Thursday, June 25, 2009

Time to Get a Larger Hammer

After two days of running perfectly, I found yesterday morning that the amps were down and the FET had blown again. This really stumped me because I thought I had done everything to keep it cool and out of the switching region.

After showing the system to the power supply EEs at work, they told me that any FET in a TO-220 case like the one I had would not handle a constant 8-amps for a long time. The heat dissipation comes from the bond wires going to the silicon inside, not the actual silicon itself. They suggested that I put multiple FETs in parallel to lower the resistance and spread the heat over multiple sets of FET bond wires. If you can lower the resistance by half, the power dissipation goes down by a factor of four.

As you might imagine, I'm getting really tired of blown FETs, so I rummaged through our lab stock and got the biggest FET I could find.


Here's a comparison of the FET's I'm using. Instead of one TO-220 case FET on the left, I'm installing THREE of the TO-247 case FETs on the right in parallel with heatsinks. I'll be taking the "on" resistance from 30 milliohms down to 3 milliohms. By dropping the resistance by a factor of 10, I can theoretically lower the power dissipation by 100 (I^2*R from ohms law).


Here are the three high-power FETs installed inside the charge detector box. Note the burnt spot in the lower right corner of the printed circuit board from the smaller, individual FET that burned up yesterday morning.

I just started charging with this new circuit about an hour ago. The charge detector box is COLD. I can barely feel a slight temperature difference between the top of the box and the vehicle chassis.

Some might think this is overkill. Those folks would be right. It probably is, but I'm sick of wasting my own time and the time of all the folks at work who are graciously helping me out. I would rather nuke this one with an overdesigned system instead of having to revisit it several times.


Here's a third schematic (click to enlarge) including the changes to Q1, Q2 and Q3 FETs on the right side.

I sure hope this issue is dead. I've got a few EV shows coming up and I'm done with dealing with this issue. Let's see what happens. Hail Mary.

Monday, June 22, 2009

Cautious Optimism

After talking with the power supply EEs at work, I came home and tried a few experiments to see if I could fix the alternative charge detector circuit.

First, I used an external 12V UPS battery instead of the AUX power output on the Belktronix charger to see if noise on the power supply was causing my problem. Nope: The circuit still behaved as if the opto-coupler latched closed.

Next step was to pull out my oscilloscope and start probing around the system. Fortunately, most of the battery pack and components are floating, so I can attach the ground of the scope to several different high-voltage points to serve as a reference.

Yowzers! There's a hell of a lot of noise in this system. Whenever the Belktronix charger starts up, I get 50V spikes at ~200Khz on the oscilloscope by just holding the tip of the input probe with my finger. Keep this thing away from your pacemaker! I'm guessing that these high voltage noise spikes are causing problems with the opto-coupler.


Here's a modified schematic. The EE folks at work suggested adding C4 (.01uf cap across the opto-coupler input) to help cut down on the noise. I also reduced C3 down to 10uf from 47uf. This reduction increases the maximum switching frequency for the FET (due to the hysteresis in the 555) from 1Hz up to about 5 Hz. While the FET switches at a slightly higher frequency, it's still so slow that the FET stays out of the linear switching region for long periods of time.

I plugged in the system and (gasp), the FET stayed on like I intended it to. After suspiciously pondering the circuit thinking that I had messed something up, I shorted out the OVP signal and the FET quickly turned off and the LED came on. The link-10 meter shows a full 8 amps going into the batteries. In a few hours, the BatMon boards should start lighting up and I'll see if the circuit responds correctly. The charge detector box is a bit warm to the touch, but definitely not hot.

I'm very excited that adding a .01uf capacitor to the opto-coupler input cleared up much of the noise. Someone over in the UK had their Belktronix charge detector blow up too with a lithium ion battery pack, so I hope this circuit can help them too.

After my heated rant last night, I received several supportive e-mails and comments from people. Thanks to all who responded for your ideas and well wishes. I feel good when I hear from folks because I don't know how many people are actually keeping track of this blog, especially since things have quieted down recently.

Let's keep our fingers crossed with cautious optimism. Cheers.

Sunday, June 21, 2009

Failure Again!

I came home this evening after a 20 mile drive with excitement to see the new charge detector circuit work. I plugged in the car and the circuit had an interaction with the optocoupler on the BatMon board and didn't see the correct signal. Nothing is burning up, but the FET is stuck in the off state for now, leaving the system to charge slowly with a really hot 3-ohm resistor.

