Strapping mosfets with schottky diodes... 18fet -> 12fet

[methods]

So in an 18 fet controller there are 3 phases
Each phase has 3 mosfets on the positive and 3 on the negative

I propose that if we swap one of those three mosfets with a high quality TO-220 schottky diode the overall temperature of the controller would go down allowing for a cooler running controller - especially at partial throttle.

The IRFB4110 has something like 4mohms rdson
Three in parallel makes for ~1.33mohms rdson
Ok - so if we pull one of those from each group of three we now have a 12fet controller with about 2mohms rdson right?

But... a great deal of the heat is really coming from the body diodes of the fets freewheeling all that nasty motor current around. This is part of why our big 100V controllers tend to run hotter at a 20A current limit than running wide open. Of course turning the fets on and off is obviously a huge part of that waste heat...

Anyway - the Vf of a Schottky is so much lower than the Vf of the body diodes that significantly less waste power will be developed on the heat sink allowing for a lot more system performance before overheat.

I recently saw this implemented in a cheap DC controller. They had just a single nChannel fet doing the PWM and a second TO-220 that was strapping it.

Anyone try this yet?

If I were not a lowlife I would try it right now.
Actually I am just sitting here driving 60A CC through my new LVC Breaker


-methods

 

[bigmoose]

The STPS30SM100S might be a candidate for low Vf. The thing to scope is the shutoff transient. The Schottky's sure shutoff quickly with near zero Trr. I would stick it in the middle slot of the 3 original FETs.

 

[methods]

I wish my good scope had not died
leaky cap syndrome


-methods

 

[scriewy]

if i wasnt a lowlife retard i would understand wus up here.

what is strapping a FET means ?

if the shotkey is in parallel with other 2 FETs, how the fact a shotkey can switch faster effects the FETs that still switch at their usual speed ?

 

[rkosiorek]

no need to buy a separate diode. just remove the gate drive resistor from one of the existing fets. then short the gate to the source of that FET. one fet now becomes a schottky diode.

rick

 

[methods]

But how good of one is the question.

-methods

no need to buy a separate diode. just remove the gate drive resistor from one of the existing fets. then short the gate to the source of that FET. one fet now becomes a schottky diode.

rick

 

[liveforphysics]

no need to buy a separate diode. just remove the gate drive resistor from one of the existing fets. then short the gate to the source of that FET. one fet now becomes a schottky diode.

rick

All the FET body diodes I test have twice the voltage drop of a schottky..

 

[Arlo1]

I have a Curtis 1204 dc controller I just modded. And this is what they do. I was really planing this for my controller build as a end goal. It really doesn't need to have many schottky diodes to work well. My theory is if the schottky diode is ~2x as good as a fet diode then 1 for every two fets will cut the diode heating in half in the fets themselves.
The big reason this is a great idea is because this means the fet body will stay much cooler and you can push them harder when wanted and like methy mentioned the part throttle cruise will not kill the controller as easy. Again this is a great thing for the hard to drive motors because the controller will be trying to current limit them even more which in turn is just harder for the fets themselves. I will be playing with this soon. IN fact I need to add 100v 150v and 200v schottkys to my next order.

 

[Teh Stork]

Check out this topic started by me. It takes up the problem with freewheeling diodes

 

[bigmoose]

The typical body diode is a minority carrier device, with a significant Trr (Reverse recovery time). A schottky is a majority carrier device with essentially a nil Trr. The body diode is a parasitic diode, caused by the epi process. They have been "tuned" over the years to have more benign characteristics.

A schottky will get better Vf (forward voltage drop) by about 60% with elevated temperature to 110 C. This is why it is hard to parallel them, but they will get "better" Vf running a little hot. They have horrendous reverse leakage currents, but the one I spec'd above is not too bad. Typical engineering you get something good, it comes with a little something bad...

 

[Lebowski]

What you're suggesting is a McGyver sticky tape kind of solution. It's a patch
which does not solve the fundamental issue.

