Got some FET temperature data after abusing my bike tonight

[zombiess]

Tonight was the first night I got ride my bicycle after having a ton of mods done such as switching to a 26" fork for more rake, having my rear swing arm lengthened by 8 inches, had a friend who is studying to become a mechanical engineer help me securely mount my new battery bag which can carry 30S3P
very nicely to the center of the frame.

This is what the bicycle looks like as of 11-22-11. Wheel base is 45" without me on it and it's quite comfy and stable to ride. Shock is a Cane Creek Cloud 9 and is very nice. I can run any size wheels from 20" to 26".

Motor is a 9C 2806 in a 20" wheel, 12 AWG phase wire upgrade.

Max recorded speed @ 100% throttle was 52mph, managed 56 mph at 105% throttle but the current didn't really like to drop off much below 50A once at top speed at 105% where as at 100% throttle it would drop down to 40A, pretty big difference. Not sure the extra 4mph are really worth the extra 10A of current draw.

Settings on the controller were 100% throttle, 47 battery amps (really 60A due to soldered shunt), 125 phase amps, 0.1S block time, peak off load voltage was 124.4V and would see up to a 4V sag under load, probably because it's was a bit cold. Battery config is 30S2P Hobby King 6S 20C Lipo. Testing method was several high speed runs, then several dead stop and instantly going WOT (could get a tiny power wheelie for a few feet on the 100% setting), 8" longer swing arm combined with 140mm cane creek cloud 9 air shock is bad ass.

So onto the data.

This is how my sensors are mounted in the controller

Inside ambient temp sensor which is near the 3 shunts, but not directly over them

FET temp sensor. It's not directly on the tab as I would like but it's between the cases of the two FETs and has thermal paste in contact with them. It's mounted on the green phase wire bank, not sure if this bank suffers switching losses or just resistive (if it has switching losses the temp should be hotter on this bank)

Outside temp, 12.5C
Highest logged ambient case temp was 34C and it averaged 29.5C while riding the entire time
FET Temp max was 42.7C for 1 second (3 seconds later it was down to 39C and dropping fast) and I only have a few points where the temp would get over 40C.

Average temps on the ride was 25.5C for the FET and 29.5C for the ambient case temp. This was a 10 min ride (just over 600 data points at 1 sec intervals), logging started after I had already been playing around with 105% throttle to get everything up to working temp, then switched to 100% throttle for the test.

While the controller was in current limiting mode the current draw held steady at 60A (shunt has been calibrated). Peak amp draw was 73A. Cycle analyst reported 79Wh/mile I wasn't being very easy on it for those 10 mins.

I did a quick check of the winding temps when I pulled back into the garage and got a max of 104C which proceeded to drop down to 90C within about 3 mins so it was probably about 110C before I got the probe into the coils.

The bike is an absolute blast to ride, but needs less voltage sag. 4V seems like a lot of sag, but during the day when it's warmer I don't get quite as much sag usually 2.5-3V, I'll keep an eye on it.

This controller was modded by myself following markcycle earlier post. Each 3 FET bank has it's own buss bar made from 10 AWG stranded wire connecting the three tabs of each bank in parallel to the B+ as well as each phase bank connected by tab in parallel to the phase wires.

See this post for details on the mods done http://endless-sphere.com/forums/viewtopic.php?f=2&t=33847

Observations, I don't think I'm really stressing this controller after making these mods and could go up on the current quite a bit, but I don't see any need to as the bicycle accelerates quite fast and is tons of fun. As soon as my Greyborg Hub Monster is laced into it's 20" wheel, it's getting tested next. Wish it had a faster winding than 9.5 KV, but at least it's going to be easier on my 18 FET 4115 controller.

Things to try, buy a Lyen 18FET 150V controller like the one I have, mount temp probes an see how much my mods have helped (probably quite a bit). Run the current setup in hotter weather to see what happens to the FET temps. Seems like right now they peak 30c over ambient and run about 17C over ambient average during high speed abuse. Get cars to drag race from stop lights

Hope you guys like this data.

 

[nicobie]

Thanks for the info Zombie

 

[grindz145]

Sweet. What did you use to log?

It's pretty hard to get a good contact on a to220 package though. Especially after it heats up. The numbers seem reasonable though.

 

[ZOMGVTEK]

You are using nylon screws to secure the TO-220's to the rail? Sounds a bit adventurous. What FET's are those on that controller?