I'm so angry at this charging system! It's a good thing I just got back from my mindfulness meditation session or I would throw a wrench at the car. God, I'm pissed! It worked so well with the batteries near a full condition, but the circuit fails when the batteries are discharged. I suppose this is a good data point for further analysis, but I'm rapidly running thin on patience.

I'm taking the circuit into a bunch of EE experts tomorrow at work (they always like working on these problems instead of their real jobs) and we're going to get to the bottom of this. I'm running out of energy to get this working. Damn, this sucks.

A New Hope

After the utter failure of the Soneil and the Joule chargers, I did some more research on individual battery chargers. Ideally I was looking for the following:

  • Power factor corrected
  • Weather-proof for mounting under the hood
  • Isolated to work with a series pack of batteries
  • 6-8 amps charging capacity
  • low power enough to put all 12 chargers on one 15 amp circuit (possibly 20 amp circuit)
  • somewhat affordable
I did some more searching and found the BatteryMinder 12248, the Xantrex Trucharge 10, and the Waeco PerfectCharge IU1012. The Waeco units seemed really good but were from Germany with no distributors in the USA. The Xantrex and BatteryMinder units were $150 apiece (ouch!) and didn't meet all the criteria above.

I decided to go back to the drawing board and see what I could do with the Belktronix charger. In concept, the charging system was great. It's a nice power-factor-corrected series charger that dumps out a solid 8 amps and it's sealed so that bugs and dirt don't get in.

Since I've moved to the Synkromotive motor controller, there aren't any more mid-pack taps off the battery pack, so the batteries tend to drift out of balance more slowly. Arguably, the resistive shunts with the Belktronix charging system should be able to handle this quite well.

The problem in the past has been the charge detector. Without going into all the details of operation, the OVP lines from all the BatMon modules would pulse and cause the FET inside the charge detector to turn on and off. Since this pulsing was frequent, the FET was often in the switching region and easily overheated.

The FET inside the charge detector determines if the large 3-ohm 180-watt series resistor is inserted in the charging circuit. If the FET is on, then the 3-ohm resistor is shorted and the batteries get a full 8 amps, which is great for fast charging. This brings up a problem when the BatMon boards detect a "full" condition on the batteries. The small 3-ohm shunt resistors start to burn up if they have to sustain the 8 amp current flow. Thus, the charge detector needs to turn off the FET to insert the large series 3-ohm resistor to limit the current. If you turn off the FET too early, it takes forever to charge your batteries and the large 3-ohm series resistor gets really hot. Turn off the FET too late and all your shunt resistors burn up, causing massive smoke and potential fires in your EV. (ugh)


Here's a picture of the inside of the blown charge detector. The FET is clearly blown in half with burn marks all the way around it. Note the melted region on the case cover in the lower right.

The FET in the charge detector is a very good one with a very low turn-on resistance. If we can limit the amount of FET switching and turn it on/off hard to keep it out of the linear region, it should stay relatively cool, even with 8 amps flowing through it. So, I put on my EE hat and designed a completely new circuit to put inside the charge detector box.


Here's the charge detector box with the new circuit inside of it. The circuit (shown at the end of this post) uses the comparators and flip-flop inside a 555 timer (Radio Shack special!) to drive the FET with hysteresis on the input to limit switching.


Here's my old Radio Shack 555 timer handbook. It's a crying shame they don't sell these anymore. I guess it doesn't make a profit and people just aren't into dinking around with 555 timers anymore. As you can see (click to enlarge picture), this tiny chip contains two comparators, an RS-flop and an output driver. This is just what we need to observe the pulsing OVP signals from the Batmon boards to determine if we should turn the FET on and off to short the large 3-ohm series charge resistor.


I tend to go overboard with these things. This is my kitchen table with all the soldering/test equipment on it.


Here is the simple circuit I put inside the charge detector box to control the FET that shorts out the large 3-ohm series charging resistor. Note that this lacks a few safety features that would prevent a blow-up if the user hooked things up in reverse, so this is not a product-worthy circuit. It's simple enough and you can get most of the parts (except the FET) from Radio Shack. If you click to enlarge the picture, you can see the theory of operation and get more details.

I'm going to try this out for a few days and see if I can charge faster without blowing up the FET. The potentiometer still probably needs adjustment to properly set the duty-cycle detection on the OVP pair, but this is a good start. If this works, I'll have fast charging without setting the small shunt resistors on fire. Here's to hope...

Friday, June 19, 2009

A Very Dark Day with the Joule Chargers

This whole experiment with the individual chargers has been a very humbling experience.

