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

Infineon controller REVIEW+tests with 18x IRFB4110

Hey Doc..

I think its 22awg i used .. I wanted to do this with thicker wire .. but i would of had to cut the circuit board holes bigger .. i would of been happy with 18 AWG .. ..


I tried the controller sucessfully at 133v @ 32amp and it pulls hard with litterly tiny bit of heat...


I've added a shunt wire to give me 133v @ 46 amps.

I'll let yout know how she runs soon..

-steveo
 
Hello Doc,

I ordered the 2 Infineon kits. Besides the FETs, was there any other miscellaneous parts that you had to get (Caps, etc?)

DK
 
CamLight said:
Doctorbass said:
The compact desing of that controller' mosfet make mica difficult to place. The only way to put some would be to get large sheet of mica that cover all the cooper bar... but i dont know any source of that kind of mica dimension!
Doc
Doc, check out McMaster-Carr (http://www.mcmaster.com). They have 4"x8" sheets of .004" thick mica for about US$6. They have larger sheets too.
Once you get the sheet, see if you can split it into two thinner sheets. Not only will you double the amount of mica you'll have, the thinner sheets will have a lower thermal resistance. :mrgreen:

:mrgreen: :mrgreen: :mrgreen:

Excellent John!! Thanks.. I also thought about Mmcarr but did not searched for :oops:

I'll see and order some.. But i think there is some problem to order for people that have no business or commercial account?

I'll try anyway! I want 150A !! :mrgreen:

Doc
 
Doctorbass said:
CamLight said:
Doctorbass said:
The compact desing of that controller' mosfet make mica difficult to place. The only way to put some would be to get large sheet of mica that cover all the cooper bar... but i dont know any source of that kind of mica dimension!
Doc
Doc, check out McMaster-Carr (http://www.mcmaster.com). They have 4"x8" sheets of .004" thick mica for about US$6. They have larger sheets too.
Once you get the sheet, see if you can split it into two thinner sheets. Not only will you double the amount of mica you'll have, the thinner sheets will have a lower thermal resistance. :mrgreen:

:mrgreen: :mrgreen: :mrgreen:

Excellent John!! Thanks.. I also thought about Mmcarr but did not searched for :oops:

I'll see and order some.. But i think there is some problem to order for people that have no business or commercial account?

I'll try anyway! I want 150A !! :mrgreen:

Hey Doc,

I made up my own business name LOL ..

thats how i got my parts from

IRF
AVNET
MOUSER


=D

video to come tommorw for the bike .. it runs sweet!

-steveo
Doc
 
Doctorbass said:
Excellent John!! Thanks.. I also thought about Mmcarr but did not searched for :oops:

I'll see and order some.. But i think there is some problem to order for people that have no business or commercial account?

I'll try anyway! I want 150A !! :mrgreen:

Doc
Doc, you should be OK. You can pay by credit card instead of having an account so they don't seem to care who orders (persomal or business) as long as they pay. :mrgreen:
 
CamLight said:
Doctorbass said:
Excellent John!! Thanks.. I also thought about Mmcarr but did not searched for :oops:

I'll see and order some.. But i think there is some problem to order for people that have no business or commercial account?

I'll try anyway! I want 150A !! :mrgreen:

Doc
Doc, you should be OK. You can pay by credit card instead of having an account so they don't seem to care who orders (persomal or business) as long as they pay. :mrgreen:

I built up a 700$ order with them last summer since they stated that they shipped to canada on their internet site, and I even created an account with my canadian address and all was OK... but once I sent the order in, they said that they don't ship to canada!!! Arrrg! A place with sooooo much nice technical stuff all together, but they don't ship here anymore because of the political security paranoia. At least that's what they told me. Let me know if that's changed.

Another alternative that might be better yet than mica is loctite 3873. It's something like good thermal paste, but contain micro beads that assure a very thin but uniform shim between the part and heatsink while filling all voids. The 3873 has a Coefficient of Thermal Conductivity of 1.25 W/(m·K), and shims to a ~5mil gap (0.127mm). Mica is about 0.71 W/(m·K), so for the same thickness would be poorer than the 3873. From what I understand you can use the hardener or not depending on what you want as a result.

*EDIT: Changed mica's CTC to 0.71 from 1.25 --> oops!
 
Another way to do this might be to cut the thermal spreader bar into sections, such that the FETs with common tab connections are each on their own spreader bar. Using that McMaster Carr mica, you could then insulate each spreader bar from the case. That way, the hottest parts (the FETs) get the best thermal connection and the thermal spreader bars, with their much greater area, get to stuff the heat through the insulators perhaps a bit more effectively. An added bonus is that it beefs up some of the high current connections, too.


