If you need assistance, please send an email to forum at 4hv dot org. To ensure your email is not marked as spam, please include the phrase "4hv help" in the subject line. You can also find assistance via IRC, at irc.shadowworld.net, room #hvcomm.
Support 4hv.org!
Donate:
4hv.org is hosted on a dedicated server. Unfortunately, this server costs and we rely on the help of site members to keep 4hv.org running. Please consider donating. We will place your name on the thanks list and you'll be helping to keep 4hv.org alive and free for everyone. Members whose names appear in red bold have donated recently. Green bold denotes those who have recently donated to keep the server carbon neutral.
Special Thanks To:
Aaron Holmes
Aaron Wheeler
Adam Horden
Alan Scrimgeour
Andre
Andrew Haynes
Anonymous000
asabase
Austin Weil
barney
Barry
Bert Hickman
Bill Kukowski
Blitzorn
Brandon Paradelas
Bruce Bowling
BubeeMike
Byong Park
Cesiumsponge
Chris F.
Chris Hooper
Corey Worthington
Derek Woodroffe
Dalus
Dan Strother
Daniel Davis
Daniel Uhrenholt
datasheetarchive
Dave Billington
Dave Marshall
David F.
Dennis Rogers
drelectrix
Dr. John Gudenas
Dr. Spark
E.TexasTesla
eastvoltresearch
Eirik Taylor
Erik Dyakov
Erlend^SE
Finn Hammer
Firebug24k
GalliumMan
Gary Peterson
George Slade
GhostNull
Gordon Mcknight
Graham Armitage
Grant
GreySoul
Henry H
IamSmooth
In memory of Leo Powning
Jacob Cash
James Howells
James Pawson
Jeff Greenfield
Jeff Thomas
Jesse Frost
Jim Mitchell
jlr134
Joe Mastroianni
John Forcina
John Oberg
John Willcutt
Jon Newcomb
klugesmith
Leslie Wright
Lutz Hoffman
Mads Barnkob
Martin King
Mats Karlsson
Matt Gibson
Matthew Guidry
mbd
Michael D'Angelo
Mikkel
mileswaldron
mister_rf
Neil Foster
Nick de Smith
Nick Soroka
nicklenorp
Nik
Norman Stanley
Patrick Coleman
Paul Brodie
Paul Jordan
Paul Montgomery
Ped
Peter Krogen
Peter Terren
PhilGood
Richard Feldman
Robert Bush
Royce Bailey
Scott Fusare
Scott Newman
smiffy
Stella
Steven Busic
Steve Conner
Steve Jones
Steve Ward
Sulaiman
Thomas Coyle
Thomas A. Wallace
Thomas W
Timo
Torch
Ulf Jonsson
vasil
Vaxian
vladi mazzilli
wastehl
Weston
William Kim
William N.
William Stehl
Wesley Venis
The aforementioned have contributed financially to the continuing triumph of 4hv.org. They are deserving of my most heartfelt thanks.
Registered Member #146
Joined: Sun Feb 12 2006, 04:21AM
Location: Austin Tx
Posts: 1055
Amazing sparks! If you don't mind my saying so, the whole set up looks rather... eh... 'ghetto'? Perhaps a short appeared somewhere and killed the IGBT?
Sure, why would i waste too much time making something really nice before i know it works reliably? In fact the important parts were electrically quite good. There is no chance that something shorted out, all the connections are actually solid and proper.
I switched over to using a fixed oscillator, but again the IGBT fails. Ive measured the peak voltage to be less than 800V for sure (more like 500V), and the TVS dont get warm (they really are only there just in case) so im satisfied with my voltage divider measurements on the scope to be relatively correct. Im certain that im exceeding the junction temp for this small IGBT, so either my heatsink clamp needs improvement, or maybe im losing my nice class E switching at higher powers. I noticed the RF envelope on the CE junction starts to "blow up" after some Vin, but its been too hectic to get a good scope reading of whats going on, its possible that im getting non-ZCS at the worst possible time, and that its all just a tuning issue.
Registered Member #1232
Joined: Wed Jan 16 2008, 10:53PM
Location: Doon tha Toon!
