Welcome
Username:

Password:


Remember me

[ ]
[ ]
Online
Guests: 24, Members: 0 ...

most ever online: 333
(Members: 3, Guests: 330) on 06 Jun : 15:15

Members: 3396
Newest member: ghilinta.daniel
Members Birthdays:
All today's birthdays', congrats!
Reaching (32)
ryanshow (30)


Next birthdays
07/23 Will (30)
07/23 Arlecchino (33)
07/24 Jim_VE7UV (57)
Contact
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.
Forums
4hv.org :: Forums :: Electromagnetic Projectile Accelerators
<< Previous thread | Next thread >>   

Another Coilgun

Go to page  1 2 3 4 [5]
Author Post
DerAlbi
Wed Dec 14 2016, 01:53PM
Registered Member #2906
Joined: Sun Jun 06 2010, 02:20AM
Location: Dresden, Germany
Posts: 524
Its about 1kg of plastic equal to 50€ worth of filament.
Yes i am using copper for the prototype because aluminium is extremely expensive because it would be custom made with minimum 10kg per wire diameter which is quite a lot of wire considering its aluminium. For the prototype its not an issue. It will be uncomfortable heavy but it will work.
Square aluminium is also only available with >=0.5mm "diameter". My first coils would need 0.25mm so this prototype now also explores how i can adjust to thicker wire through PWM switching in the first few stages and what it does to the projectile and efficiency.

The prototype in the beginning exploded at stage 16 due to mechanical failure of a diode in the half bridge. Look at the beginning of the thread for photos of exploded parts, It was the end of the big prototype. Since then i did never shoot again.

Building the Coil-Winder now adds a bit delay to the project but being able to wind the coils is important and it also has the benefit that i am developing software for the microcontroller architecture that i am going to use in the end.

Meanwhile i have implemented the stepper motor drive and acceleration profiles and i am confident that this coil winding machine can do up to 3 turns per second. Its quite fast compared to hand winding however its quite slow considering its a machine.
The problem is the stepper motors loosing torque as they become faster. The fact that i added a 5:1 gear is therefore a problem, however i need it to ensure i have sufficient torque. As consequence i would need a way bigger stepper motor for direct drive of the winder, which means more cost and also much more difficulties driving the motor since there is no payable of-the-shelf solution anymore.
Anyway. first experience with a coil winder. prototype. good enough.
Back to top
DerAlbi
Mon Jan 09 2017, 07:27PM
Registered Member #2906
Joined: Sun Jun 06 2010, 02:20AM
Location: Dresden, Germany
Posts: 524
First very primitive test inductor.

Its just a single layer of 0.2mm wire to see how precise the steppers and the wire tension algorithm handle the job.
Its denser than hand wound for sure. Not by much however.
Wound with 2 turns per second. 90% of the turn is completed until the horizontal correction is done. i will have to learn about the different behaviors. i guess with such fine wire this is ok, with thicker wire maybe not. But its in no way a limitation of the machine, its just the way i programmed it to do the job. i can also go gradually side-wards of course.
The start and the end of the inductor is not messed up. i programmed it that way to start and finish with a few turns of larger wire distance.
Now i need a desktop program to make more complex movements.
The white between some turns is not bad wire distance.. its resin that comes through.. light does the rest. (i wanted to do a finished coil just for fun.. therefore i glued it in place directly - taunting few minutes. )
Back to top
DerAlbi
Sat Feb 18 2017, 09:20AM
Registered Member #2906
Joined: Sun Jun 06 2010, 02:20AM
Location: Dresden, Germany
Posts: 524
So its time for another update again. There was not much progress made to be honest because i am catching up on tooling here. In a sense this much is progress but in terms of the actual gun coming to life.. hmmh. As i posted above i am still building the coil winding machine. The PC-Control program is basically done and ready to use and what is still missing is a good coil former and actual knowledge how to wind a coil. Layer-changes are a bitch.
For the coil former i actually bought a CNC milling machine off Ebay (CNC3020T with 800W Spindle) which was a lot of work to get operational and of course its quite time consuming to get into all the software associated with it - tool path generation and so on.
However with this now i am quite capable of doing a hole lot of stuff in combination with the 3D-Printer. I honestly feel amazing right now. I feel the difference in quality i am able to produce if i can make the proper tools for the job. 3D-Printing the coil former was not an option btw, because the layered (surface-)structure of the print disturbs the coil turns. The only solution is to sand the surface which in turn will produce dimensional errors. So milling is indeed the right way to do it.
Here is a side wall of a milled coil former:



