Gate drive wiring / General Science and Electronics / Forums

Forums

4hv.org :: Forums :: General Science and Electronics

« Previous topic | Next topic »

Gate drive wiring

Move Thread

LAN_403

Marko

Tue Jul 17 2007, 10:12PM

Registered Member #89 Joined: Thu Feb 09 2006, 02:40PM
Location: Zadar, Croatia
Posts: 3145

I'm in dilemma again, and my last thread is still corrupt

I need to provide gate drive to 8 mosfet's (2 paralleled half bridges).

After lots of digging trough archives about hardcore PCB and wiring design, I think I can do better than a bunch of twisted-pair wires :p

I kept the GDT on control board since I think it will be very benefical in matter of primary leakage inductance.

My general fashion is to use ribbon cables for interconnection. When I bridge all connector pins in line I have current flowing in opposite direction in each next wire making an excellent low-inductance connection. It also looks very neat and is easy to wire.

But now, I have a rather special bridge design built into a homemade heatsink made f aluminium plates, and I can't mount connectors on bridge side but would have to solder or screw them which loses all the elegance when I need to do it with 80 thin wires.

Also, as I never preferred to use less than 10-wire cable per device I fear the huge bunch of cables sticking out everywhere.

Now I'm curious, how much better coaxial cable is inductance-wise in this case than ribbon?
Should I prefer coax?

Now, another problem, I can solder the cables on bridge side, but how to have budge of 8 coaxes connected to control board and still have them relatively easy to disconnect?

Soldering is not an option since I don't want to desolder them all each time I need ot take the board off. SInce board is perforated screw-pads also won't work too well.

I had little of idea how to do that while retaining (at least to a point) neatness and elegance.

Now, any ideas? suggestions? flames? (flames are good as long as I haven't started anything)..?

Sulaiman

Tue Jul 17 2007, 10:39PM

Registered Member #162 Joined: Mon Feb 13 2006, 10:25AM
Location: United Kingdom
Posts: 3140

I think the most obvious answer is to move the GDT to the bridge.

Better a little primary leakage inductance (not much if you use twisted-pair or co-ax)
than several secondary leakage inductances.

Marko

Tue Jul 17 2007, 10:56PM

Registered Member #89 Joined: Thu Feb 09 2006, 02:40PM
Location: Zadar, Croatia
Posts: 3145

Sulaiman wrote ...

I think the most obvious answer is to move the GDT to the bridge.

Better a little primary leakage inductance (not much if you use twisted-pair or co-ax)
than several secondary leakage inductances.

That's just not possible because of the way bridge is designed (dug into aluminium with a small hole for output connections.).

If I moved the GDT I would still have to make (ugly) cable connections between it and each mosfet, same as I do now, and it would be even more work.

There would be little difference except GDT would be 10cm farther from control board.

Even on my last bridge I ran into same problem and that's why I want to keep GDT on control board now. I can also keep gate resistors and diodes on board that way...

ConKbot of Doom

Tue Jul 17 2007, 11:18PM

Registered Member #509 Joined: Sat Feb 10 2007, 07:02AM
Location:
Posts: 329

Firkragg wrote ...

Now I'm curious, how much better coaxial cable is inductance-wise in this case than ribbon?
Should I prefer coax?

Coax should be a good bit better then ribbon cable.

RISI makes exploding bridge wire detonation systems, Low inductance is definitely critical for those, and according to that paper there, using their coax vs their 'twinlead' cable you can achieve 3x the distance because of less inductance. The twinlead is pretty much high voltage zip cord, nothing too fancy, not twisted, etc... So yes, coax would have a good bit less inductance then a ribbon cable. A return path on either side of the conductor in the ribbon cable would be an improvement over just a pair though.

Dr. Slack

Wed Jul 18 2007, 07:36AM

Registered Member #72 Joined: Thu Feb 09 2006, 08:29AM
Location: UK St. Albans
Posts: 1659

The bottom line is that ALL connections are transmission lines. It's just that coax is a well controlled characterised line, twisted pair and ribbon cable are less so (but still not bad), and single hookup wire changes every time something moves.

You only have to consider the transmission line aspects when the wire length exceeds perhaps 1/10 wavelength, so for TC use that's the secondary wire, every other compoenent and connection can be considered to be lumped.

Transmission aspects aside, it is worth using coax in certain places because the impedance, and thus the inductance per unit length and the pacitiance per unit length are well defined, and usually lower than you can acheive with open wires. A lower impedance line will have lower inductance. 50 ohm is ubiquitous, 35 ohm can be had, much lower is pretty rare. However, if you put several lines in parallel, the total impedances behave like Ohms Law, so 4 paralleled 50 ohm lines look like 12.5 ohms, as every high power VHF amplifier designer knows from their line-wound transformers.

You can make your own transmission line connections, designing it from scratch for a particular impedance. I came across some plastic coated copper tape that was used for the screen turns on power transformers. At 1" wide and 50um plastic thickness, a ground/signal/ground stack would give stupidly low impedance.

Steve Conner

Wed Jul 18 2007, 08:30AM

Registered Member #30 Joined: Fri Feb 03 2006, 10:52AM
Location: Glasgow, Scotland
Posts: 6706

I've tried a bunch of different things for transmission lines. My personal favourite for gate drive is the ribbon cable that Firkragg mentioned. I think a ribbon cable pair has a Zo of about 100 ohms, so 10 parallel pairs should be around 10 ohms, which is 5 times better than coax at 50 ohms.