Posts: 881
Hi Steve,
The work you have done here is very interesting, and you have probably got it working far better than many would have thought possible with a single IGBT.
This is a generalised statement, but here goes anyway: "IGBTs aren't good for Class-E circuits."
There are two main reasons for this. Firstly, Class-E (and other flyback topologies) work by closing the switch, ramping up a current in the drain (or collector) choke then opening the switch to create a damped ringing waveform. IGBT's in this situation take a pounding because they are interrupting the collector current right when it is at its maximum value. Even with modern high-speed IGBT's turning off large currents is still going to generate high switching losses. In contrast, a MOSFET with good gate-drive can turn off this current in a few tens of ns.
Secondly, Class-E is designed to make the switch turn-on with zero-current through it and zero-voltage across it. This is important for MOSFETs as they generally have high Cdg and Cds capacitances so incurr high turn-on losses from VxI operlap, and energy stored in Cds at turn-on. In contrast, similarly rated IGBTs typically have lower Cce capacitance and can turn-on very quickly indeed.
The behaviour you noted where the Class-E ringdown waveform is only correctly damped at one point in the envelope is common were Amplitude Modulation is employed. The corona at the top of the secondary more or less acts like a constant-voltage clamp. That appears as a constant-current load at the output of the Class-E amplifier. Unfortunately a constant-current load is not constant-impedance, so the operating point (damping) of the Class-E amp changes throughout each RF burst. To a lesser extent the resonant frequency of the TC detunes with corona, and the switching device's capacitances also change during each RF burst, also changing the damping. The device capacitance variation is a lesser effect, but is observable in high-power Class-E amps used in AM radio transmitters. Even though the load is a well behaved 50 ohms in this case, the Class-E waveform can still distort slightly as B+ is modulated to invoke AM.
If you can get the Class-E "soft landing" right at any point in the RF envelope it wants to be at peak mains voltage for a SSTC, since this is where switching losses would be highest, and it is also where the mains waveform lingers longest. Conversely, for an AM transmitter, or audio modulated SSTC you want to the Class-E waveform to be optimal at the quiescent carrier level as this is the operating point at which the circuit spends most of its time.
Registered Member #146
Joined: Sun Feb 12 2006, 04:21AM
Location: Austin Tx
Posts: 1055
Richie, Thanks for reminding me about the IGBT turn off losses! Doh! I actually forgot about that, and its probably a big reason why this thing is failing so often from overheating.
This is just one of those projects that became addicting after seeing some good first results . But now i see even more clearly why the half-bridge is such a wonderful method (compared to this Class-E stuff) for higher power lower frequency stuff where you have the option of *not* using class-E.
Registered Member #89
Joined: Thu Feb 09 2006, 02:40PM
Location: Zadar, Croatia
Posts: 3145
Hi guys,
Those are great results Steve, makes me feel frustrated when I recall how I could barely achieve 16 inch spark with whole handfuls of IRFP450's. (mostly thanks to my poor thermal and mechanical design though). If I see it correctly, you aren't using a RFC, but just the primary between + and drain, with shunt cap across mosfet, right? Are you going to implement the RFC?
Mates built similar coils based around 1200V IGBT's if I recall, although he didn't seem to bother with tuning too much with tuning - he just used several tens of TVS in oil to clamp off any excess overvoltage... also he had a whole bunch of motor run caps in series with AC input to drop the voltage (kind of weird idea, I'd consider it unreliable but it apparently worked).
Still now for a while I'm actually failing to see benefits of class E for tesla coils over typical bridges. Only benefit I can actually see is no need for high side gate drive, and then only drawbacks:
- requires very careful tuning - in tesla coil duty needs constant load and constant supply voltage (pulsating waveform is bad), drawing arcs isn't probably a good thing either - switch voltage rating needs to be more than pi*supply voltage
Regarding switching losses, I now absolutely fail to see any benefit of class E over bridges.
They are both ZVS and both do the hard turn-off of current.
Theoretically looking class E can only have lower efficiency because voltage rating of the switch needs to be higher.
