My new 18 FET TO-247 layout riding video page 10

[Futterama]

I'm thinking I'll be using the ACPL-333J which already have optocouplers built-in. It's a bit expensive I think, but I just have to realize my original budget was a bit on the low side and get a better controller with a slightly larger budget

 

[HighHopes]

lewbowski, agreed
but for gate driver & motor drive power bridge, i would reword "this node is what I call 0V and I measure all the others with respect to this one"

to "this node is what I call reference and I measure all the others with respect to this one"

i trust you understand the subtly of the change

live4physics, yep & yep.
it is possible, my feeling, to build a ... moderately reliable, non-robust, cheap bootstrap controller up to 10kW if you knew what you were doing and expert level on layout for power electronics & power systems. i beleive it is hopeless endevour beyond 10kW. i/m talking continuous rating here.

 

[Alan B]

Earlier posting here was in wrong thread.

 

[liveforphysics]

if you want to go the 6 isolated supply route, make sure to combine this with 6 opto couplers for the PWM signals !

So many isolated gate drive chips exist today my friend.

 

[Alan B]

So I'm searching for info on inverter snubber caps, and one of the first papers that comes up is:

http://www.cde.com/tech/design.pdf

"Design of Snubbers for Power Circuits" by Rudy Severns

Which is interesting because I know him from his Ham Radio articles and work on Antennas. Sometimes it is a small world.

This paper looks a little dated, but the basics are probably unchanged.

How's the layout going??

 

[HighHopes]

all that info is not too useful for our purposes here. only figure 14 a is the type of snubber to use for >10kw. if you need other type of snubber because noise is too high or stress is too high, then your design/layout/geometry should be fixed before changing to other type of snubber.

on a more general note.. motor drives and especially gate drivers are not well known. you have to be very mindful of where you get your info from and who you trust. most app-notes out there on this subject are not useful or worse, misleading.

 

[zombiess]

So I'm searching for info on inverter snubber caps

Snubber cap should be polypropylene and pulse rated. On Digikey you can find them under application type "snubber" and "High Pulse DV/DT". Most of them will be from Epcos and Kemet. You'll need 3 of them, one per phase. I typically use 1.0uF because I have a lot of high quality ones on hand. Cost is 0.50 and up in small qty.

 

[Alan B]

Good to know. Thanks for the input. It looked a bit dated.

Generally multilayer ceramic capacitors are good at ripple current and high frequency response. Why are they not useful for snubbers at our lower voltages? Many of the snubber type caps are rated for voltages we don't require.

 

[zombiess]

Good to know. Thanks for the input. It looked a bit dated.

Generally multilayer ceramic capacitors are good at ripple current and high frequency response. Why are they not useful for snubbers at our lower voltages? Many of the snubber type caps are rated for voltages we don't require.

I'll try my best to answer from the little I know about caps.
Ceramic caps usually have a lower dielectric strength than polypropylene. I have not seen a pulse rated ceramic and a snubber has to handle high di/dt situation. I have not seen a ceramic cap with an inductance rating either, but pulse rated polypropylene caps will usually specify this in the data sheet, so I believe ceramic might have higher (or at least an unknown which is bad when trying to meet EMI spec) inductance which is bad for stopping EMI. Ceramic caps can also be fragile when subjected to thermal stress due to expansion and contraction (remember Methods experienced this when trying to use them as snubbers on the Xie Chang boards?). Polypropylene caps also exhibit a self healing ability allowing them to continue working after a non catastrophic fault in the capacitor.

I think I read something about ceramics having too low of an ESR which can contribute to ringing. Dielectric constant properties or something.

I guess the short version is Polypropylene caps have the electric properties most desirable for pulse operation.

 

[Alan B]

Thanks for your comments.

I have been running into a number of mentions in various papers that indicate MLCC caps may be a good or even better capacitor than film for these applications.

Here's one example:

http://www.murata.com/articles/ta06d2.pdf

I recall something about Methods getting some really big MLCC's for snubbers, not sure what happened there, or if the problem he had was related to the configuration which may not have been ideal.

Perhaps some MLCCs are not appropriate, the Murata units appear to be somewhat different than the usual bypass caps. I'll have to see if I can find more about Methods experience.

