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 #2906
Joined: Sun Jun 06 2010, 02:20AM
Location: Dresden, Germany
Posts: 727
female chicken wrote ... I know this is a rather expensive solution, but have you tried one of the new SiC high voltage schottky diodes?
I mean i know it exists, but "tried" no. As you said, its expensive and the price difference between one SIC-diode vs PN-Junction diode with the same ratings covers more than the BOM-cost for an active clamp. Not to mention that you need two diodes per halfbridge Another issue is then, that you still have to fight ESL parasitics if you suddenly can switch this fast.. with SIC you spend a lot of money to switch faster in order to reach the next practical boundary which then needs active clamping again. Nothing (relevant) gained, just money lost. There is really no need to switch insanely fast in a coilgun anyway.. Switching losses matter in periodic waveforms and are part of thermal design considerations which arent applicable in a coilgun. It has such a negligible duty cycle and switching amount that every thought about optimization is unnecessary. As long as one needs active clamping one switches fast enough
Registered Member #11734
Joined: Thu Mar 21 2013, 08:44PM
Location: Brno Czech republic
Posts: 35
DerAlbi wrote ...
Thanks for the response Unfortunately there is too much bad information out there (mainly on youtube) so we wont change the world Considering your snubbers: the coil it self will oscillate when the current gets back to zero. To prevent this reaching the IGBTs, you can simply put some diodes in series. (Some beefy/cheap P600G from ebay will do, parallel them if you like. A snubber across the coil will help a lot. If you want to protect the IGBTs from a C-E-voltage spike however, a snubber capacitor wont so the job. There are several reasons: A true snubber does have a series resistance. Since a snubber is directly stressed with the turn-off current in a halfbridge, the ESR must be low enough to not reach the 600V limit during turn-off. Such a low ESR will however result in an extremely high turn on current. So nothing gained in the end. A RC-snubber does not work in a coilgun, not even a RCD-snubber because the diode is the bottleneck.. The problem is (i learned this "recently"), that the voltage spike on the collector is not caused by ESL, but much much more by the diode forward recovery behavior. (A diode does not become forward conductive immediately). The only way to combat this is to use active clamping: In your case, use a much larger gate resistor (you really dont need to switch faster than your diodes can handle the resulting turn on, i use 47R as gate resistor, thats still low enough to make the IGBTs faster than my diodes) and put a TVS diode (2 in series to get 450V < V_breakdown < 600V) between collector and gate. This will conduct any excessive C-E-spice towards the gate so during turn of a gate voltage plateau forms that will keep the IGBT conductive until the half-bridge diodes are ready to take over the current from the IGBTs. So you arent fighting parasitics like ESL but you are most likely fighting semiconductor effects in the diodes. If you think about using faster diodes - forget it. The only way around the diode forward recovery would be using schottky, but they have poor (over-)current rating, so you are bound to silicon junctions.
You really should measure this at low voltages before you try to optimize this with snubbers or what not. You have an oscilloscope! Use it
Well I am pretty sure that 1uF non inductive snubber cap across IGBT is completely enough because substrate diodes inside my IGBT modules are rated for trr 200ns and thats pretty fast. Actual IGBT's are twice much slower than substrate diodes even with 3.1Ω gate resistor. You must consider that these modules are rated for switching inductive loads in power inverters at high frequency and what I am trying to do with coilgun is pretty much normal operation for them. We use this setup at my university for even more powerfull aplications (electric car chargers, teslacoils, driver for train motor and etc.) without any problem ;)
Registered Member #2906
Joined: Sun Jun 06 2010, 02:20AM
Location: Dresden, Germany
Posts: 727
Dear Skynet trr (reverse recovery) is something completely different - it characterizes the diode when going from forward to reverse bias. I am speaking of forward recovery. That means going from reverse bias to forward biasing. The reverse recovery time is meaningless in a coilgun, because the current reaches zero smoothly (no step involved). However the turn on for the diode is hard, because you go from 0A to xxxA in no time. I have diodes that have a 30ns trr rating but they need well over 100ns to take steeply applied current while they allow a forward voltage of several 100V for that time! (instead of the usual 0.6V you would expect). During this time you need active clamping. You have to realize, that if you put a low ESL/ESR cap across C-E on turn on of the IGBT, you short circuit the capacitor during turn on with/by the IGBT. This will result in a dramatic current spike in the IGBT which results in at least 50mJ turn on loss at 450V. This might be more than any turn off loss that you try to minimize. Use active clamping. There is no way around it. You have the benefit of using only 450V and having a 600V IGBT.. use bidirectional TVS diodes and you are fine. If you are unsure about it, i had a post here that contained this image: Use only the TVS (put 2x 250V in series) and dont use D8. Use depicted 22R gate resistance. Part numbers would be 1.5KE250CA. I looked them up, they are easily available on Ebay. Active clamping is the only counter measure to a voltage spike that ensures minimum switching time and minimum possible switching loss while perfectly protecting the IGBT. Brute forcing a capacitor somewhere where you see a voltage spike issnt the right answer every time. But you can try... its your IGBTs, of course