On my DRSSTC, I used gate drive transformers with primaries made of RG-213 miniature coax. I also used the coax to connect the GDTs to the driver board. The shield was just being used as a Faraday shield to protect against EMI and flashovers, though. It wasn't part of the transmission line.

The only downside to ribbon cable is the voltage rating. For my OLTC2 I managed to find some coaxial speaker cable with a 2.5 sq.mm. centre conductor and an equally heavy outer braid. I used 12 runs of this in parallel to connect the pulse generator bit to the primary coil.

Marko

Wed Jul 18 2007, 02:02PM

Registered Member #89 Joined: Thu Feb 09 2006, 02:40PM
Location: Zadar, Croatia
Posts: 3145

Hi steve

You guys keep confusing me about these cable impedances. What is the 50ohm impedance of a coax cable anyway? It's characteristic impedance, sqrt(L/C)? Since I'm driving a large capacitive load I think capacitance of cable is unimportant here, and inductance is value I need to know.

I've tried a bunch of different things for transmission lines. My personal favourite for gate drive is the ribbon cable that Firkragg mentioned. I think a ribbon cable pair has a Zo of about 100 ohms, so 10 parallel pairs should be around 10 ohms, which is 5 times better than coax at 50 ohms.

Hey, this shifts favor for ribbon again.

I somehow intuitively thought that coax has to have much lower inductance that ribbon because of it's construction, but 10-lead ribbon actually beats coax by 2.5?

And where is the mini coax in comparison to this? I would expect it to have lower impedance than 50 or 75 ohm coax.

Another thing, I maybe wasn't clear about: I was recommended by everyone, to use a bifilar SSTC primary coil with each on it's own separate bridge to assure current sharing. I don't know how much sense it makes since mosfet's are PTC, but I still redesigned the bridges that way.

But, since bridges are now separate I see I also need to split the gate drive to each mosfet to prevent dU/dt differences from blowing the gates up.

That's why I in the end use 8 gate drive signals. I'm not sure how much this makes sense in overall, though. I hoped to see some ''proffesional'' opinions

Dr. Slack

Thu Jul 19 2007, 08:53AM

Registered Member #72 Joined: Thu Feb 09 2006, 08:29AM
Location: UK St. Albans
Posts: 1659

You guys keep confusing me about these cable impedances. What is the 50ohm impedance of a coax cable anyway? It's characteristic impedance, sqrt(L/C)? SInce I'm driving a large apacitive load I think capacitance of cable is unimportant here, and inductance is value I need to know.

Yes. The characteristic impedance is sqrt(L/C). For oridanry plastic cable, the speed of light is in the 0.7c region, which fixes the LC product. For cable built to 50ohm dimensions, the C per m is about 100pF, and the L per metre about 250nH. So if you know the impedance of a plastic insulated cable, you know the inductance per unit length without any more calculation.

As your driving frequency is low, the line is therefore short, it is more appropriate to consider it as a lumped C, which you add to your load, and a lumped L in series, which you need to drive.

And where is the mini coax in comparison to this? I would expect it to have lower impedance than 50 or 75 ohm coax.

Mini coax tends to be made for the same 50ohm impedance as regular size coax. We use some stuff only 0.5mm O/D for microwave baluns, but it's still 50ohm, so you can design with it and interchange it with larger sizes. The impedance depends on the ratio of inner O/D and the outer I/D - go google some transmission line theory. You may well find 35ohm and 25 ohm lines available - used in making matching transformers to and from 50ohm, but it's not common.

Flat ribbon cable, using ground - signal - ground - signal etc is probably the most practical and low cost way of building a line of arbitrarily low and repeatable inductance, though the width will get excessive if you push it. My recollection is that a ground - signal - ground triple is nearer to 100ohms, at least in the 20 conductor/inch stuff I used a while back, but the memory is hazy. Ribbon is not actually specified for impedance as coax is, so different constructions will vary a bit.

You could build your own low impedance coax. Start with a large diameter inner - perhaps some coax or some #0 cable, and then strip the outer off (the coax will be more flexible). Then add the thinnest layer of insulation you can over the bare conductor, 1.1 turns of sellotape or cling film perhaps. Then add an outer, alli foil would be a good choice. The resulting structure will be very sensitive to air spaces between the electrodes, so binding down the foil to compress the layers into good contact with stretchy tape or elstic bands would be good. Having finished, what would the impedance/inductance be? You could do it by calculation from the log(D/d) formula (google), but residual airspaces will push the impedance up.

By far the easiest way to reduce stray indutance when driving the gates is to notice that cable is generally inductance per unit length, so reducing the length of the cable by 3 acheives the same as tripling the width of the flat ribbon, just a thought.

Marko

Thu Jul 19 2007, 12:49PM

Registered Member #89 Joined: Thu Feb 09 2006, 02:40PM
Location: Zadar, Croatia
Posts: 3145

Hey, thanks a lot Neil. And Steve.

I'l go for a 10lead ribbon (5+5) for my mosfet's.

By far the easiest way to reduce stray inductance when driving the gates is to notice that cable is generally inductance per unit length, so reducing the length of the cable by 3 achieves the same as tripling the width of the flat ribbon, just a thought.

I've taken care of that, and I'll only have several (10 at most) centimeters of cable length. What I fear the most is inductance at connection points, which may be order of magnitude larger than cable itself if not taken care of. That's why I am going to solder the cables directly on bridge side and use connectors on the other. If this works I'll have really superior gate drive over here.

Moderator(s): Chris Russell, Noelle, Alex, Tesladownunder, Dave Marshall, Dave Billington, Bjørn, Steve Conner, Wolfram, Kizmo, Mads Barnkob