fet suppliers

[eP]

With the stock 4710's, it runs much hotter, all other things being equal. This leads me to think that the on resistance is the most significant sorce of loss.

Yes it is. But because switching freq. is lower than 50 kHz at low rpm hub.
The situation would be quite different if you try replce 1 irfb 4110 by two stp120NF10.

Switching loss seem to be very small. Conduction loss during freewheel would be high, but the duty cycle for that is quite small, so overall it doesn't seem to be a big factor.

It is very small at low switching freq. But when you go to 6000 rpm high pole low inductance bldc motors the situations would be quite different.

So wait and see what will cadstarsucks calculate for us.

 

[bobmcree]

i recently posted that irtronix had put the 4110 online for 3.19 and i suggested that the online price for single quantity was bound to be significantly higher than the 100 piece price when cadstar... insisted that my 2.60 price was not repeatable.

i just called my contact at irtronix, mike o. (sorry i said mike n before) and ordered 100 pcs at 2.60 which i will have in about a week. mike told me that i was welcome to refer my friends to him and that he would give anyone calling him the 2.60 price for 100 pcs with delivery about 10/17.

they did find 150 pcs at the factory to cover my order and to put another 50 online for low volume purchasers at the 3.19 price, which is still the best price around.

it is very common in the industry to charge folks buying one part a lot more than those buying 100, so the 20% price differential seems quite acceptable. these parts are often on allocation from the factory and even the large suppliers who charge higher prices run out, so the need to wait a month to get the next batch is not unusual.

all the fets i have in stock are assigned to controllers i have waiting to upgrade, so if anyone needs theirs done it will be another week before i have the parts. if i get a few more in the queue before oct 17th i will place an order for another hundred or more to keep up

i understand some of the 35A controllers are now being shipped with a single 4310 replacing the stock pair of 4710s, with empty locations. this is probably ok for many users, but the extra thermal conductivity from two fets to the heatsink would make them more reliable. a single 4110 would probably be fine, but i am replacing all 12, making the controllers virtually bulletproof even with the current limit increased to 50A or more.

it is important to consider that most of the current in these controllers is carried by the very heavy layer of solder on top of the pcb traces, and that going much over 50A limiting might overly stress the interconnects, even if the fets can easily handle it. bus-bars could be added to improve the interconnect current handling capacity.

since i can still get the fets at this great price, i will keep the price at $80 to replace all 12 in working controllers or those with just blown fets and no other damage. i feared i would need to increase the price to cover the higher parts cost, and am glad i can still pass on this great price.

 

[fechter]

You're willing to upgrade the FETs in working controllers for $80?

Sounds like a deal to me. I can probably steer you some business.

I don't really have time to do those jobs, but get a lot of requests.

Xyster... are you listening?

 

[bobmcree]

You're willing to upgrade the FETs in working controllers for $80?

Sounds like a deal to me. I can probably steer you some business.

I don't really have time to do those jobs, but get a lot of requests.

Xyster... are you listening?

yes, i am doing it for several vendors and have a stack of controllers on my input shelf right now. with shipping the fets cost me about $36 and it takes a couple of hours to mod and test, so i am content with $20 an hour for the labor, since i am officially retired.

my retirement income does not provide quite enough to support the lifestyle i want, so i am happy to do this extra work and to provide the service for my friends and fellow ebike enthusiasts. i am working on a new controller design for next spring, but so are justin and several others, so i do not expect to get rich from that project.

i can turn around a typical controller in a week or less, and will not charge over the $80 unless there is damage besides just blown fets. i am working out the best option and prices to provide increased or adjustable current limiting as an option.

at this time i think the best option for adjustable current limiting is just to raise the controller limit to 50A and use a cycle analyst for adjustable throttle based limiting.

 

[xyster]

You're willing to upgrade the FETs in working controllers for $80?

Sounds like a deal to me. I can probably steer you some business.

I don't really have time to do those jobs, but get a lot of requests.

Xyster... are you listening?

Yes! Already on it. I knew you were doing me a huge favor, Fechter, by agreeing to mod my controller. It sounded like Bob was looking for business and something to do, so seeing this, I emailed him about it. Got a reply already. Consider my business steered.

 

[Jozzer]

Some customers may not have a CA, why not put a pot in the side of the controller, so your solution is stand alone?