Edit: I just captured a turn off (35A) on my halfbridge for you. The capacitors are charged to 403V at this point. As you can see the turn off takes around 600ns until its all finished. The overshoot is about 130V. This overshoot is purely due to forward recovery of the diode which struggles to take over the current immediately. For my circuit the active clamping does not kick in because its set to around 630V. It is possible that your setup may behave equalish you mine and you wouldnt need clamping because you have 150V headroom to the 600V igbt limit. But i guess you will turn off much harder current than 35A.. the forward recovery voltage will be much higher in such cases. When i get 130V @ 35A think about how your diodes will struggle at 200A or whatever.
Registered Member #11734
Joined: Thu Mar 21 2013, 08:44PM
Location: Brno Czech republic
Posts: 35
Hi Albi, in theory it sounds quite perfect :D But as I know using of active clamping is more risky for my transistors than putting caps across them. When active clamping engages there is possibility that transistor would go to linear mode (just few volts on gate) and you don’t want to have current flowing though IGBT's when its hundreds volts across them. That would kill them for sure ;)
The correct solution is precise geometric placement of components to reduce stray inductance’s and using snubber caps. That’s the way how it should be done. Inside IGBT module you have this sorted by manufacter so nothing to worry about. If it were true what you say no motor regen would be possible... What we have there hmm... current going from zero to hundred amps and it works like a charm without any active clamping. Interesting isn't it? If you use discrete components its even more crucial to minimize every possible cm of wire that you don’t need. Also you have to count with some voltage spikes in design. So my actual voltage wouldn’t go over 350V on these IGBT's. Its rule number one that you use components rated for at least twice much voltage that is used in circuit. So if i ever wanna go for full power it wouldn’t be on these boys ;)
Registered Member #2906
Joined: Sun Jun 06 2010, 02:20AM
Location: Dresden, Germany
Posts: 727
Ah i i know now where your thought process comes from! Thanks. -> Your "correct solution" is unfortunately flawed. First of all, a physical well done build will not reduce semiconductor effects. As i said, actual ESL might be way less significant than the forward recovery (which mimics inductive behavior of a diode voltage wise, but resistive behavior power loss wise). There for you can shuffle and optimize your components for ever never becoming better than your components allow it. So you are currently overthinking parasitics when the problem might actually be somewhere else. -> Also your "linear operation" argument is flawed. Think about it.. a capacitor can only store energy in order to not let it heat up the IGBTs. At 450V a 1uF capacitor hosts 101mJ of energy. (at 600V its 180mJ). So your snubber only helps as long all the energy in your assumed stray inductance is less than this. And yes, during turn off then the IGBT is protected from this stray energy as long as the capacitor eats it all. But where does the energy go? You can dissipate it it away in a parallel resistor but that takes time and introduces a static current. As soon as you turn on the IGBT again, the 101mJ are simply dissipated at turn on, because the capacitor is short circuited then. And the heat is generated inside the IGBT. So overall you dont solve the problem you mention, you simply shift it from turn-off to turn-on. Nothing gained here. -> Then again think again about what amount of energy you are talking about! Your capacitor can only handle 180mJ switching loss anyway, this is something your IGBT silicon can handle too and even more - this is why active clamping wins in every position: it dissipates the stray energy during switching, which also happens in your solution too (just not at turn-off, but at turn-on). Additionally active clamping works even in a very bad build, while your capacitor solution is restricted to 180mJ stray energy capability.. -> Btw, 180mJ heat dissipation is equivalent to 112.5A for 1ms (@1.6V VCEsat) on your IGBT (roughly 250A peak current on a 1ms saw sooth current wave form). Given your setup, i guess you will exceed this anyway (either current wise or time duration wise, likely both). So if you think that this energy can destroy/harm your IGBT, you have a way bigger problem.
