vintage tantalum capacitor replacement

[David Hess]

What voltage rating is required?  I've heard to derate them by 50%.  I've heard 60%.  None of that resolves the issue of infant mortality from the micro shorts.  I did some research into this and the tantalum makers recommend initial power to be applied with a current limit to allow the micro shorts to self heal (or burn out or whatever they do) without overloading the cap with a major current surge through the short.

I think it was Kemet who had the best white papers on this topic.  The fact that you haven't seen a problem doesn't mean it doesn't exist.  It means you just have not designed a circuit that was a challenge to the cap.  Do you know what parameters that you designed right?

There are other limits to surge current so voltage derating by 1/3rd to 1/2 will almost always prevent surge related failures.  The thicker oxide prevents the self healing mechanism from experiencing thermal runaway.

[Tantratron]

Ok now just to make sure I understand your recommendation if I go the route of stop using tantalum NOS and only replace by Nichicon electrolytics. Say I want to replace a K100-20V tantalum (+15Vdc bus), shall I replace by Nichicon 100uF of 50V or do you need to derate by 2 its capacitance as well (i.e. 200uF and 50V) ?

To get a comparable impedance at high frequencies, an aluminum electrolytic capacitor needs to have like 4 times the capacitance of the solid tantalum capacitor it replaces.  This is often reflected in old application notes which will recommend a 1 microfarad tantalum capacitor or 10 microfarad aluminum electrolytic.

Since you know both TG501 and PG506, let's see the attached pictures of 4 cases where i've added yellow arrow pointing on the failed tantalum. Ok the +5V bus rail seem to survive and I think it is because tektronix used 10V rated tantalum. However the +16.5V or the +15V are failed (shorted) where basically they put tantalum rated 20V, clearly not enough margin.

Now replacing by electrolytic with 10x capacitance cannot work, there will be no room to install them. Furthermore all these tantalum caps do have in parallel fast smaller non-polarized capacitor so these do take care of HF.

There must be a middle path of either using Al electrolytics or using modern tantalum with higher voltage rating but keeping same volume to fit the compactness of these PCBs once they're assembled together.

You know these TM500's section are very compact, I guess tektronix had no choice to save space of using these tantalum. Some voltage bus were properly rated but as I've observed so far, the problem is on the +15V or +16.5V with crazy 20V capacitor.

[David Hess]

Now replacing by electrolytic with 10x capacitance cannot work, there will be no room to install them. Furthermore all these tantalum caps do have in parallel fast smaller non-polarized capacitor so these do take care of HF.

That is right; a physically larger aluminum electrolytic replacement is not always feasible.

There must be a middle path of either using Al electrolytics or using modern tantalum with higher voltage rating but keeping same volume to fit the compactness of these PCBs once they're assembled together.

My middle path is simply to use a higher voltage new or NOS solid tantalum replacement.

You know these TM500's section are very compact, I guess tektronix had no choice to save space of using these tantalum. Some voltage bus were properly rated but as I've observed so far, the problem is on the +15V or +16.5V with crazy 20V capacitor.

The TM500 form factor is pretty small but I think Tektronix used solid tantalum capacitors for their better performance.  They perform better at high frequencies than aluminum electrolytics.

I do not have any relevant photos of my PG506 but below are some measurements of the fast rise output on my sampling oscilloscope and 100 MHz 2232.  Ignore the jitter on the sampling oscilloscope measurement; it was part of a deliberate test and would not be visible under normal operating conditions.

I have not gotten around to rebuilding my TM501 yet.

[gnuarm]

There are other limits to surge current so voltage derating by 1/3rd to 1/2 will almost always prevent surge related failures.  The thicker oxide prevents the self healing mechanism from experiencing thermal runaway.

It didn't for me.  I specified a 20V T491X157K020AT from Kemet for a 12V supply that is brought up at a controlled rate.  In fact, my board is a daughterboard mounted 8 to a motherboard.  When my customer first installed them these caps so slowed rise time on the 12V rail the sequencing logic thought there was a problem and shut down.  They had to change a timing capacitor for them to work.

Meanwhile the test fixture only mounts one powering from a 400 mA wall wart.  This is when the caps can fry.  The first CM was loosing these parts often without telling me.  At another CM they never said they saw lots of bad caps, but they had one that burned!  At that point I dug in and learned about conditioning with a controlled current.  No problems since.