 

[giveahoot]

I was about to place an order at IRTronix, but they wanted $33 to ship 6 tiny mosfets from CA to NY via UPS Ground. Damn.

 

[bobmcree]

Some customers may not have a CA, why not put a pot in the side of the controller, so your solution is stand alone?

that is certainly an option, but it is not quite as simple as just adding a pot. the current limiting in these controllers is a bit unusual as they use the error amp in the pwm chip for that and low voltage detection to save components and money, and there are very small voltage changes that result in large current changes.

 

[bobmcree]

I was about to place an order at IRTronix, but they wanted $33 to ship 6 tiny mosfets from CA to NY via UPS Ground. Damn.

i expect a large part of the $33 is a minimum order penalty. they only charge me $10 to ship 50 or 100 pcs. i think if you increase the order to $25 worth of parts the shipping cost will go down.


ordering 6 would indicate to me that you are either modifying a 20a controller or replacing 12 with 6 in a 35/40a unit. i'm not sure i would recommend either option. using only 6 in the bigger controller cuts the thermal path from the fets to the heatsink, and it should work ok, but is not optimal.

in all these controllers they turn on the low side fets for each phase drive cycle and do the pwm switching only in the high side fets. in the 35/40a units they follow the high side drive of the ir2101 driver with a transistor stage that provides enough drive current for the much greater gate load of the 4110, but in the 20a ones they do not. i would be concerned that the ir2101 cannot supply enough current to drive the higher capacitance of the 4110 gate, and that could result in slower switching time or possibly not even turning the high side fet all the way on in the time provided. both of these could make the controller operate with higher power losses than the original fets, and eliminate the advantage gained by using the better devices, or keep it from working properly.

i have not tried driving the 4110s directly driven by the 2101, but the numbers in the data sheet do not make it sound like a good option.

 

[Jozzer]

Interesting, I've been doing my 20A controllers with 4310's, is that a better match? (it does seem to be working well, but I have been thinking that it runs warmer than it did with stock fets)

 

[bobmcree]

Interesting, I've been doing my 20A controllers with 4310's, is that a better match? (it does seem to be working well, but I have been thinking that it runs warmer than it did with stock fets)

the 4310 is probably a better choice than the 4110 for the 20a controller, as it has significantly lower gate capacitance. you really should look at the gate with a scope and see how fast it is switching with the replacement vs the original fet. the required drive current is directly proportional to gate capacitance and time to switch, so if there is not enough drive the fet will never turn all the way on.

the way they lower the rds-on is to use more cells in parallel, so the lower resistance devices have much more gate capacitance

i

 

[giveahoot]

I was about to place an order at IRTronix, but they wanted $33 to ship 6 tiny mosfets from CA to NY via UPS Ground. Damn.

i expect a large part of the $33 is a minimum order penalty. they only charge me $10 to ship 50 or 100 pcs. i think if you increase the order to $25 worth of parts the shipping cost will go down.


ordering 6 would indicate to me that you are either modifying a 20a controller or replacing 12 with 6 in a 35/40a unit. i'm not sure i would recommend either option. using only 6 in the bigger controller cuts the thermal path from the fets to the heatsink, and it should work ok, but is not optimal.

in all these controllers they turn on the low side fets for each phase drive cycle and do the pwm switching only in the high side fets. in the 35/40a units they follow the high side drive of the ir2101 driver with a transistor stage that provides enough drive current for the much greater gate load of the 4110, but in the 20a ones they do not. i would be concerned that the ir2101 cannot supply enough current to drive the higher capacitance of the 4110 gate, and that could result in slower switching time or possibly not even turning the high side fet all the way on in the time provided. both of these could make the controller operate with higher power losses than the original fets, and eliminate the advantage gained by using the better devices, or keep it from working properly.

i have not tried driving the 4110s directly driven by the 2101, but the numbers in the data sheet do not make it sound like a good option.

You're right. I just tried now. Seems they have a handling charge below a $50 order. I would buy $50 bucks worth, but they only have 8 4110s in-stock. I'm going to give ebikes.ca a try, at $5 a pop. And thanks for the concerns and information. I was going to change out the blown IRF3205 55V fets in my 48V 20a controller (running 4x12V 12ah SLAs) with these. From what you're saying, perhaps I should go with a 75V part instead or go with the 4310s -- and beef up my heatsink.

