External recovery diode / General Science and Electronics / Forums

Forums

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

« Previous topic | Next topic »

External recovery diode

1 2 3 next

Move Thread

LAN_403

nrhoades

Fri Nov 09 2007, 06:37PM

Registered Member #610 Joined: Wed Mar 28 2007, 09:44PM
Location: Middletown, RI
Posts: 110

Quick question:

I see people adding external reverse recovery diodes on their FETs, WITHOUT including series diodes above their drain. Isn't this bad practice, seeing as how the intrinsic body diode in the FET is going to current-hog when it is hot? Of course this is assuming the forward voltage drop of the external Schottkey is lower then that of the body diode.

Steve Conner

Sat Nov 10 2007, 01:05PM

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

Yes, it is bad practice, but people seem to get away with it.

I did a series of experiments a while back that proved that the body-drain diodes do indeed current-hog and the external diodes do nothing, when the schottky is left out. I then built a SSTC with no external diodes at all, to prove that you can get away with it under some conditions.

My PLL driver started as an attempt to track the resonant frequency of a SSTC well enough that the power factor would always be unity or lagging, so that the diodes could be left out without worries. It became irrelevant when everyone switched to IGBTs.

Marko

Sat Nov 10 2007, 02:30PM

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

I thought that you really don't need diodes in soft switched system like SSTC...

But there are things that still confuse me, I hope they are related enough:

Most people seem to use numbers of primary turns so low that their bridges would die instantly from massive magnetizing current if ran without secondary.

What I don't understand is how actually secondary *increases* this impedance and makes magnetizing current unfeelable even though coupling is relatively poor.

System could basically be split into resonator, ideal transformer and primary leakage inductance... now, through some flawed understanding I can't see how secondary creates some higher apparent Z in *series* with primary which appears resistive, and ultimately reactive current is heavily reduced; I thought that this impedance could only decrease in any case.

....?

Steve Conner

Mon Nov 12 2007, 10:50AM

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

wrote ...
What I don't understand is how actually secondary *increases* this impedance and makes magnetizing current unfeelable even though coupling is relatively poor.

It doesn't. All it can do is reflect sinusoidal current at a leading power factor back into the primary, to cancel out the magnetizing current at the zero crossings. This of course makes the impedance lower and the peak current higher, although zero current switching is achieved.

Also, it can't reflect enough to do this once it's loaded by streamers. Richie and I have scoped primary current under streamer loading, and it looks kind of like a square wave with slow rise times, hard switching almost the full peak current. I always thought of this as the sum of a sinusoidal reflected current and a triangular magnetizing current.

So, a SSTC with few enough primary turns that it would die from magnetizing current, would still die from overcurrent with or without a secondary. I think Richie and I agreed that hard switching of magnetizing current is what limits the performance of single-resonant SSTCs. The DRSSTC fixes it.

Marko

Mon Nov 12 2007, 02:06PM

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

Hey, thanks for this... I feel like I really needed to figure these things out.

If you and me were right, then how is this:

If you remember your dirty SSTC, you seemed to have used primary 20turns around plant pot densely wound with unknown dimensions... I guessed 1*10cm (probably exaggeration), but that is still just 80uH at most and gives about amps of magnetizing current for 130

In comparison to this, steve w. mini SSTC uses primary (is that 75mm pipe?) which isn't any more than 2uh in any case - that's really small inductance! I don't know the frequency, but for a guess of 500kHz it's 12,7 amps of just magnetizing current.

His SSTC 2 doesn't look much better either? That primary when tapped fully up looks like about 2-3uH too, but let's give it 5 and for calculated 214kHz secondary frequency without toroid magnetizing current is 25 amperes!

I have no clue how can one survive with such designs in CW. If what you are saying is true then all this should blow up instantly without regrets.

Looking after other's designs bought some fat speaker cable I intended to use for a larger CW SSTC and figured out that my magnetizing current will be 16 amps - Doh!

I can post about the project if anyone is interested.

Now, really, what's about diodes? Wasn't everyone saying how SSTC's can work without diodes?

But it is intuitive that power factor will get poorer as I reduce the coupling or load the secondary down.

Diodes may not although conduct at all even in that case because there is no significant deadtime. Does this save it?

So, should I use them or not? I really do care since they cost several times more than mosfets themselves.

Used calcs

Steve Ward

Mon Nov 12 2007, 07:41PM

Registered Member #146 Joined: Sun Feb 12 2006, 04:21AM
Location: Austin Tx
Posts: 1055

I think your figures are probably correct Marko. I recall measuring one of my SSTCs with 24" plasma running about 30A of primary current (with 170VDC input). This is of course despite the fact that the line current was maybe 16A, which is why you dont select power transistors for the real average power they deliver! Its almost always worse, unless the load is purely resistive. Having good load power factor is nice, but high Q circuits still gonna give you more circulating current.

My larger "5kVA" SSTC runs around 40-50A in the primary circuit at 350VDC... and this is CW

. Drawing arcs off of it messes up the nice sinusoidal primary current, just as Conner says, but the 80N50 mosfets i use are just damn robust and put up with it. It produces about 28" of plasma in CW mode, and 36" in pulsed DC (full-wave or half-wave DC).

Marko

Mon Nov 12 2007, 08:38PM

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

Ok, but what with the diodes? It that they aren't needed simply because they never really conduct since there is no deadtime? (when you are drawing arcs),

And will only conduct if entire thing is loaded enough for voltage drop on the mosfets to reach diode drop? (at which point adding diodes may give some more efficiency).