 

[HighHopes]

Alan, it really maters a lot if you are talking <10kW or >10kW. for less than 10kW, you can do whatever you want (ok, not quite but pretty close) and you will end up with something functional. use MLCC if you want (i have before, worked fine and is in flight today).

when we're talking >10kW, the margin of error is tight. it is not worth a conversation to say "i read this paper by a manufacturer that does not develop for this application where they said some stuff they know nothing about". who cares for that? it is mostly garbage what they write if they have no experience on high performance 3-phase inverter greater than 10kW. it is better to stick with what is proven and known to work (functional + reliable). don't waste your time reinventing the wheel, at least not until AFTER you have designed your working prototype, then you have the authority to do what you like. but until then, stick with the guidance of those that have success in this industryy. the margins are tight and it is not easy. if you go off and experiment with unproven concept you better know what you are doing or you will fail. how do you know this is true? look around the net, tell me how many DIY controllers are out there open source, high performance, >10kW ? there are not hundreds of examples, there are not even 10, i doubt there is not even 3. this is becaues it is NOT EASY to get something to be both functional AND reliable. So if it is not easy, then don't go off on your own tangent until AFTER you know what you are doing.

that is my advice, not that you were asking :wink:

 

[Alan B]

I appreciate the comments. Working on learning a bit beyond what works and into the why.

Working on a couple of different designs, will have something to show later on.

I think my first power level target will probably be what I can get from six 4468's. Keep it fairly simple to start with. I burned a 24 FET recently, and am running a 12 FET backup that is getting hot. But it appears to me that a reasonably well done set of six 4468's should do as well or better than a 12 FET Infineon, at least a stock one. So I'm aiming for a bit lower power than Zombiess is here. At least to start with. I'm also tinkering with a 12 each 4468 design but not happy with the options for handling the paralleling current.

Anyway, wondering how Jeremy's project is doing.

Thanks again!

 

[zombiess]

1 IRFP4468 = limited by package. Economy of scale comes into play based on device characteristics vs package limits vs price. It's gets really interesting to run the numbers for different voltage/current specs. HighHopes and myself have played this game a little bit.

I generally work of the package limits because it's usually pretty easy to find devices under 200V that have silicon that can exceed the package capabilities. The thermal path from the die to the outside world can have a large effect on this since not all devices have the same design, some having better thermal paths than others.

As for my own project, it's sitting on the shelf right now, power stage and drivers are built as shown in my pictures. I'm now waiting on boards to be manufactured to run Lebowskis chip to test everything. I needed something small enough to fit in my enclosure, so I had to do a complete redesign.

 

[Futterama]

A few posts ago you were talking about snubber caps. These are used with a resistor in series, right? From phase out to Vbus- and from phase out to Vbus+.

So the caps that goes directly from Vbus- to Vbus+, what are they called? There are 2 types, the big aluminium capacitors with high µF rating that supplies the current to the MOSFETs when they switch on, and the pulse rated film capacitor that takes up the energy spike when the MOSFETs turn off.

 

[HighHopes]

for >10kW, high performance, you would only use one style of snubber cap (if at all) and that is a poly cap with sufficient dV/dt rating and placed Vdc+ wrt. Vdc- as close to upper mosfet drain wrt lower mosfet source legs as possible. one per phase leg (not two in parallel), and never NEVER with a resistor, a diode, a... whatever, we want JUST the cap. you do not need the resistor if you have good layout and quality cap, the resistor just adds weight, volume, cost, heat, and lowers effectiveness of the snubber.

for DC link bus, you should read this thread if you have not already: http://endless-sphere.com/forums/viewtopic.php?f=30&t=31804

but not happy with the options for handling the paralleling current.

yep. that is a major concern when going to higher power using discrete mosfets.

 

[Futterama]

HighHopes, I'm just trying to understand what people are talking about when they use these expressions because they are new to me.

I have Lebowski's low inductance output stage in my mind, and the big aluminium cap is the "DC link" cap, right?

So the blue poly film capacitor Lebowski also added, is that what we would call the snubber cap? And from what you say, there should "really" have been 3 of them where Lebowski only added 1?

I have read the thread about input capacitance, but I haven't tried to calculate my needs, I just used the stuff about ESR and ripple ratings to select a capacitor and then I looked at how much capacitance a controller known to be reliable uses.

