor. But yes, it would still be good to get it further away.
I like the snapoff load winglets. Maybe add some pads on the main section just in case you ever need to reattach a removed winglet. A couple of pads on each side would give you something to solder some bridging wire to.
Good idea. Thanks.
[Good idea, keeping a lot of the copper intact on the winglets, should help with heat dissipation. Acting as a heatsink.
What isn't shown in the above 3D images is that there are similar heatsink copper pours under the 8 bypass resistors surrounding the piezo on the main part of the board.

Also not obvious from the 3D views is that we not only have snapoff bypass winglets, we also have snapoff terminal-bolt winglets. I don't know why the row of holes isn't shown on them, like it is on the bypass winglets.
One of the fathers of MeXy the electric MX-5, along with Coulomb and Newton (Jeff Owen).
7circle
Senior Member
Posts: 540
Joined: Thu, 04 Feb 2010, 02:18
Real Name: Ken
Location: Melbourne

Weber & Coulomb's LyteFyba BMS

Post by 7circle »

Nice improvements fellas

So many cell sizes:
Image

New factory development linked to Winston in Russia
http://liotech.ru/productionen
Image

The Bendy board bridges would need to be pointing the same direction on both sides of a CMU to allow alternate rotations.
   ///
O+[XX]-O
   \\\
O-[XX]+O
   ///
O+[XX]-O
   \\\

If that txt picture makes sense.

Do the RX/TX Tabs in the manufactured position function without being snapped off and repositioned with two-pin connectors?
ie .. do they have tracks across the break line to a RX+/RX- or TX+/TX-
(the way the schematics suggests)

Is the same Snap-off section/piece used for both RX and TX fibre connection?

Are the row of holes for snap-offs just holes or vias with plated-through-hole structure or just bare-holes. Wondering about strength And if they were soldered over to add strength if needed.
It's a concept I'm not familiar/wise with.

What kind of vibration can they withstand and problems of im-britlement causing failure. If the completed boards are sprayed with a a coating that might help.
Would an extra interconnect on the sides to the terminal bolt "tail" be worth including,
and even between CMU's like the squiggle links.

Say a Battery Shop has the LiteFyBa CMU's ready on the shelf fully assembled by the box full. They supply them with cells and they or the user trim off the the bits not needed. Using some side cutters and soldering iron tools they:
- fit the parts to Fiba RX or TX modules as needed
- position them to suit optic fibre harness.
- And remove un-wanted BY-PASS tabs.
- And extend Bolt/Strap tabs or discard extra length.

So simple enough, no need to get out the Dremmel or hand-file to make it work.
  • Some large cells have multiple terminals per electrode.
    Would it be worth monitoring all straps.
    Could diodes be used as an OR gate to pick up a high potential on extended and extra snap-off bolt/strap detectors.
    Problem will be more detectable as a high temp at the bolt.
    Just brain-storming issues and ideas for other cell formats.
Perhaps you may invest in building a 4 cell large Ah 12V battery to test out the design with extended bolt/strap tabs.

You could use it in the MEXY as Auxiliary battery and have it linked in somehow via Fiba cables to the main BMS network, with Alarm monitoring.

But the challenge of having the CMU'd 12V AUX Battery in the system is a challenge that most EV's will need to consider.
As a broader use CMU concept, buses and trucks with 24V systems may have more than the HV traction battery pack that could use the LiteFyba CMU's.
Imbalance on 12V systems is more obvious and can be managed by just DC/DC controller Hi and Lo Limits. But 24V systems it can be hard to pick up a cell failure.

This sounds like a dedicated smart HV-DC/DC-12V with 4 cell monitoring would be a better approach.

Also would the CMU's cope with higher cell voltage chemistries?
Say 4.5V max. Winston have the LTHPxxx Series with higher cell voltage.
Or lower types like the Lithium/Sulphur? (still on the WINSTON's partnered site Generplus Italy - Winston LSP )

Did you have ideas for a EV that needed a charger interface.
Do I remember correctly someone mentioning that CMU #1 could monitor system stress and control a charger like the ELCON?
Or would this need a separate brain PCB?

On the FUSE defusing. What would cause this?
Should the Fuse be before the diode?
Would a 1A "poly-switch" that self-resets be an option?
It would make the board more "rugged" to failure and rework repair.
(If it suites the failure mode)
But is the FUSE for final voltage isolation in some "hammer" test for safe operation under crash conditions?
After thinking/and rethinking any event that drew more than ex;ected current (1A) would be a issue that a blown fuse would need serious investigation.
A self reset is the lazy approach/band-aid solution.