Doing it properly means always either the high or low side is on, with
the diodes only conducting during the short deadtime (few 100 ns max)
both FETs are off .

 

[methods]

In my experience "doing it properly" never happens.

and if we are lucky... one time out of a hundred someone can actually get the McGyver tape to stick. To me this hacker with the sticky tape is a hero.

My standards are low friend

So if we can do anything to these 18 fet controllers to get a 20% increase in performance then I will actually start doing it on my production line... instead of just talking about the theoretical optimal solution that will never happen.

The prior is my goal and the latter is just intellectual masturbation.

-methods

What you're suggesting is a McGyver sticky tape kind of solution. It's a patch
which does not solve the fundamental issue.

Doing it properly means always either the high or low side is on, with
the diodes only conducting during the short deadtime (few 100 ns max)
both FETs are off .

 

[bigmoose]

To me this hacker with the sticky tape is a hero.
My standards are low friend

Perhaps not so low, my friend! While sitting in a conference room last week, I heard the following statement (actually a proposal if pearl harbor happened in our little box): "Kapton tape is space's Duct Tape!"

... we were discussing if adding a layer of kapton would protect board B, after board A went thermal and blew a hunk of silicon through the conformal coating... sh_t happens, and everyone needs some sticky tape!

Methy I love your get it done now, with what I have on hand, or die attitude!

 

[rkosiorek]

so just so i understand this.

1. the body diode of a FET actually make a terrible diode.
2. these diodes are supposed to help with the shoot thru problem, but because the body diode is so crappy it may actually contribute to it at part throttle.

the best solution would be to carefully tune the dead time between the high side and low side fet on times. but as a bandaid brushed controllers add a separate schottky in parallel with the fets.

the question being asked is whether or not this same bandaid can be used to help brushless controllers.

i have tried understanding the discussion, but i'm shy on a lot of the theory. i cannot determine if the question was ever answered. also why would not commercial controller designs include such diodes if it does work?

would it be work adding such diodes? or would the answer depend on knowing more specific information about the controller design and motor being used?

rick

 

[dnmun]

http://en.wikipedia.org/wiki/Power_MOSFET#Body_diode

 

[Arlo1]

so just so i understand this.

1. the body diode of a FET actually make a terrible diode.
2. these diodes are supposed to help with the shoot thru problem, but because the body diode is so crappy it may actually contribute to it at part throttle.

the best solution would be to carefully tune the dead time between the high side and low side fet on times. but as a bandaid brushed controllers add a separate schottky in parallel with the fets.

the question being asked is whether or not this same bandaid can be used to help brushless controllers.

i have tried understanding the discussion, but i'm shy on a lot of the theory. i cannot determine if the question was ever answered. also why would not commercial controller designs include such diodes if it does work?

would it be work adding such diodes? or would the answer depend on knowing more specific information about the controller design and motor being used?

rick

You almost have it. So in a brushed controller they dont always use an H bridge the simple controllers methy and I are talking about are for series wound motors and reversing the output of the controller will not reverse the motor it self so all the fets inside are parallel, and parallel to the fets there is some schottky diodes to help keep the stress off the fets them selfs.

Now I think I understand this correctly in a H bridge on a brushless controller you can turn on the opposite set of fets during the off times in PWM to save the diodes some stress. Example If you have the Hi side fets driving the motor with PWM the low side in the same H bridge can turn on when the hi side turns off. If you don't do this then the diodes in the low side fets will flow the current through the fets so its better to turn on the fet it self.

 

[Alan B]

Would the Schottky diodes, due to their high speed, add to the noise/peak current from the caps?

Would you have to add low L caps to handle the Schottky currents?

Will those 30A schottky diodes handle enough current for this application?

 

[methods]

This is where Fechter should pop up and remind us that one test is worth a thousand opinions.

