1994 US Electricar Repair Blog by Mike

Fritz put up his fet board for sale today. It’s a beauty! It’s made from 3 oz copper. It should handle the current spikes nicely. I have to decide to either have Shawn make me one of his chassis for the welder or just use a short PC tower that are super cheap but probably require some labor. Maybe a cool transparent case would look sweet. I am only waiting on the fets to arrive from Hong Kong and the board to arrive from Fritz. I still need to order the 4 farad capacitor too.

Today I wanted to see if I could devise a way to use my 3 axis cnc to wind 41ga wire onto the current transformers from the Dolphin charger. They tend to fail if the fets go up in smoke. So there are better grades of fets that have helped the chargers live well. But now I found that I don’t have a single spare charger. They all have open circuited secondary windings. After unwinding a bad one it looks like it holds about 407 feet of 41ga wire. But it is even more important that the secondary gets 3000 turns, I think.

Just as I was firing up the computer to see how to write the g-code for making my mill wind these coils, the hardware started popping fuses, and fets. When I installed the 60 volt 10 amp power supply in my rack cabinet, I must have rotated the knob on the variac to full output. Normally it’s set to about 55 volts ac. After rectification it ends up being about 75 volts. So at 120vac the rectified voltage had to have been much higher. That’s what killed the fets, fet drivers and the diodes.

I used some old 1407 fets to sub for the correct ones just to help me debug the system. The Z and Y channels worked out eventually. At first so did the X channel, but it then decided to smoke again. This time several of the fast recovery diodes were hit too. Now that I don’t have any diodes, I have to wait until I place a major order with Newark before I can get the parts. At least 2 of the 3 channels work.

I still need to figure out what code to find/make to wind the transformer. The idea is to mount the transformer to the vice. Then chuck the spool of wire with some tubing to the collet of the mill. The table will then move the transformer around in a circle to wind itself. It will be a very slow process, but I think simple to construct.

The second idea is to chuck the transformer into a collet and spin it to wind the wire up. Not that difficult as I think about it now. Just have to make the z axis go up and down to evenly distribute the wire onto the transformer. The Z axis is very slow and not too precise. I could move the wire up and down by hand. I’m trying to automate this as much as possible.

Either way I have to wait to get the cnc system restored to operation and installed back in it’s cabinet. That will take getting the parts. For now I can use 2 axis’ to just work on either of the 2 winding concepts.

Some wonderful people sent me a pair of Dolphins to repair. Someone before them
had been into both of them deeply and did some serious damage. Pins were not put
back into connectors properly. The fets on the charger board were not even
parallel to the base plate. Both main boards and chargers were toasted. Even the
GFI hardware was missing from both units. Parts were missing from one of the
dc-dc boards. So when these systems arrived they had one board out of 6 running.

The one charger was the very latest revision from USE as it had a new board and
revision I’d not seen before. Only one fet was shorted, the other was fine. My
standard rebuild is to replace both fets, both relays and the diode weather they
are bad or not. So that’s what it got. I was very happy to see that the T1
current transformer survived the onslaught of failure.

One main board had the standard shorted bridge rectifier that goes along with
the shorted charger fets. The fet driver pwm output was fine however. But it had
happily open circuited the traces that usually just get shorted to other things
and cause a whole other set of issues. So some jumpers were added to the new
bridge rectifier. The board made it through my long QC list of tests and runs
well in the truck. Both boards got the regen upgrade (I modify the Dolphin to
increase the regen output by 40%) and the Classic Dropout mods for vastly
improved reliability and a better driving experience.

Once the charger was supposedly rebuilt, and the main boards repaired, It was
time for a test run. The Dolphin booted fine. But as soon as the throttle was
pressed the Dolphin faulted with a sizable thunk and gave an IGBT fault. It
turned out that the large diode on the charger board, when shorted, ties the
motor’s neutral line to the pack negative. That’s essentially a short across the
output of the Dolphin. It was odd at the time that none of the Dolphins IGBT
fault circuitry was triggered. All of it was normal. Yet I got an IGBT fault. It
turns out that the software looks for a reasonable load at various throttle
positions. Since the load was nearly infinite at low throttle it protected
itself by disabling the IGBT’s by not energizing them rather than the IGBT
circuitry sending a fault.