A 1A fuse will be at its limit with 3.3 ohm by-pass at 4.2V-0.8V.
But your calcs have it at 800mA.

Also..
Would it be worth having Vdd on the Jtag Header P5?

Do you ever contemplate if gnd and its symbol should be replaced with a triangle and COM net label.
I still find it miss-leading using GND for things that are never meant to be grounded to earth with a spike in the ground.
Is there an international standard for this?
Perhaps "wonday" will have Moon and Mars etc ground symbols too.
I should stop wondering off topic.

Does the Blue LED need a by-pass cap to stop noise injection from Vdd?
Or is C1 close enough.
Could it allow reverse current back to Vdd if no power applied to Vdd from cell.
Just thinking this could negate the 1F Cap purpose if activity was displayed in such an unlikely event. (ie By-Pass active but no cell voltage present)
May occur if programmed for bottom balancing ... but who'd want to do that....
Or the unlikely event the FET fails Short/circuit and blows the fuse and the 1F is used to allow time for detection of the failure into Brain alarm Log.
How much reverse leakage would there be...
Error active .. so 2.5V onto reversed Blue LED (1V) to RES(39R) to 3R3 shunt...
- Say est 1V/39R = 25mA .. but Blue LED wouldn't cope well with this,...
Time to stop my obscurity wondering ...

Looks good guys.
You've stretched me far for "constructive criticism".

7C Image
User avatar
weber
Site Admin
Posts: 2936
Joined: Fri, 23 Jan 2009, 17:27
Real Name: Dave Keenan
Location: Brisbane
Contact:

Weber & Coulomb's LyteFyba BMS

Post by weber »

Wow! Thanks 7circle. I will address your suggestions in another post. But you will see I've changed the fuse to 1.25 A. Thanks!

Here's the artwork, which should answer some questions. If it raises other questions, or suggestions, please make them soon. I need to get this off to fabrication ASAP. I've got an off-grid solar job in which I plan to use them.

Astroboy, you will see that I have used a smaller piezo buzzer as you suggested, although I still wasn't able to move the 100 uF electro very far from the bypass resistors. But I'm afraid the pads for re-attaching bypass winglets were just too hard, as they would eat into the heatsink area.

This shows both isomers. You can see that they are not simple rotations of each other. Only the wings and the section directly between them rotates. The LEDs do not rotate, nor do the comms and JTAG terminals or the parts related to them. The north and south areas stay the same while everything between them rotates. And we only have to maintain one isomer. The other one is generated from it by a few simple operations. Pretty tricky huh?

Image

Image

[Edit: Oops. I forgot to flip the arrows near the plus and minus signs on the north isomer. They point to where the high-current link or strap should be.]
Last edited by weber on Mon, 13 Oct 2014, 17:04, edited 1 time in total.
One of the fathers of MeXy the electric MX-5, along with Coulomb and Newton (Jeff Owen).
User avatar
weber
Site Admin
Posts: 2936
Joined: Fri, 23 Jan 2009, 17:27
Real Name: Dave Keenan
Location: Brisbane
Contact:

Weber & Coulomb's LyteFyba BMS

Post by weber »

7circle wrote:The Bendy board bridges would need to be pointing the same direction on both sides of a CMU to allow alternate rotations.
   ///
O+[XX]-O
   \\\
O-[XX]+O
   ///
O+[XX]-O
   \\\
If that txt picture makes sense.
I understand your picture. But you will have seen now, that it is in fact:
|  ///
O+[XX]-O
   /// |
O-[XX]+O
|  ///
O+[XX]-O
   /// |
Do the RX/TX Tabs in the manufactured position function without being snapped off and repositioned with two-pin connectors?
ie .. do they have tracks across the break line to a RX+/RX- or TX+/TX- (the way the schematics suggests)
We call those 3 bendy bridges "squiggle joins" and yes the leftmost one has tracks across it connecting RX to TX. These squiggle joins are only of use on cells which are 46 mm thick (most manufacturers' 40 Ah cells). For all other cells they must be cut off (aviation snips work well) and replaced with soldered wires, or headers and crimp or IDC/MTA connectors.
Is the same Snap-off section/piece used for both RX and TX fibre connection?
Correct. We only have these snap-off IFO daughter boards on every second CMU. So there are enough to have IFO connectors on both ends of 12 volt batteries (4 cells).
Are the row of holes for snap-offs just holes or vias with plated-through-hole structure or just bare-holes. Wondering about strength And if they were soldered over to add strength if needed.
It's a concept I'm not familiar/wise with.

