controller PCB, work in progress

[bigmoose]

I got a quote in today on the MMIX1F420N10T "super package" Fets. Minimum order quantity is 20 and they are going to run around $26 apiece + shipping and probably a packaging fee for information. I don't think the package is worth a fet cost of $180 for a really nice 3 phBLDC controller... I was hoping they would be in the $9 to $13 range.

[Arlo1]

I got a quote in today on the MMIX1F420N10T "super package" Fets. Minimum order quantity is 20 and they are going to run around $26 apiece + shipping and probably a packaging fee for information. I don't think the package is worth a fet cost of $180 for a really nice 3 phBLDC controller... I was hoping they would be in the $9 to $13 range.

Yeh they look awesome but not worth that. I still think bang for buck the IR 4468-4668 series are the way to go. I can get them ~$4 each. But I wonder it might be worth setting up a program and a fet test bench with one of my pic18f chips and a fet driver to learn how to test to the limits. It might be an easy way I could wind up an inductor = to the motor I want to test and just pump current into it.... Maybe on a Simple brushed sample code... Hmmmmm

[Lebowski]

2d.jpg (233.02 KiB) Viewed 307 times

3d.jpg (164.05 KiB) Viewed 307 times

[Arlo1]

Looking great man. I have a feeling this will a liquid cooled heat sync and 4568 fets and my BMX could make more power then my 24 fet thats on it with 24 4110 fets. All of that in something 1/4 the size!!!

[liveforphysics]

If your caps aren't local to the source/drains, you may as well not have them.

Just like effects at the Miller plateau, it's another thing that is totally invisible until the moment you're switching current.

[Jeremy Harris]

If your caps aren't local to the source/drains, you may as well not have them.

Just like effects at the Miller plateau, it's another thing that is totally invisible until the moment you're switching current.

That mirrors my experience. I found that I needed to fit the caps right at the FETs on each phase, as close as I could physically get them. Even then I still had spikes, which were only reduced to a manageable level by putting higher frequency capable, lower value and ESR, caps virtually at the FET pins (the 0.1 µF caps on my boards are 4 mm from the FETs, the electrolytics are just under 10 mm away). Having the caps a few tens of mm away from the FETs means more stray inductance, and probably explains why the snubbers were needed (my view is that the snubbers may well not be needed if the caps are right up tight to the FETs and the tracks are nive and wide/thick).

[liveforphysics]

Agreed.

Making it look pretty on the scope when not under current is kinda a pointless exercise. Perhaps setup a dummy load, maybe just a hubmotor with it's rotor locked (or removed), then switch and you will have something worth watching on the scope.

[Arlo1]

2d.jpg

3d.jpg

It would be realy easy to reposition the fets to have a cap inbetween the Hi and low fet on each H bridge. It would make the controller WAY more better'er

[texaspyro]

It would be realy easy to reposition the fets to have a cap inbetween the Hi and low fet on each H bridge. It would make the controller WAY more better'er

Mostest definitely...

[Lebowski]

I ordered all the components today to build 2 controllers. One is going to be the equivalent of what I have
on my bike now, 150V components and 20A current sensors.

The second one is going to be 150V but will have 150A current sensors To get rid of the dissipation
in the (6 milli-Ohm) 4568's I also included a blower in the order. I got some ideas about how to build
my own heatsink to use in combination with the blower, it'll be cool (I hope ). Still, for 120A amplitude
of phase current the dissipation will be 130W, I hope the blower and homemade heatsink will do the trick...

Once the components are in I can measure all the wires thicknesses (of the more exotic components like the power
Schottky's, current sensors etc). With that info I can check all the footprints on the PCB, fill the top and bottom
with groundplane and send the PCB's off to be made.

[gensem]

I ordered all the components today to build 2 controllers. One is going to be the equivalent of what I have
on my bike now, 150V components and 20A current sensors.

The second one is going to be 150V but will have 150A current sensors To get rid of the dissipation
in the (6 milli-Ohm) 4568's I also included a blower in the order. I got some cool ideas about how to build
my own heatsink to use in combination with the blower, it'll be cool (I hope ). Still, for 120A amplitude
of phase current the dissipation will be 130W, I hope the blower and homemade heatsink will do the trick...

For instance, what would be the dissipation of 120a phase using 4468 fets?
Either way a good computer cooler should be able to handle the dissipation, but you ll need something big for 130w.

[Lebowski]

I ordered all the components today to build 2 controllers. One is going to be the equivalent of what I have
on my bike now, 150V components and 20A current sensors.

