"Thumper" / Projects / Forums

Forums

4hv.org :: Forums :: Projects

« Previous topic | Next topic »

"Thumper"

first previous 5 6 7 8 next last

Move Thread

LAN_403

Steve Conner

Sat May 30 2009, 01:23PM

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

Hi guys,

Finn, I didn't mean to dis your implementation of the N/P circuit. I think it's pretty good as implementations of it go: I've seen a few builders of it who didn't understand any of the issues raised in this thread, and came to the (old) forum with questions like, "It draws 12 amps from the supply, I lose AM radio reception and the MOSFETs catch fire, what am I doing wrong?"

It's the circuit itself I have the issue with, not your implementation. If you keep tweaking it carefully, I'm sure it'll work fine.

About the Odin gate drivers, they're not going anywhere

Finn Hammer

Sat May 30 2009, 07:33PM

Registered Member #205 Joined: Sat Feb 18 2006, 11:59AM
Location: SkÃ¸rping, Denmark
Posts: 741

Steve, all,

Thanks for saying that, Steve.
But hey guys! I think we are having a blast here, good shop humor. I am not offended in any way, only pretending in an attempt to play funny. It's ok with a slightly rough tone, I can handle that.

Now, The Ghetto driver, the hack!
Present status is as sitting on the bench, screwed onto a presumably good CM600

, and undergoing examination of key waveforms.
The tek P6021/134 is monitoring current between MosFets, which means it monitors shoot trough when gate voltage drops (as if there was any), and current into gate when gate voltage rises.

And as such, here is the big picture:
Yellow is gate voltage on CM600, -10V/+20V, 1Volt overshoot, no gate resistor.
Magentha is signal into optocoupler.

But let's have a closer look.
On the rising edge of the CM600 gate voltage, you can see that voltage is being ramped up from -10V to -2V over 200nS, and then it shoots up to +20V in 100nS. No ringing but 1V overshoot. No gate resistor.
No miller plateou
Current is green, 4A/div, peaks @10A.
Cyan is signal out of opto.

The delay is massive, 300nS for the opto to react, another 300nS for the totem pole to switch, total: 600nS from gate signal into opto, before the CM600 is on. More about this later.

On the falling edge:

Again, no ringing but delays are:
400nS for the opto to react, 300nS for the totem pole to switch, in all 700nS delay.

The delays are the drawbacks of this driver, and would render it useless in any other DRSSTC but this one: I can dial this delay in when I set the predikter.
The neat thing is that it switches very cleanly, there are no ringing worth mentioning, no overshoot and no gate resistor to tame things. A word like synergy comes to mind, byt I am probably just talking rubbish.
During the day, I have logged 5 hours of continous CW on it, and it runs slightly hot under CW conditions, which means that in pulsed mode it will run stone cold.
I think I'l use it despite the delays but due to the clean switching and because I designed it myself.

Cheers, Finn Hammer

Dr. Drone

Sat May 30 2009, 07:44PM

Registered Member #290 Joined: Mon Mar 06 2006, 08:24PM
Location:
Posts: 1673

hvguy

Sun May 31 2009, 08:08AM

Registered Member #289 Joined: Mon Mar 06 2006, 10:45AM
Location: Conroe, TX
Posts: 154

Those are some great looking waveforms, and that's a nice dense layout. I can't wait to see how this all plays out once the coil is running. What did you have in mind for a high side PS? Also, I have to agree with the Dr., I use 200V FETs where ever possible simply for higher reliability and transient ruggedness. I know your using 100V parts here, they will probably be fine, especially with such tame switching speeds.

Just for the purposes of comparison, here are a few waveforms I pulled out of my "archive" from my off line gate driver driving a CM1000.

Blue trace is drive input
Green trace is the gate of the CM1000 at 5V/div

Note the overshoot from the small, but still there, leakage L of my custom 9:1:1:1:1 GDT. This is a problem you certainly don't have

Here we see the rise time, about 100ns to zero, 25ns to the miller, and another 75ns to the top (just barely clipped on the screen). The delay appears to be about 150ns. The circuit was designed specifically to keep this delay as small as possible since the parts where already so slow.

Not quite as pretty as yours, but considering the 220nF gate capacitance, I was very happy with these results

I should also note that these screen shots where taken while the system was running at low power (about 5KW).