What kind of vibration can they withstand and problems of im-britlement causing failure. If the completed boards are sprayed with a a coating that might help.
They are just bare holes. To DesignSpark they are pads of zero diameter with 1.2 mm unplated holes spaced at 1.9 mm centres. On the terminal wings they will be drilled through the copper pour as well as the FRP. FRP is very strong stuff. Based on the force required to snap off the IFO daughter boards on the previous version (with no copper across), we think you'll need a vise and/or pliers to break off the terminal wings.
Would an extra interconnect on the sides to the terminal bolt "tail" be worth including, and even between CMU's like the squiggle links.
I don't understand what you're suggesting here. For unusual widths, provided they don't use larger than M8 bolts, we've provided strain-relief holes and solder pads on either side of the dotted line so both sides of the PCB can be reconnected with wires after the piece has been snapped off.
Say a Battery Shop has the LiteFyba CMU's ready on the shelf fully assembled by the box full. They supply them with cells and they or the user trim off the the bits not needed. Using some side cutters and soldering iron tools they:
- fit the parts to Fiba RX or TX modules as needed
- position them to suit optic fibre harness.
- And remove un-wanted BY-PASS tabs.
- And extend Bolt/Strap tabs or discard extra length.

So simple enough, no need to get out the Dremmel or hand-file to make it work.
That's the plan. Simplify inventory. No need to decide how many of each of maybe 4 types to stock, like other cell-top BMSs. And a cell-top BMS avoids the safety issues of a "spaghetti BMS" where dozens of fine, unfused, lightly-insulated wires are run across the tops of cells, and these harmless-looking but lethal wires are even taken outside the battery box in some cases.
Some large cells have multiple terminals per electrode.
Would it be worth monitoring all straps.
Could diodes be used as an OR gate to pick up a high potential on extended and extra snap-off bolt/strap detectors.
Problem will be more detectable as a high temp at the bolt.
Just brain-storming issues and ideas for other cell formats.
Interesting thoughts, but diode-ORing will not work in that way, as it stands, and anything more complex would increase the cost.
Perhaps you may invest in building a 4 cell large Ah 12V battery to test out the design with extended bolt/strap tabs.

You could use it in the MEXY as Auxiliary battery and have it linked in somehow via Fiba cables to the main BMS network, with Alarm monitoring.
...
This sounds like a dedicated smart HV-DC/DC-12V with 4 cell monitoring would be a better approach.
We have two 12 V batteries made from the same 40 Ah CALB cells that are in MeXy's traction pack. We haven't put one into MeXy yet. We could have the BMS simply switch a relay on and off between the dumb DC/DCs and the 12 V battery.
Also would the CMU's cope with higher cell voltage chemistries?
Say 4.5V max. Winston have the LTHPxxx Series with higher cell voltage.
Or lower types like the Lithium/Sulphur? (still on the WINSTON's partnered site Generplus Italy - Winston LSP )
Yes. We designed them to work with both LiFePO4 and LiPo cells (4.2 V end of charge), and anything in between. And we recently determined that the same PCBs could be used on Lead-Acid cells by changing some components and replacing the blocking diode with a link. They will not work below 1.53 V or above 5.0 V.
Did you have ideas for a EV that needed a charger interface.
Do I remember correctly someone mentioning that CMU #1 could monitor system stress and control a charger like the ELCON?
Or would this need a separate brain PCB?
You remembered approximately right. We now have code in the CMUs that will control an Elcon/TC Charger (CAN bus model), not using CAN bus, but using the 2400 baud serial interface on the charger that their CAN module plugs into. However it is not CMU #1, but CMU #N (the last one) that does the job.
See viewtopic.php?title=weber-and-coulombs- ... 0&start=62
On the FUSE defusing. What would cause this?
This is the flip-side of cells or terminal connection going high-R under load, namely going high-R during charging or regen. In that case the CMU spanning the high-R will see the full difference between the charge source o/c voltage and the battery o/c voltage, in the forward direction. The diode blocks up to 900 Vdc in reverse. The fuse can break max 50 A at 125 Vdc in the forward direction.
Should the Fuse be before the diode?
Makes no difference that I can see, except that the diode gets better heatsinking the way it is now.
Would a 1A "poly-switch" that self-resets be an option?
We looked into that way back, and came to the same conclusion as you.
A 1A fuse will be at its limit with 3.3 ohm by-pass at 4.2V-0.8V.
But your calcs have it at 800mA.
Oops! Thanks for catching that. Bumped to 1.25 A now.
Would it be worth having Vdd on the Jtag Header P5?
For testing for fuse blow? Not a bad idea. But you can check for Vdd (with occasional pulsations) on either the Blue LED anode (Act_R net) or the Fibre LED Anode terminal (FLA net).
Do you ever contemplate if gnd and its symbol should be replaced with a triangle and COM net label.
I still find it miss-leading using GND for things that are never meant to be grounded to earth with a spike in the ground.
Is there an international standard for this?
Perhaps "wonday" will have Moon and Mars etc ground symbols too.
I should stop wondering off topic.
I'm just used to it I guess. Actual earth earths are usually marked P.E. (Protective Earth) and have a kind of pitchfork symbol. A down pointing triangle is typically a supply that is negative wrt "GND".
Blue LED [issues] ...
Coulomb and I have thought about some of these issues before, and I don't think the cost of any extra parts to protect the LED are warranted. It isn't near C1, but it can take 5 V reverse and 150 mA for 100 us forward (30 mA continuous). And I don't think the blue LED can cause any problems for anything else. The 1000 V blocking diode stops the cell from actively pulling Vdd down. We already have the blue LED connected to Vdd in the CMUs that are in MeXy. There has been no problem. The only difference is that it no longer has its own dedicated processor pin.
Looks good guys.
You've stretched me far for "constructive criticism".