The second one is going to be 150V but will have 150A current sensors To get rid of the dissipation
in the (6 milli-Ohm) 4568's I also included a blower in the order. I got some cool ideas about how to build
my own heatsink to use in combination with the blower, it'll be cool (I hope ). Still, for 120A amplitude
of phase current the dissipation will be 130W, I hope the blower and homemade heatsink will do the trick...

For instance, what would be the dissipation of 120a phase using 4468 fets?
Either way a good computer cooler should be able to handle the dissipation, but you ll need something big for 130w.

For me it's a mute point as hjns (for whom I'm building the 2nd controller, we'll try it out on his bike) likes
to ride around with 30s (so > 120 V). But based on the 2.6 mOhm versus the 6 mOhm of the 4568's, less than half
of the 130 W.

P_diss_for_each_FET = i_amplitude^2 * R_ds_on / 4

4468's are logic level though, I'm not a big fan of that as they'll be more difficult to keep off during switching events.

I'm hoping lots of airflow through a square pipe with internal laminations will do the trick but it'll be a case of slowly
upping the phase current and seeing what happens....

[Teh Stork]

Uhm, I use this for each fet:

P_dissipation_each_fet = I_rms^2*On resistance

I_rms = I_phase /1,41.

Plugging in the values gives me 110W of dissipation total for a 6 fet bridge with 2,6mOhm on resistance running 85A battery current (120A phase).

I've been quite uncertain as to what to plug in, but running FOC this seems to reflect the reality nicely. I might also mention that running a FOC algorithm over a six-step commutation decreased my current ripple by 60-70% depending on load.

[Lebowski]

I reason that the I_rms^2 * R_ds_on is the power dissipated in both FETs connected to the
specific motor phase, so for each FET it is halved. The momentary current from a motor phase
is either going through the high side or the low side FET but never through both at the same time....

[Arlo1]

I reason that the I_rms^2 * R_ds_on is the power dissipated in both FETs connected to the
specific motor phase, so for each FET it is halved. The momentary current from a motor phase
is either going through the high side or the low side FET but never through both at the same time....

But its going thought a Hi side or low side on another H bridge at the same time.

[Teh Stork]

I reason that the I_rms^2 * R_ds_on is the power dissipated in both FETs connected to the
specific motor phase, so for each FET it is halved. The momentary current from a motor phase
is either going through the high side or the low side FET but never through both at the same time....

As Arlo is saying - lets say carrying Current I across one motor winding will have to go in with one fet - and out the other. If one fet delivers current to two motor windings - it has to deliver twice the current.

Say you have:
101
010

Each of the upper fets would carry the current going through the windings. The bottom fet will carry twice the current.

I could be wrong as I've seen several different "loss equations" in different application notes.

[Lebowski]

I reason that the I_rms^2 * R_ds_on is the power dissipated in both FETs connected to the
specific motor phase, so for each FET it is halved. The momentary current from a motor phase
is either going through the high side or the low side FET but never through both at the same time....

But its going thought a Hi side or low side on another H bridge at the same time.

Well, yes and no For the typical china stuff, you're correct. But in a 3-phase system with 120 degree
phase differences where all 3 wires are powered it is common to look at it differently. In a Y configured
system the 3 voltages are said to power the 3 loads, with the return current (from the Y center point)
being 0. A Delta configure system works the same but for the understanding it's difficult to define
a central current return point.
The way I look at it, each full bridge is powering one coil with no return current path, like in the
3 phase power grid. The full bridge always has one of the FETs on, it effectively generates a
voltage equal to the dutycycle * V_bat. For no power delivery the dutycycle is 50% (on all phases).
looking at one phase, for a small positive signal the dutycycle is 60%, so 60% of the time the current goes through the
high side FET and 40% of the time through the low side FET. 180 degrees later the dutycycle is
40%, now the high side conducts the current 40% of the time (60% for the low side). On average
each FET carries current 50% of the time.
Therefore the losses of one motor phase are shared over both FETs.