Then there are the mice that got in these Dolphins. They love eating insulation
off of wiring. Especially teflon wire. Then of course they have to use the HV
section of the Dolphin board as the bathroom.

After all of the repair work, and the passing of the many bench tests, it was
time for a test drive. The first board had dropouts so badly, that it could not
even pull itself back into the garage. I would say a dropout every second. The
next board dropped out every 15-20 seconds. After much studying and testing over
4 years, this really bad board made it a little bit easier to diagnose what was
causing the classic dropouts. Thankfully I figured out the dropout issue as
these boards were destined for the scrap bin! I pulled a 3rd board out of
mothballs that had bad Classic Dropouts, and it too was cured of dropouts with
the mods I had made to the other two boards.

After performing the Classic Dropout mod, I took both Dolphin boards for another
ride in my truck. Zero dropouts at very high regen or very high acceleration.
Nice improvement! I went for a walk at the park to enjoy the rare sunshine. I
noticed a bad coolant leak as I walked back. The bottom plate on the Dolphin
chassis is just .125″ thick aluminum glued and screwed on. The glue is very
brittle. When I used my cooling system pressure tester on the radiator, the
Dolphin cooling plate squirted coolant all over. It’s almost impossible to get
the tiny red-loctited screws out. It takes quite a while to scrape all of the
adhesive off too. I’m about 60% done. Makes me think this is another ticking
time bomb. This is the second Dolphin I have had that has cooling plate leakage.

Time to get to it!

Better charger fets were ordered today. The wattage rating is double that of the original fets. The rds of the IXFH44N50P-ND is .140 ohms. Rds on a IXYS IXFH26N50Q-ND is 0.20 ohm vs 0.25 on the original IXFH21N50.

The original MR826 charger diodes are too hard to get. So I found an equivalent at Mouser. 625-GI828-E3, 5.0 Amp 800 Volt.

I found that all of my spare charger boards have a bad T1 transformer on them. The secondary is open circuited. Rick had rewound his 5 years ago. So I am ordering wire and going to rebuild all of mine as well. They hold 407 feet of either 40awg or 41awg. Rick rebuilt his with 40awg. My math shows that it should be 41awg.

Seriously. It’s freakin’ solved after 4 years of research….

One of the boards that arrived for repair last week, was by far the worst I’ve
ever seen for dropouts, and it flashed the fault indicator in a very radical
fashion. It is now perfectly drivable!! No dropouts under acceleration or regen!
I was in total disbelief!! It works? Must be a mistake!! So I grabbed the second
board in for repair. Dropped out every 15-20 seconds. Made the same
modifications to it. Now it’s smooth as glass too! Zero dropouts under
acceleration or regen!! I had a third board laying around for years who’s only
crime was having massive repetitive dropouts. The modifications completely
stopped all traces of dropouts on it too. All 3 of these boards drive perfectly
now! Am I dreaming!!?? Feels like the Twilight Zone around here!!

Three boards saved from the junk pile. I have seen many more just like them.
That feels so good!! Many owners have this problem right now.

The Dolphin hardware has component values that are just barely within spec. With
miles of driving, time, and thermal cycles, the hardware goes out of spec, and
the Dolphin starts faulting, usually without any faults shown in Dolcom/Dol7.
I’ve posted this in the past. But the faults are felt as a sudden shudder in the
drive train. An instant loss and then instant return of power lasting only a
split second. It can rattle your fillings loose!! Some do it rapid fire, one
dropout after another. It’s all the same problem with the hardware going out of
spec. I found the faulty hardware!! Finally!!

During this research, I found an old quality control document from USE. It was a
long list of things that needed to be checked on their test fixture after each
board was assembled. The items on that list were very telling. So I have figured
out the procedure for checking most of the items on that list on my test
fixture. This ensures that the board gets thoroughly checked. I use it on every
board, just like the factory did.

Since it’s still very cold over most of the country, now is the best time to
have the offending boards with dropout issues repaired. The cold always makes
the dropouts worse. It’s the best season to ensure that the boards work in the
worst conditions of the year. Dropouts are less common in the warm months.

Send me your board and I’ll make the classic dropout modification and put your
board through the checkout list on my test fixture for $50 + shipping. Pack your
board very carefully, like it was your first born baby.