7C Image

Thanks 7TauRadians. Image

Below is a table I made, of LiFePO4 cell dimensions for CMU purposes, based on datasheets here:
http://www.lithiumstorage.com/2012-08-2 ... ata-sheets
http://www.thunderstruck-ev.com/Manuals ... Manual.pdf
http://en.calb.cn/
http://evworks.com.au/index.php?category=5

But I find that different datasheets for the same product, presumably published at different times, often give dimensions that vary by 1 mm or more from each other. So use the following at your own risk.

SP = Sinopoly (black/cyan)
CA = CALB (grey)
SE = Sky Energy (blue)
TS = ThunderSky (yellow)
WB = Winston Battery (yellow)
70 Ah or less use M6 bolts.
Greater than 70 Ah up to 200 Ah use M8 bolts.
     \  Terminal spacing
Cell  \  60   64 mm         81   82   83 mm       102   106 mm    120 mm
thickn \----------------------------------------------------------------
 36 mm |                    SP60b            |
 41 mm |      CA60                           |
 46 mm | TS40 SE40                           |
 46 mm | WB40 CA40                           |
 47 mm |      SP40                           |
 50 mm |           (M6)     CA70             |
 56 mm |              ,----------------------'   CA130?
 57 mm |              | (M8)                            SE130
 60.5  | SE70         |
 61 mm | WB60         |          WB90
 61 mm | TS60         |     SP90     TS90
 61 mm | SE60 SP60a   |     SP100
 62 mm |--------------'                                 TS100
 62 mm |                                                WB100
 65 mm |                                         TS160            WB160?
 65 mm |                                         WB160?
 67 mm |                    SE100
 67 mm |                    CA100    WB100a
 68 mm |                             TS90old
 71 mm |                                                TS160old
 71 mm |                                                SE180
 71 mm |                                                CA180
 71 mm |                                                SP180
 71 mm |                                                SP200
 93 mm |                                                CA200?
Last edited by weber on Tue, 14 Oct 2014, 03:43, edited 1 time in total.
One of the fathers of MeXy the electric MX-5, along with Coulomb and Newton (Jeff Owen).
User avatar
weber
Site Admin
Posts: 2936
Joined: Fri, 23 Jan 2009, 17:27
Real Name: Dave Keenan
Location: Brisbane
Contact:

Weber & Coulomb's LyteFyba BMS

Post by weber »

Here are the top and bottom solder masks, so you can see the rectangular pads on the snap-off terminal wings, for joining them with wires, for wide or non-standard terminal spacings such as these HiPower and ZycTech cells.
http://www.electriccarinternational.com ... teries.php
http://www.zyctech.com/product/showprod ... g=en&id=10

In these cases you should snap off and discard the negative wing extension and should bolt the CMU to the negative terminal, and snap off and extend the positive terminal wing with wires. That's because cell temperature-sensing is on the negative wing, via a thermal pad on the microcontroller.

You might think that these CMUs can't be used on cells thinner than 41 mm (the north-south dimension of the main part of the PCB), but because they can be staggered east-west they can be used on cells as thin as 27.5 mm.