Side note: Are you reasonably close to 'the strip'? I'm going to CES in January...

 

[zombiess]

You are using nylon screws to secure the TO-220's to the rail? Sounds a bit adventurous. What FET's are those on that controller?

Side note: Are you reasonably close to 'the strip'? I'm going to CES in January...

Yup, I'm trying out nylon screws. I know guys who use them in the engine bays of their cars for some things and they hold up fine there. Most hood insulator are also held on by nylon fasteners and they don't melt unless their is a fire.

http://www.matweb.com/reference/deflection-temperature.aspx
Shows that at 66psi the nylon didn't deflect until 160C for nylon 6, I'm not clamping anywhere near 66psi. Not sure exactly how this correlates to the screws I'm using but I was surprised how tight I get make them before I could break one (tried it just for fun). I'm actually not sure what nylon these are made of, but if I have problems, I'll be sure to post. The head of the nylon screw has a brass lock washer between it and the tab so the entire head is not exposed to the temp. It's an experiment. I had one engineer and another guy who used them in his car tell me I'd be OK so I figured I'd give it a shot. I've heard of others using these on to-220 packages with success as well. I also re-tapped all the holes to 6-32 from 3mm so I'm using much larger screw than normal just fits through the tab hole.

I'm about 15 mins from the strip in a suburb called Green Valley which has some bike paths if you bring your bike and want to go for an ebike ride. I've always had to ride alone and have a nice 10+ mile loop I like to do, mostly on paths.

Sweet. What did you use to log?
It's pretty hard to get a good contact on a to220 package though. Especially after it heats up. The numbers seem reasonable though.

This is what I'm using to log the temps.
http://www.amazon.com/gp/product/B0039LAVEM/ref=oh_o00_s00_i00_details
As I showed in the picture, I lightly squished the probe between 2 FETs with a small amount of thermal paste. Might not be the best spot, but I am getting data.

 

[zombiess]

Cliff notes: Careful with regen, it makes things get a lot hotter. Updated first post with picture of bike taken tonight.

I updated the picture of the bike so you guys can see it's current config, pic taken tonight, damn I love this thing, just need to get a chain on it really soon, then the Croatian Hubzilla.

Got some more data tonight, couldn't get regen to work on Lyens controller but after a few hints from him I set it for 63/126V through hex editing and it started to work great but doesn't stop as well as I thought it would even on setting 2, then again, I was going 40+. I need to wire it into the brake lever with a small switch so it's automatic, very nice to have since I currently have no rear brake. I have it wired into one of Lyen's moped cluster switches which feature a push on/off button, a momentary button and a 3 speed selector switch all in a small neat package. If you don't have one, get one now, so convenient and neat, cheap too. Shameless plug for Lyen for helping me out so much. Never saw him advertise them but he mentioned it to me so I grabbed 2 of them. It does require knowledge of wiring, it is not plug and play, but if you can use a continuity tester, you can wire this up, I even have diagram now as to what wire does what on it

Anyways, onto the data.

Logging time was just shy of 10 mins by a few seconds with about 8 mins of super hard abuse on a fully charged 125V pack 30S3P. Peak current of 73A and a normal battery current of 60-62A on the CA (controller programmed for 47A 125A phase).

Was 12.5C outside. Both probes started off at 13C with no previous riding
With regen now I can really get the FETs and motor to warm up

FET peak temperature was 57.4C which occured after 6 mins of really abusive riding, speed up to 40+ hit regen down to 15mph, do it again 2-3 more times in a row. Average running temp for the 10 min log which included 2 mins of me parking in the garage was 33.0C. After it peaked at 57.4C which took about 3 hard stops with the 2nd one creating a 51C peak and the 1st one somewhere around 48C, I continued riding but was almost done and wasn't abusing it anywhere near as hard. 1 min later it dropped from 57.4C to 41.6C, another min later (but still using regen at lower speed) it was down to 35.8 because I was done and pulling into the garage.

Now that I'm done abusing it at high amps I dialed it down so from a programmed 47A (actual 61A) to 43A (actual 56A) battery and phase from 125A to 115A since I feel I was getting close to the saturation point or close to it since more amps doesn't seem to give much more acceleration. Not sure if the phase needs to be multiplied by the 1.29 factor the same as the battery current does.

