• Howdy! we're looking for donations to finish custom knowledgebase software for this forum. Please see our Funding drive thread

More info on 4115 vs 4110 FETs I have learned

zombiess

10 MW
Joined
Jan 31, 2011
Messages
3,048
Location
Oklahoma City
So I've been continuing to experiment with the Xie Chang based controllers and have a strong affinity for the IRFB4115 FETs because of being able to safely run 100V battery packs.

Tonight I did some testing and tried out one of the EB318 controllers with matched FETs that I built and will be selling. This is not a for sale thread so don't talk about it or PM me about it (there will be a thread in the for sale section when I'm ready), I'm just mentioning it because it has caused me to look closely into these controllers and how to make them survive better. Good news is I have succeeded in getting one of my IRFB4115 18 FET controllers to run at 100A battery, 135A phase. Bad news is I blew it up first running it at 115A battery, 180A phase (yes, I was stupid and actually tried it this high for my maiden voyage... made it 50 ft :lol: )

Little back story to explain why I started off so high. I received a blown to hell (literally had exploded FETs) EB318 controller board in a set of 3 broken ones I purchased from someone on here to keep around for spare parts. I looked at it and said... I can fix that. 7 hours of work later, rebuilding traces and killing myself working with SMD parts that had evaporated traces... it lived. I did not want to waste IRFB4115 FETs on it in case it didn't work so I used some IRFB4110's that I had left over. I had just enough to build 6 closely matched sets of three. Since then I have been flogging the crap out of this controller using a 75V 18AH nanotech pack on one of the hardest to drive motors I know of, my 18.6Kv Cromotor. It has an inductance of 30uH which is less than half of what the controller killing Crystalyte 5302's had.

Anyways, I build all my controllers the same way. Polished heat sink bars (compliments of my wife), individual bergquist insulators, properly applied thermal paste between heat sink bar and case, matched FETs, beefed up traces, good low esr electrylytic caps and then I add on pulse rated polypropylene low esr caps at each end of the power rail and one between each phase for a total of 4. These 18 FET controllers use 0.22uF caps in addition to the 0.1uF caps that come stock.

So this IRFB4110 controller I built from literal scrap has been running 150A battery, 170A phase at 75V power this controller killing Cromotor for around 100 miles now without any issues. It even runs much cooler than I thought it would (at least externally, I'm lacking the internal probe at the moment). Max speed is limited to 99% so it is always in PWM on the high side. I scoped the low side of the controllers and the EB318 controllers only do block commutation on the low side, just like the EB2xx boards do. The EB3xx boards appear to have some way to measure and limit phase current from what I can tell so far but haven't been able to prove it yet.

So since I am doing a batch of 20 IRFB4115 18 FET controllers I need to know what they are capable of handling so I can rate them for people. Since I had such great luck with my IRFB4110 controller I decided to start off high on the 4115 controller for shits and giggles just to see what happens. I have a thermocouple installed to monitor the high side FET temperature directly on the body of the FETs. Settings were 115A battery, 180A phase. It worked fine on low (33%), dead stop, hitting it hard, no problems. Then I tried 99% and felt that familiar problem of something just went wrong. Took it inside and ripped it apart. Low side FETs of Phase A were dead. Not exploded, just dead with one showing a tiny pin hole and cracked case but no flames. Other 2 looked ok but were toast. Unfortunately it also took out the low side driver NPN transistor as well. Replaced everything and got it back up and running. Used my brain this time and turned the settings down. 100A battery, 134A phase. Battery at 74V and away I went. Watched the current spike to 105A then hold 98-102A during acceleration and taper off to 85A once I got up to around 48MPH. Top speed was GPS'd at 53MPH. I rode it WOT for just over 1/2 a mile and stopped under a street light to read my temp meter. It said 84C was the max recorded. Turned around and went back the opposite way for about 3/4 of a mile non stop. Max temp 100C. Did another pass after a 30 sec break to check the meter and when I stopped again I managed a max temp of 102C. I sat for about a min, checked the external temp of the controller with my hand which was quite warm, much warmer than the IRFB4110 controller runs at +140 amps. Made another 3/4 mile pass just to see if the controller would blow up or heat up more but it appears to have leveled off temperature wise and when I stopped the temp started dropping from 84C pretty quickly. Called it quits at this point because I had faith in the controller. Unfortunately for some reason I no longer have regen working after it blew up. When I hit regen it engages for a second, then cuts out like the FETs popped the first time. Need to trouble shoot this or just disable it all together since everything else is good.