But to use them with pouch cells or cylindrical cells would require a different PCB layout. It's possible that a single long thin PCB, say 120 mm x 10 mm (no wider than an IFO connector) could both match the thin-ness of many pouch cells and fit in the triangular space between many cylindrical cells. There's a job for someone else. Collect the dimensions of the most popular pouches and cylindricals, install the free DesignSpark PCB software, get our PCB files from GitHub and go to it. It's all open source. http://dkeenan7.github.io/LyteFyba/

Image

Image
Last edited by weber on Fri, 31 Jul 2015, 21:18, edited 1 time in total.
One of the fathers of MeXy the electric MX-5, along with Coulomb and Newton (Jeff Owen).
User avatar
weber
Site Admin
Posts: 2936
Joined: Fri, 23 Jan 2009, 17:27
Real Name: Dave Keenan
Location: Brisbane
Contact:

Weber & Coulomb's LyteFyba BMS

Post by weber »

Here's how the terminal hole spacing works. You can compare it to that table of common spacings above.

Image
One of the fathers of MeXy the electric MX-5, along with Coulomb and Newton (Jeff Owen).
User avatar
weber
Site Admin
Posts: 2936
Joined: Fri, 23 Jan 2009, 17:27
Real Name: Dave Keenan
Location: Brisbane
Contact:

Weber & Coulomb's LyteFyba BMS

Post by weber »

Thanks for your help guys. Thanks Coulomb, for reminding me about strain-relief holes for soldered wires going to the RX and TX pads.

I just sent the files (CMU rev 61) off to PCBcart. 10 boards (80 CMUs) for AU $415.00 including postage, so $5.20 per CMU. This included a one-off setup fee of $153, so the next 10 will only be $262 or $3.30 per CMU.

I paid a little extra for ENIG (gold flashed) instead of HASL (tinned), and yellow silkscreen on black soldermask just for fun. If they don't find any problems with the gerbers and drill files I should have them in about 2 weeks.

Now to order the parts.
One of the fathers of MeXy the electric MX-5, along with Coulomb and Newton (Jeff Owen).
User avatar
PlanB
Senior Member
Posts: 471
Joined: Sat, 16 Jan 2010, 15:24
Real Name: Kris McLean
Location: Freemans reach
Contact:

Weber & Coulomb's LyteFyba BMS

Post by PlanB »

Is there a single 'go to' document that lists the salient features of this pack management system? I know there is a thread that details development but it's hard to sort through looking for functionality (like cell temperature monitoring).
I've been looking at scalable BMS from the likes of lithiumbalance.com but that web of wires snaking across the pack make me nervous.
User avatar
weber
Site Admin
Posts: 2936
Joined: Fri, 23 Jan 2009, 17:27
Real Name: Dave Keenan
Location: Brisbane
Contact:

Weber & Coulomb's LyteFyba BMS

Post by weber »

PlanB wrote: Is there a single 'go to' document that lists the salient features of this pack management system? I know there is a thread that details development but it's hard to sort through looking for functionality (like cell temperature monitoring).
I've been looking at scalable BMS from the likes of lithiumbalance.com but that web of wires snaking across the pack make me nervous.

Hi PlanB. Please see http://dkeenan7.github.io/LyteFyba/ and feel free to suggest other information we should include in such a document.

I suppose a proper user manual showing how to wire them up and listing the available commands and operating modes would be a good idea some day soon. We have a quick reference guide to the available commands, which we keep updated in the GitHub repository as "LyteFyba/trunk/common/definitions.txt". But at present, it's only designed for reminding the people who wrote the commands. Image
Last edited by weber on Fri, 31 Jul 2015, 21:23, edited 1 time in total.
One of the fathers of MeXy the electric MX-5, along with Coulomb and Newton (Jeff Owen).
User avatar
PlanB
Senior Member
Posts: 471
Joined: Sat, 16 Jan 2010, 15:24
Real Name: Kris McLean
Location: Freemans reach
Contact:

Weber & Coulomb's LyteFyba BMS

Post by PlanB »

Thnx Dave. The old celltop modules from EV works have some of the circuitry potted, do you feel the fine pcb track spacing of your boards will be OK without conformal coating?

In terms of the document I'd be really interested in some detail on how a semi retired old guy might access stuff like your fuel gauge & stress monitor with a Rpi
User avatar
weber
Site Admin
Posts: 2936
Joined: Fri, 23 Jan 2009, 17:27
Real Name: Dave Keenan
Location: Brisbane
Contact:

Weber & Coulomb's LyteFyba BMS

Post by weber »

PlanB wrote: Thnx Dave. The old celltop modules from EV works have some of the circuitry potted, do you feel the fine pcb track spacing of your boards will be OK without conformal coating?
No. We conformally coat them. We used Electrolube HPA on the ones in MeXy.
In terms of the document I'd be really interested in some detail on how a semi retired old guy might access stuff like your fuel gauge & stress monitor with a Rpi

All Raspberry Pi configurations have a couple of pins which provide a UART function. You would configure these for 9600 baud and would adapt them to the IFO connectors with a series resistor to limit the LED current to 20 mA and a pullup (or pulldown depending on the pin's polarity) to limit phototransistor current to 5 mA.