-> Now lets continue: 180mJ allows for 9uH stray inductance @ 200A or 2.25µH @ 400A. You can see that this is excessive!! You will never have stray inductances that big. So lets think about about the turn on, where your CE-Capacitor is short circuited under normal conditions: you will inevitably dissipate 101mJ there (=1uF @ 450V). Even 101mJ is equivalent to a 1.26µH stray inductance @ 400A. You will never ever have this! The same argument holds true for 350V... Therefor your capacitor solution will increase your switching loss, instead of minimizing it. Even worse: the turn-on peak current due to the schort circuit can even latch up your IGBT, turning on its parasitic SCR, and you are just exploding your IGBT right there.
I can repeat my self, active clamping is the way to go, there is no more optimal solution.
wrote ... Also you have to count with some voltage spikes in design. So my actual voltage wouldn’t go over 350V on these IGBT's. Its rule number one that you use components rated for at least twice much voltage that is used in circuit.
Argument wise, you are shooting your self in the foot here. Active clamping will !!ensure!! that your voltage spikes will never reach your IGBTs Vce-rating, no other solution does that for you, as you seem to realize yourself. Your rule number one is fundamentally wrong: its fear driven and not specification/engineering driven. It wastes half of your IGBTs rating for something thats easily avoidable by choosing the right circuit/topology. Your IGBT can handle 600V. Period. I push my 650V IGBTs to 630V clamping voltage. No fatalities yet
So in conclusion you are adding a capacitor in order to *increase* your switching losses, *risking an SCR latchup*, while *not gaining confidence* in your build that you will adhere to your voltage specifications while making the build *even bigger* with *heavier* components, all in order to avoid using 2 diodes that would solve all your issues. Every point you wanted to make better in your build you actually worsened; and its easily shown with just energy conservation arguments, no special math or simulations involved. DONT.
However if you can show calculations that show why you choose a 1µF capacitor, i would love to hear an explanation
Just another remark: i will stop bothering you changing your topology. I made my point, you made yours. I can destroy your position in every respect, but its about you to choose to listen. The whole argument is about some mJ energy amount that is suboptimal. Going for your 1µF @ 350V will likely work, if the IGBT does not latch up which is a serious risk. I just want to help you here, and as i said, there is a lot of bad information out there specially regarding coil guns. Your misconceptions about your betterment of your circuit would make another example where something was simply not thought through. Yet it will be a (hopefully) working solution, but "working" means only "sufficient", not "good". Yet the "working" will make you feel that it is "good" leaving you with a fundamentally flawed, false experience with gained confidence in decision what can be objectively proven wrong. I just dont want you to be another person who, in the end, didnt actually know what he was doing while feeling confident that he is on the right track.
Registered Member #11734
Joined: Thu Mar 21 2013, 08:44PM
Location: Brno Czech republic
Posts: 35
Oh man... Calm down :D At first my design is not focused on effiency at all... My coils are hand made garbage, my barrel have thick walls so I don’t have big expectations. Some mJ in switching loses are meaningless for me. And I am trying to build powerful coilgun (I just need to get to specific muzzle velocity at all costs because it would be replacing big air canon at my university) so if it would be big, heavy and complicated it’s okay...
Yes, I know that with snubbers I will dissipate some energy to my transistors but it’s just fraction of what they could survive without any problems. About my IGBT modules... Diodes inside these modules are designed to take over maximum current of transistors. => It just way how it works in every frequency converter, motor driver and etc. without using active clamping. Foward recovery exist but it's not as big prolem as you think. IGBT's in modules are latch up proof (yes manufacturers can make them nowadays) so again => nothing to worry about
About my rule for design of VCE rating... I don’t force you to use it. And man... It’s not fear at all its just failsafe you never know what could go possibly wrong in your circuit. Some protection could fail and if you have your design at its limit for normal operation what it will do in fail condition? The cost for more VCE room is acceptable for more beefy circuit.
