Potential problems with series capacitors?

[d3x0r]

Can I do something like this?
It's hard to find such low voltage zeners... might be better to just use like 4 diodes in series (0.7*4 = 2.8 )
or 3 900mV  at $0.40...
or 1 2.7Vf  at $1.60 http://ixdev.ixys.com/DataSheet/DSEI12-10A.pdf  (1000V 10A)  (actually a less than 10A the voltage drop is down to 2V... )


Hmm maybe I'm reading the spec sheets wrong... for voltage less than 2.7, there's still current?  They're all curvy, so they're not really X voltage drop, just at a specific voltage you get a certain current....

[MarkE]

Your idea looks like it should work.  Those ICs you found seem ideal as long as an 80mA limit works for you. 

Digikey has Zeners down to 1.8V, but they get leakier and leakier at low voltages. 2V MMSZ4679's are cheap, and SOD-123's are reasonable to solder even with an ordinary iron.  How much current do you plan on passing through your stack?  What kind of transients do you expect to encounter?  http://www.onsemi.com/pub_link/Collateral/MMSZ4678T1-D.PDF


Ordinary diodes will not be very square.  If you go that route there will be a lot of slop in the voltage.  Four PN's in series could get you anything from about 2V to 2.8V at room temperature depending on the current.

[d3x0r]

Your idea looks like it should work.  Those ICs you found seem ideal as long as an 80mA limit works for you. 

Digikey has Zeners down to 1.8V, but they get leakier and leakier at low voltages. 2V MMSZ4679's are cheap, and SOD-123's are reasonable to solder even with an ordinary iron.  How much current do you plan on passing through your stack?  What kind of transients do you expect to encounter?  http://www.onsemi.com/pub_link/Collateral/MMSZ4678T1-D.PDF


Ordinary diodes will not be very square.  If you go that route there will be a lot of slop in the voltage.  Four PN's in series could get you anything from about 2V to 2.8V at room temperature depending on the current.

Ya I went searching for mosfets that might do the job; but they also aren't very square... and those that do turn on around 2.5-2.7V are at very low current themselves until they hit 4 or 5V...
Working on implementaion of lasersaber's joule looper thing... so low current is probably OK... and they only require 80mA when they are misbalanced... or get charged way over; but was going to run a separate 9V zener with an LED in series to indicate full charge.
I found an article from july 11,2014 that said 'ALC recently released...' regarding the balancing IC.  The PDF says it starts conducting within 0.4V of rating ... so effective top is only going to be 2.3 then (9.2 total instead of the 10.8 expected)
got the caps before considering that series caps can actually have problems...
Was experimenting with some other supercaps I already have, and 2A is kinda low for charging 100F caps... that is it takes quite a while.... and my 3000F cap will take like 2 hours to charge at 2A.  (at low voltage 0.280V even it only goes up only 0.001V/sec... and they get slower and slower as they get more voltage...)


So another question though relating to these... if I use a higher voltage (12V for instance) to charge the 2.7V cap so I can maintain the amps input, is that like overvoltage?  Or if I stop before it gets to 2.6 something V is that ok?  or like an impulse charger like a bedini with something like 400V pulses?

[MarkE]

I am assuming here that you do not have a bench supply that has a solid adjustable current limit.  That would be a very good thing to have.  At least get a hefty current limiting resistor.  I would not charge either the stack or any single capacitor with a voltage above where you want to end up.

The rate at which the capacitors charge will be inversely proportional to their individual capacitance values. 

You can make a very square 2.5V or higher regulator with an LM4040D25.  That can operate with 60uA bias.  I don't know how low an operating current you want to realize with your JT, but 60uA is probably a lot of your budget.  You can always use a switch or jumper to disconnect when you are done charging.  Or you can precharge them separately and then connect them in series.  You will want to find out first what their relative capacitances are which you can do by charging them in series with a resistor and sampling all the voltages periodically.  Use the results to limit the voltage proportionately.  For instance if you have 3 caps:  A, B, C.

Charge to:  0.5V A, 0.4V B, 0.55V C in series.
Then charge C to 2.7V, A to 2.7V * 0.5/0.55, and B to 2.7*0.4/0.55

[d3x0r]

I am assuming here that you do not have a bench supply that has a solid adjustable current limit.  That would be a very good thing to have.  At least get a hefty current limiting resistor.  I would not charge either the stack or any single capacitor with a voltage above where you want to end up.

The rate at which the capacitors charge will be inversely proportional to their individual capacitance values. 

You can make a very square 2.5V or higher regulator with an LM4040D25.  That can operate with 60uA bias.  I don't know how low an operating current you want to realize with your JT, but 60uA is probably a lot of your budget.  You can always use a switch or jumper to disconnect when you are done charging.  Or you can precharge them separately and then connect them in series.  You will want to find out first what their relative capacitances are which you can do by charging them in series with a resistor and sampling all the voltages periodically.  Use the results to limit the voltage proportionately.  For instance if you have 3 caps:  A, B, C.

Charge to:  0.5V A, 0.4V B, 0.55V C in series.
Then charge C to 2.7V, A to 2.7V * 0.5/0.55, and B to 2.7*0.4/0.55

I do have a bench supply... the current limit seems to work pretty well... 30V-3A max..
well once these get to 1.2V or so at 2.7V limited input they allow less than 1A charging; which is why i was hoping to increase the voltage to maintain the amp input for testing with the bench supply... but I can understand that charging with extra voltage might stress the outermost layers (the parts immediately associated with the electrodes)... the bedini idea was like using a highly inductive charging system that without a capacitor it generates I dunno 20-100V... when adding a cap, it doesn't go nearly that high... my actual JT draws a lot of current (6 windings to base, 6 to collector and a secondary that's 140V) it generates 100V easily across LEDs, but not that much across a bridge rectifier to a cap... (that's a JT not JL)

My version is actually higher current draw (10mA...well shows as 0.01A on the power supply) at 10V.  under 10V it reads 0.00; but the meter might be rounding up, (haven't actually put another meter inline to see if it's a truncate or round)  was going to do some more testing with some CSR inline with the power...
LM4040D25 looks like a good part. But, again, isn't its current rating going to only apply when they are misbalanced?  "Wide Operating Current Range...45 μA Typ to
15 mA"... not when using the power?

The precharging example was just experimental idea to demonstrate worst case... generally I expect them to be in-balance...other than I guess percents difference between them...  and charging with human power or maybe solar (small 12V solar panel says 400mA for example)