In software you would first separate received bytes with the high bit set (status bytes sent automatically 15 times a second) from the others (command responses). The stress level is in the low 4 bits of the status byte. The RPi can display this and use it to back off the current of either the charger or the motor controller if any cell is stressed beyond level 7.

To obtain a fuel gauge reading you would need an IMU in the optic fibre comms chain with your CMUs. The IMU (which has comms ID = 0) sits on top of your current shunt and samples the current 15 times a second. The RPi would send a 'g' (for gauge) character every 30 seconds or so, and would receive a response like '\000:g 153', meaning 15.3% depth of discharge, which it would use to update its display. Both the command and the response must be followed by a simple XOR checksum and a carriage return, unless you turn off checksumming with the 'kk' command.

If you used a Tritium EVDC instead of an RPi as your EVCU, the hardware mods and C++ software for this are already in the LyteFyba GitHub repository. But we'd be happy to help you with the RPi provided you agreed that we could include its hardware adapters and software in the open source repository as well.
Last edited by weber on Fri, 31 Jul 2015, 21:24, edited 1 time in total.
One of the fathers of MeXy the electric MX-5, along with Coulomb and Newton (Jeff Owen).
User avatar
PlanB
Senior Member
Posts: 471
Joined: Sat, 16 Jan 2010, 15:24
Real Name: Kris McLean
Location: Freemans reach
Contact:

Weber & Coulomb's LyteFyba BMS

Post by PlanB »

Already using the uart for comms with another RS232 gizmo, might be able to move that to a USB port to free up uart for LiteFyba? How is cell 'stress' determined?
User avatar
weber
Site Admin
Posts: 2936
Joined: Fri, 23 Jan 2009, 17:27
Real Name: Dave Keenan
Location: Brisbane
Contact:

Weber & Coulomb's LyteFyba BMS

Post by weber »

PlanB wrote: Already using the uart for comms with another RS232 gizmo, might be able to move that to a USB port to free up uart for LiteFyba?
Or add a USB-to-RS232 or RS422 or RS485 device.
How is cell 'stress' determined?

Each CMU calculates its local stress based on 5 different stressors: Undervoltage, Overvoltage, Undertemp, Overtemp, Bolt voltage. For each of these we've defined a zero-stress point and a unit-stress step size. It uses the maximum of these 5 stresses as its local stress. Then it transmits the maximum of its local stress and the stress it receives from the previous CMU in the chain, unless it is CMU #1 in which case it just sends its local stress.

Any CMU that hasn't heard from another CMU for some time will take over the job of CMU #1 and send its local stress, while setting a comms error bit in the status byte, and it will send a comms error message containing its ID every 17 seconds.
One of the fathers of MeXy the electric MX-5, along with Coulomb and Newton (Jeff Owen).
User avatar
weber
Site Admin
Posts: 2936
Joined: Fri, 23 Jan 2009, 17:27
Real Name: Dave Keenan
Location: Brisbane
Contact:

Weber & Coulomb's LyteFyba BMS

Post by weber »

Most of the parts arrived last week, and the PCBs just arrived. They appear perfect as usual from PCBCart.

We had some slight concerns about our gerber files because we had to go to a lot of trouble with the cell terminal pads and their holes, to convince DesignSpark to allow us to have different nets on opposite sides of the PCB, despite the fact that the holes were not to be plated through. And the circular holes _had_ to be drill holes, not routed, because there is a massive jump in price if you have more than 30 routed slots per PCB, and we already had 14 between the squiggle-joins and 16 for the outer (oval) cell terminal holes. Although I wouldn't be surprised if they actually rout such large holes anyway.

The colour scheme is a little disappointing. I was going for a bumblebee aesthetic,

Image

but their "black" is more of a dark green-grey and their yellow leans towards the green as well (possibly due to its semi-transparency), so it comes off as more of a military aesthetic. First world problems.