After the whole set of failures with the Soneil chargers, I learned from the distributor that the 1214S model doesn't support series strings of battery packs:

Tim,

Well, based on this email I sent an email to Soneil. Apparently there has a
been a design change in some of their chargers (these included) which
prevents you from being able to charge while in a "series" configuration.
This means that the half with no LED are dead. This also means you can't use
that charger type with that configuration. I apologize for this huge
inconvenience, I have no idea why they would change the design. The closest
12v Soneil charger we have that does not have this design change is the
1212SR, but it is a 5amp and is $89.95ea. We can refund your order or
replace the same value with 1212SR chargers, I'm not sure what else to offer
you. Both issues you have found have been somewhat unbelievable, I apologize
for the "luck" on this. For the electric bike application, both charger
types work great. The electric car application has a few unexpected
differences....

Thanks,
Mike (from ElectricRider)

After hearing the above, I returned the Soneil chargers and asked ElectricRider to ship me a set of their in-house Joule chargers (model JJ12060). Their technician insisted that the chargers should be fine for a series connected battery pack.

About a week later, the box full of Joule chargers arrived on schedule.


Here's the pile of twelve Joule chargers after I stripped off some insulation and added ring terminals for the batteries. I'm really excited because these are smaller than the Soneils and have integrated cooling fans. Despite the 35-40mA current draw, I'm full of hope that these will solve my pack imbalance issues.


After installing several of the chargers, I observed the following:

  • One failed to even start up (LEDs were flickering)
  • Six didn't have their fan come on. I thought that the fan might come on after the charger heated up, so I left one on for 20 minutes. The unit overheated and ceased to work. One of these six had its 100% charged LED stuck on, possibly indicating other issues.
  • Five powered up correctly; however, the fan on two of them seemed to get up to speed very slowly, possibly indicating fan bearing issues.
In short, I'm very disappointed with these chargers. They have a one-year warranty, so I could send the bad ones back for replacements. I purchased thirteen of them, assuming that one would be bad. The > 50% fallout rate greatly concerns me. In addition the PFC (power factor correction) value for these chargers is about 0.6 which roughly means only 60% of the input power gets to the battery and the other 40% gets dissipated as heat in the charger. No wonder these things get so hot!

I'm taking a day to ponder my alternatives. The Belktronix charger has a lot more "spaghetti" to it as well as hot shunt resistors. On the flip side, it has a very high PFC value which means that the primary charger is quite efficient and it doesn't heat up much. The charger is also sealed, making it much more resistant to dirt and bugs.

One option I'm considering is replacing the shunt resistors on the Belktronix charger with 50-watt halogen light bulbs. This means some of the shunted energy gets converted to light instead of heat. I'm also looking into a circuit to replace the charge detector in the Belktronix system.

EV Awareness Day here in Portland is only three weeks away and I've got to get this beast charging correctly. Here's to being stubborn. Stay tuned.

Sunday, June 7, 2009

Sleuthing the Soneils

Against my better judgment, I took apart one of the failed Soneil chargers yesterday afternoon to see if I could find out anything about the failures. There weren't any burned components or obvious damage. After powering up the unit with the cover off, I could verify that the isolation transformer between the AC side and battery side wasn't getting pulsed, and thus passing no power to the battery (or the LED). It seems that the UC3842A chip DC-DC converter chip used to drive the FET which pulses the transformer was dead. It's power supply had a resistance of about 3 ohms to ground (bad).





After mulling over the circuit in my head last night. I tried to compile a list of facts I know about the failure:

Facts:
  • The batteries were all fully charged in the morning, despite the charger failures
  • The batteries with working chargers had ~13.7 volts on them, the failing ones, ~13.4
  • There were far more failures in the batteries in the front of the car
  • The front batteries are closest to the negative pack voltage
  • I plugged in the rear batteries first when testing the chargers
  • The rear batteries typically have slightly more charge on them because they get warmer
  • I started all rear battery chargers simultaneously on a power outlet strip
  • I started all front battery chargers one-by-one by plugging them into power cords with three outlets on the end
  • The components crossing the isolation barrier in each charger are: a transformer, an opto-isolator, and three ceramic capacitors (.01 uF at 250V)
Hypotheses:
  • Since all batteries were at nearly full charge, the failure must have happened at the end of the charge cycle, perhaps when the charger transitioned from acceptance charge (14.7V at 3.5A) to float voltage (13.7V)
  • Since the front batteries typically finish charging last and I also plugged them in last, perhaps a surge from the rear battery chargers (when shutting off) caused failures in the front battery chargers
  • Is suspect that the capacitors at the isolation barrier in the chargers failed because they have a breakdown voltage of 250V. When charging one battery, this is fine. When charging a long series string of batteries (oh, say, 144V), the charger output stage can float +/- 144V relative to the AC voltage on the input stage. 120V AC really becomes 170V DC when rectified. Add this to 144V and you get 314V, which is clearly above the 250V rating in the isolation capacitors.
Based on this, I don't think sending the chargers back for replacements would really help the situation. If all the chargers have the same 250V isolation caps, then they would all fail eventually. I suppose I could take all the chargers and replace the isolation caps with similar values but give them a 1KV voltage rating.