I wont go deeper in to its features because this part is not finished yet an the next iteration of the design is milling right now... i mill in POM plastic which is good for not adhering to epoxy.. the surface is really smooth but the material is a bitch to work with. Its too ductile and therefore hard to cut cleanly. The tool marks you see are purely optical.

As for the whole coil winder side project i want to share my doubts with you. I catch my self loosing my focus to such side project which now results in the whole CNC stuff and so on. I get side tracked a lot which does not help the project at all. Its sometime really depressing how much work is ahead of me when i realize how much of the work i dont even realize. Its always like a factor of pi more.
My problem with the whole project is that i really want to achieve something new and like Barry said - "next gen". Its impossible to invent stuff in this field because everything thats viable has been done.. its only possible to improve the current state of the art by just doing everything the absolutely right way or at least as good as i can do.
This results for example in a more complex coil shape which in turn takes time to perfect. I try to keep telling myself that all for a greater good, so when my prototype is finished i have something to show - and not only that.. i also extent my skill set so if the future ever brings me to a point where i will beg for money to make the gun bigger people can trust me. Yet i realize that conveying this trust is impossible and cheap marketing bullshit would be way more effective.
I also do not know of any alternative to how to make my desired coil shape except having everything custom made.

Anyways. I am not far from making my first coil fully automatically and in a repeatable way. I will share the results with you. Time will tell if the machine could improve the L/R-time constant of my coils which is currently 1.7ms. Actually i think it wont change much.
On the one side my coil geometry becomes worse because the inner layers have less turns to accommodate an air cavity inside the coil (this is required by my projectile sensor system) which reduces power density. On the other hand i really optimized the inner shape of the coil to reduce the air gap and to reduce overall wire length by a few percent.
Additionally the overall design (i hope) will experience an energy density increase of >10% because the coils are only separated by 1.0mm thick PCBs instead of 4mm wide spacers which hosted those unreliable light traps.

Edit: i have HUGE problems with my router:

cncreport.pdf
Back to top
DerAlbi
Thu Mar 30 2017, 07:12PM
Registered Member #2906
Joined: Sun Jun 06 2010, 02:20AM
Location: Dresden, Germany
Posts: 524
I give up.
Not with the project but with coil winding. I manage to get good results for the first few layers, but my complicated coil shape is a bitch. Its not worth the trouble and the hunt for the perfect coil is wasting my time - i think my wire is not reliable enough (tolerance in diamter..). I have decided to continue to work on the electronic side of things and when the PCBs are ready i will wind coils again and i will live with the ulgy results.

More details: the coils we want as coilgunners are called "orthocyclic. It is well known that the first layer of such coils is the most problematic one and if there is the slightest error there it will amplify in later layers. Also the layer change on the side of the coil has a certain behavior that is sometimes hard to deal with. In addition my coil former is not perfectly round - it has rather sharp corners.. those corners make things worse in terms of orthocyclic winding. Turn crossing that usually moves a bit jumps there instead leaving a gap that will disturb later layers.
This is a systematic problem so i will never get pleasing results.
Now i have to think about what "pleasing results" are actually are: i have wound a lot of coils for my bigger prototype that started the thread. OF course there were good ones and bad ones, but in terms of electrical properties the ugly kids never performed any worse. So there is a lot of aesthetics involved thats more important to a human as to the actual electronics. As is it often the case.
Also.. is it worth it? When i think about the future of the project i see square aluminum wire as the way to go. This wire will have totally different properties. It wont have the grooves to auto-guide it self and normal helical winding should be possible. I think the worst that can happen is that a layer change twists the wire for 90°.. and i actually dont know if that would be so bad. So getting everything right now is not indicative for the future of the accelerator i think i should care less. The fact that i do not use square aluminium wire right now is purely cost related. I cant get it for cheap and i cant get it in small quantities.
However i learned a lot.
- I made a coil winding machine ( i wouldnt build it again that way which is progress enough.
- used stepper motors for the first time[/list]
- I used STM32 for the first time in a custom project which gave me the chance to get a good hardware abstraction layer in software (this is also the microcontroller of choice in my final design)
- I also used QT as GUI for the PC
- I also finally got my milling machine/router (i never know when to use those terms) running and i changed my CAD from SolidWorks to the free Fusion360 which also increased the broadness of my abilities[/list
All in all it was a huge gain not only for my work shop but also for me, but i have to move on.
I do know now how my coils will look like, which width they will be and at least i know that it can be done.

