First SSTC design - Need some critique / Tesla Coils / Forums

Forums

4hv.org :: Forums :: Tesla Coils

« Previous topic | Next topic »

First SSTC design - Need some critique

first previous 11 12 13 14 next

Move Thread

LAN_403

DerStrom8

Sun Jul 27 2014, 12:04AM

Registered Member #3704 Joined: Sun Feb 20 2011, 01:13PM
Location: Vermont, U.S.A.
Posts: 92

Dr. Dark Current wrote ...

Three words: Die temperature ripple.

Calculate the peak power dissipation in your transistors and then from the transient thermal impedance graph find out the longest ON-time you can run. The ripple should never be larger than 30 Â°C.
When running higher duty cycles, also check the average dissipation.

In other words, you can destroy the bridge running at 1% duty in one case, where it will run happily at 10% in another case.
Besides peak current; lower break rates, higher switching frequencies and hard switching all limit the duty cycle you can run.

This is very helpful, but I want to make sure I'm doing this right.

I'm calculating power by the following:

Tj = 150C
Tc = 27C (estimated)
Rth = 0.43C/W

P = (Tj-Tc)/(Rth) = 286W

First of all, does this sound normal? 286W seems quite high, especially for a TO-247. Here's the datasheet I'm looking at:

Then, if I understood you correctly, I use the transient thermal response graph to find the pulse duration from the thermal resistance? In which case, I'm finding it to be around 2mS (thermal resistance is 0.43 degrees C per Watt).

Just checking my work here. This is my first time working with IGBTs in an actual project.

Thanks,
Matt

Hydron

Sun Jul 27 2014, 12:40AM

Registered Member #30656 Joined: Tue Jul 30 2013, 02:40AM
Location: UK
Posts: 208

You're not going to be able to keep the case anywhere near 27C in steady state at 286W, unless it's outside in the arctic or something. Bump it up a bit and the max dissipation will get much more sane. That said, I've seen far higher numbers in TO-247 datasheets - look up the one for FGH60N60SMD.

As was stated before, the steady state temperature is often not the limiting factor, especially with pulsed operation. While you could spike the die temperature up to 150C briefly without killing it, try doing it too many times and you'll snap the bond wires or something with fatigue from the thermal cycling. Die temperature ripple must be a few 10s of degrees C max - some of Steve Ward's posts talk about this and the number to aim for.

DerStrom8

Sun Jul 27 2014, 12:54AM

Registered Member #3704 Joined: Sun Feb 20 2011, 01:13PM
Location: Vermont, U.S.A.
Posts: 92

Hydron wrote ...

You're not going to be able to keep the case anywhere near 27C in steady state at 286W, unless it's outside in the arctic or something. Bump it up a bit and the max dissipation will get much more sane. That said, I've seen far higher numbers in TO-247 datasheets - look up the one for FGH60N60SMD.

As was stated before, the steady state temperature is often not the limiting factor, especially with pulsed operation. While you could spike the die temperature up to 150C briefly without killing it, try doing it too many times and you'll snap the bond wires or something with fatigue from the thermal cycling. Die temperature ripple must be a few 10s of degrees C max - some of Steve Ward's posts talk about this and the number to aim for.

Woops! I forgot I was switching around the temperatures to see how the values changed. Originally I had 70C plugged in, which gave me 186W.

The heat sink, in case I haven't mentioned it yet, is this one:

To give you a feel for the size, each floorboard is about 2" wide, if I remember correctly. I'm hoping I'll be able to dissipate a fair amount of heat by using this heat sink and a push-pull PC fan setup on either end for air cooling.

I forgot to ask in my last post how to calculate the temperature ripple. Dr Dark Current mentioned that in his post but I guess I'm unclear as to how to determine the ripple from there.

Cheers,
Matt

Sigurthr

Sun Jul 27 2014, 06:36AM

Registered Member #4463 Joined: Wed Apr 18 2012, 08:08AM
Location: MI's Upper Peninsula
Posts: 597

Thermal aspects aside; either you hard code the arduino with on time limiting (a defined pulse sequence) or you use a retriggerable one-shot to do the actual pulses and use the arduino for the trigger pulses, thus the duty cycle of the arduino output is irrelevant. You cannot use the built in PWM or tone output functions of the arduino.

for example:

void Pulse1KHz(){  // 1KHz @ 10% duty - note the 100uS on time.
    digitalWrite(3, HIGH);
    delayMicroseconds (100);
    digitalWrite(3, LOW);
    delayMicroseconds (900);
}

void Pulse500Hz(){ // 500Hz @ 5% duty cycle - note the 100uS on time.
    digitalWrite(3, HIGH);
    delayMicroseconds (100);
    digitalWrite(3, LOW);
    delayMicroseconds (1900);
}

*note; precise timing using delay is not guaranteed. One often has to use a clever If Micros() loop where a set number of uS has to have passed since Micros() was saved to a variable in order to do the timing. The above subroutine is just for getting the idea across.

