[Coulomb]

So i'm thinking that there would be one master and 3 slave units to make up the 6kW unit. And if you want to split them, then one of the slaves must be configured into another master.

Yes, agreed. If you keep with 24 V chargers, you'll need two separate chargers for each half of the pack. One half could be at a different voltage than the others, so I don't think it would work all that well having only one master.

You're likely not going to be able to copy the firmware from the master to one of the slaves.

You say you're in Brisbane, but I assume that's not Brisbane Australia (there are a couple of Brisbanes in the USA, I believe).

They look like the older boards for which we have schematics, http://www.diyelectriccar.com/forums/showthread.php?t=89470,

Yes, that's confusing, if you bought them recently. Maybe Elcon don't sell many 6 kW models? The three white opto couplers strongly suggests a pre-2014 manufacture.

But our resident expert on these chargers is "Coulomb" from Brisbane, and he has more experience with those big monster chargers.

I have only a little experience with larger chargers; mainly a dual unit (2 x 2.5 kW).

If you wired up the slave outputs backwards, i.e. with reversed polarity, then when the output relays were commanded closed by the master, the slaves would be drawing power from the cells thru the inductors and the final stage rectifier diodes.

I would think that this would have blown the DC fuse. It's rather confusing as to what has happened. Are you running the charger from 240 VAC or 120 VAC? Elcon seem to have firmware that runs the chargers hard at 120 VAC. It could be that the wiring isn't heavy enough for 6 kW at 120 VAC (that would be over 50 A total, 12.5 A through each charger unit. But they are supposed to be derated when run at 120 VAC.

The AC wires to the slave getting hot seems to be an overcurrent condition, such as the diode bridge is shorted, but there is a fuse on the mains that should blow.

One thing i noticed is no green ground wire running to the 4th unit to the far right in the photo? For that matter, why are the grounds not connected to the input AC Ground wire and why is that wire so tiny--it should be the same size as the input at least?

Is the photo the "before" picture of how it was rec'd from the vendor, or is it showing wiring changes that you made?

How/where are you getting your 240vac--is it a generator on board, or do you connect a land line to the mains?

Nice looking boat, funny blog: could you not use the B-word for one hour...[/QUOTE]

 

[Coulomb]

They look like the older boards for which we have schematics, http://www.diyelectriccar.com/forums/showthread.php?t=89470

On closer inspection, this appears to be intermediate between the pre-2014 models and what I've been calling the 2014 model. The daughter board seems like pre-2014: the 3 white optos above each other, the white 5-pin programming port at the top, and the power supply is not on the daughter board.

But the main board looks 2014-ish: there is the gap between the AC input relay and the heat-sink with the bridge rectifier, including the row of 6 evenly spaced vias along one side of this component free area. The output connectors are stand alone, unlike the 3-terminal block on the pre-2014 models. The power supply transformer is larger and more square, as per the 2014 models, but it's located on the main board, like the pre-2014 models.

So perhaps this is an "early 2014" layout, and what I was calling 2014 or 2014+ should be called "late 2014+". Unfortunately, I can't find a date on either the main or daughter boards on mine, but chips have a 2014 date code. There are version strings however: "HQ-E-1K5W-BL" and "VER 5.3" on the main board; "HQ-E-1K5W/2KW-CAN-CON" and "VER 3.52" on the daughter board. Bj_porter's charger seems to have "TCCH-1K5W-BL" (not sure of the "B" in "BL") and "VER xx" (xx isn't readable, but could be 5.2) on the main board.

 

[Coulomb]

From what Elcon is telling me, this is not the case. As long as I get the communications wiring right (which is hard to screw up) they can stay as one charger unit and no second master is needed.

Well, that's as a single 24 V charger in two separated parts.

[ Edit: and that's exactly what you want. The rest of the post is based on an incorrect assumption. ]

What I'm saying is that surely you would want two separate and independent 24 V chargers, so that the BMS can control each half of the pack separately. But on further thought, all you really want is for the charger to run full power until one ore more of the cells requires the power to be backed off, which is really like one charger anyway.

Plus, it's finally penetrated my thick skull that this is a CAN model, so there aren't any special "curves" built into the firmware, so assuming that the white programming connector on the daughter boards implies the old 8051 compatible microcontroller, it might still be possible to make one of the CAN slaves into a CAN master. There might have to be a slight re-wiring so that responses from the new CAN master come back to the BMS (from poor memory, the CAN slaves don't send anything on the CAN bus, they just do as they are told and don't talk back). But that has never been attempted, so I think we'd want to leave that as a last resort.

So to summarise: I think that you are right with what you are trying to achieve, but must have made some sort of error to see the overheating issues you are getting. Google maps says you're about 50 km (31 miles) from me, so a visit isn't impossible.

 

[Coulomb]

A final thought.

[ Edit: This thought is based on a wrong assumption; see next post. ]

If the daughter boards are not merely very similar but in fact identical to the older daughter boards (e.g. "VER 1.7" daughter boards), like the image attached, then the serial signals that are derived from the CAN signals are not isolated from the battery terminals.