In Stork's example, what's actually happening is that one phase has 100% dutycycle while both
other ones have 25% dutycycle (assuming a center point which is standing still)

[Lebowski]

got a big box full of goodies from Digikey yesterday Next job is to measure the more exotic
components to make sure the footprints and drill sizes are OK, finish up the PCB and send it off
to be made.
I'm kind of stubbornly holding on to the design I have done so far. I think with the snubbers, extra caps
practically on the FETs are not necessary, even under load. I will test this with the high power version
I'm building. If things DO go wrong without extra caps I can do a detailed investigation and learn what
is going on, improve my own knowledge. Plus, I can always add extra cap on the bottom side of
the PCB so it will not be a waste of an expensive PCB.
And who knows, I may be right and have it work with the design I have now

[Jeremy Harris]

The Chinese controller do OK with poor capacitors dangling on long leads some distance from the FETs, but then they have pretty slow turn on and off times, as they are only driving the FET gates with around 125 mA at the most (the gate drive resistor on the Xeichang controllers is typically 91 ohm, with a nominal 12 V bootstrap voltage and a series diode drop to the FET gate).

My comments about having commutation caps as close as possible to the FETs were based on a fairly high power design (80 A at 60 V, so around 4.8 kW) intended to drive a fairly low LR motor (the 80-100 180 Kv) so may not be as relevant for a controller delivering less power to a higher LR motor, like a hub motor. It's also a pig to get the big capacitors in between the FETs and still keep the size of the board down, I found.

[Ratking]

The Chinese controller do OK with poor capacitors dangling on long leads some distance from the FETs, but then they have pretty slow turn on and off times, as they are only driving the FET gates with around 125 mA at the most (the gate drive resistor on the Xeichang controllers is typically 91 ohm, with a nominal 12 V bootstrap voltage and a series diode drop to the FET gate).

My comments about having commutation caps as close as possible to the FETs were based on a fairly high power design (80 A at 60 V, so around 4.8 kW) intended to drive a fairly low LR motor (the 80-100 180 Kv) so may not be as relevant for a controller delivering less power to a higher LR motor, like a hub motor. It's also a pig to get the big capacitors in between the FETs and still keep the size of the board down, I found.

A bit O/T but how do my cc hv160 survive driving a 80-100-130kv without letting out the magic smoke?
I've got two and they seems to handle it well. The controller itself is tiny and the fets are tiny. Nothing on that controller seems solid enough for driving anything else than a propeller. But it does, and fairly well too.

[Jeremy Harris]

A bit O/T but how do my cc hv160 survive driving a 80-100-130kv without letting out the magic smoke?
I've got two and they seems to handle it well. The controller itself is tiny and the fets are tiny. Nothing on that controller seems solid enough for driving anything else than a propeller. But it does, and fairly well too.

I believe that the CC ESCs use phase current sensing, plus the FETs are physically arranged very close together with the commutation caps as close to the main buss on each of the power boards as it's possible to get. Worth noting that the CC controllers use lot's of paralleled FETs, too. I'm not sure about the HV160, but it's not at all uncommon for RC ESCs of this size to have at least 32 FETs, sometimes more, to get the on resistance right down and limit resistive heating.

[texaspyro]

I'm kind of stubbornly holding on to the design I have done so far. I think with the snubbers, extra caps
practically on the FETs are not necessary, even under load.

Ahhh yes... the ole no time to do it right, always time to do it over.

[Lebowski]

I'm kind of stubbornly holding on to the design I have done so far. I think with the snubbers, extra caps
practically on the FETs are not necessary, even under load.

Ahhh yes... the ole no time to do it right, always time to do it over.

have a little faith plus if it does go wrong it'll be a good learning expierence for me.
Don't forget I use snubbers which the advocates of super-duper capacitors do not have. Maybe its
a case of one or the other will do the trick equally well.

Or Luke has to invite me over to have a look at his setup (hint hint)

[texaspyro]

have a little faith plus if it does go wrong it'll be a good learning expierence for me.

Moi? Have faith? Surely you jest?

It will work, but could probably work better. And, I've never met a snubber that wasn't out to kill me (or at least what it was trying to protect). They are all evil, sneaky, slimy little bastards just waiting to let your magic smoke out at the worst possible time.

[Lebowski]

I think I'm done monday the files go to the PCB manufacturer for 2 PCB's.
I measured all component legs for hole diameters. DRC and LVS correct !

2d.jpg (218.24 KiB) Viewed 260 times

3d.jpg (138.47 KiB) Viewed 260 times

front and back layers together

front_back.jpg (189.09 KiB) Viewed 260 times

front metal layer plus (in pink( the solder island area. Notice the large windows for adding thick copper wires to the lines between the FETs and the current sensors / motor wires

front_window.jpg (201.51 KiB) Viewed 260 times

back metal layer with large solder island openings for the main battery supply lines. Note how the power ground and processor ground are star-connected at the battery connection.

back_window.jpg (192.49 KiB) Viewed 260 times