Infineon controller REVIEW+tests with 18x IRFB4110

Doctorbass

100 GW
Joined
Apr 8, 2007
Messages
7,495
Location
Quebec, Canada East
= 16 march 2009 =

I just received the two dissassembled controller without mosfet today.

CRAZY FAST SHIPPING !!! 3 days China to Canada! Thanks Bob and Keywin for your help !

The packaging look nice and clean.

Once i'll receive the 40 x 4110 fets from bob Mcree and the nine continent hub motor from Jason (ebike kit)

I'll assemble the controller and then plan to test and try different asjustement on it... the secod controlelr will be used to debug the first controller even if i blow it..

For sure i dont plan using 150A 100V on the nine continent :lol: !!.. but my X5 will probably be :eek: ......... the candidate :shock:

First observations on the board.. No Hi and low side IC driver.. it seems that it's smt transistor that do the job

It also have 3 shunt.

Hi res pic to come of the board and different area on it.

Pics of the shippment arrived will be posted tonight!!


Doc

... happy!

:D :mrgreen:

Doc
 
Doctorbass said:
Is it really a 18 fets infineon on that Stealth Bomber ??

Good idea for the club!

Doc

Yes it was but the standard spec for the bike is the Xlyte digital 72v48A or 50A, whatever it is called.

It was the only 18fet infineon he had, a sample. Now it's mine :twisted:

It is in the post now, I will open and tell you what fets are in it when it arrives!
 
Ok.. I took the pics as i promissed..: the package arrived, the board, parts.. etc

Doc
 

Attachments

  • P1070279_800x600.jpg
    P1070279_800x600.jpg
    114.4 KB · Views: 4,930
  • P1070280_800x600.jpg
    P1070280_800x600.jpg
    97.7 KB · Views: 4,886
  • P1070281_800x600.jpg
    P1070281_800x600.jpg
    122.8 KB · Views: 4,886
  • P1070282_800x600.jpg
    P1070282_800x600.jpg
    115.3 KB · Views: 4,880
  • P1070283_800x600.jpg
    P1070283_800x600.jpg
    128.8 KB · Views: 4,884
  • P1070284_800x600.jpg
    P1070284_800x600.jpg
    150.6 KB · Views: 4,869
Doc,

If it's not a rude question, how much were these, including shipping?

(I'm after a couple as well)

Jeremy
 
Now here is the closer vire of the PCB

NOTE THAT THE LAST PIC IS A COMPARAISON WITH THE PCB OF MY ACTUAL CRYSTALYTE PCB THAT TAKE 90A 100V (original 72V 35A analog)
 

Attachments

  • P1070291_800x600.jpg
    P1070291_800x600.jpg
    107.6 KB · Views: 4,836
  • P1070292_1024x768.jpg
    P1070292_1024x768.jpg
    134.7 KB · Views: 1,989
  • P1070294_1024x768.jpg
    P1070294_1024x768.jpg
    134.7 KB · Views: 1,956
  • P1070295_1024x768.jpg
    P1070295_1024x768.jpg
    116.6 KB · Views: 1,957
  • P1070298_1024x768.jpg
    P1070298_1024x768.jpg
    171.9 KB · Views: 1,981
Jeremy Harris said:
Doc,

If it's not a rude question, how much were these, including shipping?

(I'm after a couple as well)

Jeremy


130$ for two shipped to canada 3 days fast shipping!.. (but without connectors and mosfet as i asked for)
 
Thanks for that, Doc, it looks to be a really good price.

I'll see if I can chase him up to let me have a couple just like you have.

Jeremy
 
Doctorbass said:
Jeremy Harris said:
Doc,

If it's not a rude question, how much were these, including shipping?

(I'm after a couple as well)

Jeremy


130$ for two shipped to canada 3 days fast shipping!.. (but without connectors and mosfet as i asked for)

Wow....This is a real deal! (I was going to buy one for 249$ US and then take it appart...)

Thanks for the tip!

Robin
 
This morning i did some update on the board about the two High current rail.

The stock controlelr is rated 55A and the trace only had a 18 gauge cooper braid ! :shock: .. and Stain is not a real conductor as well.. so it needed a little DOCTORBASS mod !

But before to solder them i had the idea to check every of the 40 mosfet i received from Bob and measure their max Breakdown voltage before they begin to conduct. so i'll be able to match and inly install the beat of the 40 fets!..