Image

Image

Image
One of the fathers of MeXy the electric MX-5, along with Coulomb and Newton (Jeff Owen).
User avatar
PlanB
Senior Member
Posts: 471
Joined: Sat, 16 Jan 2010, 15:24
Real Name: Kris McLean
Location: Freemans reach
Contact:

Weber & Coulomb's LyteFyba BMS

Post by PlanB »

Isn't it nice when the holes line up with the thing they are meant to engage with? I always like to sleep on a file before I send it off, its funny how the sleeping brain wakes up with the realisation of a stuff up the awake brain couldn't see 11pm the night before.
As for the military aesthetic, they would need to cost a hundred times as much as well to really achieve that I guess? Nice job anyway.
User avatar
coulomb
Site Admin
Posts: 6357
Joined: Thu, 22 Jan 2009, 20:32
Real Name: Mike Van Emmerik
Location: Brisbane
Contact:

Weber & Coulomb's LyteFyba BMS

Post by coulomb »

We started building the first Lite Fyba CMUs today. Here is a set of 8 CMU PCBs resting in our reflow soldering station:

Image

What's that you say? It looks like a frying pan? I don't know what you're talking about Image

What you can't see unless you are really very observant, is that the curvature of the bottom of the pan, err, reflow station, keeps it about 8mm off the bottom of the pan, which is some 5 mm above our 3 mm aluminium heat spreader.

[ Edit: actually, you can notice the gap by the distance of the shadows (reflections?) of the bolt holes. ]

Oh well, these won't be going on to 40 Ah cells, so they'll all have to be separated eventually, so we just cut the strip of 8 CMUs into two strips of 4. It would be ambitious to get 4 made the first day in any case.

We found the old truncated giant funnel, which is ideal for funnelling parts removed from cut tape into our plastic storage bins.

I quickly found that the solder paste, now several years old, is really well past its use by date, and in any event it was running low in actual quantity. Weber has been working hard on the forum database translation, and didn't get time to order new solder paste. I wondered whether Jaycar actually sell solder paste, and whether it would be available from a G20 accessible store. (Last Thursday there was a tank running down Elizabeth Street, squads of police bikes, and more; it's pretty full on in Brisbane this weekend.) It turns out that Weber was able to purchase the last tube they had in the Aspley store. The web site indicates that it has a shelf life of about 6 months if stored in a fridge. Indeed, the sales assistant had to fetch it from upstairs; they really do store it in a fridge.

So we were able to get the first set of 4 Rev 61 (!) CMUs assembled and tested today.

Here they are on the test board:

Image

The miniature beepers are a bit of a disappointment; they're just not all that loud. The software needs modification to handle the way we've juggled the LEDs and I/O ports to be able to read three bolt voltages. But otherwise they are working fine.

[ Edit: I meant to point out that the only time we can use 8 CMUs joined together would be for 40 Ah cells, and for these we would be snapping off the extended "ears" as well as the snap-off extra bypass area, in which case the strip of 8 CMUs would fit our reflow station as before. So the larger size really isn't a big problem. ]
Last edited by coulomb on Sun, 16 Nov 2014, 17:10, edited 1 time in total.
MG ZS EV 2021 April 2021. Nissan Leaf 2012 with new battery May 2019.
5650 W solar, 2xPIP-4048MS inverters, 16 kWh battery.
Patching PIP-4048/5048 inverter-chargers.
If you appreciate my work, you can buy me a coffee.
User avatar
weber
Site Admin
Posts: 2936
Joined: Fri, 23 Jan 2009, 17:27
Real Name: Dave Keenan
Location: Brisbane
Contact:

Weber & Coulomb's LyteFyba BMS

Post by weber »

coulomb wrote:The software needs modification to handle the way we've juggled the LEDs and I/O ports to be able to read three bolt voltages. But otherwise they are working fine.

We will read three voltages, yes, but only two of them are read off bolts (called BoltV+ and BoltV-), the other (CellV) is read off the positive link, and all are measured with respect to the negative link.
One of the fathers of MeXy the electric MX-5, along with Coulomb and Newton (Jeff Owen).
User avatar
coulomb
Site Admin
Posts: 6357
Joined: Thu, 22 Jan 2009, 20:32
Real Name: Mike Van Emmerik
Location: Brisbane
Contact:

Weber & Coulomb's LyteFyba BMS

Post by coulomb »

Weber kindly took a day off his Solar electric job to help me cook up a batch of 8 Lite Fyba Cell Management Units for my 24 V (lighting) system. (I'll need a batch of 16 for the main 48 V system later).

Here they are in mid-sized configuration (one snap-off ear removed):

Image   Image

What happened to be on the screen when I took the photo was a p (pain) command, and a "tick v greater" command which shows the highest cell voltage (in millivolts, here cell 3 at 3.471 V).

Image   Image

In the last photo you can see one of the IFO (Industrial Fibre Optic) connectors to talk to the string of CMUs. In the main image, you can see the white Novus RS-485 box with its small PCB that adapts to fibre connectors. This is the same unit we used to talk to the CMUs in the MX-5.