So back to electronics. I cant really go into details because i wasnt in touch with the whole circuit and PCB design stuff however i know for certain a lot of my electronics is driven by the desire for quality yet still designing on edge. The quality part of this sometimes causes decision that arent cheap or easy to implement. I think this has to change at least for the electronics for the acceleration stages.
It may very well become another fail, however its worth exploring if caring about every parasitic effect is actually worth it or if i can get away cheap. i just have a lot ideas in my head.
Every coil has a SCR in series.. this SCR has to be turned on. But how to do that in a asym half bridge with all the voltage swings... the clean way is a gate drive transformer for every SCR. But this sucks. Lets see what i come up with

Again: sorry for all who wanted to see a complete coil wound... i can still make a video if you like.. its not like its not working but the tolerances mess things up pretty quickly.
Back to top
DerAlbi
Sat Apr 01 2017, 07:20PM
Registered Member #2906
Joined: Sun Jun 06 2010, 02:20AM
Location: Dresden, Germany
Posts: 524
Ok, last post left you guys hanging... not nice. I got the SCR-Fireing visualized and i am.. idk.. i think it might work.
Have a look


The usual way to fire the SCR in that situation is to use a small transformer. In this way the control electronics can be completely independent of the violent things that goes on in the half bidge. In reality the ground of the control circuit is connected to the ground of the half bridge, however this ground connection is far from reliable. The problem with this configuration is the transformer involved. its just expensive.. and in a multi stage design you pay with the number of stages in this part of the circuit.

My alternative solution uses a PMOS to pull the SCR gate high. When the low-side IGBT is turned on the gate is basically ground referenced, thus the gate should fire. The issue here is again the bad ground connection. There is a lot of inductance in the wires and a lot of ringing involved when the SCRs turn off. So the "ground" of the SCR is by far in any way clean. This rises two concerns, one of which is actually really interesting:
First the "dont care"-issue: due to ringing or some interference the SCR could fire too late or refuse to fire for some time. i dont regard this as a big problem because its assured that the SCR will eventually fire. And the time scale is way too small to care... i mean even 5µs delay means 200kHz eigenfreiquency of the ground loop and i thing this should be way higher - even in a bigger gun.
The second issue is interference and the implied parasitic triggering of the gate. If an SCR turns on by accident its certain death to the IGBTs. I hope the capacitor and pull down resistor at the gate is enough to manage the problem.
Back to top
hen918
Mon Apr 03 2017, 01:12PM
Registered Member #11591
Joined: Wed Mar 20 2013, 08:20PM
Location: UK
Posts: 423
I would go for the conventional circuit. There are so many things that could go wrong with mismatched ground references, apart from the severe consequences if anything goes wrong. It could save you a lot of head-scratching later on when things would otherwise be exploding unexpectedly.
Back to top
DerAlbi
Mon Apr 03 2017, 07:54PM
Registered Member #2906
Joined: Sun Jun 06 2010, 02:20AM
Location: Dresden, Germany
Posts: 524
Thats not the reassurement i need!
Still thanks for that - this forces me to further explain myself - because i thought about the situation a lot during schematic level design..
If I use some 60V PFET in the new design I actually have immunity up to -48V on the low side of the coil because only when the mosfet avalanches there can be a high enough current flow to accidentally trigger the gate.
Positive ground swings arent so bad either because the diode protects the circuit from current flow. All that can happen is some parasitic current flow which is easily compensated by the 22nF (its a capactitve voltage divider xx pF vs 22nF). The more i get into the circuit the more i like it.

