| Miniproposals | ||||||||||
|
| Operators | |
| Session leader(s): | Steve Fairfax |
| Physics operator(s): | Steve Horne |
| Engineering operator(s): | Unknown |
| Engineering Operator Run Comment |
| test modifications to oh2 regulator |
| Session Leader Plans |
| Physics Operators Plans |
| Session Leader Summaries |
| Physics Operator Summaries |
| Entered: Jul 7 2004 04:36:51:647PM |
| Author: To Be Determined |
| Phys op report, Run 940531 -- Fairfax (Session Leader), Horne (Phys Op) MP 57 -- test oh2 supplies w feedback loop broken. This run was dedicated to testing whether we could break the OH2 internal voltage feedback loop, which relies only on integral gain. The objective was to try to speed up the power supply response, which should allow us to run higher elongation plasmas without loosing vertical stability. We hoped also to understand the source of the oscillation in OH2 voltage. It was discovered while setting up this run, that the integral gains on the oh2 supplies were different -- oh2L had roughly twice the gain of oh2U. This might be a clue as to why oh2L oscillated worse than oh2U. Starting with shot 4, the loop was broken, and the hybrid controlled the gate drive boards. After a dud (44), a gain tweak and a fizzle (5), we adjusted the gate drive boards to roughly match the old gain, (output volts / hybrid demand) just before commutation. After a couple of tweaks we began getting consistent fizzles. Typically, the flux plot showed a very nice null at breakdown, but there seemed to be a problem involving the OH1 and OH2 balance later in time. Careful examination of the oh2 voltages showed that while just before commutation the gain was about one, just after it was much higher -- perhaps 4. We corrected the preprogrammed voltage and got a ferocious runaway shot (13) , another fizzle, and a disruptive shot (15). Ths scatterplot showed that the initial current rise was very slow on these shots compared to friday, implying something wrong with the initial 100 msec or so of programming. After some thought (see note on shot 17) we decided to punt running with no regulation. After some cogitation it's clear that what kills us in that mode is the I-R drop in the supply, which is very substantial (50 V at commutation?). Zero demand will give output voltage = -IR, not zero. In principle we could compensate it out with the hybrid but it's definitely non-trivial. So we went on to see if pure proportional gain would work -- shot 18. Quite low gain gave wild oscillations in voltage; decided to try equal integral gain on both OH2 coils, + smaller proportional gain. After a few attempts we settled on "50 ohms" of proportional gain, which seems qualitatively to make no obvious difference to anything. The last couple of shots were an unsuccessful attempt to move the strike point that forms near the divertor nose; due to a sign error in the programming we converted it to an x-point instead of eliminating it. There is an imbalance in OH2 currents which is causing this asymmetry; moving the plasma up 1 cm (shot 20) didn't remove it. I suspect there is still some sort of gain imbalance in the supplies. The current state of the supplies is: oh2U and L integral gain set to where Oh2U used to be, "50 ohms" proportional gain. Shot list: 1 load 940527034, remove Gerry's glitches. zcur PID= (1.5,1,3) Gas 1.2e-5. Power supply problem. OH2 L commutation supply not turned on; switch never closed, various other problems. 2 fizzle. add 15 A to ef4. 