Natural selection they say! :twisted:

Doc
 

Attachments

  • Updated rail traces_800x600.jpg
    Updated rail traces_800x600.jpg
    102 KB · Views: 4,721
  • Mosfet ready to solder.jpg
    Mosfet ready to solder.jpg
    105.4 KB · Views: 4,704
hi doc
think I should add one of those to my collection, in the shipment I just got myself there was one but it is for a Team Hybrid customer I did get 2 12fet boards like yours ready for 4110 fets to be put in.
Most of the pads /holes have now been identified and we now know how to use them if you have a problem post on the infineon dummies thread for info, we all know your no dummie that is just the best thread to use at the moment.
as an after thought I am going to be running a front 5303 motor in a 26" rim with a 13ah 72v NiMh battery pack,the strength is no problem I am getting torque arms cut by my local laser cutters nice 8mm thick stainless steel with nut retainers to stop the nuts from coming off. I have no knowlege of the powewr of the 5303 though what sort of current it will draw and speed it can get, the choice of controller is still up for debate all I have is infineon 12fet controllers with 4310 fets good for 50A will that be enough and what sort of speed am I to expect from it, the design is a chopper along the lines of the american chopper TV series.

Geoff
 
geoff57 said:
hi doc
think I should add one of those to my collection, in the shipment I just got myself there was one but it is for a Team Hybrid customer I did get 2 12fet boards like yours ready for 4110 fets to be put in.
Most of the pads /holes have now been identified and we now know how to use them if you have a problem post on the infineon dummies thread for info, we all know your no dummie that is just the best thread to use at the moment.
as an after thought I am going to be running a front 5303 motor in a 26" rim with a 13ah 72v NiMh battery pack,the strength is no problem I am getting torque arms cut by my local laser cutters nice 8mm thick stainless steel with nut retainers to stop the nuts from coming off. I have no knowlege of the powewr of the 5303 though what sort of current it will draw and speed it can get, the choice of controller is still up for debate all I have is infineon 12fet controllers with 4310 fets good for 50A will that be enough and what sort of speed am I to expect from it, the design is a chopper along the lines of the american chopper TV series.

Geoff

Hello Geoff,

Thanks for the advices about the other thread for the infineon.. i'll se for more technical info on it. I spoked to Keywin this morning and he sent me a pics with all the connections hole described.. he did that in 10 minutes for me.
That is a great guy!

As for the 5303, I mounted one on a holligan MTB.. it's also a front wheel configuration. It run at 48V and it get a max speed repetable of 61km/h with the 72V 35A crystalyte controller with preset limit to 50A.
see it here: http://www.evalbum.com/2198

The torque arms will be helpfull !!.. 8mm thick stainless !! wow!..

how much for identical torque arm? :mrgreen:
 
Hi Doc,

I'm enjoying this mod thread.
Is that a custom made aluminum plate you bolt on the FETs? (first pics show more holes) If you are chasing performance I would suggest you to use copper plate, as it has almost twice better thermal conductivity and this is the critical part for heat distribution.
CopperBar.JPG
By the way do the FETs need isolation from the plate?
Can't see the shunts. Are they somewhere under the two big capacitors?
Zsolt
 
kZs0lt said:
<snip>I saw you made a custom aluminum plate you bolt on the FETs. If you are chasing performance I would suggest you to use copper plate, as it has almost twice better thermal conductivity and this is the critical part for heat distribution.<snip>
Zsolt
A good idea but just a couple of things to consider....
- While copper does conduct much better than aluminum, both metals can up and eventually reach the same temperature, i.e., reach thermal equilibrium, if the MOSFETs are creating much heat. They both have approximately the same ability to shed heat to the air so the copper doesn't offer much of a benefit unless the controller is used at a very low duty cycle (so the plates never heat up). In that scenario, copper's superior conductivity will help to keep the MOSFETs cooler.
- Copper's extra weight can be problematic if the heat sink bar is not firmly anchored. Any vibration that shakes the bar will cause a lot more torque on the MOSFET's legs and metal fatigue will raise their resistance until they burn out or just snap from the fatigue.

[Edit] Hmm....I think I see mounting holes on the stock MOSFET bars but not on Doc's. Perhaps he just hadn't drilled them yet. :)

- The bar's surface smoothness and flatness is much more important than the differences in thermal conductivity between the two metals. The surface must be both smooth and flat across the entire MOSFET mounting surface.
 