I roll the bike back inside and watch the internal temp drop down to around 55C and level off there. The external case temp had risen even more at this point, quite hot, even burned my finger on one of the heat sink screws when I touched it. From there it took about 3-4 mins before the case cooled down and was just luke warm.

A while ago I had hypothesized that the IRFB4115 would run similarly in performance to the IRFB4110 based on switching loss calculations, but run hotter under WOT conditions because of it's higher RDSon value of 11mOhm vs 4.5mOhm for the 4110's. Turns out I was correct, but it ran a lot hotter than I though it would vs the IRFB4110. I will be installing a temp probe in my IRFB4110 controller in the same location an run the same settings in it to gather temperature data from it so I can compare the two to see just how much difference there really is.

I will also be switching motors back to a regular 9.6Kv Cromotor in a 16" wheel which has an inductance of 120uH so it's much easier to drive to see what happens with the temperatures of the FETs and my ability to run higher currents based on the temps. I believe they will run a bit cooler based on my old Lyen IRFB4115 controller that I have done similar mods too as well. It has survived 105A battery, 135A phase with a Cromotor in a 20.5" diameter tire at 125V and the FETs did not get as warm as they did in this controller (I have a thermocouple on the high side FETs in it as well, but it also has some additional heat sinking from me tying the tabs together with a copper buss bar so it's not a true 1:1 comparison).

Anyways, at this point I feel pretty confident saying IRFB4114 FETs, when matched, can sustain 35A for each one in parallel (so 3 can do 105A) for bursts of 30 seconds. For continuous use it is probably best to derate them to around 20-25A each depending on your ability to cool and keep the temps below 80C. The IRFB4110's appear to be able to handle around 50A each for 30 sec bursts and I would probably limit them to 30-36A each when paralleled for a more continuous type application.

Please note that these FETs are all Miller plateau matched so they have really close turn on times. They are within -/+ 5% of each other which is critical to keeping them alive. Yes I know we have all see people run crazy amps through controllers at their absolute max voltage ratings. You are of course perfectly free to do that, but don't complain if you blow it up. These numbers I'm posting are from my own research, but I figured there would be some interest by others on here. Also keep in mind that the motor I am running is VERY hard to drive for this type of controller and the typical eBike hub motor is exponentially easier to drive due to having much higher inductance.

I scoped out the low/high gate drive on this controller and will post the scope pictures tomorrow for those who are interested in seeing the EB3xx drive setup. I need to compare it to the EB2xx drives to see if it's any better.

My keeping the high side at 99% PWM is an attempt to reduce the chance of blowing the high side FETs as they are typically the ones that pop in most controllers, but tonight with my low side going boom due to me going WOT from a low speed at 99% throttle I am going to try setting them at 100% throttle (after looking at the switching times) to see if they live and hopefully run cooler.

I'll be posting more info as I gather it. Scope pics should be up tomorrow.
 
It was the low inductance motor that killed it in 50ft wasn't it?

I'm glad you're testing the 99% vs 100%. I played with it and couldn't tell any difference in controller temps. I can only see it benefiting those rare tunings with a surge in power at mid speed during WOT. I believe that with more typical tuning, by the time the controller gets to full duty (not the same as asking for full duty with WOT) the controller is in the safe zone of operation. I've blown a lot of controllers, but 0 at full duty. Maybe the results are different with more typical Kv hubbies.

It's good to hear that the EB3's are different in a significant way. I've never had an EB3, and every EB2 has run hotter than some other brands with my speed wind lower inductance motors.