Also the conventional circuit does have issues too. The problem is there in the first few stage where inductance is high - the current does not rise fast enough to reach the latching current during the very short time the SCR is fired by the 4.7uF pulse capacitor. The pulses last only some micro seconds there, so its also not the most reliable circuit - however its failure behavior is kind of better.

If you find problems in my logic please argue with me - this is important ^_^
Back to top
DerAlbi
Thu Apr 13 2017, 08:51PM
Registered Member #2906
Joined: Sun Jun 06 2010, 02:20AM
Location: Dresden, Germany
Posts: 524
I was bored.
I opened up a coil - just to see how high the fill factor is. Here is the picture.



1) obvious: the conductor fill factor is quite high. I think its as high as it can get - meaning ~90% (i think the limit of orthocyclic winding is 91%).
2) a defect propagates though the layers. You can see a diagonal line at the left side. There was one turn just a liiittle out of place and it never got back to normal. Keep in mind that i wound the coil manually and hand-corrected every turn.
3) the second layer is on top of the first layer. Its the region where the wire goes from one gap to the next one.

The most devastating implication is the high fill-factor. previously i thought that i have quite a bad fill factor and thought that when switching to aluminium using a rectangular wire can make up for at least some of higher resistance. But it seems rectangular wire has no benefits when it comes to fill factor.. i guess its just easier to wind (which is also important).
I have to revisit my old thoughts on how efficiency develops with a worse conductor.. because aluminum will hit a lot harder than i guessed.
Back to top
DerAlbi
Fri May 05 2017, 03:00AM
Registered Member #2906
Joined: Sun Jun 06 2010, 02:20AM
Location: Dresden, Germany
Posts: 524
All PCBs are finished. 10 different PCBs, 500eur order.
I am currently redesigning the case. I have imported the outlines of the PCBs and the most critical connectors to the CAD-Tool (Fusion360) to see if everything lines up as planed.
The whole thing is full of stuff. Totally crowded, every volume used.. bigest clearance inside the case is like 5mm.


The accelerations path with its sensor-PCBs is not shown. Also the plastic components are hidden except the trigger section (transparent) to give a reference what you are looking at.
I also included some component information. (If anyone gives a f.)
I just realize that the USB-connector that is mentioned is not shown. Its on the other side; it protrudes behind the HB-Power supply, the connector is soldered on the bottom side. Sry.
Back to top
DerAlbi
Mon May 15 2017, 06:36AM
Registered Member #2906
Joined: Sun Jun 06 2010, 02:20AM
Location: Dresden, Germany
Posts: 524
Did i mention that i redesign the case? What a shitload of work. Also i decided to go with the flow and try to make something presentable - meaning adding funky outlines beyond of whats functionally important. My biggest issue here is that i am not really an artist. More an autist Hehe, sry.
This is also the week where i expect some PCBs to show up. As far as i am informed the first batch will be all the 70um PCBs which are the Cap-Bank and the Halfbridge. I hope the other PCBs will follow soon. Man... guys. You dont believe how pumped i am. This is the climax of 2 years of concept work. Yes, it has been 2 years since the last projectile was accelerated. Since then i worked hard, refined, tested, dismissed and accepted solutions... everything thats inside the gun has been evaluated - it should work. However nothing was ever assembled - it always was just concept work with a lack of real world context. Now everything comes together and i hope for the best fearing the worst. I know that some concepts never worked in my tests, but i learned from the failures and i was confident that this iteration will work. However confidence seems to be a volatile thing as the PCBs become reality. Hoooly fuck. I suffer from PTSD. Just this time it means Pre-traumatic-stress-syndrome.
Anyway. The fun part:









So i want go go for a 2 color scheme. Black is just neutral enough and does not become dirty or filthy from handling the plastics. I plan to use a green or orange mid-line which gives some kick to the design. I also opted for showing my balls - meaning the caps arent just encapsulated in a box, but they show through for all the sexiness like a really tight thong .
I am currently not sure if i will add this barrel stub thingy in the front. i am also working on giving the top side some structure, so the shape is more interesting to look at. However this is quite a challenge since modeling such stuff takes hours at a time and the motivation is low if you just dont know how the idea turns out. i currently dismissed 2 designs already.