Back to first, so I'm simply going to have a series of inductor and resistor (secondary) with impedance sqrt(r^2+X^2)?

Why did you use such low turn count btw? You look fr a match between inductive and resistive part... but wouldn't it be just easier to use higher coupling?

Regarding your 5kW (not 50 :p) coil, I don't want to badger but it is something I'd *really* like to see on your site. I remember you posted pics somewhere (as a part of your site) on forum but I can't find them now. I don't know if you have idea how much I liked and learned from your site

I think you also mentioned something about this when you used too low primary turn count and killed your mosfets.

Now I wonder if there's a point where it would just be easier to use a large ferrite transformer and a magnifier configuration rather than primary with huge amounts of reactance. Especially at low (<100kHz) frequencies.

OK, I may have gone way too much from the real thread topic (couldn't resist after steve's post);

As I think, in various ferrite-transformer SMPS's (forward converters) there is only a tiny amount of reactive power on the bridge, only that from primary leakage inductance.

The output buck inductor appears resistive because it only stores but never returns energy to primary circuit, so diode requirement is quite minimal in that case.

Putting series schottky's may actually create much more loss than possible benefit of faster recovery time, so most designs seem to go without them.

With BJT's everyone seems to use small diodes usually rated like 5-10x less current than devices themselves.

But when you are driving a poor PF load like flyback transformer things are different, and at that point you'll need diodes probably as big as your mosfet rating.

BTW, wasn't steve's pre-SLR CCPS supply dying of poor power factor under arc loading combined with hard switched turnoff?

..oh, make sure to correct me if I'm wrong about anything!

Steve Ward

Mon Nov 12 2007, 11:11PM

Registered Member #146 Joined: Sun Feb 12 2006, 04:21AM
Location: Austin Tx
Posts: 1055

Ok, but what with the diodes? It that they aren't needed simply because they never really conduct since there is no deadtime? (when you are drawing arcs),

Internal diodes appeared to be "good enough". I usually depend on those first unless i see some reason not to. Many mosfets actually have fairly good diodes, only some of the older mosfets have crappy slow diodes that should be bypassed for optimal performance.

And will only conduct if entire thing is loaded enough for voltage drop on the mosfets to reach diode drop? (at which point adding diodes may give some more efficiency).

That would be a pretty difficult condition to achieve! If the diodes are conducting, they are taking energy *out* of the load system (i think this is always the case but maybe im missing some subtle cases).

Why did you use such low turn count btw? You look fr a amtch between inductive and resistive part... but wouldn't it be just easier to use higher coup!illling?

Regarding your 50kW coil, I don't want to badger but it is something I'd *really* like to see on your site. I remember you posted pics somewhere (as a part of your site) on forum but I can't find them now. I don't know if you have idea how much I liked and learned from your site

Higher coupling is always limited by voltage flash-over which was in fact a big issue. The only other solution is more volts/turn to get more current, so off with those extra turns! BTW, it is a 5kW (not 50) coil. You can see some pics and whatnot here:

Now I wonder if there's a point where it would just be easier to use a large ferrite transformer and a magnifier configuration rather than primary with huge amounts of reactance. Especially at low (<100kHz) frequencies.

And at that point, how big of a ferrite do you need?

I wouldnt say its "easier". See the DRSSTC... it helps with this problem. I have in fact made 10 foot long "SSTC" sparks from my DRSSTC, but it consumes a few hundred joules of energy per 10mS shot... not something thats feasible to power from even a 240v 60A line at higher pulse rates.

This is really a tesla coil thread now.

BTW, wasn't steve's pre-SLR CCPS supply dying of poor power factor under arc loading combined with hard switched turnoff?

Indeed! This used IGBTs with excellent reverse diodes as well. Just too much hard switching apparently. Im not entirely sure what the failure-mode was but id guess it was likely avalanching that pushed the dissipation too high. Even "really good" diodes will give a recovery such that large amounts transients are produced. Good diodes doesnt = good design :P.

nrhoades

Tue Nov 13 2007, 05:10PM

Registered Member #610 Joined: Wed Mar 28 2007, 09:44PM
Location: Middletown, RI
Posts: 110

Comparison of MOSFET body diodes

IRFU12N25D	IRF740

VDSS		250V		400V
RDS(on)		0.26		0.46
ID		14A		10A
Qg (nC)		23		35

dv/dt (V/ns)	9.3		4
trr (ns)	                140		370
Qrr (uC)	                0.71		3.2

Despite the fact that the older 740 can handle higher voltages, the intrinsic body diode is much slower. The 740 would need atleast 400ns deadtime in order to eliminate cross-conduction from ever happening. Though tesla-coils are not really mission-critical applications.

*PS*

I noticed that tab characters don't work on this forum? Can that be fixed?
*mod edit: nope doesn't look like it :(

Steve Ward

Wed Nov 14 2007, 02:42AM

Registered Member #146 Joined: Sun Feb 12 2006, 04:21AM
Location: Austin Tx
Posts: 1055

The 740 would need atleast 400ns deadtime in order to eliminate cross-conduction from ever happening.

That isnt quite right. *NO* amount of deadtime will solve the problem unless the current in the recovery diode just happens to stop flowing for some other reason. For the case of tesla coils, or any load thats inductive, there will usually be these reverse current flow when the MOSFETs shut off, and the reverse flow only stops when all the energy stored in the inductance is gone. Typically this is impractical, and we must switch before then. Recovery time also varies with the diode current, so ZCS can be helpful here.

1 2 3 next

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