John
 
I strongly suspect it died in the 1st 50 ft on high power because of the low inductance and me commanding max speed which at low speed can cause a large current spike before the controller takes over and starts limiting. If I would have waited until I was going 20+ mph it might have lasted longer. With such a low inductance motor the current spike rises 4x faster than a motor with 4x the inductance. This di/dt rise time can allow a spike that goes beyond what the FETs can handle. The controller was probably just too slow to limit the current in this case. By changing the phase current from 180 to 135a it causes the controller to try and limit at an earlier point. Therefore it sees the limit crossed sooner and tries to react. It could very well be overshooting to 300a before the limit brings it back to the safe zone. Without getting my bike on a dyno and verifying the level of overshoot its hard to say.

This failure was minor compared to most. When I opened up the controller all I could smell was a faint burnt smell we all know well, but no charring anywhere. I used a dmm in resistance mode to find the suspect phase. The FETs failed softly compared to normal. The motor did not cog at all. After I removed all 3 from the low side and measured them they all failed open vs shorted. One had a crack and tiny pin hole in its case. Sadly the jolt killed on of the low side driver transistors but luckily I had a spare from the dead boards I bought, it was even the same part number.

To me the interesting part of this failure was that it was low side. I suspect that if the low side was under pwm as the high side is it would not have blown up because the 16.05 kHz pwm would have limited the di/DT rise from the low inductance. I can simulate the current rise with bigmooses motor inductance spreadsheet to get some idea of how high the current could have gone. At this point though it is pretty moot. I found out through empirical testing that the controller can live at 135 phase amps 100a battery amps with me hammering on it like a monkey. I averaged 92w/hr on my 5.5min test session where I went 2.7 miles. Sometimes one real world test can be worth 100s of simulations :)

Just need to find my regen bug now. I will post up scope shots comparing an eb224 vs this eb318 later. I have both high side and low side run under no load (I really wish I had a dyno). I have yet to compare them myself
 
Thanks for doing all this and sharing!

Given that you did all this with the higher voltage fets I wonder what is the limit of the 3077's.

Living in Florida I found that for the majority of my hubs I don't need a lot of voltage since here the terrain is almost flat. So I tend to use packs that are 15S or less. I wonder what the true current handling capability is of a 3077 fet. Being that it has less RDSon. Of course to really know matching them into a controller and testing is the best way.
 
migueralliart said:
Thanks for doing all this and sharing!

Given that you did all this with the higher voltage fets I wonder what is the limit of the 3077's.

Living in Florida I found that for the majority of my hubs I don't need a lot of voltage since here the terrain is almost flat. So I tend to use packs that are 15S or less. I wonder what the true current handling capability is of a 3077 fet. Being that it has less RDSon. Of course to really know matching them into a controller and testing is the best way.

I don't understand what you mean. High voltage is best on the flatlands where you can go fast. Absent lower gearing in the hills is where the high voltage can be a problem, because you run at lower throttle and that partial duty mean more switching and higher phase currents. Most of my controller failures, other than a few just going too high with current, have blown climbing hills where I get stuck behind traffic going slower than I wanted. Once I learned to maintain at least 40-50% of top speed while going up hills, I stopped blowing controllers...almost 3 years now, knock on wood.

Zombiess is doing great work, because economical yet dependable high power controllers at operating voltages above 95V are sorely needed.

John
 
nicobie said:
I get so tired of folks guessing what will happen and then arguing about it. Testing is the real thing.

Thanks...

That's why I sent 3 IRFB4115's and one SOT23 NPN transistor to semiconductor heaven last night. This regen bug is really annoying me now. I want to find out why regen is broken on this controller. Maybe I have another FET on the same phase on the high side that is just slightly damaged. I'll be taking apart this controller again this week and making more measurements to see if I injured something else. Having a known working identical controller to make resistance measurements on parts when troubleshooting a controller with a problem makes finding the culprit many times easier.

I figured I could sacrifice some stuff for the sake of testing. I know I'm not exactly doing things scientifically, but it's more than I've seen others done. I'm just so short on time with all the work I have to get done.
 