Mike

The board layout for the 160ah Thundersky/Sky Energy lithium cell BMS is finished. I’d like to scale it down to also fit the 90-100ah cells, but that takes another board layout session since the 160ah cells are so large compared to the 90-100ah cells. For now getting the thermal paths, thermal cycling, and basic operation tested is a higher priority. Making a board that will fit both the 90-100ah and 160ah cells would be the next task. At this point it’s setup to bypass 3 amps. 5+ amps is possible.

The BMS will control the charger output as well as the motor control if any of the set points are hit. I’ll get the BMS to Dolphin interface board going as well.

One of the most sobering tests performed by a guy making his own BMS is here. His experiment showed that using the wrong parts can in deed cause a fire. Several designs out there use these same parts.

In another post, this poor fellow had his whole car burn to the ground. What a shame. He did a lot of work on it.

Here is a Prius fire story using A123’s. A bad connection was suppose to be the culprit of that disaster.

Yesterday started out poorly. I only slept for 4 hours. After waking up at 4am, I headed over to my computer. There was a notification from Ebay about some US Electricar stuff that went up for auction. As it turns out, it was the same hardware that one of the guys in the USE group had mentioned was for sale. We thought it would be a private sale. But here it was on Ebay. I immediately posted that fact that it was on Ebay to the group. The boards were all suppose to be 65kw. But only 2 were 65kw, the other 6 were 50kw boards. So I looked at the name of the seller, and it looked vaguely familiar. I searched my email archives and found that I had corresponded with the guy in the past. So I wrote him a plea to not sell them on Ebay since some random buyer could end up with these important parts. He agreed to pull them from Ebay and sell them to me. Sometime later this week they should arrive. Also included were 2 charger boards and a dc-dc board. There was also 4th board that I have never seen before.

What I have done in the past with two other batches of boards I have received,
is to go through these boards carefully with my test fixture, and see if they
work or what they need to make them work. One 50kw board looks to have some
serious damage to the Mach220. I have replaced a couple Mach220’s already. But
that serious kind of damage may also have toasted some board traces. Worst case
the board becomes a parts donor. I hope the DSP chip is ok as it could be used
in another 50kw board. Almost every chip from the 50kw boards are missing. Since
the DSP chip cannot be copied, then the boards will always have to borrow a DSP
from another board.

The 65kw boards will go onto my test fixture to see if they can be lit up or
not. At least they have all of their chips. I’ll see if the 65kw chips are any
different than the 50kw chips. At least one of the 65kw boards has a red arrow
on it indicating that it needs something.

5 out of 6 of the 50kw boards are 2nd generation. They have a socket on the
Mach220 plus some other additions. I think the factory finally figured out that
when the Mach220 gets damaged, that unsoldering it from the board is dangerous.
At least with the sockets the Mach220’s can be swapped in and out for testing. I
have the code for the Mach 220’s so duplicating them is easy.

There is also a pair of charger boards and a dc-dc included. There is a 4th misc
board that I have never seen before. No idea what it does.

I know of 2 other sources that need boards. So if any of the 50kw boards are any
good, 2 of them will get shipped out asap. Usually when the chips are missing
the boards have issues.

All of the boards that run will have to be test driven to see if they exhibit
any of the classic dropout issues. As I was discussing the dropout issue last
night with a fellow member, he pointed out that every board we have could end up
with severe drop out issues just simply from driving them long enough. That is
exactly what has happened with a couple of the boards that I have. To me this
issue is 50% resolved. The last 50% will be tough.

So the next morning the pack was still soft. The voltage sagged right away. So I electrically removed about 14 cells from the pack so the charger could keep the current level up around 1-1.5 amps even at full voltage. This will overcharge the nicads and help equalize them. Flooded cells are the only ones you can get away with this. So the charger was programmed to stay on an extra 5 hours after full pack voltage, 400v, was reached. The pack was charged like this all week.

A couple of days ago, I noticed that the resting pack voltage after the charger shut off during the night, was very low. About 306 volts. This concerned me. The first day the resting pack voltage in the morning was 325v. But today it occurred to me that I may have finally gotten what I was trying to achieve. The pack’s cells are now all at the same state of charge. So the cell to cell voltage is closer. With 237 cells in circuit (14 removed), at 306 volts, that’s 1.29v each cell. That’s a bit low. Usually the cells are about 1.35v to 1.4v after sitting for hours. I am guessing that the cells hit the knee in the curve as a group, and the voltage starts to go down as the cells are overcharged. This is some fairly serious overcharging. But it shows that the cells now are more equalized.