 

[bobmcree]

[
You're right. I just tried now. Seems they have a handling charge below a $50 order. I would buy $50 bucks worth, but they only have 8 4110s in-stock. I'm going to give ebikes.ca a try, at $5 a pop. And thanks for the concerns and information. I was going to change out the blown IRF3205 55V fets in my 48V 20a controller (running 4x12V 12ah SLAs) with these. From what you're saying, perhaps I should go with a 75V part instead or go with the 4310s -- and beef up my heatsink.

i think the 4310 is a better choice for the 20a controller. justin paid digi-key prices for his fets so his price is not unreasonable for the 4110. i wish i had told him of the irtronix price earlier; he knows about it now, but he has a couple hundred he paid more for that he has to sell first.

 

[giveahoot]

i think the 4310 is a better choice for the 20a controller. justin paid digi-key prices for his fets so his price is not unreasonable for the 4110. i wish i had told him of the irtronix price earlier; he knows about it now, but he has a couple hundred he paid more for that he has to sell first.

Yeah, you're right. I just checked out the data sheets, and the 4310 seems to match up nicely. Thanks. I only wish I had access to a scope. And I certainly don't begrudge Justin for trying to turn a profit. I just wish Irtronix had more of the things I needed in stock.

 

[bobmcree]

i think the 4310 is a better choice for the 20a controller. justin paid digi-key prices for his fets so his price is not unreasonable for the 4110. i wish i had told him of the irtronix price earlier; he knows about it now, but he has a couple hundred he paid more for that he has to sell first.

Yeah, you're right. I just checked out the data sheets, and the 4310 seems to match up nicely. Thanks. I only wish I had access to a scope. And I certainly don't begrudge Justin for trying to turn a profit. I just wish Irtronix had more of the things I needed in stock.

when i bought 100 4110s today mike bought the 150 they had in stock at the factory, so i expect the other 50 will show up in online stock next week, and they will have another 2k on 10/17.

one must be cautious when substituting parts, and consider all the specs. it does no good to use a fet with a very low on resistance if the driver cannot charge up the gate capacitance in the time provided, just like it did me no good to put the 850 cfm Holley double pumper 4 barrel on my 57 chevy 283 until i put the cam and headers on it so that it could actually burn the gas and then get the exhaust out. (no i don't have it any longer; i sold it when i lived in san francisco after dozens of parking tickets made it too expensive to keep 2 cars there).

 

[fechter]

I was worried about the gate charge with the 4110's in the 35 amp controllers, but scope testing indicated there was no noticable slowdown of the gate signal with 4110's. It might be worth a test putting them in a 20 amp controller.

As I recall, the low side runs directly from the 2101 on both models, so those should be no problem. The high side also runs directly off the 2101 on the 20 amp controller, but the 35 amp model has a transistor booster in line. Since the 20 amp only needs to drive one FET, the 2101 might work fine to drive a 4110.

If the 4310's have sufficiently low on resistance that the controller doesn't get warm during normal operaton, then using the more expensive 4110's would be sort of a waste.

 

[Jozzer]

Do both controllers use te same driver? My 20A uses a 2101s, i was sent a 2101 by mistake from RS, and was surprised to see its twice the size as the 2101s.

 

[bobmcree]

both use the 2101s which is the surface mount part. without the s you probably got the dip mount part.

 

[bobmcree]

replacing one fet with another requires analysis of many parameters, and one cannot always replace another just because it has the same voltage rating. in our application it is of course preferable to have the lowest on resistance possible, but lower on resistance means a larger area and higher gate capacitance, requiring more current to charge the gate in the time provided.

the stock 4710 has a gate capacitance of 6160 pf, the 4310 has 7670, and the 4110 has 9620. the designers of our controllers used the cheapest configuration possible, and they found the need for a buffer stage to drive 2 of the 4710s with a total load of 12,320 pf.

the 4310 with only 15% higher gate load can probably be substituted for the 4710 in the 20a controllers without the buffer stage, but the 4110 with 50% higher capacitance is not quite as obvious as a drop-in solution. i have not heard from anyone who has measured the risetime of the gate on a 4110 driven directly by the 2101, but from a quick glance at the data sheets it looks marginal.

the drive capability of the 2101 will vary somewhat with temperature and voltage, so a single measurement does not necessarily indicate tht it will work under all conditions. in the 20a controllers the 4310 is probably a better solution, until someone does a detailed analysis of the performance.

i will not replace the fets in a 20a controller with 4110s at this point. in the next few weeks i will try to find the time to do one and report on the results if someone else does not beat me to it. if the fet switches more slowly there could be power losses that negate any benefit of the lower on resistance.