The image attached has had the black gunk removed from all the components, so it may look dramatically different in real life. [ Edit: the late 2014+ models gave up on the black gunk, so yours might not have the gunk either. ] If present, try to imagine the black gunk as being transparent to compare with the attached photo. Ignore the coloured rectangles except for the light brown one in the bottom right corner. It shows R20 and R21 (with a capacitor between them). I think if you had an incident due to non-isolation of the serial comms lines, then I think that one or both of R20, R21 will have had half the pack voltage across them. Call it 25 V, the resistors are 100 Ω, so that would be 250 mA coming out of the battery. The resistors would fry pretty well instantly, because that's 25 * 250 / 1000 = 6.25 W in a resistor that can dissipate 0.125 W (a 50x overload). Not at all a "massive load coming out of the battery", but perhaps they went low resistance (that can happen with overload sometimes). If this happened, you would expect to see some sort of disturbance to the black coating under these resistors (like the black coating completely being blown away, but it might be more subtle). However, this would not have had to wait for the output relay to connect, or even for the chargers to be turned on; it would happen as soon as the battery was connected to the charger output (assuming that the two half-packs were still joined at that point). I think that damage to R20/R21 also implies damage to U8, the shunt amplifier, and possibly a few other components and maybe even PCB tracks.

But hopefully, when you look at the version number a little north west of the middle of the daughter boards, they won't say "VER 1.7", and there will be no damage to R20/R21.

 

[Coulomb]

A post-final thought

The bank is 32 x 180Ah CALB cells in a 24V arrangement (4P8S)...but that's less important. ... What I got was one massive 6000W charger 220V in, 24V out, with CAN.

Sigh. When I read 6000 W and 32 cells, I immediately assumed 48 V. The inductors, which change appearance with voltage, look exactly like the 48 V charger I have here at the moment. Even when you said 8S, I was still thinking 24 V. Even when you said you received a 24 V charger, it still didn't click. I assumed the voltage mismatch was just part of the mix-up with ordering, and that you'd be wiring 3000 W of charger on one half of the pack, and the other 3000 W to the other half (splitting that in two so that you could charge at a lower rate than 6000 W total). Sorry for the bonehead confusion!

So all that stuff about the outputs not being isolated... it is now irrelevant, and can't be the cause of your problem.

It's even more mysterious what the actual problem could be though.

 

[Coulomb]

We opened a 3kW unit and found that they don't run the mains ground wire to the slaves, so that may not be related to this issue, although it is a poor practice.

I think that they assume that the case will conduct the ground to the other PCB. In this case, with separate halves of the charger, I'd use a green/yellow or pure green wire to the other chassis.

Is it a poor practice? It's arguable. Certainly a cheap one.

 

[Coulomb]

I can't see the harm in running the [ground] wire though. Wires...I have lots of.

Yes, I think it's a very good idea to run the earth wire. Then the slave-only half is safe even when not bolted to the hull. The cheap practice I mentioned was for the manufacturer not to do this.

 

[Coulomb]

Thanks BJ, for the kind words and your hospitality. It was a quite interesting day for me.

On the ferrites designed to cut out noise... of course that' noise in the SSB radio that is coupled via electromagnetic waves, so electrical noise, not acoustic noise (granted, resulting in acoustic noise from the speakers). So deliciously ironic.

On the technical front, I post my quick and dirty picture of the daughter board. Rev 1.8, I actually can't pick any differences from the 1.7 boards. Perhaps if I had two boards side by side I could.

I've attached two images, identical apart from the size (sm for small and med for medium).

Here is my slightly whacko theory. They the trialled the 2014 model with the power supply on the daughter board, ARM processor, parts on both sides of the daughter board for a while, and hated it. Perhaps it was a nightmare to troubleshoot in the factory, or they had a few for repair and didn't like it, or maybe the original developer came back from extended leave and said "what have you done you fools!". Who knows. The attendant main board has the power supply back, but with a non-Viper chip, and the larger transformer. I don't know if they get the extra isolated output to the daughter board. I think the new transformer has 8 pins, one primary and three isolated secondary, compared to the original with 6 active pins, one primary and two isolated secondaries. I forgot to check if there were extra pins after pin 32 (the 2014 model has 42 pins from main to daughter board).

Ah, maybe these are just different models, the HQ (2014+ model) and ones like these. Or maybe Elcon decided they didn't want to strand their customers that need the 7-pin round connector, and insisted that they continue manufacturing essentially the old models, with just a few modernizations. Elcon doesn't seem to stock the HQ models. TC Charger continues to show only the HQ models in 700 W and 2 kW, and the old TCCH models for 3 kW and larger.

 

[Coulomb]

It appears that it's all connected properly.

Yes, it all seemed OK, with the minor provisos BJ mentioned.

One odd thing i noticed is that there are no output fuses on the Positive Outputs on any
of the boards--usually there is a fuse soldered onto the boards right next to the B+ terminal...?

I think they've given up on those fuses... too expensive to provide DC rated fuses, and they never seem to blow anyway. They provided blade fuses (eww, but OK for 24 V and maybe 48 V), but none of the four modules there had a visible DC fuse. The 2014 HQ model here is the same.