So really, what is special about class E, apart from a bit easier gate drive???
The only speculation I can give about why bridges get less efficient at higher frequencies is, that they lose ZVS because deadtime gets too short and there's not enough time for output capacitance to drain.
But if it's just that, why don't we just use birdges anywhere we would use class E? Where are actually the benefits?
Registered Member #1232
Joined: Wed Jan 16 2008, 10:53PM
Location: Doon tha Toon!
Posts: 881
Steve W wrote:
Richie, Thanks for reminding me about the IGBT turn off losses
No worries. I actually tried to do exactly what you did several years ago with some slower first generation IGBTs and saw massive device heating. So what you have achieved with yours is very impressive.
now i see even more clearly why the half-bridge is such a wonderful method (compared to this Class-E stuff) for higher power lower frequency stuff where you have the option of *not* using class-E.
Yes, a pair of IGBTs running with a slightly leading power factor will give ZCS which is much better suited to IGBT's switching characteristics. There is a term called "switch utilisation" which is basically the watts of RF you get out divided by the sum of all the semiconductors VxI ratings. Even though a half-bridge requires two devices it still has a better device utilisation figure because the device voltages are bounded by the supply rails.
Marko wrote:
So really, what is special about class E, apart from a bit easier gate drive??? Regarding switching losses, I now absolutely fail to see any benefit of class E over bridges.
Class-E is a topology specifically developed to allow efficient operation of MOSFETs (with intrinsic device capacitances) in HIGH-FREQUENCY HIGH-VOLTAGE switching amplifiers. The benefits include:
1. Zero voltage at turn-on. This eliminates turn-on losses due to discharging Cds into the channel. (This loss could be very large for a large die device running at 20MHz!) 2. Zero current at turn-on. This eliminates turn-on losses due to Rds dissipation as the device transitions through the linear region. (ie No VxI overlap.) 3. No "Miller effect" during turn-on because the drain voltage has already fallen. 4. Capacitive snubbing of the shunt capacitor reduces turn-off losses by extending voltage rise time. (i.e. minimises turn-off VxI overlap) 5. Circuit makes use of device capacitance, and stray capacitance and inductance that would hinder hard-switched circuits at these high frequencies. 6. Extremely high efficiency so little heat is generated 7. Simple design requiring only a single high-speed switch (ie reduced cost) 8. Eliminates bridge cross-conduction problems found at high frequency. 9. Low EMI and harmonic output because of the smooth drain voltage profile in Class-E. 10.Elimination of free-wheeling current through slow anti-parallel diodes when tuned correctly. 11.Some tolerance to miss-tuning afforded by the soft-landing of drain voltage at the device turn-on point.
Basically the drain voltage is meant to land at zero volts with a first derivative (gradient) of zero exactly when the MOSFET is about to turn on again. The "zero volts" part eliminates capacitive turn-on losses, and the "zero gradient" part means that it lingers at zero for some time giving some latitude in the exact instant when the device is turned on. The "zero gradient" requirement also means that there is zero current flowing at that time. Meaning that there is no displacement current flowing through the device's Cds capacitance at the switching instant which would have to be re-directed across the die and can be the source of excessive dissipation in the device's bulk silicon at very high frequencies.
There are other subtle advantages that happen due to device's voltage-dependent capacitances but the points above are the main advantages. For MOSFETs operating in the "middle ground" of frequency but at higher voltages there is also Class-DE which uses two switches - It gives most of the advantages of Class-E but combines them with the two main advantages of the half-bridge. Those being, supply bounded device voltages, and quasi-square wave output (which is sometimes desirable.)
If a design calls for a switching amplifier using high-voltage devices at high-frequencies driving a well-behaved load, then Class-E will do it with outstanding efficiency. For lower voltages, lower frequencies or variable loads, most of the advantages of Class-E are quickly lost.
If you want to know more, I would recommend anyone to read the papers by Sokal.