 

[eP]

replacing one fet with another requires analysis of many parameters, and one cannot always replace another just because it has the same voltage rating. in our application it is of course preferable to have the lowest on resistance possible, but lower on resistance means a larger area and higher gate capacitance, requiring more current to charge the gate in the time provided.

the stock 4710 has a gate capacitance of 6160 pf, the 4310 has 7670, and the 4110 has 9620. the designers of our controllers used the cheapest configuration possible, and they found the need for a buffer stage to drive 2 of the 4710s with a total load of 12,320 pf.

the 4310 with only 15% higher gate load can probably be substituted for the 4710 in the 20a controllers without the buffer stage, but the 4110 with 50% higher capacitance is not quite as obvious as a drop-in solution. i have not heard from anyone who has measured the risetime of the gate on a 4110 driven directly by the 2101, but from a quick glance at the data sheets it looks marginal.

You are wrong Bob. Gate input capacitance is not so critical as much as Miller one.
Lets look at Miller charges at marked green fields and lets look at rising /falling times.
The IRFB4110 is the same fast as 4710 and faster than 4310.

 

[bobmcree]

the miller charge determines how fast you can turn the fet OFF. i did not mention it because, as you pointed out, there is not a big difference in that parameter.

the driver chip still has to be able to supply enough current to drive the gate capacitance in the switching time provided. as i said, there are multiple parameters that must be considered when replacing one device with another.

have you replaced the fets in one of the 20a controllers with 4110's?

 

[Roy]

Thanks Bob!

Just a note from an owner of a 20A controller. I'm running at 48v and want to move to 60v and expect I'll decide to replace the stock FETs. This discussion is very informative.

Roy in Santa Clara

 

[bobmcree]

i will be the first to admit i do not know everything. sometimes when i read my own posts i am afraid i come off as a know-it-all who is just lecturing to hear myself talk. that could not be further from the truth. i am just trying to share the bits i do know in a forum where i will get feedback.

if you are going to 60v you might only need to use 75V fets like the irfb3077 which at only 3 milliohms has an even lower on resistance than the 100v 4110 we have been discussing, but i did not see them at irtronix and digikey wants over $6 for them. you are probably better off with the 4310 which i understand has been used successfully in these controllers, and at 20a the minor differences in on resistance will not be that big a deal.

irtronix has the 4310 for 2.42 in single piece qty but they do have a $50 minimum. maybe you can get several guys to go in together to get the deal on them.

 

[fechter]

I was worried about the switching speed when I first installed 4110's in the 35 amp controller. Looking at the gate signals with my scope showed it was still pretty fast. In actual operation, the dramatically reduced heating indicates the switching losses are not bad.

I have tried substituting FETs on other controllers in the past where the gate charge and capacitance were clearly too much for the driver and the switching waveform was seriously 'rounded' on the edges. This also resulted shoot through since it was a synchronous circuit and the FETs took too long to turn off. Thank goodness for current limited power supplies.

I think somebody's just going to have to try it and see what happens.

 

[eP]

the miller charge determines how fast you can turn the fet OFF. i did not mention it because, as you pointed out, there is not a big difference in that parameter.

No Bob. You are wrong again. The miller charge determines OFF and ON times as well.

It is a big difference. Very big DIFFERENCE. Look at the table above !!

the driver chip still has to be able to supply enough current to drive the gate capacitance in the switching time provided. as i said, there are multiple parameters that must be considered when replacing one device with another.

have you replaced the fets in one of the 20a controllers with 4110's?

Have you idea what are you doing when you replacing fets ?
Didn't you wrote before:
i agree the beta is low, but with a microsecond inductive switching time they were the best thing available at the time, and are fast enough for a switcher running at 100 kHz easily.

And now we see 200 ns switching time is poor for 17 kHz switcher.

So look againg at the table for miller charges and input gate capacitances.
You will see there are big differences from one fet to another.