Just a thought...............

Jeremy
 
Jeremy Harris said:
Another way to do this might be to cut the thermal spreader bar into sections, such that the FETs with common tab connections are each on their own spreader bar. Using that McMaster Carr mica, you could then insulate each spreader bar from the case. That way, the hottest parts (the FETs) get the best thermal connection and the thermal spreader bars, with their much greater area, get to stuff the heat through the insulators perhaps a bit more effectively. An added bonus is that it beefs up some of the high current connections, too.

Just a thought...............

Jeremy
Very good thought indeed Jeremy! This is the way I went. I am still not sure what the best way to couple the FET case to the copper spreader is though. So far I use thin copper foil and prepare all surfaces, and just use pressure to assure a good connection.

Other than this I have been thinking about using a few silver foil sheets instead of copper foil (ordering them soon), and will also try some high end silver filled epoxy. This type of epoxy is really good when compared to regular heat goo (coefficient of thermal conductivity of up to 2.0W/m·K), but I wonder about how thin we can apply this epoxy. I would imagine we could get less than the ~0.005mils for mica or the loctite 3873 I mention above, which would make this solution much better than an electrically insulated FET connection. Specially for Doc's application - the drag race setup. (EDIT: loctite 3880 mentions 0.127mm thickness which is the same as the other non-conductive loctite).

Pure silver foil is very thin, and has a CTC of about 400W/m·K ---> over 200 times better than mica or other such solutions!! This is even better than copper, second only to diamonds for thermal conductance! I figure that silver is also softer than copper, so should compress well between the copper surfaces of FETs and spreader. I have yet to try this, so let me know any pros/cons you think about using silver foil like this.

Pat
 
Doc, you don't want anything between the MOSFET case and the heat sink/spreader other than a thermal compound to fill in the air pockets.

Every layer offers another opportunity for micro air pockets to form, causing a huge increase in the thermal resistance. The foils won't help to transfer heat any better than having the case of the MOSFET directly contact the heat sink/spreader....even if it was diamond. No matter what the substance, it still has some thermal resistance and you'd be adding that thermal resistance to the ones already there (junction-case, case-to-sink, sink-to-air).

Take out the copper foil, leave out the silver foil, just use a good compound and (I completely agree with Jeremy) separate the bar into independent electrical sections. You've now got a huge spreader than can efficiently pull the heat from the MOSFETs. Do not use that Loctite 3873 if you're separating the bar into sections! That 5mil gap that the compound forms is a HUGE gap and is much. much worse than using a good thermal transfer (unfilled) compound like Aavid UltraStick or Wakefield 126.

To electrically insulate the bar sections from the case, you can use any good insulating (filled) compound or a good Sil-Pad or split-Mica. Do not use any foil between the bars and case either! No matter how conductive it is, it has more thermal resistance than using nothing at all. Since split mica (and good thermal transfer compound (unfilled!!!) on both sides can be a lot thinner than that 5mil gap for the Loctite 3873 epoxy, try the mica if you can. Otherwise, check the thermal conductivity ratings for the other gap fillers like SIl-Pad. I remember some having ratings up to 1.6W/m·K. That's useless if it's 20mils thick, but if the they're thin enough, they might be worth using (in sheet form between the bars and case).

IIRC, you used a dremel polishing wheel to smooth the copper spreader. While a smooth surface is great for enhancing thermal conductivity, the small size of the wheel can easily cause the polishing to make the surface uneven. That is, the surface is smooth, but not flat. I strongly urge you to get a sheet of 600 grit (wet) sandpaper, place it on a very flat table or other surface, and place the MOSFET mounting surface of the bar down on the sandpaper. With gentle pressure, move the bar in a circular motion just a few times and then look at the mounting surface. If the abrasion of the sandpaper hasn't covered 100% of the bar's surface (that is, you can see lots of untouched spots), then your bar isn't flat. And this can be a lot worse than a bar that isn't smooth.

A flat, but "rough" bar still has thousands of metal points touching the back of the MOSFET as it is pressed down. The thermal grease then fills in the low spots. Not ideal, but not bad.

A bar that is not flat, but is mirror smooth, causes entire sections of the MOSFET to hang in midair with absolutely no direct touching to the heat sink. Even though the thermal compound fills in the big air pockets (even a 1mil deep pocket is huge), this can be a much worse situation than the rough flat bar.