57.4C means the FET junction temp was 80.4C and the RDSon value was 16.9 mOhm per FET, but with 3 in parallel that means it was somewhere around 5.6 mOhm and the bank of 3 was dissipating 553 watts for that bank of 6 fets. I have a spread sheet I'm using to estimate where the FETs will fail if run continuously and at these settings it looks like if I keep the FET case at these settings below 70C it's OK. It could probably tolerate a short burst over that though. As I stated, I was abusing this to get to these temps, my normal riding style and speed would probably never get them this hot unless it was summer time. I have the parallel FET's derated to the equivalent of 2.25 of them in parallel instead of 3 since they never share current perfectly.

Spread sheet estimates I could probably run 80A actual battery current and 170 phase amps continuously if the FET case is kept under 66C, which is pretty easy to do once the motor is in it's operating range. Just need to make sure to roll into the throttle vs just jam it wide open, but it would probably take the short burst no problem as long as block time is really short like 0.1S. If tying the tabs together helps with them switching more evenly then derating by 2.5 means it would sustain 72C continuously before having issues. Unfortunately I have no idea how close my spreadsheet is to reality, but it's working great at 60A battery 125 phase so far.

I finally got the motor hot enough to smell it! Yay! Got it to (I only logged 123.8 because it took me 1.5 mins to get my gear off and stick the probe into the winds so I figure it was a little warmer and am estimating it was 126C. Regen loves to make things HOT!, but so does dumping lots of amps at high volts into a tiny motor like I am No smoke, no discoloration, just a little odor that it got warm from me purposely trying to make it hot. I monitored it for another 5 mins and the temp had dropped down below 90C, cooling holes for the win.

Controller peaked at 39.4C around the same time as the FET peak and had an average running temp of 37.6C

Top speed recorded was 51mph in a 25 zone... uhhh oops

There are still even more ways to improve these controllers, but I think pulling from the tabs by making buss bars like I have has got to be one of the best fairly simple mods.

Another thing I noticed is the bike is much smoother to ride and the throttle very easy to modulate. Before the swing arm riding this thing was a pure death wish.

Nerd mode off for now, hope this data is useful to someone. I might add some small fans to see how that effects the temps in the future, but for now I'm just going to enjoy a nice bike ride Major thanks to my friend Tim (who is a future mechanical engineer) for coming over an helping me figure out how to mount everything really securely. There is no movement in the mounting of the battery bag or the controller. The big bag also allows me to grip it with my knees much like a motorcycle if I wish or even pedal when going low speeds.

 

[CamLight]

<snip>
57.4C means the FET junction temp was 80.4C and the RDSon value was 16.9 mOhm per FET, but with 3 in parallel that means it was somewhere around 5.6 mOhm and the bank of 3 was dissipating 553 watts for that bank of 6 fets. I have a spread sheet I'm using to estimate where the FETs will fail if run continuously and at these settings it looks like if I keep the FET case at these settings below 70C it's OK. It could probably tolerate a short burst over that though. As I stated, I was abusing this to get to these temps, my normal riding style and speed would probably never get them this hot unless it was summer time. I have the parallel FET's derated to the equivalent of 2.25 of them in parallel instead of 3 since they never share current perfectly.
<snip>

Be careful about using temperatures measured between two FETS with against their epoxy shells, and not in very tight thermal contact. The "case temperature" of a FET (necessary for calculating junction temp) must be the temp of the center of the FET's "rear" metal plate, not the epoxy shell. That 57.4C measurement you made is much lower than than the actual case temperature and you'll need to factor that into your calculations. You can come within a few degrees of the actual case temp by carefully measuring the temp point where the tab enters the epoxy shell (on the "front") or drill a small hole into the heat sink alongside the FET, near the center of the epoxy shell, and measure there (pressing the thermocouple against the metal and using thermal paste or epoxy to fill the hole).

Coupling the thermocouple very tightly against the metal surface and using thermal epoxy (or, in a pinch, thermal paste if wrapping all around) and then covering it with tape or something else to prevent any fan or ambient air breezes will help to prevent cooling of the thermocouple and the adjacent wires which would result in too-low temp readings). Even a little bit of epoxy/paste between the thermocouple and the surface being measured will also result in too low a temperature reading.

Thanks for going through the trouble to get this data though!! The more we have the better.