CamLight said:
kZs0lt said:
<snip>I saw you made a custom aluminum plate you bolt on the FETs. If you are chasing performance I would suggest you to use copper plate, as it has almost twice better thermal conductivity and this is the critical part for heat distribution.<snip>
Zsolt
A good idea but just a couple of things to consider....
- While copper does conduct much better than aluminum, both metals can up and eventually reach the same temperature, i.e., reach thermal equilibrium, if the MOSFETs are creating much heat. They both have approximately the same ability to shed heat to the air so the copper doesn't offer much of a benefit unless the controller is used at a very low duty cycle (so the plates never heat up). In that scenario, copper's superior conductivity will help to keep the MOSFETs cooler.
- Copper's extra weight can be problematic if the heat sink bar is not firmly anchored. Any vibration that shakes the bar will cause a lot more torque on the MOSFET's legs and metal fatigue will raise their resistance until they burn out or just snap from the fatigue.

[Edit] Hmm....I think I see mounting holes on the stock MOSFET bars but not on Doc's. Perhaps he just hadn't drilled them yet. :)

- The bar's surface smoothness and flatness is much more important than the differences in thermal conductivity between the two metals. The surface must be both smooth and flat across the entire MOSFET mounting surface.

Thanks for both comment kZs0lt and Camlight!
As for the mounting holes, yes they are on the bar, but you dont see them cause i installed the grey insulator with only the holes for the fets screw holes.. the backside have the mounting holes :wink:

Cooper and aluminium. have different weight and since i'm constructing a drag ebike i want the lightest possible components trade off with burst performance.. so the heat inertia seems to have not much difference in both alluninium or cooper for that 20sec...

As for the flatness, i ensured that that aluminium bar is very flat like a mirror before to put mosfets on it... But they are permanently installed. Here is why:

I am wondering if that could be interesting to test every of the 40 fets i got to screen al those to get the 18 best that have the best breakdown voltage?.. I mean i'll drive at 100-101V.. exactly like my actual crystalyte mooded with 12 4110 fets. is it a miracle if it never blown even though the rating of the fets is 100V and that i drive 100.8V full charge?

So if i could test their breakdown voltage to get only the 18 best high value, could it worth?

I found this simple test method somewhere:
 

Attachments

  • Mosfet%20Voltage%20Breakdown%20Test.JPG
    Mosfet%20Voltage%20Breakdown%20Test.JPG
    21 KB · Views: 3,985
hi doc
if you are going for 100v plus are you going to change the 100v caps onthe infineon board for 150v caps like the x'tlytes use? I'm sure I saw a few of the 15v brown capsin one of the pics you posted before.

Geoff
 
Doc,

Since this is for the drag bike, then copper would definitely be the better choice unless the specific heat capacity of the bar comes into play, since copper's is less than half that of aluminum. Assuming the bars are the same dimensions, then copper will win all around due it's 3 times greater density making it's heat capacity higher on a volume basis. There are far better ways to save more weight than that 150 grams or so.

Don't forget to ice your motor down very well before the drag, since 15-20 seconds will definitely be long enough to heat the windings and decrease performance.

A question about the 100m and 1/4 mile drags; Is pedaling ok? Since we're talking about bikes, I'd think that human assist is well within the rules. Any other "rules" I should know of in case I want to participate?

A question regarding modifications to the stock controller. My 15 fet Infineons are more similar to your 18's than they are to the 12 fet versions. Do you have to change anything in the MCU to run higher voltage and current, or is it just a matter of physical changes, ie adding more capacity to the traces, better fets, and modifying the shunts? Those seem to be things even a novice like me can do as long as I'm careful.

John
 
Doc - That FET breakdown V test method was posted by Wayne (wrobinson), but I don't think it will help you much in this case. Your FETs should be from the same batch anyways (bought in a tube from a good source), so should have little variance between parts. One important thing to remember is that FETs breakdown voltage rises with temperature, and IR's V ratings are at done -20oC or -40oC, so this gives you some play. At 20oC I think it's around 107V-108V for 4110's. But watch out if you're doing winter tests!! Your caps might be OK since 100.8V is your absolute max, but to be sure you might want to put your scope across them to check how high the voltage spike is during your FET switching instant (the top of the ripple after the output switches to batt+). Or just wait to see if they POP!

For high end controllers, I think that they mostly try to match Rds (on resistance) and gates threashold voltages since these are what determines how much the paralleled FETs share the current equaly. However, I don't think this will be an issue for you since the circuit's layout is not optimal anyways. I'm quite sure that the parasitic resistances and inductances of such a circuit would be much bigger than Rds and other variances you would encounter inside a batch of MOSFETs.

And +1 for what John in Costa Rica said about using copper for your heat spreader plate. Since you want to use it for short bursts, copper will help your FETs survive by being able to absorb much more temporary heat during your drag racing. This will let you set higher current settings if you need them. But remember what Frodus said about the FET-->Heat spreader thermal contact being very important. I sand my FETs using #600 sand paper, and do the same with the copper plate they sit on, just to be sure there are no voids. Thermal paste is terrible at transfering heat when compared to copper, so the less the better. Thermal pads are even worse.