I'm trying to figure out the best option for now. To drive the car, I'll probably need to re-install all the Belktronix resistive shunt balancer modules. The car would be driveable but it would still take a long time to charge. I don't like the idea of hand-modifying twelve Soneil chargers, but that would probably be what it takes to get them working.

I could send all the Soneil's back and get the Joule chargers again. That is, after checking with the designer that they support 1KV of isolation. The Joule chargers draw 35mA of current when sitting there unplugged. This hypothetically translates to 2857 hours of discharge from 100Ahr batteries or about 4 months until the batteries are completely dead. If I went on a long vacation, I guess I would leave the chargers plugged in on float charge anyway. The Joule chargers also only charge with 6A, which means (despite a bad PFC rating), I might be able to plug in all chargers to a 15 amp AC circuit (The Soneils pull about 18 amps for all 12 chargers).

This is a really bad time because we have so many electric vehicle events happening now. Having a dead EV is not too appealing. I'll ponder this for a bit.

Happy Sunday morning.

Saturday, June 6, 2009

A dark day with the Soneils

This morning was a dark morning.

I received the Soneil chargers yesterday around noon and stayed up until 11:30 last night to install them. After installation, I powered them up and all chargers lit up with an orange LED. Yay. This morning, I got up and half of the chargers had a green lit LED and the other half were dark. I used a Kill-a-Watt meter to determine that the dark units were pulling no current at all.

I've heard rumors that Soneils don't like to be attached to a series pack.

It looks like my EV is dead for the near term. Putting the Belktronix system back in with all of its shunt resistors and battery boards does not sound appealing right now.

Ugh.

Friday, June 5, 2009

Fixing the Potbox, Soneil's Arrive

Over the past few days, the car has still been lurching around despite the potbox parameter adjustments that I've made inside the controller. I contact Bob Bath, the original CivicWithACord builder who sold me my PB5 potbox. It was brand new from KTA Services and not used/worn-out. Yesterday in stop-n-go traffic, the car was lurching again and the controller faulted when I haphazardly floored the accelerator with the car out of gear. After expressing my frustration, Ives at Synkromotive highly recommended that I install the factory throttle position sensor (TPS) from the original Civic instead of using the PB5 potbox.


Here's the factory Civic TPS installed. I mounted in the same place as the PB5 potbox, but used a piece of 1.5" angle iron 4" long to mount it. Unlike the PB5 which only has two wires, this TPS has three which allows for higher reliability and more accurate signalling to the controller.


Here is the rear side of the installation. You can see the top of the grey-painted angle iron that I bolted to the engine mount side. I had to chamfer one corner to prevent interfering with the engine mount rubber. The actual electrical potentiometer is black at the far left of the picture, while the cable attachment is on the far side of the large butterfly intake valve. There were some extra bolt holes in the throttle housing, so I just used a piece of all-thread and bolts to suspend a small piece of 3/4" angle iron to support the fixed point on the throttle cable.

After a test drive, the system was sooooooo much better. I can creep forward with the car now and the lurching is gone except when I really stomp on the accelerator. I'm now completely convinced that the PB5/PB6 potboxes are complete crap compared to the factory throttle position sensor. It's too bad the TPS doesn't go down to zero ohms, so it can't be used for many other controllers like a Curtis.



After I arrived back home from the test drive, the FedEx truck showed up and dropped off 13 brand new Soneil 7amp chargers. I've been waiting for these for over a month now and am happy they are here. This will replace the Belktronix charging system and allow me to remove all the spaghetti associated with the shunt resistors and battery monitoring boards.

This is a good day. The car is running smoothly again, and I'm psyched to get these new chargers put in. Have a great weekend everyone.

Cheers, Tim