And 1uF snubber is just commonly used value in power circuits for my size of IGBT modules but you can calculate it. Some manufacturers like Infineon released equations how to precisely determine value but it's not worth spent time if 1uF works flawlessly ;)
Registered Member #11591
Joined: Wed Mar 20 2013, 08:20PM
Location: UK
Posts: 556
An interesting discussion. Assuming an modern IGBT (if latchup isn't really a problem anymore, I have never experienced it, but I rarely design circuits with IGBTs), the old fashioned snubber seems the way to go. The switching frequency of most coil-guns is *very* low compared to even a large SMPS, so 0.1 J of energy dissipation seems negligible if you are switching several kJ. If you require ultimate efficiency, small size, precise current control with PWM, and fast fire rates, maybe active clamping is the way to go. Remember that Skynet's Coil-gun has very different specification to yours DerAlbi.
BTW, the hen in my name is short for Henry. Why I choose hen918 I don't know. I've used it for most of my life.
Registered Member #2906
Joined: Sun Jun 06 2010, 02:20AM
Location: Dresden, Germany
Posts: 727
"Skynet" wrote ... IGBT's in modules are latch up proof (yes manufacturers can make them nowadays) so again
Henry the chicken^_^ wrote ... Assuming an modern IGBT
No worries, i see where your ideas come from, but lets see the datasheet: 1) "not recommended for new designs" = old 2) no short circuit specification = no proof of latch up resistance
Stating that "other IGBTs are latch up resistance, then mine is too" is.. well what is it like? Maybe like jumping off a bridge without a rubber band after you saw bungee jumping on TV. Never EVER assume. When there is high power involved, only facts count. Also if others use some capacitor somewhere, it doesnt mean, its the right solution to you too. You also assume that forward recovery is not an issue when i went the extra mile to provide hard evidence to the contrary. I know that forward recovery is not a very often discussed topic, yet it is real and in fact maybe an issue.
Hen, I dont think you interpret my concerns the right way.. its really not about efficiency.. because that does not come from the IGBTs in the end.. and i am actually interested in what efficiency will be achieved - if its bad its also an interesting information to me. I realize that his setup is very different to mine, thats whats makes it interesting. But it does not mean, that we have no common problems to solve. My concern is about doing the right thing regarding the circuit design. This is the design of a coilgun switching stage, not an SMPS. Its really about educating about misconceptions that seem to be prevalent here. Specially when its obvious that not even the most basic calculations were made during reasoning about the topology.. If Skynet says, in a SMPS, IGBTs are used this way [with 1µF], and his conclusion is that its ok to also add an 1µF capacitor, thats a fundamental problem! I am not sure if he understands that maybe the SMPS implements ZVS where the capacitor isnt even a snubber. Are you sure, its not a resonant cap instead? Should i really shut up? Better safe than sorry when it comes to high power switching. I also would like to ask you why you would prefer a snubber capacitor instead of the much more universal active clamping circuit - specially when its smaller, cheaper, simpler and more reliable than a capacitor... is it because.. its unknown maybe?
It’s not fear at all its just failsafe you never know what could go possibly wrong in your circuit
..that is fear. If you dont know what will happen, why prepare for the worst... because of.. optimism?
Registered Member #11734
Joined: Thu Mar 21 2013, 08:44PM
Location: Brno Czech republic
Posts: 35
Albi... Before I stared to mess with power electronic I’ve studied microelectronics and as I know my IGBTs are trench gate design which is latch up proof. And why prepare for the worst? Maybe if some of my protection fails (and yes it could happen to everyone) and I was and dumb that my voltage ratings are at its limits in normal operation? I don’t know what they learn you at your university but we are just used to have comfortable room. It doesn’t give you confidence in your design but chance of catastrophical failure would be less possible ;)
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.
My school coilgun poject
wrote ...
Another issue is then, that you still have to fight ESL parasitics if you suddenly can switch this fast.. with SIC you spend a lot of money to switch faster in order to reach the next practical boundary which then needs active clamping again. Nothing (relevant) gained, just money lost.
Every point you wanted to make better in your build you actually worsened; and its easily shown with just energy conservation arguments, no special math or simulations involved. 