Have fun!
Pat
 
I've been doing some calculations on the thermal qualities of controllers, FETs, heatsinks etc today, mainly because I want to try and work out just how much current it's possible to handle with an 18 FET controller.

The thermal conductivity of the heatsink only plays a modest part in things, by far the biggest factor is the ability of the device to effectively conduct heat to the heatsink and the ability of the heatsink to dissipate heat to the surrounding air. All the chat about the thermal conductivity and specific heat capacity is a bit of a red herring, as at the end of the day it's the surface area of the heatsink that is far and away the biggest factor.

I managed to find a very useful "rule of thumb" for estimating the thermal dissipation capability of a heatsink. Apparently, a reasonable estimate can be obtained, in deg C per Watt, by dividing 50 by the square root of the surface area (in square cm) that is available to dissipate heat to the surrounding air. A rough estimate for the little 6 FET infineon gives an effective surface area (including the fins but assuming the base is bolted to something insulating) of around 250 cm². This makes it equivalent to a heatsink rated at about 3.2 deg C per watt with an air temperature of 25 deg C and no forced air cooling.

The critical temperature is the junction temperature of the FETs, which has to be kept below 175 deg C, and preferably has to be a lot lower than this for reliability. There is resistance to heat flow between the junction and the FET case, between the FET case and the insulating pad, between the insulating pad and the mounting bar and between the mounting bar and the case. All together these add up to around 2.5 deg C per watt, so even with a perfect heatsink (one that has 0 deg C per watt), the FETs cannot handle more than about 70watts dissipation without overheating. Adding in the case gives a total effective heatsink capability of 5.7 deg C per watt, so allowing for errors and to give a margin for safety let's assume it's 6 deg C per watt.

6 deg C per watt allows a maximum dissipation from losses in ALL the FETs, plus the power supply regulator, of just 29 watts. If we ignore the power supply losses and just look at the FETs, then each FET can only be allowed to dissipate about 1/6 of this figure, say around 5 watts, before it overheats. The standard FETs in the 6 FET controller are STP60NF06s, which have an Rdson of 0.016 ohms max. Just considering resistive losses in the FETs, this means that the maximum current they can take is 17.7 amps. Once we add in an estimated duty cycle factor, perhaps a WAG that only one third of the FETs will be conducting at any one time (a combination of phase timing and PWM) then we can multiply that theoretical maximum current by about 3 to give a thermal limit of around 53.1 amps maybe.

Changing the FETs for something like the IRFB4110s, makes a big difference. The Rdson of these is just 0.0045 ohms, which raises the current allowed to about 33 amps. If we add in the estimated duty cycle factor then the thermal current limit goes up to nearly 70 amps. This ties in fairly well with what we already know, that this little controller is good for around 50 amps with 4110s without getting too warm.

The bigger controllers probably scale up pretty linearly, with the 12 FET being around twice as capable as the 6 FET and the 18 FET being maybe three times as capable as the 6 FET.

Adding a better heatsink will make a difference, but one quick win would be to replace the silcon rubber insulating "washers" under the FETs with old fashined mica washers. Mica is much better at conducting heat and would reduce the junction to controller case coefficient to around 1.5 deg C per watt (from around 2.5 deg C per watt). Painting the controller case matt black would also make a worthwhile improvement, as it might reduce the case thermal coefficient by maybe 0.5 to 1 deg C per watt. Doing both of these might increase the thermal current limit by around 15% to 20%, which equates to a direct increase in thermal current capability, or reduced junction temperature (and hence improved reliability).

Improving airflow over the controller would also make a significant difference, maybe another 15% to 20% improvement. All told, I'm reasonably sure that these 18 FET controllers "should" be OK for up to around 200 amps, with very low Rdson FETs and decent thermal transfer provision from the FETs to the case.

Jeremy
 
geoff57 said:
hi doc
if you are going for 100v plus are you going to change the 100v caps onthe infineon board for 150v caps like the x'tlytes use? I'm sure I saw a few of the 15v brown capsin one of the pics you posted before.

Geoff

Geoff,

Yes i'll change the capacitors for sure!! i dont want to have electrolyte spred explosion in the controller !! lol

I have 4 great quality 390uF 220V.. but i think the capacity would not be adequate... the actual total capacity is 3000uF and i wonder if increasing the current on this controlelr would requier more capacity.. so more volt + more capacity = very big capacitors!!!!

i'll see wich would enter in the controler

Does anyone kbnow if i just need to calculate the ratio to know if the capacity is suffisnet?... ex: 3000uF rating for 55A = 9000uF for 150A??

