My Christmas day 9T246 restoration! Almost.....

[DaveWM]

yes, and the power factor as well. For the most part the test well over like 50%, if it was under I would reject it outright. I like to compare them to new caps, they generally test higher value, about the same in power factor, and up to about 2 to 4 times the leakage (but its still very low). I used to have a chart that was used for acceptable leakage.

It would not be uncommon for me to take more that 24hrs for one section. Is this really needed, maybe not but it cant hurt to try and besides like I said before it appeals to my interest in experimenting.

What I would really like would be a test setup that I could program to step up the voltage based on current, say have it ramp up to try and maintain a set constant leakage current (selectable), then graph that voltage over time and set the max voltage to end the test. The resulting curves would be fun to examine.

you could do it with a constantly variable DC power supply and a current meter (analog would be easy to watch) but I dont really want to stand by the PS and work that for however long it may take.

[DaveWM]

all the sections seem to have responded well to reforming. I will set it aside and check it in a couple of weeks.

[Zenith6S321]

Originally Posted by DaveWM

What I would really like would be a test setup that I could program to step up the voltage based on current, say have it ramp up to try and maintain a set constant leakage current (selectable), then graph that voltage over time and set the max voltage to end the test. The resulting curves would be fun to examine.

I threw together what you described above from Radio Shack stuff. It uses an Arduino Uno and a pulse width modulated motor controller to drive two 12.6V to 120V filament transformers to step up to 240VAC. Then a voltage quadrupler. This results in a programmable 6-900 VDC supply, The Arduino software reads the analog inputs to raise the capacitor voltage while limiting the current and max voltage, and then monitoing the current decline to the required minimum current. Anyway, here are some plots from four capacitors. I set the current limit to 5 mA, and the voltage limit to the cap working voltage. The first cap is a 630TS 40uF 450V, the second a Sprague 16uF 450V cap added as a repair, the third a 630TS 80uF 150V cap, and the last a Nichicon 82uF 450V cap. My Sencore LC75 says the RCA caps show a lot of dielectric absorption, the Sprague is toast, and the Nichicon was the best (of course, its new) but did also show some dielectric absorption. I thought it would be interesting to plot the computed power going into the cap as well as its computed effective resistance along with the measured voltage and current. Although its now controlled through USB, it could be standalone with the addition of a small two line LCD panel and some switches. I think I could save some of my 630TS caps, but some are completely unresponsive to reforming.

[DaveWM]

interesting graphs, would like to see both graph at 200 seconds. for 1 and 4

Also if you reran the test on the old caps does the graph change (reformed) so the low current target at rated voltage is reach sooner.


I looks like the 82uf took that long to hit max voltage, perhaps for reforming you could ramp down the current to say 1ma max. Very neat project.

[Zenith6S321]

Originally Posted by DaveWM

interesting graphs, would like to see both graph at 200 seconds. for 1 and 4

Here is a plot of 1 up to 250 seconds. Is that what you mean?

[Geist]

Hi All;
I would like to see a circuit diagram ?? So, I can build one as well ..
THANK YOU Marty

[Zenith6S321]

[QUOTE=DaveWM;3058394]Also if you reran the test on the old caps does the graph change (reformed) so the low current target at rated voltage is reach sooner.
QUOTE]

I ran the reform process again on the two old RCA caps and attached are the plots. Yes, they got a better result from a second reforming. They both still show a lot of dielectric absorption, as seen by the LC75.

[Penthode]

[QUOTE=Zenith6S321;3058457]

Originally Posted by DaveWM

Also if you reran the test on the old caps does the graph change (reformed) so the low current target at rated voltage is reach sooner.
QUOTE]

I ran the reform process again on the two old RCA caps and attached are the plots. Yes, they got a better result from a second reforming. They both still show a lot of dielectric absorption, as seen by the LC75.

I see from your graph that full rated voltage is applied after only a minute or so. I believe this is too fast. I generally reform much more slowly and I use a limit the initial current to no more than 3mA. I will then leave the capacitor and return, say 15 minutes later, up the current again to 3mA step-by-step until the rated voltage is reached and the current reduces below 0.1mA.

I have reformed quite a few capacitors with a simple variable DC supply. (I first used an old variac, rectifiers and series resitance but now use my Sprague T06 capacitor checker).

I assume by diectric absorption you are referring to leakage? Certainly the older capacitors leak more than newer capacitors. But that was normal. What is important is the leakage is kept under control and that it remains low. I find gently reforming and lengthened testing results in a higher percentage of saved capacitors.

The leakage current scale on the graph is too high which does not allow you to see the resultant leakage current accurately. You must be able to ascertain that the leakage is below 0.2mA inn order to determine whether the capacitor is good or bad.

The resistance indication on the graph is interesting. Perhaps the fluctuation is the dielectric rebuilding process?

[Zenith6S321]

Originally Posted by Penthode

I generally reform much more slowly and I use a limit the initial current to no more than 3mA. I will then leave the capacitor and return, say 15 minutes later, up the current again to 3mA step-by-step until the rated voltage is reached and the current reduces below 0.1mA.

I searched the internet for capacitor reforming current limits and found a wide range of values. The highest value I saw was from my Sencore LC75 manual that describes your supply setup and says to limit the current to 50 mA! I thought that value was way too high so I used 5 mA. To make it visible on the graph I multiplied the current value by 100, so 500 for the current on the graph is actually 5 mA (sorry for that confusion factor). I will try using your 3 mA value on the next caps I reform.

Originally Posted by Penthode

I assume by diectric absorption you are referring to leakage?

Until I bought and started using the LC75 I had not heard of dielectric absorption. Here is a good description that echoes what is said in the Sencore manual:
http://en.wikipedia.org/wiki/Dielectric_absorption
It goes on to say that all capacitors have some amount of this and its generally not a problem for capacitors used in power supply circuits.

Originally Posted by Penthode

The leakage current scale on the graph is too high which does not allow you to see the resultant leakage current accurately. You must be able to ascertain that the leakage is below 0.2mA in order to determine whether the capacitor is good or bad.

I was not able to find a way to show multiple scales on the graph to make it more readable. I started out using a value of .2 mA for the target final leakage value. I have since found formulas in the LC75 manual that they got from EIA RS-395. The manual provides a table of acceptable/expected leakage values calculated from those formulas. I ended up using those formulas in my program so that it calculates the target leakage value and terminates the reforming when it is reached. During the reforming process the program prints out, on the Arduino serial monitor, the voltage and current applied to the capacitor. That data is what I cut and paste into a spreadsheet to make the plots. The Arduino analog inputs have a 10 bit (0-1023) resolution so the voltage sensing resolution is 1 volt (0-1023 volts) and the current sensing resolution is .01 mA (0-10.23 mA).

Originally Posted by Penthode

The resistance indication on the graph is interesting. Perhaps the fluctuation is the dielectric rebuilding process?

The resistance shown is calculated by taking the sensed voltage and dividing by the sensed current. So it is an equivalent DC resistance due to the leakage current (not ESR). Some of the fluctuantion probably is the rebuilding process, but I suspect a good bit of it is from the limited resolution on control of the pulse width modulation used to drive the two step-up transformers shown on the schematic this thread: http://www.videokarma.org/showthread.php?t=256919
The pulse width has a range of 0 to 255 with a value of 127 producing a 50% duty cycle waveform that I use as the upper driving limit. When the leakage current gets low a single increment in the transformer drive can make a sizeable change in the quadrupled output voltage.