Finn Hammer

Mon Jun 01 2009, 08:09AM

Registered Member #205 Joined: Sat Feb 18 2006, 11:59AM
Location: SkÃ¸rping, Denmark
Posts: 741

Aron,

Thanks for posting your gate waveforms. I have often missed those to compare my work with the reality of others.
Are you averaging the signal to clean the trace up? Mine are a bit more fuzzy.
Going to be interesting to see how they react to a CM600 under load.
High side PSU?. I guess a Mazilli and a flyback core, like on the CCPS. It works well, and is not too stiff, so the V-Reg. Zeners stay cool.
Daniel is going to get the first prototype predikter board for a coil he is building for his work at Aalborg University, and it will be driving an intermediate full bridge, to drive the SKM 400GB124D power bridge. So we may be seing the first results on that coil.
I have just started drilling the heatsink for the CM600 bridge, so it will soon be available for preliminary testing.
Jeez, can't wait to be able to pass 1+kA without worrying about the bridge failing due to over current.
I am also awfully tempted to buy one of these:

So if you would please all step back and don't bid against me

Cheers, Finn Hammer

hvguy

Mon Jun 01 2009, 10:18AM

Registered Member #289 Joined: Mon Mar 06 2006, 10:45AM
Location: Conroe, TX
Posts: 154

No problem, Finn. I often whish people would post more waveforms as well. Hopefully the proliferation of inexpensive (relatively) DSO's will help with that in the coming years.

Thatâ€™s a good question. I do not have anything about it in my notes, but I may have had the bandwidth limit on.

I figured you would go that route for the high side; that driver is just sooo simple, it's really perfect for the task.

Ya, big systems are nice. I have been so much happier with my coils since I started running only 1200V parts and only running them within their DS ratings. I haven't had an IGBT failure in a DRSSTC in at least two years (I think the giant coil was the last one. That $1,100 brick hurt!).

Good luck with that monster inverter assembly. I can only imagine what the came from

colin heath

Sun Jun 07 2009, 04:03PM

Registered Member #123 Joined: Fri Feb 10 2006, 12:58PM
Location:
Posts: 162

Very informative thread. I think I have understood a little of it at least;-)

Your work is always very thorough Finn and you always strike me as a very organised thinker. (based on your help with my DRSSTC lash up that year at derby)

Keep up the good work while I try and make sense of it all

Dalus

Sun Jun 07 2009, 04:24PM

Registered Member #639 Joined: Wed Apr 11 2007, 09:09PM
Location: The Netherlands, Herkenbosch
Posts: 512

It's indeed a very good informational thread. Think I'll borrow a few things from this thread for my own DRSSTC. Starting to understand the various gate drives and the pros and cons of each type. Though I'm curious to see how much you'd need to slow down the gate driver when driving large slow bricks. Fast isn't always the best option.

But I'm really curious to see how well the delay circuit works and how sensitive it is.

You've done a great job so far Finn Hammer

Finn Hammer

Sun Jun 07 2009, 08:48PM

Registered Member #205 Joined: Sat Feb 18 2006, 11:59AM
Location: SkÃ¸rping, Denmark
Posts: 741

Thanks a lot folks, you are being very kind.

I am riding high on a surge of gumption and creativity. Today, I have been making the bus for the bridge, which is now starting to take shape:

In my urge to tuck the lytics as close as possible into the bridge, I had to do a lot of machining, and next time, the buss plates are going to be extended to both sides, and the caps are going to be mounted vertical instead of horisontal.

The gate drivers have been completed, and they open the gates "blixt schnell" 200nS and we will see if it is good or bad. Pls. note, that with the "Predikter", there is no need to slow down the turn on for the traditional reason: to allow the other IGBT to shut off. This design features separate timing for the turn off as well as the turn on, and the turn off happens at current zero crossing, right?
So a completely new situation here.

The air core coil forming the predictor information in the load for the current transformer has now been replaced by a 1ÂµH high class magnetically shielded part from Epcos, and it is working flawlessly.

Nice little 10mm x 10mm footprint SMD part.

I've got 3 days off, next week, so lots of progress is anticipated.

Cheers, Finn Hammer

Dalus

Sun Jun 07 2009, 09:15PM

Registered Member #639 Joined: Wed Apr 11 2007, 09:09PM
Location: The Netherlands, Herkenbosch
Posts: 512

Your H-bridge looks great nice and compact. It would give me nightmares if I had to use single IGBT modules. Halfbridge modules together with some angle iron give a way simpler layout. I also had the urge to mount my lyctics sideways and almost on the igbt's

first previous 5 6 7 8 next last

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