I am pretty happy how the cap-section/sextion turned out... here is a 1:4 model i printed over night. Notice that the printer fucked up (because the model wasnt really made for printing this small) and it gave the thing some horizontal structure on its upper surface. I think this will be my next option for improving the looks.



And yes, a small show off. Small albi being proud, sry.




So not sure about the size. I think for the specs (20J, 40m/s) it ended up pretty bulky. On the other hand i will be able to empty my 6-proj-magazine in the time most youtube videos realize they should edit/cut/remove the cap charging process from the video while still being smaller and not much weaker. This form factor is not the best advertisement imho. But if you think about scaling up.... and when you realize that its just about adding stages and adding a few caps and keep in mind that a bigger gun will inherently have more space for electronics, i can imagine that this can be attractive.
Back to top
DerAlbi
Sat Jun 03 2017, 06:58PM
Registered Member #2906
Joined: Sun Jun 06 2010, 02:20AM
Location: Dresden, Germany
Posts: 524
Time for some texting.
The PCBs were populated and i had a good time testing the things. It turns out that for example the power management PCB is working great - except a flipped footprint i had no problems with it. This means that the accelerometer, inductive projectile sensor for the magazine reload, the touch-channels for people-sensing and all the other features seem ok.
I am currently working on the halfbridge, there the sun issnt shining anymore
Not only that i fucked up my schematic in such a way that the IGBTs are per default all ON, but also the clamping circuit for the inductive peak does not behave.
Remember the IGBT active clamping i showed one page earlier?


Well it turns out that this suffers from some oddities i did not expect. This is literally the first time where diode FORWARD-recovery is an issue. Have a look:



Yellow is the Collector, blue is the Clamping voltage between C2 and R4. The cap bank voltage is 50V. The current should be around 55A. You are looking at the low side IGBT.
It seems like the gate driver turns the gate off very fast. The IGBT reacts after some time and the Collector voltages spikes. (This spike needs clamping). This spike should in theory make D3 conductive, therefore keeping the gate high by conducting the pulse through C2. However it seems like the diode is too slow. It take its around 20ns to become conductive and another 20ns until the gate has the right voltage for the active clamping. Those 20ns are catastrophic as you can see.
Anyway, if you want to read datasheets: the IGBT is
The clamping diode is unluckily the forward recovery time is not specified, but i never would have guessed it can be that big.

Currently the resistors R2 and R4 are both 6.8Ohm. I guess i have to change that to 33Ohm for the gate and 0Ohm for the clamp. Hopefully this will give slower switching and time for the clamping diode to function properly.

EDIT: Woooow.. thank you forum for being a good listener. By linking in the datasheets i was motivated to see if someone else could find something i have missed Well. the diode has a non repetitive peak current of 1A (pathetic!) but.. well.. with my 6.8R i currently stress it to 2A during clamping. I do not think that this has any lifetime implications since its in the ns-range, however in terms of recovery (forward or reverse) this is of course important. I guess the plan to increase gate resistance is a good one. The 33R will result in only 400mA.. i will post the new waveforms soon ^_^

Edit2: I changed the resistors, but unfortunately the behavior does not change. oh oh

Edit3: Last thing to do is to change the diodes. And indeed, it seems i have picked one of the worst ones. See how the others perform: (blue is now the direct gate voltage!)
This is the RFU02


And this is a much beefier CD1408


It seems like the bigger diode performs noticably better. I am not sure why however. The bad thing is now that this is indeed a matter of testing. None of the diodes have their forward recovery behavior specified. I guess the bigger diode persorms better mainly because its acting much more capacitive near the point of action.
I need schottkys! (since they do not have recovery beahvior at all). But 600V and really small. Not sure if they exist.
What a setback.
Back to top
Finn Hammer
Sun Jun 04 2017, 08:48AM
Registered Member #205
Joined: Sat Feb 18 2006, 11:59AM
Location: Hou, Denmark
Posts: 725
DerAlbi wrote ...