If I had not unwisely given all of my 18 fet controllers to liveforphysics I would just wire one up and do a side by side test. We could just load each of them for 5 minutes with a load profile then measure delta T. Seems the results would surface quickly. One would get hotter than the other.

I have a batch of 18 fets on order - when they arrive we will swap out the middle fets (good idea Moose) and just run the experiment.

I am surprised that nobody else has tried it yet.... thought I was going to get off easy.

-methods

 

[Farfle]

Excuse my ignorance, but does putting in the diode require removing the fet?

 

[dnmun]

double down on the ignorance here. where does this current or voltage come from?

how do you know what voltage diode to use and current carrying capacity. in the mouser catalog the schottky diodes are also axial and could you use an axial diode jumpered above the mosfets with the legs spanning the FET?

along with the 'rectifying current' the diode specs in the mouser catalog also include a big transient current spike in the spec and i wondered how is that used?

what are you calculating to arrive at that number? thanks for explaining, this is real EE.

 

[Thud]

I am "all in" on the ignorance card!

Link me to a suitable T0-220 schotkey & I'll order up a set.....I have 5-18fet kits sitting in the box & have been waiting for time to assemble them up.....I tend to lean to the 3077 fet's though...they seem to be better for < 75v applications & I have no need to run 100v in my higher kv motors....but I was planning to build at least 2 of them with 4110's to play with at 20 cells.

Now lets see 5th street. (I am pretty sure I am the fish at this table)

 

[methods]

double down on the ignorance here. where does this current or voltage come from?

So in a really simplified explanation... lets look at a 6 fet controller. Picture one of the windings in the motor. Say it is connected between the YELLOW and GREEN phase wires. The controller has the ability to send DC current from Yellow to Green, and from Green to Yellow. It does this by having 2 fets on each of the phase wires. One fet can tie the phase to VCC and the other can tie the phase to GND

So normally all the fets are off
If you want to hook VCC to the yellow wire and GND to the green wire then you turn on the respective fets... that would be the fet between VCC and Yellow and the fet between Green and Ground

You can then reverse the direction of the current by turning on the fet between VCC and Green and the fet between Yellow and Ground.

So in a nut shell - that is how a brushless controller works. The computer decides the timing and the phases are hit in different combinations in different directions... the details of which are not that important here. What is important is the basic principle that once a DC current starts flowing through an inductive path it does not want to stop. So... if you turn on one of these phases and then instantly attempt to turn it off you get a huge backlash voltage. This can be seen by simply taking 10 feet of wire and using it to short circuit a 100V Lipo battery. WHen you attach the wire nothing will happen - maybe a tiny spark. WHen you attempt to remove the wire you will pull a HUGE spark that can jump an inch or more.

Same way a car works to create a spark in a sparkplug

So this backlash voltage can easily be 10X the battery voltage (or much higher) and we have to do something with it. A parallel example is when you drive the coil of a relay - you are always supposed to hook a "freewheeling diode" across the relay. This is so that when you attempt to stop driving the relay coil the backlash voltage does not damage your drive circuit. This diode in parallel gives that voltage a virtual short circuit back to its reference and it quickly dies out.

Now our fets have what are called "body diodes" - these are parasitic diodes (a byproduct really) - that happen to work very well for us. THey allow these big backlash currents to circulate and they protect the components in our controllers. The problem is that they are not very good diodes... they have a large forward voltage... so as these currents flow a great deal of power is dissipated. V*I=W

Most people do not realize that a great deal of the heat developed in their controllers is actually from these body diodes trying to short circuit all this "noise".

This is part of why setting a current limit low - at say 20A - can sometimes result in an even hotter controller than setting the current limit higher.