3 Better fizzle. Go in and break voltage reg feedback on OH2 supplies. Gwinn claims we'll get within about 25% of our current gain. 4 dud. oh2 voltages a tad low; raise on oh2 from 4 to 6 5 fizzle. At about -.2 sec on shot 4: ask get oh2U -70 -30 L -68 -30 raise hardware gain on power supplies by factor of 2. Drop oh2 gain back to 4 6 Oscillation in voltage on oh2U. One theory is that hybrid is driving the oscillation. Looks to me that hybrid is fighting; possible phase lag still there causing oscillation? If it's a system instability, should be sensitive to BR gains. Set BR gains (-.8,2.),-.5,3),(-.4,4), (-.2,5) for next shot. 7 Oscillation gets worse with gain. Try br P gain =3 from -.5 on. 8 nice null, breakdown, but oh1 seems not right. BZ positive too fast? Move oh1 V point at 5 msec from +90 to +95. 9 Better. Very nice null, early breakdown. Move same point to 100V. 10. Looks good at breakdown. 15 A more neg on ef4, drop gas to 10e-5, try again. 11 even better. Examine oh2 voltage carefully; note that just before commutation (neg voltage) the oh2 gain seems about right -- just after, gain seems much higher. Drop the positive voltage step on oh2 at 5 msec from 20 to 10 v. 12 better. Take it to -10 v. 13 Lots of Xrays -- screen full of snow! 14 raise gas to 1.2 fizz 15 no change, nasty disruptive shot. 16 br tweak., fizzle 17 voltage tweak on oh2 L only, fizzle It's clear that the small signal gain for + voltage is too high. A plot of output vs demand for the supplies shows that the gain isn't constant as the voltage goes through zero - gain is higher for positive demand (physics convention) by large factor than for negative. We tweaked the power supply gain to be right for precharge, so it's way high just after commutation. We can't run this way; try using proportional gain. 18 reload shot 1; power supplies have proportional voltage gain at ".2 x max knob rotation." OH2 faults at -.8 sec. Decide to try integral gain again, but fix gains to be = to OH2U value. 19 Plasma!. Disrupts at about .6 sec. 20 move up 1 cm, to get power off the divertor nose. Add "50 ohms" proportional gain to the existing integral gain. Dud. Br too negative? try 1 mt +. Also set power supply gain to "100 OHMS" 21 plasma 22 reduce ef4 current at .3 sec from 1000 to 800 (or so) to reduce elongation. 23 Helped only a little -- still lost at .65 or so. xpoint still there. For 24, reduce oh1 weight in rcur controller from -6.7e-3 to 5.e-3 at 100 msec. 24 Longer, but still not full shot. for shot 25, drop Z P gain from 1.5 to .5 at .4 sec to reduce oscillation. Also change rcur controller oh1 even further, from -5.e-3 to -3.e-3. 25 changed rcur controller oh1 maxed to -8.e-3 ; ef3 from 4.e-2 to 3.e-2 all at 100 msec. Dud. |
| Session Leader Comments |
| Physics Operator Comments |
| Engineering Operator Comments | ||||
| Shot | Time | Type | Status | Comment |
| 1 | 09:16:47:420AM | Plasma | Bad | oh2 com fail |
| 2 | 09:40:40:830AM | Plasma | Ok | efc breakers off |
| 3 | 09:53:34:770AM | Plasma | Ok | |
| 4 | 10:06:16:250AM | Plasma | Ok | first shot with oh2 feedback loop broken |
| 5 | 10:21:35:460AM | Plasma | Ok | dud, adjusting controller gains |
| 6 | 10:39:22:880AM | Plasma | Ok | fizzle. gain increased to 2X for this shot |
| 7 | 11:03:07:720AM | Test | Ok | |
| 8 | 11:18:32:740AM | Test | Ok | |
| 9 | 11:40:59:560AM | Test | Ok | |
| 10 | 11:52:13:190AM | Test | Ok | |
| 11 | 12:05:12:920PM | Plasma | Ok | godd shot, power test |
| 12 | 12:26:17:770PM | Plasma | Ok | good shot, test pulse |
| 13 | 12:44:56:720PM | Plasma | Ok | plasma current no faults. |
| 14 | 01:00:52:230PM | Plasma | Ok | No faults. |
| 15 | 01:13:01:110PM | Test | Ok | No faults |
| 16 | 01:32:42:970PM | Plasma | Ok | Dud No faults |
| 17 | 01:57:40:090PM | Plasma | Ok | NO faults |
| 18 | 02:33:53:410PM | Plasma | Bad | comm fault due to OH2 Proportional gain |
| 19 | 02:49:26:300PM | Plasma | Ok | NO Faults |
| 20 | 03:08:49:110PM | Test | Ok | No faults |
| 21 | 03:27:49:990PM | Plasma | Ok | good shot, plasma |
| 22 | 03:48:34:100PM | Plasma | Ok | Fizzle No Faults |
| 23 | 04:02:51:710PM | Plasma | Ok | Plasma shot no faults |
| 24 | 04:20:59:410PM | Plasma | Ok | NO faults |
| 25 | 04:33:37:110PM | Plasma | Ok | No Faults |
| 26 | 04:48:56:810PM | Plasma | Ok | DUD No faults |