Midweek I reduced the overcharge time from 5 hours to 2 hours.

This weekend I’ll reattach the 14 cells that were taken out of circuit and see how the whole pack is acting. Although the 14 cells did not get the huge overcharge, the pack should show improvement. I could add some overcharge to the 14 cells before closing the pack up just to try and bring them into the same state as the rest of the pack. I’ll actually make sure I do that.

The bottom line is that with 3 different versions of the same cell, they are getting worn out as exhibited by their impedance being high. This allows the pack voltage to rise artificially high and make the charger think the pack is full when in fact many of the cells are not fully charged, resulting in a sagging pack.

Isn’t debug fun? It sure separates the boys from the men, or me from my sanity Wink

I’ll be on vacation starting Monday. That excuse one for not having my BMS installed. #2 is that the second truck has taken all of my time to get corrections made to it’s various electrical systems. It’s basically done. Just needs a new pack. So the #1 truck is the tough vehicle to get a shielded cable from the pack to the interior of the truck. Once the cable is mounted through the wall of the pack, then adding the BMS is easy. Now I have Thundersky’s that I have been testing as well as 3 new 50ah Hi Power cells to test.

To look for noise it’s helpful to determine if it’s conducted or radiated noise. Steps that I use.

Conducted Emissions

1) I like to use clip on ferrite beads for initial debug. Easier and faster than adding components. The kind you see attached to a wall wart power supply or your pc monitor cable. 25-30mm long. 15-20mm in diameter. When picking them up from surplus locations, get a few different versions as you don’t know which RF material they are made from. Clip them onto both ends of the master bus cable. One at the master board and one at the first slave. Check for changes. Does the system work better? Verify with a scope, always.

2) Now repeat step one with the Slave bus, with all of the master bus ferrites removed. The slave bus could easily carry EMI/RFI and cause issues.

3) Repeat step one with ferrites on Master and Slave bus.

4) The shotgun approach is to just do both Master and Slave bus at the same time. But you won’t know the sources of noise as well.

If 1-4 improve function then it’s the noise is mostly conducted emissions. Although radiated emissions from the vehicle could turn into conducted emissions due to the huge amount of wiring we have to use.

5) Always twist pairs of cables to 4 turns per 25.4mm(1 inch) between slaves. Cordless drill works great for twisting wires together.

6) Use shielded cables with twisted pairs for Master to first slave cables.

Radiated Emissions

Here is a fantastic article on how to make a home made probe for sniffing EMI/RFI with your scope from a piece of coax a tiny ferrite bead, and some sandpaper. When I showed the crew at work this article, everyone had me making these probes for them.

http://www.elecdesign.com/Articles/Index.cfm?AD=1&AD=1&ArticleID=7491&pg=1

Here is the probe I made from the article.

Below are scope shots using my probe on a project that has so much EMI that it smoked the processor and other drivers badly enough that my head with a full face helmet, hit the ground hard and made me unconscious. Had a concussion for a year. So EMI and I have become great friends!

There are EMI and RFI (aka E field) probes. An E field probe is just the ground of your scope probe tied to the probe tip. It shorts it in a dc sense. But for RFI, it’s a path to joy and harmony. If you make your probe ground lead a bit longer and coil it, the probe becomes much more sensitive to weak signals.

Both EMI and RFI probing examples of my concussion making machine (a home made self balancing scooter) are shown here.

These scope shots showed me clearly at the time where the emissions were coming from. Each output from the processor got a 1k resistor to isolate it from the drivers. A 1k resistor was put on the output of the driver chips. These two steps did not reduce EMI, but did reduce it from getting into the sensitive parts. The next step was to reduce the EMI itself. I had used 250mm long ribbon cables for connecting the master board to two slaves (control board to 2 H-bridge boards). Normally the ribbon cables are about 50mm long. Reinstalling the 50mm cables did the trick. The EMI probe showed me the way. As you can see from the scope shots I also found EMI from the power supply inductor as well that I could follow with my handy EMI probe along a ground trace. That’s right. EMI was following the ground!!

So assume nothing, and measure everything!!