So I don't have and EB218 board built up to do a direct comparison.. yet. I have a "broken" EB218-AS-1 which I will eventually repair and measure to get a better understanding of the differences. So for some rough comparisons I am using some scope traces I took from my EB236 controller I built up a while ago.

EB236 high side rise
eb236-01 High Side Rise.jpg
EB318 high side rise
EB318-01 High Side Rise.jpg

Only thing of note here is that the EB318 driver appears to be slightly slower at turning on the FETs vs the EB236 driver, this could be due to variances in the FET's gate charge. As expected the EB236 hangs on the Miller Plateau for much longer. The EB318 board transitions through it in around 0.5uS vs the EB236's 2uS transition.

EB236 high side fall
eb236-02 High Side Fall.jpg
EB318 high side fall
View attachment 2
The EB318 controller discharges the FET gate to an off state in about 1uS with only hanging on the Miller Plateau for around 250nS. The EB236 controller takes around 2uS to transition to off while hanging on the Miller Plateau for around 500nS.

EB236 low side rise
eb236-06 Low Sides Rise.JPG
EB318 low side rise
EB318-06 Low Side rise.jpg

This part I find interesting the most. The gate driver design didn't change much at first look, but as you can see on the EB318 low side there is no visible miller plateau at either turn on or off. The EB236 low side drive exhibits the typical Miller Plateau as one would expect.
The EB318 controller has the low side FETs on in around 1.25uS and the EB236 has them on 1.75uS, but the EB318 transitions right through were the Miller Plateau is expected and into full saturation.

Enjoy, nothing really major here, but I thought the comparison was interesting enough to show. It also illustrates how the gate charge effects the driver when the driver design is not good enough to supply the necessary current for quick switching. Double the FETs ~ double the switching time :)
 
Turns out my regen issue was caused by myself. I'm so use to running high voltage that I often change the R115 resistor to a 510 ohm from the normal 330 ohm. At 50+ volts this has never been a problem for me on the EB2xx boards I have built, but on the new EB3xx boards I am finding out the power supply gets starved for current. It worked perfectly fine at 75V for riding, but when I would engage regen there would be a small 2-3v spike put into the power supply side which caused the oscillation of the SMPS to stop. Changing R115 back to a 330 ohm resistor cured the problem.

I detected the issue by measuring the voltage drop from ground to the SMPS side of R115. With just hall sensors and a throttle connected it would drop to 18V with a 47v supply, that's 29V! I hooked up one of the original 330 ohm resistors and it read 40V and 38V with all accessories connected. Now regen works fine.

Cause of problem... user error.
 
I wanted to add this so that anyone who reads it in the future gets a better picture of the differences.

I have found that the limiting factor on the IRFB4115's in the 18 FET configuration to be thermal limited, not component limited. While several 30 sec back to back runs at 100A on an 18 FET IRFB4115 controller can cause the FET body temp 100C, I did a comparisons against a 24 FET IRFB4115 controller. With the 24 FET controller pushing 130A I was unable to get the FET's much over 55C. I had to up the power to 160A to get them to hit a peak of 72C for just for a second and that peak was reached with around 30 sec of WOT to 50mph, high regen, back to 50mph, regen, you get the point.

So at 160A battery side and abusive riding that is 40A per IRFB4115 and still they stayed under the 80C danger zone. I believe I can safely push 200A battery side on a 24 FET controller for short bursts due to their superior cooling.
 
Hi, I'm just 'dipping my toe into the pool of knowledge' about the workings of controllers. Thanks for this post. Keep the data coming.
 
Alan B said:
Nice info Jeremy! I'm planning to run an 18 FET 4115 controller at 60A or so, seems like it should be no problem at that level.

If the FET's are matched and it gets airflow no problem but without matched FET's it's up to fate on if it will survive 60A.
 
zombies said:
Watched the current spike to 105A then hold 98-102A during acceleration and taper off to 85A once I got up to around 48MPH. Top speed was GPS'd at 53MPH.
You are way over the 134 phase amp setting. You need to get a fast reading current sensor and mesure a phase wire.
You say they can handle 35 amps each based on this but.... When pulling 105 amps off the line the phase amps are insanely high.
Let me word it this way with a phase amp setting of 350 amps which I measured with my scope and 350 amps is realy where it was hitting at the start of a dyno run my DC amps were ~25 amps yes only 25 dc amps at the start when the throttle is phase amp based (torque throttle)!