I need schottkys! (since they do not have recovery beahvior at all). But 600V and really small. Not sure if they exist.
What a setback.





Finn Hammer
Back to top
Website
hen918
Sun Jun 04 2017, 01:46PM
Registered Member #11591
Joined: Wed Mar 20 2013, 08:20PM
Location: UK
Posts: 423
I was about to suggest SiC diodes. Very good. I use them for rectifing high frequency stuff as the fast reacting-ness makes switching losses negligible, (at least according to Cree, there are some relating to the junction capacitance, although this of course doesn't prevent it recovering instantly)
Back to top
DerAlbi
Sun Jun 04 2017, 03:12PM
Registered Member #2906
Joined: Sun Jun 06 2010, 02:20AM
Location: Dresden, Germany
Posts: 524
Thanks for your search guys The problem with the linked diode is that it is 7mm wide the other one is even bigger ; i need something close to a 1206 package.. (thats like 3.4mm) I guess diodes with such capacitance will avoid the whole issue due to miller effect leading to slow switching and unnecessary IGBT heating. On the other hand, the above waveforms show, that no matter what, i do not need to switch faster than 200ns since thats the clamping time period anyway. maybe i should increase the gate resistance even further? I mean the less i ask of the diode the less forward recovery time it will have. maybe.

I just tested some theory: maybe a bigger capacitance diode might help. (due to introduced miller capacitance close to the region where the diode may become conductive anyway) So i paralleled two diodes to increase their perceived capacitance.
It does not help. The peak is quite the same i guess. Then i wanted to know what determines the peak size, so i increased the current to 100A. (i think so, cant really tell)



The higher gm of the IGBT makes the circuit oscillate! Or is it the gm of the diodes that increased... very possible.
Now thats something. So Gate resistance to maybe 47Ohm and clamping resistance to 22Ohm ? Let me check. I hate changing those resistors -.-

47R gate resistant | 22R clamp: all nice again but the switching times went up a lot. (count the squares from the edge of the gate to the tip of the collector)



Its interesting to see how miller capacitance seems only to matter a really low Collector-Gate.voltages. Its also interesting to see what the gate resistance actually does. It does not shorten the clamping time, its just slowing the initial collector rise time. very interesting. I think i will half both resistors now.

24R gate | 11R Clamp:

notably shorter switching. I will now remove the paralleled diode and see how the peak behaves.

"smaller" diode (not 2 in parallel anymore)



more prominent peak So capacitance helps. Or actually the forward recovery voltage decreases if the current is shared between two diodes.... not sure which effect dominates.

So i used 47Ohm for the gate and 47Ohm for clamping. I increased the gate resistance because i analysed the images again. Even with 47Ohm the switching seems slower, the avoidable lossy part of the turn off (the miller plateau before clamping kicks in) is pretty much insignificant even it changes by 50% in duration.
So now i tried the clamping circuit in a more realistic coilgun scenario, meaning a repetitive switching in a short time. (i switch 8x in unter 2ms)


As you can see the clamping voltage rises a lot because the 10nF capacitor gains voltage. What is also interesting is that the clamping time becomes shorter while the voltage rises . I ask myself what happens if the caps are charged to a higher voltage, since the clamping duration should decrease even further its quite possible the whole problem vanishes to thin air.

I just edit the crap out of this post -----
The diode found by Fin Hammer seems to be symbolic for an interesting discontinuity within the technologies. There seems to be a real problem finding small high voltage diodes based on a schottky junction. I ask myself what the problem is or why there is no demand for such parts. Schottky is nice for fast switching and SiC can even go up to reasonable high voltage. I dont see why manufacturers try to satisfy a high power marked as soon as high voltage is involved... what about the people who just need a few 100mA.