 

[Lebowski]

my 2 rappen: the blue one is the 'decoupling capacitor' .

A snubber typically is an RC series circuit (sometimes there's a Diode in there too),
so a snubber cap would be part of an RC(D) network designed to change the damping of a LC circuit.

P.S. my 3kW recumbent has the same setup as shown above, so only one 220uF (the black can) and one 470nF (the blue block)

 

[Futterama]

Black 220µF in the picture is "DC link", and the blue is "decouple"? So you have no snubbers as the ones HighHopes are talking about (the ones without resistor or diode)?

 

[Alan B]

Terminology varies, inverter terminology considers the blue pulse cap to be the Snubber cap. The snubbing resistance in inherent in the device and circuitry (very low).

 

[HighHopes]

ya i think Lebowski uses decoupling cap where i use snubber cap, different words referring to the same blue rectangular.

Lebowski shows the layout correct. do you see how the DC+/- cables come in? you can almost picture the energy flow goes from battery down cables and then first to the DC link capacitor (large electrolytic black capacitor) and then to the snubber and then to the mosfet. that is correct. more specifically, if the blue & black capacitors were swapped in location, it would not be correct anymore and the overall performance would suffer.

for future reference, the large electrolytic capacitor, i refer to as "The DC Link Capacitor" because it "links" the battery to the inverter. it does this linkage through an energy exchange, long math story short, it is the link of the reactive energy and also the supplier of quick pulses of current (batteries are too far away, so cap has to fill that purpose too). for quick pulses of current, i'm not talking about the current that goes to the motor, that is slow, i am talking about spikes of current that are much faster edge rate and is usually the reverse recovery diode and other physics things.

it's good if we can talk the same language, helps with learning & discussing different ideas.

 

[zombiess]

I did some testing of the current my gate driver can deliver. Setup is as follows.

I setup a 104mOhm shunt (10x 1.0 ohm resistors) in parallel with the gate drive output hooked up to a 1.0uF snubber cap. The boost stage has 2 10uF caps in parallel feeding it. The 1.0uF cap was used because it was handy and low enough ESR to get a high peak current. I need to parallel several small caps to see how quickly the current can rise. With a 0.1 Ohm gate resistor I managed 14.4A out in 570nS and 13.6A drained from the gate in 960nS which doesn't seem too bad for a giant capacitive load. Now I have some idea about the current rise rate (still needs more testing for rise time). It's no IXYS IXD-614 for speed, but it has some kick ass safety features integrated vs just being a bare gate driver.

On (purple trace is current and is channel B which is my math function.

 

[heathyoung]

Very interested in how this develops, and if it is possible to order boards from you Zombie - can't find anyone locally who can do a 4 layer board like this.

I'd probably want to go 200V parts for my application (20Kw+), not sure what the best FETs would be for this.

 

[zombiess]

Very interested in how this develops, and if it is possible to order boards from you Zombie - can't find anyone locally who can do a 4 layer board like this.

I'd probably want to go 200V parts for my application (20Kw+), not sure what the best FETs would be for this.

I now have a new buss design that fixes one of the major issues with this one not having a laminated buss. It also increases the current carrying capacity while maintaining > 90% overlap of the DC bus. I'm confident in this design based off of what I've seen work in commercial consumer setups around this same power level which have major design issues (but usually work OK).

There does not seem to be a lot of information available on how to do this well, I've looked all over. The information is very scattered or just non existent.

 

[HighHopes]

healthyoung, just out of curiousity, when you say you want 20kW+ is that a peak value for 30 seconds? if so, what is your desired continuous rating?

 

[heathyoung]

Peak value for acceleration - rather than continuous. With the existing controller I see this for about 8-10 seconds, usually by this time its going fast enough to warrant letting off the throttle!

Pack voltage is 144V, and sags badly to 100V@200A. New batteries will resolve this to about 135-140V @ 200A+.

Continuous is about 12Kw @ 105Kph, worse up hill or into a headwind.

I'm considering building one of these to replace the controller in a Vectrix, the IGBT's are a little temperamental and expensive when they go bad ($500). The takeoff current is nothing too exciting 50A battery side. I think this is due to the phase current limits of the IGBT @ 275A due to current multiplication at low rpm.