The loop can be implemented similarly to:

void loop (){
    currentMillis = millis();
    digitalWrite(3, HIGH);
    if (currentMillis - previousMillis >= OnTime) {
        previousMillis = currentMillis;
        digitalWrite(3, LOW);
        delayMicroseconds (OffTime); // since offtime is much longer timing isn't an issue
    }
}

*note; preceding code is untested and I'm tired. Again, it's just to illustrate a point.

Dr. Dark Current

Sun Jul 27 2014, 07:22AM

Registered Member #152 Joined: Sun Feb 12 2006, 03:36PM
Location: Czech Rep.
Posts: 3384

The DRSSTC transistor power loss calculation goes as follows:

1. Peak conduction loss. Look at the output characteristcs graph and read out saturation voltage for a current somewhat lower than your peak current (lets say ~70% Ipk). Calculate the peak dissipation in one transistor as Ipk * Vsat / pi.

2. Peak switching loss. You need to determine your switching timing. If you use some form of delay compensation ("phase lead" or PLL), you tune it slightly to the inductive side, so calculate with a TURN-OFF current of lets say 20% your Ipk. If the feedback is just direct amplification, you will need to calculate with a TURN-ON current which is relatively large and depends hugely on other factors. It could be in the range of maybe 20-80% of your Ipk. Higher frequencies and slower feedback makes it worse.
Then you read out in the switching energy loss graph (for high Tj) your switching loss and multiply by frequency.

Total peak power loss = peak conduction loss + peak switching loss

DerStrom8

Sat Aug 02 2014, 05:30PM

Registered Member #3704 Joined: Sun Feb 20 2011, 01:13PM
Location: Vermont, U.S.A.
Posts: 92

Got another question for you guys--

I have become doubtful that my 20nF 5kV tank capacitor will be enough--20nF is awfully low, and I would prefer to have a capacitance of closer to 100nF or higher. I've been thinking about just blowing past my budget by around $70 and buy some 942C-series capacitors from EasternVoltageResearch:

I have also considered picking up 20 of these:

I really didn't want to spend too much money on this project, but I feel I really shouldn't cut corners with the capacitors--besides the IGBTs, I'm thinking they're one of the most important parts.
What are your thoughts on the above options?

By the way, I have been re-considering using my design, as it is unable to ensure soft-switching of the IGBTs and there are a variety of other things that concern me. I think I'll probably end up going with Steve Ward's DRSSTC-1 design, as shown here:

EDIT: I just purchased 8 of these:

Really hoping I didn't spend $30 for caps that won't work. I'm a bit concerned about the Irms. Based on the following formula:

$3add0ab352a43691cf6576d58491261b$

I calculate my coil, at 220 bps with a 1mS on-time, will have an Irms of 16.91A, but the caps are only rated for 10.1A according to the DS. I misread it when I was looking up the part number, which is why I bought these capacitors--I thought they would work. I may have to resell them and buy some better, 942C- ones.

Peak current of these caps is 216A, I calculate my ipk to be around 35A, but if I am not mistaken it is the Irms that I really need to worry about. Any suggestions?

loneoceans

Tue Aug 05 2014, 01:18AM

Registered Member #4098 Joined: Fri Sept 16 2011, 09:26PM
Location:
Posts: 236

I see you're now trying to sell your 940C caps and buy 942C caps instead in the buy-sell - I'd recommend just sticking with the 940C which also work well. And at this point I think the driver design and geometry are more crucial.

I'm not sure where your 1ms drive time comes from. I think you might be still confusing DRSSTC with SSTC operation. In DRSSTCs, you'll only usually run your drive for a few cycles, around 5 to 15 cycles or so. So if your coil is operating at 100kHz, that's an ON time of 100us for 10 cycles. Your coil is running at 240khz so the ON time will probably be even shorter. Therefore if you use your tone library for DRSSTC drive, using your example, you'll drive your coil for 1.14ms, and in DR mode this will probably lead to some magic silicon smoke being released somewhere

While it is nice to have quick connect transistor slots, the current way you're doing looks really lossy. You can find those green connectors with direct leads at the bottom which you can solder, which will be much better than the stacked way you have.