There was only one minor issue; I don't have straps on this battery, just a bunch of yellow ring connectors. These aren't as thick as a strap that would be needed in an EV, so the CMUs would have been much closer to the top of the cells than usual, and some of the larger components under the boards (the fuse and diode) would have pressed against the cell vents. To prevent this, I put the spring washer under the CMU, instead of the usual position just under the bolt head. I hope that this doesn't lead to issues in years to come.

Last edited by coulomb on Wed, 24 Dec 2014, 12:33, edited 1 time in total.
MG ZS EV 2021 April 2021. Nissan Leaf 2012 with new battery May 2019.
5650 W solar, 2xPIP-4048MS inverters, 16 kWh battery.
Patching PIP-4048/5048 inverter-chargers.
If you appreciate my work, you can buy me a coffee.
tonyw
Groupie
Posts: 56
Joined: Thu, 26 Jun 2008, 01:08
Real Name: Tony Wyatt
Location: Sydney

Weber & Coulomb's LyteFyba BMS

Post by tonyw »

Hi guys, been silent for a while, just getting interested again.
Good to see the progress with technology. I love the new designs for CMUs and BMSs.

I just have one niggle - I seem to recall that in Oz, at least, we are bound by AS3000 for all wiring. As I read it, that means that it is contrary to the regulations to have the tightness of the cell connections dependent on the non-metallic PCB material.

I quote from AS3000, para 3.11:

3.11.1 General. The connection of all conductors at terminals shall be made by clamping the conductors between metal surfaces in such a manner as will not allow the connection to slacken off under normal conditions of use. Clamping shall not depend for its effect on pressure on non-metallic materials or on metals of insufficient hardness for the purpose, which may flow or deform under the clamping pressure...

Now you can read that to mean only that the conductor must not come away if the clamp becomes loose, therefore we are OK using a PCB as part of the clamp. However, an inspector might take it to mean that the connection must not have any non-metallic materials within the clamping area.

Fortunately, the remedy is simple, although it will increase the height somewhat: replace the bolt by a stud and assemble in this order, from the bottom:
Cell terminal;
Strap to next cell/cable lug;
Spring washer;
Half-height nut, tightened to hold the conductor firmly;
PCB;
Washer;
Spring washer + nut or (better) Nyloc nut.

Now the clamping of the conductor (and consequently the contact resistance) is not dependent on the PCB maintaining its original thickness. The temperature measurement is unaffected and it is no longer necessary to measure voltage drop across the sides of the PCB.

cheers

tony

User avatar
weber
Site Admin
Posts: 2936
Joined: Fri, 23 Jan 2009, 17:27
Real Name: Dave Keenan
Location: Brisbane
Contact:

Weber & Coulomb's LyteFyba BMS

Post by weber »

Thanks tonyw. That must be an old version of AS3000. But your point stands because I eventually found a very similar paragraph, which I wasn't aware of, in AS/NZS 3000:2007 (and unaffected by the two more recent amendments).
3.7.2.4 Mechanical connection devices
...
Such devices shall--
...
(b) not be dependent upon compression of insulating material for an effective electrical connection; and
...
And yet, this is done all the time inside power electronics equipment such as inverters. But note that they always use belleville washers (on top of flat washers) to maintain loading for as long as possible, given creep in the glass-fibre-reinforced-epoxy printed circuit board. And so we recommend bellevilles with our CMUs, and have used them in the MX-5.

I guess inverters escape by not being electrical installations but electrical equipment, and hence not being within the scope of AS3000. Can we claim that a battery made up of multiple cells is "electrical equipment". Even if we do, what really matters is: Is it safe?

Well, that's why we added the bolt volt-drop measurement capability (which as yet has no software and so is untested).

The method you suggest, with a stud and two nuts at every connection, would certainly meet the requirements of AS3000, however there are other reasons for bolt loosening besides PCB creep (which is slow compared to typical (non-reinforced) insulator polymers such as PVC). In an EV the primary cause of loosening is typically vibration, combined with inadequate mechanical clamping-together of the cells. By putting the CMU outside of the bolted joint, you forgo the ability to detect loosening from those other causes. So it's a bit of a Catch 22.
One of the fathers of MeXy the electric MX-5, along with Coulomb and Newton (Jeff Owen).
T1 Terry
Senior Member
Posts: 1952
Joined: Thu, 30 Sep 2010, 20:11
Real Name: Terry Covill
Location: Mannum SA

Weber & Coulomb's LyteFyba BMS

Post by T1 Terry »

The other method would be to use electrical terminals and cable so the board could then be remote mounted, thus also moving the heat generation problem away from the cell.

T1 Terry
Green but want to learn
Post Reply