So having said that you should find a way to measure the true phase amps and tell us what they are....
I bet you are well over 100 amps per fet during the launch.
 
Arlo1, I really want to measure the phase amps. Guess I should spring for a $30 400A hall effect bi directional 10v unit from bigmoose and make a quick and dirty el cheapo data logger from a PIC that correlates battery (through the on board shunt) to phase amps vs RPM. I've been curious about this relationship myself; mainly because of how differently the EB3xx boards handle phase current compared to the EB2xx boards. I am 90% sure modding the shunt on an EB3xx board does not alter the phase current. I know on EB2xx boards it does, but with all my work on these boards and shunt modding the one thing that seems to stay constant on EB3xx boards is phase current and regen current no matter what I do to the shunt. I seem to be doing more testing on these boards than anyone else so I might as well do this too so we can get some answers. P.S. I'm almost done with my brain board for Lebowki's setup which I'm beyond excited to start on (but need to finish projects that make $$$ first... almost there and my design just keeps getting better but I'm 95% now so it's almost go time). Then I need to build a motor to motor load cell dyno so I can do real testing under load. The projects just never end, but this little dyno will save a lot of wear and tear on my back compared to what I do now. Sucks I have to use someone like Bigbluesaw to make it which won't be cheap, I don't know any fabricators near me :(

BTW, did you catch the note I made about the thermal limit causing my rating of 35A each for the 18 FETs. I have my test 24 FET unit up to 160A which is 40A per FET and can't even get it close to the 102C I managed to get the 18 FET to pretty quickly and I was flat out purposely hard core abusive on something like 6 back to back passes with the 24 FET just to get it to peak at 72C. I drained just over 500W/Hr from the 1300W/hr pack in less than 5 mins of running. 25-50W of that was probably into the wiring/connectors though, fixing that next. Motor was only slightly warm after sitting for a few minutes. I guess 30 strands of wire wrapped 2 times per tooth can handle a decent amount of current :mrgreen: I LOVE this high KV motor. High KV + Mid drive would be the bomb if I had fabrication skills, but even as a hub motor it's way fun and I'm not even close to pushing this motor hard yet. I'm guessing once I go to 100V, at 250A it will be a whole new world. Don't worry, I have a custom frame being made soon for stress testing, then my current testing frame is going up for sale after I clean it up a little bit so it's not quite so dodgy looking. It's very stable but I don't want to take it into the +70mph club.
 
Jeremy great work and info! Did you ever post the phase resistance for your high speed cro motor? ... perhaps I missed it. I have noted you said the phase inductance for it is 30 uH.
 
bigmoose said:
Jeremy great work and info! Did you ever post the phase resistance for your high speed cro motor? ... perhaps I missed it. I have noted you said the phase inductance for it is 30 uH.

Measured it at 30uH inductance 37mOhms resistance phase to phase. "30x2" winding written on stator. It's hard for me to get this motor warm, even at 150A and hard pulls, wonder what 30 strands is equivalent to.

Normal Cromotor is 130uH inductance and 75mOhms resistance phase to phase. "18x4" winding written on stator.
 
Thanks for posting this zombies. So its good to see if you want more amps use a 4110 but if you can make use of more voltage with less amps then a 4115 is ok.
PS I usualy scope drain to source for turn on times the gate charge means nothing.
 
Hi,
I apologise for hijacking your threads, but I'm assuming you are pretty familiar with these controller boards and hope you may be able to help me out identify and source a component.
I'm repairing an EB318-A-1 board from my Crystallite 72V50A controller and have a shorted driver transistor which did nice things to the FETS.
All I can see is Y2 printed on it and it's board ID is Q2A.

Any ideas what type of transistor this is, or where I can source one?

Thanks
Max

Blon FET and Track.JPG
 
Back
Top