----
Anyway. I discussed the problems here with my colleagues today because this is the first time forward recovery bit me and it thing thats interesting to share. While discussing the issue i was forced to describe why i need clamping in the first place and basically my answer until now was "ESL". (changing the current direction inside the caps will create a voltage spike due to the ESL)
Hooooowever that seems wrong My cap-bank is built very compact. Additionally the electrolytics are supported by foil capacitors which form the interface to the half bridge directly (see my cad model above). Measuring with the oscilloscope reveals that there is indeed no voltage spike across the caps - the foils do a very good job. The only step you see is from the ESR, which seems around 5V/100A.
So why the need for clamping? Its the forward recovery again! The diode of the half bridge suffers from the same problems as every other diode. The problem here is, that its inherent and i cant really change that because schottkys will not work there due to their limited overload capability. If there was no forward recovery the IGBT could switch faster and i wouldnt even need the clamping, however as it seems the forward voltage can happily exceed 200V or <50ns for the Diode . With my next order i will try a and see if the "high speed" in the datasheet is worth anything. The forward recovery is actually specified for the diode. My hope is not to avoid the clamping ciorcuit, but to minimize its interaction.
Specially with my current thought about PWM-ing the coils to control the current more tightly easier reducing switching loss is a significant contribution towards a more reliable half bridge.

Well.. conclusion time: its an interesting time to be a coilgun builder
Back to top
DerAlbi
Sun Jul 02 2017, 11:12PM
Registered Member #2906
Joined: Sun Jun 06 2010, 02:20AM
Location: Dresden, Germany
Posts: 524
Albi in tears
Guys i learned something the hard way, so i want to share it.

Everytime you build a cap-charger, be aware that the output rectifier diodes might fail. One way to limit the damage is to avoid that the full cap bank energy discharges in your circuit.
What helps is to add secondary diodes between the charger and the cap bank.
So put just as small as possible output capacitors after the rectifiers (i got 220nF)... only connect to the cap bank from there by using additional diodes.

It would have saved a lot. I just destroyed all chips in my build. Huge setback. One day gone just for the repair and still not finished.
I was so close to having something presentable for a new post here.. but now.. nooh
The way the damage occurred was that one of my output rectifiers died, and the cap bank discharged through it by blowing the hell out of the PCB... This lead to a disconnected cable that went flying over the PCB while the cap bank held some residual charge. With a flying cable under voltage going across the PCB... well.. enough said.


Edit: i got it back up now and i isolated the problem. It seems that under full power the stupid voltage control circuit does not work anymore. This is a really interesting problem. I havent looked deeper into it, but its scary. This explains how the output rectifier exploded and caused the devastation. I just had 850V on my caps again while the target/stop voltage was set to only 200V. Kind of insane. I luckily reacted faster because i was on the edge of my seat while trying the full power test this time.
The current solution is that i enable the charger just for 450ms. With a starting voltage of 25V on the caps this yields 430V final voltage (again with 200V target voltage ) getting there with only 10V input voltage.
This equates to just 135W of average power. I am not satisfied with that however the maximum power is not tuned in (i have to change timing resistors for that)
The average power will also increase once the caps start at 100V since low output voltage reduces power delivery. (this is what i expect in a real scenario with repeated shots).
Anyway... its a good sign. The problem is identified and the performance is promising. It also will not explode anymore. I am pretty sure... even if everything that prevents it is a software intervention.

Getting hold of the problem... i am not sure if this will be easy. I suspect this has something to do with a massive ground voltage gradient across the PCB since the very high currents.
Also the primary inductance is just 500nH.. so every nH of PCB parasitic will act as an inductive voltage divider. Adding the switching frequency of 600kHz... its shakes the poor little electrons quite a lot.
Backside



The whole setup:


Left: magazine control and power conrol unit. i think i spoke of that before.
Middlle: the power thingy. The thin part is voltage regulation (12V sepic, 3.3V stepdown). The upper part is obvious i hope.
Right: 2x 15A stepdown.. some bay stuff. I delivers what the battery will later deliver. Set to 10.2V and combined 25A. (My power supply does not deliver enough power at 10V, because of 10A current limit)
And yes before you ask: everything fits beautiful into the case. Its just not practical to put it in yet.