Ideally, you want smooth and flat. If the sandpaper test shows an uneven surface, press down the entire bar firmly and evenly and use a circular motion to continue to wear down the surface until it's even across the entire bar. Then switch to an even finer paper and "polish" the surface (all at once, face down on a smooth, flat surface) until it's even again. Wipe clean and it's about as good as you can get it. Polishing it further, by hand or machine, would just make the surface uneven again.

Also, and very importantly, all of this is for naught if the case can't efficiently remove the heat generated by the MOSFETs. You have to make absolutely sure that enough heat is dumped to the ambient air to prevent the MOSFETs from overheating. That case is a huge "thermal wall" and will, by far, be the biggest problem in removing the heat. Even if the case feels cool, it can be a big problem. In fact, the case feeling hot can be a great sign. It can mean than the MOSFETs are efficiently moving their heat the case instead of sitting there burning up. :mrgreen: :mrgreen:
 
You could probably go to an overclockers computer site and learn alot. They are big on how to polish CPU heat sinks..etc
 
Deepkimchi said:
You could probably go to an overclockers computer site and learn alot. They are big on how to polish CPU heat sinks..etc
Be careful though. They often know a LOT less than they think they do. There's an incredible amount of bad science and misinformation floating around on those forums. Let the browser beware. :mrgreen:
 
What then if I want both the best thermal and electrical contact at once? The thin thermal compounds like the Aavid UltraStick are not electrically conductive, so this leaves me with either using nothing or potentially much thicker silver filled epoxy. Or what else?
 
wrobinson0413 said:
Maybe solder them in place. :wink:

ZapPat said:
What then if I want both the best thermal and electrical contact at once? The thin thermal compounds like the Aavid UltraStick are not electrically conductive, so this leaves me with either using nothing or potentially much thicker silver filled epoxy. Or what else?
I've tried this before on my previous controller, but decided against this option because correctly soldering FETs to thick copper spreaders is a PITA. I did start looking into using special low temp solder (like indium based solders) for this purpose, but have read about possible long term reliability problems related to the solder bond created. From what I understand, a well done compression contact can be more reliable is the long term than a soldered contact, specially when considering the additionnal stress to the FET caused during soldering to such a large copper mass. Not to mention the headache when we want to replace FETs! :?
 
Has anyone tried IXYS 14 amp mosfet drivers Digikey ixdd414ci-nd I don't know what their turn on/off time is with a cap load.

From a farm supply store they sell copper oxide powder- might work mixed with a silcone grease, small spacer washer under the screw area
(kapton film -or thin mica).

I did use a ground copper powder sold with thermacouples to fill hole around tc for better thermal contact
 
ZapPat said:
What then if I want both the best thermal and electrical contact at once? The thin thermal compounds like the Aavid UltraStick are not electrically conductive, so this leaves me with either using nothing or potentially much thicker silver filled epoxy. Or what else?
The only decent way to get both is to solder. Using a PCB with 4oz.-6oz. copper and large areas of copper on both top and bottom, with fan cooling if desired, you can move an awful lot of heat from the MOSFET. I don't recommend soldering the MOSFET to a spreader bar as that's going to be a LOT of heat flowing to the MOSFET for a long time as the bar cools. And hand soldering to a PCB is not recomnended for the best heat and electrical contact. It's hard to get solder under the entire tab or mounting surface of the MOSFET when hand soldering without overheating the MOSFET. Reflowing with solder paste is a much better option. But, hand soldering works great for less-than-absolute-max thermal/electrical contact.

There's just really no other way to get both great thermal and electrical conductivity at once. Best is to not depend on the tab or case of the MOSFET for electrical contact. Then you only need to worry about heat removal, which can be done very well using higher performance thermal compounds, proper mounting and a good heat sink.

Just remembered one method...
If you really, really didn't want to use one of the MOSFET pins for electrical contact, you can use a ring terminal and have it between the MOSFET tab (if it has one) and the head of the mounting screw. But, that becomes a weak point in the design as it's quite easy for that screw to loosen when thermally cycled...a very bad thing. :mrgreen:
 
JEB said:
Has anyone tried IXYS 14 amp mosfet drivers Digikey ixdd414ci-nd I don't know what their turn on/off time is with a cap load.

From a farm supply store they sell copper oxide powder- might work mixed with a silcone grease, small spacer washer under the screw area
(kapton film -or thin mica).