And if so.. should i use more smaller caps along the Hi and low side rail?? or can i put a larhe 8200uF 160V i already have external to the controller with big size wires...?


Doc
 
John in CR said:
Doc,

Since this is for the drag bike, then copper would definitely be the better choice unless the specific heat capacity of the bar comes into play, since copper's is less than half that of aluminum. Assuming the bars are the same dimensions, then copper will win all around due it's 3 times greater density making it's heat capacity higher on a volume basis. There are far better ways to save more weight than that 150 grams or so.

Don't forget to ice your motor down very well before the drag, since 15-20 seconds will definitely be long enough to heat the windings and decrease performance.

A question about the 100m and 1/4 mile drags; Is pedaling ok? Since we're talking about bikes, I'd think that human assist is well within the rules. Any other "rules" I should know of in case I want to participate?

A question regarding modifications to the stock controller. My 15 fet Infineons are more similar to your 18's than they are to the 12 fet versions. Do you have to change anything in the MCU to run higher voltage and current, or is it just a matter of physical changes, ie adding more capacity to the traces, better fets, and modifying the shunts? Those seem to be things even a novice like me can do as long as I'm careful.

John


John and Jeremy,

Your idea about cooper bar is excellent.. so i'll see in my stock if i could get this piece in cooper :wink: I still want to do the perfect Powerfull controller !!


John, for the pedaling, yes i'll may pedal.. but at that rpm and my 11 tooth gear.. I may need to stop pedaling at speed over 60-70 kmh :lol: .. the pedal speed too high for me.. !

There is no rules about pedal.. cause THERE IS STILL NO RULES AT ALL ABOUT EBIKE "BICYCLE" ON DRAG LANE IN THE WORLD !!.. I might be the first to try an ebike on the 1/4 miles.. and that's what i want to do!! :twisted: :twisted: :mrgreen:

John in CR said:
Do you have to change anything in the MCU to run higher voltage and current, or is it just a matter of physical changes, ie adding more capacity to the traces, better fets, and modifying the shunts? Those seem to be things even a novice like me can do as long as I'm careful.

I never worked with a controller MCU from now.. and i'll need your help for that.. About the shunt.. i'll upgrade it to handle higher current and not get too hot and have a drift current error... for the sensing of the shunt that limit the current i think it should be easy like the crystalyte and just to modify a resistor value.. or maybe just to reprogram the MCU with the interface.. but if that's the only way i might need help from someone that already played with that :wink:

Doc
 
ZapPat said:
Doc - That FET breakdown V test method was posted by Wayne (wrobinson), but I don't think it will help you much in this case. Your FETs should be from the same batch anyways (bought in a tube from a good source), so should have little variance between parts. One important thing to remember is that FETs breakdown voltage rises with temperature, and IR's V ratings are at done -20oC or -40oC, so this gives you some play. At 20oC I think it's around 107V-108V for 4110's. But watch out if you're doing winter tests!! Your caps might be OK since 100.8V is your absolute max, but to be sure you might want to put your scope across them to check how high the voltage spike is during your FET switching instant (the top of the ripple after the output switches to batt+). Or just wait to see if they POP!

For high end controllers, I think that they mostly try to match Rds (on resistance) and gates threashold voltages since these are what determines how much the paralleled FETs share the current equaly. However, I don't think this will be an issue for you since the circuit's layout is not optimal anyways. I'm quite sure that the parasitic resistances and inductances of such a circuit would be much bigger than Rds and other variances you would encounter inside a batch of MOSFETs.

And +1 for what John in Costa Rica said about using copper for your heat spreader plate. Since you want to use it for short bursts, copper will help your FETs survive by being able to absorb much more temporary heat during your drag racing. This will let you set higher current settings if you need them. But remember what Frodus said about the FET-->Heat spreader thermal contact being very important. I sand my FETs using #600 sand paper, and do the same with the copper plate they sit on, just to be sure there are no voids. Thermal paste is terrible at transfering heat when compared to copper, so the less the better. Thermal pads are even worse.

Have fun!
Pat

Thanks Zappat! for these advice about the breakdown.. I did not know that it is usually measured at -20 or -40 for the specs. so i'll try to keep below 101-102V and will need to put a load on my fresh charged 30s 3p A123 to remove the surface charge and sit them to 102-102V right in the area where the RI is the lowest!!! :mrgreen:

Question:
I wonder if sanding the fet back and the cooper bar is better for surface conductivity than a miror style polish?

Doc

Doc
 
Back
Top