So now to the point -> Voltages in parallel are always equal so if we place a very high quality diode in parallel with the body diodes of our fets we will have a lower forward voltage and there for a lower power loss on our heatsink and in our fet bodies. These schottky diodes also have very good speed response so they dont take long to turn on and they dont stay on longer than we want them to.



how do you know what voltage diode to use and current carrying capacity. in the mouser catalog the schottky diodes are also axial and could you use an axial diode jumpered above the mosfets with the legs spanning the FET?

Since these are not really being used like zener diodes (the way we use TVS diodes) it does not really matter what voltage you select so long as it is much higher than your operating voltage. We are going to use these forward biased. You are thinking of how we would wire in a TVS diode (transient voltage suppressing) and those diodes act like a very fast zener diode (reverse) and you would in that case select one that is just below the maximum voltage of your fets such that they will never "see" anything over their rated voltage.

You would not want to use an axial diode.
The reason is that these guys are going to create a ton of heat and you want to be able to dump that heat into the heat sink. Otherwise they are just going to end up overheating and popping.



along with the 'rectifying current' the diode specs in the mouser catalog also include a big transient current spike in the spec and i wondered how is that used?

I would just pick the biggest one I could find in a TO-220 package. Probably not one of the plastic packages either... I would want a real deal metal tab. The data to look at is really the data that shows how efficiently the diode can dump heat into its heat sink from its silicon.


what are you calculating to arrive at that number? thanks for explaining, this is real EE.

No calculations needed really. In this case bigger is better - so we want to just fit the beefiest diode we can in there and get it good and heat sunk to the case. That is the first order goal - the second order goal (if you are good like Moose) would be to carefully select the diode that has the best turn on and turn off behavior. Typically though... if you just do a little research and find the expensive ones... those will do the job quite nice Kind of like choosing an IR fet -> IR just makes good parts.

I am trying to get a sampler platter of controllers from 3 different sources right now and as soon as someone actually sends me something I can do some testing.

Anyhow - My explanation is somewhat simplified (and not 100% correct) but I hope that it illustrates what we are trying to do. Pop open a controller and confirm what I said about the fet layout, the positive rail, the ground rail, and draw a little sketch of what is going on in there.

In hind sight I should have given a DC controller example - MUCH easier to explain.

-methods

 

[methods]

How about you just send Matthew and I two of those kits

You could do some experimenting... I would suggest building two identical controllers with the exception that one of them would have the center fet replaced with really good diode. You may have to play games and get sketchy here... one the fets that go to VCC you woudl want the cathode going to VCC and the anode going to Phase. On the fets going to ground you would want the cathode going to Phase and the anode going to Ground. I have not drawn this out yet or compared it to any parts data sheets.

You would definitely want to keep the legs as super short as possible.

You know what... to do a fair test.... you might want to build three controllers:

1 built normal
1 built with only 2 fets per channel (so it would be a 12 fet)
1 built with only 2 fets per channel + Diodes (so it would be a 12 fet also)

We would then need a repeatable test you can perform - like a 2 minute ebike run - but you would have to be careful to keep it consistent.

BIg Moose offered up a pretty good part. Pumping this number into Digikey (STPS30SM100S) results in some TO-220 options. Proceed with caution though - you can t just plug these buggers in - you must identify the Anode and Cathode and do some thinking about how that will line up wit the controller PCB.

Build up those controllers first and find something that gets them hot! Then we can try this mod and see if it helps.

I have 5 cromotors in the mail right now. I have decided to use one to revive my KMX Trike. Going to do a 20" tire and 100V 100A -> It is going to frigging be insane. Matt better watch out, I am going to try and chase down his yellow buggy.

-methods

I am "all in" on the ignorance card!

Link me to a suitable T0-220 schotkey & I'll order up a set.....I have 5-18fet kits sitting in the box & have been waiting for time to assemble them up.....I tend to lean to the 3077 fet's though...they seem to be better for < 75v applications & I have no need to run 100v in my higher kv motors....but I was planning to build at least 2 of them with 4110's to play with at 20 cells.

Now lets see 5th street. (I am pretty sure I am the fish at this table)