I did use a ground copper powder sold with thermacouples to fill hole around tc for better thermal contact
I haven't tried that driver but, typically, the filled compounds result in a much thicker gap between the active device and the heat sink. Combined with the truly awful thermal conductivity of the kapton/mica/silicone insulating pads, it results in a huge thermal resistance. It's much better to go with the absolutely thinnest layer of compound you can get. Especially if the design forces you to use insulating pads. :mrgreen:
 
What about the caps? I just blew up my second C'lyte because the caps vibrated and broke the leads (I guess it was vibration - I assume as a BSME the lead was adequately designed by our EE bretheren to handle the electrical loads). I tried a heavy dose of silicone to glue the thing down but the silicone is flexible and it still moved. Are short leads with the cap flush to the board adequate? It seems better than the lying down cap in the C'lyte. I plan on drilling the board, missing any traces although there don't seem to be many over there, and strapping the d@mned thing down with a cable tie. The holes are spaced for vertical mounting but the case height doesn't allow it. The Infineon case is tall enough?

P.S. I am interested in joining the Infineon/5303 club. I did order another C'lyte 4840 to get back up and running but I feel the need for a backup controller. I'll either rebuild one of the smoked C'lytes - one may still have the driver circuits and only need MOSFETS and a cap - or order an Infineon. I see the ComCycle units on eBay but how would I go about getting a kit? I am assuming you send a request to Keywins paypal address (ecrazyman@gmail.com) but what would I ask for? Is this a good choice for a 48V setup or should I consider a 12x with the standard MOSFETS or 18x with the standard or what? My 4840's would run at 2500 watts peak and 2000 watts until the battery gave out (or the caps fell off). I would want at least that and limiting something capable of more using the Cycle Analyst appeals to me - with more Pings I may want to run higher current or even go two 48V20aH in series - 116V hot off the charger. I could also use the hack I saw on (one of) the other Infineon threads to disable the LVC and us the CA/BMS cutoffs.

P.P.S. (OK, off topic or at least into meta-topic, I'll shutup after this) A condensed technical reference document for each of the major devices would be real nice thing. It is great that all the information available on these devices is on this forum but it is getting spread among threads and some form of organization of the information the manufacturer should have provded in the first place is in order. We could start a Wikipedia entry on each device, Infineon Controller, C'lyte Controller, Kelley Controller, 5303 (530x?), Wilderness Whatever, etc. and put condensed, vetted(how?) versions of all the tech data on regen, flashing etc. I checked and got only a stub for hub motor and nothing at all for Crystalyte, for example. We could build our own Wiki but that seems like a duplication of effort - is a thread on this in order?

I shutup now...
 
Can someone please perform a test?
I have 9 infineons: 8 have no fets and one is built into a bike and is inaccessible.

As those of you who have run the Infinion at high power know, the "soft start" has made for terrible throttle response (at least on my modded 6 fet 50A unit)
I have nearly a 1 second delay between full throttle application and the power "hitting" when I am riding fast. Totally over damped response.
This lag is multiplied by the higher current limit because when the power does come on it jerks unexpectedly and at different times.
This could just be an artifact of using a larger cap in the 48V 15A unit... Perhaps the 18fet uses a smaller capacitance for less effect.

the test-
On the 18Fet Version:

The throttle line (SP) goes through a 2.2K resistor (R8) then a cap (C8) ties it to ground before going to the uC.
I suspect that this C8 is part of (or all of) the soft start problem (thanks for the tip Fechter)

If anyone out there is in experimenter mode could you try opening this C8 cap to see if you get instant throttle response?
Luke and I (and others) are building super high power bikes with dual 18fet 4110 units and this soggy throttle response is a deal breaker.
We must be able to set the controller up for 1:1 throttle response.

By instant, I mean neck snapping instant.... If you crack the throttle the bike should lurch...
Probably wont make for a bike that cruises well under 20mph....

If not, I will report back after I get one of my 18fet boards built up.
Doc, I am counting on you here... Do my research for me.

Thanks,
-methods
 
That’s a good idea. Very sustainable and reversible method.

Maybe I will take it a step further and solder a 0 ohm resistor in parallel to decrease delay to 0% :twisted:
I am still not confident that the delay is caused solely by this RC circuit (need to test)
I am stuck doing family stuff for weeks so it will be a while before I can build my boards up and test.

Just wanted to find out before I invest any more time/money.

-methods
 
I took the day off work for other things so I am going to lift the cap today.
Will report back as to how much of the delay is removed via butt dyno.

-methods
 
Back
Top