Finally, your primary inductance looks really high at 28uH. I think this might because you're using a really small 20nF tank capacitor so you need a primary with lots of turns to get your freq up to 200kHz. If you keep the same frequency and increase to 100nF as you mentioned, your inductance has to drop by 5 to keep the same frequency, and your ipk will increase by 5 times assuming the same drive duration.

[Edit] I just saw that your original design used 10 turns of primary in a cylindrical form. This will result in far too high coupling between your primary and secondary, and will lead to flashovers in DRSSTC operation. With your 8 caps, you can start off with two strings of 4 which will give you 75nF which should be a good starting point for your coil. You can go for a flat primary, or even a cylindrical one with a larger diameter and fewer turns. It might also be better to start off with a simple 555 timer for your interrupter which will be one less thing to worry about and to debug :)

DerStrom8

Tue Aug 05 2014, 02:26AM

Registered Member #3704 Joined: Sun Feb 20 2011, 01:13PM
Location: Vermont, U.S.A.
Posts: 92

loneoceans wrote ...

I see you're now trying to sell your 940C caps and buy 942C caps instead in the buy-sell - I'd recommend just sticking with the 940C which also work well. And at this point I think the driver design and geometry are more crucial.

I'm not sure where your 1ms drive time comes from. I think you might be still confusing DRSSTC with SSTC operation. In DRSSTCs, you'll only usually run your drive for a few cycles, around 5 to 15 cycles or so. So if your coil is operating at 100kHz, that's an ON time of 100us for 10 cycles. Your coil is running at 240khz so the ON time will probably be even shorter. Therefore if you use your tone library for DRSSTC drive, using your example, you'll drive your coil for 1.14ms, and in DR mode this will probably lead to some magic silicon smoke being released somewhere

Hi loneoceans,

Okay, I will hold on to the 940C caps for a while and see how they work.

The 1mS comes from a (possibly incorrect) calculation of 2mS maximum on-time before the IGBTs start to suffer, so I decided 1mS was a way to stay safe. I have since considered using a 555 timer wired in monostable mode to deliver consistent 250uS pulses, which should allow a maximum frequency of 2kHz from the interrupter at 50% duty cycle. From there, the duty cycle would decrease (the on-time would remain the same). Not quite sure if I"m on the right track here, but it's what I've been thinking about lately.

Using JavaTC I designed the cylindrical primary to have a coupling coefficient of 0.131k. I was thinking that was a good value....?

The IGBT connectors won't be permanent. I'm not sure if I posted the latest version yet either--I shortened the leads significantly, so that they're just spacers now:

But you're right, I will definitely be looking for better options.

I had a feeling a 20nF capacitor was awfully small, which is why I bought the 940Cs. I re-did some designs on the fly one time and got some more reasonable values, but unfortunately I didn't save them. When I bought the new caps, though, I knew I would have to re-design the whole thing. I have been wondering if I should try to lower the frequency of the coil by increasing the topload capacitance, but someone also mentioned that this might load the coil too much and cause arcing directly from the side of the secondary.

I also need to be careful about how much current I allow through the IGBTs and the capacitors, so I'll need to be careful about the primary inductance.

And you're right, I definitely need to start off with a 555 timer interrupter. Then I can move on to the Arduino/MIDI controller. I really need to break this down more into smaller steps :D

Thanks!
Matt

DerStrom8

Wed Aug 13 2014, 06:41PM

Registered Member #3704 Joined: Sun Feb 20 2011, 01:13PM
Location: Vermont, U.S.A.
Posts: 92

Hi guys,

I've made some progress in a few areas. First of all, the interrupter using the Arduino. I ended up using the TimerOne library and Bresenham accumulators to generate a note with a specified on-time. Here's a waveform of the Arduino emitting a 440Hz (A4) note.

I also have a question about mounting the bridge to the heat sink. Which would you guys recommend: Putting two IGBTs on either side of the large heat sink, or put all four on one side (it's a big heat sink) to reduce lead length? I'm thinking the lead length isn't terribly important between the GDT and the gates of the IGBTs, to an extent. Am I correct? In which case, I should probably mount two IGBTs on either side....?

Now that I have a nice modern scope (cheap, but it works), I'll be able to continue with the actual build.

Thanks for the help,
Matt

Sigurthr

Wed Aug 13 2014, 08:33PM

Registered Member #4463 Joined: Wed Apr 18 2012, 08:08AM
Location: MI's Upper Peninsula
Posts: 597

Those bursts (pulses) look WAY too long. I can't magnify the image right now but does that say 500uS!?

Keep the secondary side leads of the GDT as short as you possibly can. That is where you'll get terrible parasitic oscillation of the gates. Primary side isn't nearly as critical.

first previous 11 12 13 14 next

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