Which AC wire was getting hot? None of the Red AC terminal lugs are touching the housing right?

Nothing like that. I could still smell the melted plastic a day later, and there was plenty of gunk on the AC capacitor across the mains. It was the mains input on the slave in the same half as the master. BJ says the green-yellow wire melted the worst. That's really odd because I can't imagine a scenario where there would be a lot of current in the green wire and the Ground Fault breaker not tripping immediately. It's possible that the original owner of the boat didn't wire the GF breaker properly. But he says that breaker does trip now and then, as you might expect on a boat.

Can you hold back the white plastic on the master upper right hand corner and get a photo to show
where the small red and black serial port wires are connected--is there a separate little circuit board
such as seen on the slaves or is there a header connector?

I think BJ just wants to put it all back together and finish it now. Yes, I did notice two small PCBs with the opto and 5.1 kΩ resistor on it on the slave unit. The boards went to the pins 2 and 6 of the 7-pin header (DGND and comms input). So yes, those wires are not CAN bus, but serial signals (not RS232, but optos with 12 V behind them).

Do you have a CAN dongle from Elcon thru which it routes the CAN messages to the charger
and does it connect to the 7-pin round DIN connector with the blue cap? i assume that is the
case but wanted to check.

I had the same thoughts, but yes, he goes through a standard CAN box that Elcon finally sent BJ after asking questions for days. I suppose that makes a little sense, since some BMS units can talk directly to the serial signals and don't need the CAN box.

Does the small red serial line from the master route to the green wires (pin 6) going to the slaves?

Yes, but via those small PCBs with the opto and resistor. If I understand your question.

In the Slave unit one board uses a yellow wire for the ground pin 2, the other looks like it
uses a black wire.

I noticed that too. Odd, but true. Some of the wires paralleling the charger outputs use black for negative and black for positive.

Those serial port grounds are also connected in common with the battery negative terminals,
so polarity is important.

Yes, it is. But it seems that the "CAN bus" signals (2400 bps serial) were never the problem. I'm not totally convinced that current ever went out of the battery into the charger; maybe BJ didn't notice the sign change (there was a constant load of 5-6 A from the batteries, so if charging, the current would have ramped down to zero then ramped up in the opposite direction).

The excess mains current was all too evident, however. The little sparkles were apparently from carbonized mains plastic settling into the power supply switching circuitry. All in all, I think BJ dodged a few bullets there.

 

[Coulomb]

That bare control board without the black coating sure would have made our past efforts easier--but if ever there were boards that needed coating, or should have had the coating, it would surely be on an ocean-going BOAT.

Yes, nice. But they still have the conformal coating.Interesting that it didn't help much with the tiny sparks and the wire residue.

Edit: the 2014 HQ models are also without the black gunk.

 

[Coulomb]

Continuity check with the cover on and the charger disconnected from everything shows a direct connection from the Positive output to the case, and to the AC ground wire.

Am I right in thinking when this thing connects to the battery the whole thing is maybe closing a completely inapproproprayte 24V circuit?

This could explain ...

Yes! it could explain a lot of things. Like how the earth lead could overheat, yet not trip the RCD. Answer: because it's carrying DC, from the battery. There must be some resistance in the circuit somewhere, or the fault current would be well over 1000 A. Perhaps that's why my brain resisted that thought; battery faults are usually more shower of sparks than silent overheating.

Edit: I'm also used to solar energy circuits or EVs, which usually have fully floating batteries. But in a boat, you might ground the negative side of the battery. Is that the case? I'm guessing it must be, to explain this problem.

 

[Coulomb]

Grounding can be a little weird in boats, since we're not actually on the physical ground. It's not like you can drive a big spike into the literal ground like if you're grounding a big HF antenna.

Yes, I chose the wrong word. I mean do you use the hull to conduct 24 V to lower current devices like lights?

Or more directly, do you deliberately connect battery negative (or positive or whatever) to the hull?

 

[Coulomb]

I'm frankly amazed that I didn't destroy any of this hardware.

Agreed! That's at least 3 bullets that you've dodged, BJ!

These are CALB 180 cells, and can charge at C, so there shouldn't be an issue with charge acceptance.

That's not the issue. The issue is getting thick enough cables so when the charger is putting out 90+ amps, the voltage measurement at the charger end isn't such that it appears that the batteries are full.

It's putting out about 90A set at 100A.

90% of requested amps isn't too bad considering the high currents and low voltage.

BTW, on one of your screenshots, I saw that the target voltage was 29.6 V. That's 3.7 VPC. CALB recommends a maximum of 3.60 VPC, or 28.8 VPC. Other manufacturers say you can go to perhaps 4.0 VPC, but there isn't much capacity above 3.5 VPC, so I'd limit it to 28.8 V. However, the lower voltage might mean the charger cuts back too quickly at high charge currents due to the inevitable voltage drops, so you might have to leave it at 29.6 V and terminate the charge a little early, or change to 28.8 V when the current starts tapering off (say when it falls to 50 A). It would be great if you can get that to happen automatically. Watching batteries charge isn't much fun after a while. This is only an issue because of the very high charge current and low battery voltage.