Registered Member #89
Joined: Thu Feb 09 2006, 02:40PM
Location: Zadar, Croatia
Posts: 3145
Hi Richie, Steve,
Class-E is a topology specifically developed to allow efficient operation of MOSFETs (with intrinsic device capacitances) in HIGH-FREQUENCY HIGH-VOLTAGE switching amplifiers. The benefits include:
1. Zero voltage at turn-on. This eliminates turn-on losses due to discharging Cds into the channel. (This loss could be very large for a large die device running at 20MHz!) 2. Zero current at turn-on. This eliminates turn-on losses due to Rds dissipation as the device transitions through the linear region. (ie No VxI overlap.) 3. No "Miller effect" during turn-on because the drain voltage has already fallen. 4. Capacitive snubbing of the shunt capacitor reduces turn-off losses by extending voltage rise time. (i.e. minimises turn-off VxI overlap) 5. Circuit makes use of device capacitance, and stray capacitance and inductance that would hinder hard-switched circuits at these high frequencies. 6. Extremely high efficiency so little heat is generated 7. Simple design requiring only a single high-speed switch (ie reduced cost) 8. Eliminates bridge cross-conduction problems found at high frequency. 9. Low EMI and harmonic output because of the smooth drain voltage profile in Class-E. 10.Elimination of free-wheeling current through slow anti-parallel diodes when tuned correctly. 11.Some tolerance to miss-tuning afforded by the soft-landing of drain voltage at the device turn-on point.
Now, I can apply almost all of your numbers to a normal H bridge and they will still be true! (apart from ten, indeed in a bridge magnetizing current will always be added to load current; switches circulate reactive power - important thing i missed in my post).
If I can't, please tell me so!
It is now clear that in typical SSTC driver bridges, there is zero current and zero voltage at turn on. Class E will turn off peak load current, while bridge will turn off (more or less) peak magnetizing current.
Class E doesn't circulate reactive power through the switch, thus gaining some efficiency, but must use lower input voltage due to Ucc*pi thing. So this looks like a main tradeoff, but not a worthy one to my opinion.
I'm not new to class E, but more I understand how it works less I understand the hype about it.
Indeed, when I sum everything it really just seems that I trade off between very difficult tuning and easier gate drive versus no need for tuning and more complex gate drive.
So can anyone explain why can I build a 10Mhz sstc with class E amplifier and not with a H bridge. In theory there's nothing obviously wrong with it, but in practice it won't work well for some reason?
Registered Member #1025
Joined: Sun Sept 23 2007, 07:53PM
Location: Czech Rep.
Posts: 566
You also forget about the practical advantages of ClassE which are also relevant. I mean the fact of easy transistor paralleling. All the trasitors can be mounted on the same heat sink without pads and in case of failure only one transitor is destroyed (in case of bridge you destroy mostly both of them).
Marko wrote ...
Mates built similar coils based around 1200V IGBT's if I recall, although he didn't seem to bother with tuning too much with tuning - he just used several tens of TVS in oil to clamp off any excess overvoltage... also he had a whole bunch of motor run caps in series with AC input to drop the voltage (kind of weird idea, I'd consider it unreliable but it apparently worked). Marko
Marko, those motor caps in my design are not for volatge clamping but are used to solve the problem with long turn-off times of the IGBT. The caps are in series with the primary of the coil and after the transistor is opened the current flows only unless the these caps are charged. Because these caps are designd with very low internal resistance the charging is a question of few hundered nanoseconds. That's why I can also PW modulate the driving pulses. Short driving pulses can be very easily placed into the ZV areas. You can check this scheme to understand what I mean.
Registered Member #30
Joined: Fri Feb 03 2006, 10:52AM
Location: Glasgow, Scotland
Posts: 6706
Marko, your comments apply only to a H-bridge operating in Class-DE, so you should probably read the Sokal papers and/or Ian De Vries' PhD thesis. (He built a half-bridge putting out over 1kW at 13.56MHz.)
Someone called Mates' circuit Class-TVS, which I found quite funny. I personally think it's an OLTC, due to the series capacitors, and the drive frequency being a subharmonic. Derek Woodroffe used to drive OLTCs at breakrates of 1/4 or 1/3 of the resonant frequency.
FWIW, I think Mates' circuit would work better if the 18uF and 370uF caps were interchanged. It would perform just the same and inject 370/18 times less RF back into the mains.
