On Monday 11 January 2016 14:44:54 John Kasunich wrote: > On Mon, Jan 11, 2016, at 02:00 PM, Gene Heskett wrote: > > On Monday 11 January 2016 13:10:10 John Kasunich wrote: > > > SSRs can fail shorted, I don't think this buys you much in terms > > > of reliability and safety. > > > > > > I think there is a little over-kill going on. > > > > > > Do you really need a fast bleeder (<1 minute to safe)? Or would a > > > five minute bleeder be OK? > > > > > > My understanding is that you have two 68000uF caps in series, > > > charged to about 130V. (If this is incorrect let me know - the > > > thread is long and rambling and refers to other thread. I'm > > > basing this on something written in the first message of this > > > particular thread.) > > > > Actually 10 6800 u-f in parallel, working at their labeled 63 volts > > ratings. > > > > But yes, you have the right idea. :) > > > > > Series caps divide, so you have 34000uF. Let's say that "safe" is > > > 20V. So you need to discharge from 130V to 20V, that is to 15% of > > > the original voltage. Ln(0.15) is -1.9, so it takes 1.9 RC time > > > constants to get to a safe voltage. Call it 2.0 for easy math. > > > If you want to get there in 5 minutes, then one time constant is > > > 2.5 minutes = 150 seconds. T = RC, solve for R, gives you R = T/C > > > = 4411 ohms. At 130V, that would draw 29mA and burn 3.8 watts. > > > Next lower 10% value is 3.9K, that would draw and burn 4.3 watts. > > > I like to derate power resistors by 50%, so you want a 3.9K 10W > > > resistor. > > > > I actually have, atm, 6 ea 5k 10 watters in parallel clipped onto it > > right now that I was going to use once the relays get here. Thats > > 833 ohms which would drain it even faster than just the one. > > The 5K's are across the entire bus, or across each cap?
The bus, since there are 4 toroids and rectifiers, (dictated by the toroids I could source for close to nothing,) is arranged as 5 caps per group, in a 2 in parallel, stacked two more on top of that so there is one 5 k R across each 5 caps. 20 of those 6800x63 caps total. I did not have a bar joining the center of the stack, still don't. So there's a possibility of a volt or so's miss-match there. > > If across the entire bus, 130V would result in 3.38 watts per > resistor. Less than the 5W rating, but more than half the rating, > which is the derate I like to use to keep things from getting too hot > and ensure long life. These are 10 watt R's. > > I'd suggest 3 of the 5K across each side of the bus. At 65V across 5K > each resistor would be running at 0.85W, nice and cool. Total > resistance of the 3-parallel 2-series arrangement would be 3.33K, RC > time constant 113 seconds, bleed-down time from 130V to 20V = 212 > seconds = 3.5 minutes. And free voltage balancing between the two > sets of caps. > > > And, somewhere > > on an undermanned galley in the pacific, is a 250 ohm 250 watt > > resistor which would do the bleed-down even faster, and the current > > peak would be nominally .5 amps in that case. I can't see that as > > being capable of welding a relay contact. > > Making 0.5A probably won't weld any contacts. Breaking 0.5A DC on > a relay designed for AC is another story. And keep in mind that > contact bounce = breaking and making. True, very true. > I'm just not a big believer in switched bleeders. Significant hassle > and risk. If fixed bleeders can meet your discharge time > requirements, call it good and move on. That, once I put in GOOD charge-pump detectors, is exactly what I'm going to do. And when I put the cover back on it, a dymo label that reminds me that the minimum safe off time is at least 20 minutes. > Context - I design large motor drives for a living. The vast majority > used permanent bleeders only. One recent design does augment them > with a switched bleeder. But that drive has 90 caps, each rated > 6800uF 400V, in a 30-parallel by 3-series array, total of 68000uF. > Nominal bus voltage is 975V DC. Stored energy is 32 kilo-joules. UL > requires that it discharge below 50V in five minutes. The fixed > bleeders would do it in 10.5 minutes, so we added a switched bleeder > to help it along. > > Your unit has 34000uF at 130V, for a stored energy of 287 joules. > > > > If you do indeed have two caps in series, they ought to have > > > balancing resistors anyway, so you could split the bleeder into > > > two sections. Make each section 2K or 2.2K and 5 watts. > > > > That too, I was contemplating, but the 51 ohmmer would need to be > > dismounted to gain access to the pcb bus connecting them all > > together. Since I have them already 2 each of the 5k's across each > > bank would seem to be about right. They are well discharged atm, > > and I could do that yet this afternoon. If my math is correct, at > > 5k and 63.5 volts=0.80645 watts per, so if I spread them out on the > > busses, the heat shouldn't hurt the caps. > > Agreed. But why not use all 6 of the 5K? I could, if I tied the center points of the stacks together, otherwise I'd need 8 so I could put 2 on each group. > > > Inrush limiting is a separate issue. Separate resistor, bypassed > > > by a contactor or relay for