Edit again:
I found the problem! That CAP_FULL indicator was able to inhibit enabling the statemachine of the charger but wasnt able to stop it. Pretty stupid error. The reason it worked on low power is that I used my power supply directly instead of the stepdowns. This triggered the UVLO when the current limit kicked in which resulted in an emergency shutdown... so it kind of auto-disabled regularly and the re-enabling was then inhibited by the CAP_FULL indicator, so it appeared as a working voltage regulation. To give you a picture of the complexity i am dealing with, here is the statemachine within the CPLD which controls everything in the CapCharger (its also ab backup for me if i ever loose my data...)
The red line was missing..




Another edit:
Finally it works. I upped the power delivery to 210W (from 200V -> 600V (660uF) in 500ms) which is great and perfectly as simulated before. The battery is struggling a bit under the load and dips down 1V under full load. I still have some cleaning up to do.. but anyway, its fine.I could even increase the power even more, but i wont bother.


Blue: cap voltage
Yellow: the battery voltage (measured through the balancing connector), initially 12V.
Back to top
DerAlbi
Sun Jul 16 2017, 12:46PM
Registered Member #2906
Joined: Sun Jun 06 2010, 02:20AM
Location: Dresden, Germany
Posts: 524
I tested my sensor system and the results are phenomenal. It works way better than expected and whats even more positive is that the accelerations coils dont even disturb the sensor. I havent tested pulsing a coil yet, but at least the sensor issnt blinded by the sheer presence of the acceleration coils as earlier tests had indicated. This is likely a result of a changed sensor topology i use the region between a parallel and series resonance to get a region of steep phase change. Remember: the phase between voltage and current across a sensor is the sensor signal.
This brings me so far that i can say that all essential systems are tested and i know they will work as specified. This is a huuuuge milestone for the project.
It also turns out that my barrel-less acceleration path design is a crazy success. The level of tolerance i achieve with the mechanical design is amazing.

This is a look through 2 acceleration coils. You see the inside of the coil is perfect - it represents the minimum length thats required to go around that geometry. This should theoretically yield best coupling (due to average low air gap) and high time constants. You also can see how nicely i have the tolerances and clearances under control. The coils are perfectly centered and that is done without any effort. (the coils have nipples formed with resin which lock into holes on the sensor PCB). One can simply press the assembly together and nothing can move and its inherently stable.

Now the embarrassing part: i am not actually able to wind good coils. I know the image above suggest otherwise however keep in mind that having a good first layer (or even second layer) is pretty easy. However if any mistakes are made the layers above start to become 'wild windings' with a pretty horrible fill factor.
And the worse thing is you dont even need to make any mistakes. The design of the inner shape itself is the problem: the sharp corners provide enough disturbance that the position of the valley-change (when you are on an upper layer and the wire crosses the lower wire in order to get to the next "gap" between turns) changes randomly. Once this happens it never recovers.
Solutions:
a) go back to the old, regular, round shape. I would need a inner diameter of 12.2mm at least which results in >1mm air gap. (in contrast to 0.3mm). Ironically this yields better time constants (L/R) than the current wild winding. (i know my old coils had 1.7ms... the once you see in the image above have 1.35ms which is quite a bad result given it should be more optimal :-/ )
b) using rectangular wire. This is the plan anyways later when using aluminum. However this becomes attractive now for several reasons.. on the one hand i really like to see if L/R improves as thought (scientific curiosity) and on the other hand it should be easier to wind since it should behave well using the easiest way of winding a coil (helical). But.... rectangular wire is hard to get. If anyone knows a source i am happy for any tip. For now i just know about manufacturers where i can order some ridicules minimum quantity i will never use.
Back to top
Go to page  1 2 3 4 [5]

Moderators: Chris Russell, Noelle, Alex, Tesladownunder, Dave Marshall, Bjørn, Dave Billington, Steve Conner, Wolfram, Kizmo, Mads Barnkob

Jump:     Back to top

Powered by e107 Forum System
 
Legal Information
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.