Registered Member #1232
Joined: Wed Jan 16 2008, 10:53PM
Location: Doon tha Toon!
Posts: 881
Marko wrote:
Now, I can apply almost all of your numbers to a normal H bridge and they will still be true... ...If I can't, please tell me so!
As Dr Conner already said some of these points do apply some of the time to a half-bridge or full-bridge depending on the load's power factor, but in the general case they dont. You really should read Ian de Vries' Class-DE paper as it makes a lot of these things more clear.
Where Class-E can be considered as "ZVS voltage profiling for a single switch", Class-DE can be considered as "ZVS voltage profiling for a half-bridge." It uses the same type of resonant behaviour as used with one switch in Class-E but applies it in a different way so that top and bottom switches in a half-bridge turn on with no voltage across them, and the body-drain diodes never conduct.
Marko wrote:
So can anyone explain why can I build a 10Mhz sstc with class E amplifier and not with a H bridge.
It is exceedingly hard to make a half-bridge of standard power MOSFETs like IRFP460 produce any significant RF power at 10MHz or above. The main problems are:
1. High switching losses due to a large part of the total period spent transitioning the linear region. 2. Very high switching losses due to repeated discharge of Cds into the channel at turn-on. 3. Mismatched turn-on and turn-off delays causing shoot-through (cross-conduction.) 4. Complete inadequacy of slooooowwwww body-drain diodes to provide free-wheeling duty at this frequency.
However, you can make two IRFP460's produce several hundred watts of power at 5MHz or so by applying Class-E techniques to shape the voltage profile in the deadtime. The resulting topology is called Class-DE. Because I know you probably won't read the Class-DE paper I'll sumarise by saying it is achieved by taking a standard half-bridge, increasing the dead-time dramatically, bolting capacitors across each switch, and tuning the output matching network in a way that provides just the right combination of resistive and inductive reactance to the bridge.
Having designed Class-DE amplifiers for RF power applications, I can say the biggest limitation to increasing operating frequency beyond a few MHz is this... The bridge leg's mid-point voltage must commutate from one supply rail to the other during the dead-time. It takes lots of current to charge and discharge the MOSFET's Cds capacitances in a short time. As you increase the operating frequency less time is available for this to take place. So you either have to increase the deadtime to the point where it dramatically eats into the total switching period and considerably reduces output power _OR_ you have to tune the output network to get more circulating current during the deadtime to slew the mid-point voltage quicker. Either of these things actually give you less power out, for a given RMS current through the MOSFETs. As you increase the frequency this issue eventually becomes un-manageable. This is the point where a single transistor topology like Class-E can continue operating up to much higher frequencies, because it is not affected by these problems.
-Richie,
PS. I have replied here at length as I feel partly responsible for the whole Class-E Tesla Coil thing. I would like to point out that I never said "Class-E" was the world's best for driving Tesla Coils. When I originally wrote that article on my web-page I set out the task of trying to apply a cutting-edge technique from RF power engineering and transmitter design to a Tesla Coil application. I had also been asked many times if a Tesla Coil could be driven directly off the output of a radio transmitter. I set out doing this knowing fine well that it would be a challange because it had not been done before. The Class-E amplifier is finicky about load conditions and is really easy to blow up with a mismatched load, and Tesla Coils are well known for presenting a dynamically varying load to the driver, so you can draw your own conclusions about whether it is or isn't an appropriate technology to use! I set out on that project as a personal challange, and never dreamt that it would inspire a whole load of dinky high-frequency solid-state coils driven by Class-E RF power amps!
This site is powered by e107, which is released under the GNU GPL License. All work on this site, except where otherwise noted, is licensed under a Creative Commons Attribution-ShareAlike 2.5 License. By submitting any information to this site, you agree that anything submitted will be so licensed. Please read our Disclaimer and Policies page for information on your rights and responsibilities regarding this site.
Class E revisited
. But now i see even more clearly why the half-bridge is such a wonderful method (compared to this Class-E stuff) for higher power lower frequency stuff where you have the option of *not* using class-E.