normal operation. Put it on the AC > > > side so you don't have any issues with DC contact ratings. > > > > That is what I am doing right now with the SSR's, one to switch on > > the power thru that 51 ohm resistor, and a second SSR about 5 > > seconds later to short out the resistor. > > > > > The main risk is that > > > if you try to run with the bypass relay open you can fry your > > > resistor. One way to avoid that is to make the "resistor" a 100W > > > incandescent light bulb. The initial turn on does not develop any noticeable heat in that 51 ohmmer. It obvious has to, but its not enough my hand can detect it in a 72 degree F garage thats probably closer to 80F at that height, 6+ feet, above the floor. > > Which rigid tapping would quickly make history. The spindle > > turnaround, even if I have it rate restricted, still pumps the > > supply high enough to blow a C7 bulb in 4 or 5 pecks. Jon's pwm > > servo amp is effectively a full 4 quadrant controller, so it dumps > > motor energy back into the supply when it reverses the motor. I > > haven't actually measured the voltage as my digital meters aren't > > near fast enough to catch the true peak. I have about a 1 second > > rate ramp in the hal file to ease that to about 1 second from 2500 > > to -2500, without that its positively brutal. But thats yet another > > problem that I should stretch the allowed time on. One setp in the > > .hal. > > We must not be talking about the same thing. I'm not talking about > putting a light bulb across the bus as a bleeder, in fact a couple > paragraphs down I specifically say that is a bad idea. I'm talking > about using a light bulb as a _charging_ resistor - in your case, it > would replace the 51 ohm part. > > Using a light bulb as a pre-charge resistor does NOT subject it to the > DC bus voltage. The bulb is in series with the incoming AC line, > before the rectifier, and can never see more than the AC line voltage. > In addition, during normal operation the bulb is shorted out by the > bypass relay. > > > > Using bulbs for bleeders is risky because they run all the time, > > > and when they burn out they leave the caps charged with no > > > indication of the risk. But using a bulb for charging is > > > different. The bulb only lights for a few seconds when you first > > > apply power. If it is burned out the caps won't charge. If you > > > try to run with the bypass open the light will glow to let you > > > know. > > > > > > A 100W 120V bulb has a hot resistance of 144 ohms. With 34000uF > > > the charging time constant is 4.9 seconds. As the bus charges the > > > voltage drop across the bulb goes down, it cools down, and its > > > resistance decreased - that is good, it make the "long tail" of > > > the charging curve go faster. The inrush during the first cycle > > > or so will be based on the cold resistance of the bulb, but if > > > your switch and diodes can handle a 100W light bulb you know it > > > can handle any size cap bank fed thru that same 100W light bulb. > > > > Some how, given the quality of 100 watt light bulbs today, I trust > > this 50 yo 51 ohm 200 watt resister a lot better. And that gives me > > a T=RC= 1.734 seconds, so at 5 seconds, the current delay in then > > .hal file, they should be well charged. > > There are two risks with that approach: > > 1) suppose Jon's drive develops a shorted IGBT or a cap fails. Its a switch mode, using a hexfet bridge. > Your > 51 ohm resistor will dissipate 331 watts with 130V across it. It is > only rated for 200 watts. Fine for the 5 seconds or less that a > normal charge takes, but going to get very very hot if something goes > wrong and the bus can't actually charge. The light bulb is designed > to take rated voltage indefinitely. > > 2) closing the bypass relay 5 seconds after power-on without actually > knowing that the bus is charged is risky. Again, what happens if > there is a short somewhere and the bus voltage didn't come up? Huge > inrush when you close the bypass device. Hopefully all it will do is > blow a fuse. Might toast your SSR and/or diodes. It will trip a 20 amp circuit breaker in the panel quicker than it can fry anything. > > I'll go put the 5k's, 2 in parallel per bank of caps, in yet this > > afternoon & report back. > > Good plan. But while you're up there, why not put three in parallel? See above, no bussed connection to the center of each stack. I'm not sure why I didn't, but... Cheers, Gene Heskett -- "There are four boxes to be used in defense of liberty: soap, ballot, jury, and ammo. Please use in that order." -Ed Howdershelt (Author) Genes Web page <http://geneslinuxbox.net:6309/gene> ------------------------------------------------------------------------------ Site24x7 APM Insight: Get Deep Visibility into Application Performance APM + Mobile APM + RUM: Monitor 3 App instances at just $35/Month Monitor end-to-end web transactions and take corrective actions now Troubleshoot faster and improve end-user experience. Signup Now! http://pubads.g.doubleclick.net/gampad/clk?id=267308311&iu=/4140 _______________________________________________ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users
