Will my capacitor recovery voltage damage SRAM?

[BrettD]

A .047F supercap in my 20 year old audio receiver failed (this is a common failure according to the net.) It's provides backup power for the user configuration settings stored in SRAM.

I didn't have one of these and ordering is going to take time. So I fished out a old 4700 uF 25V electrolytic in my parts drawer. It obviously won't provide the same duration backup as the original part but it might be sufficient to prevent annoyingly losing my settings all the time until I get the correct part.

I tested the cap by charging it with a 9V battery then checking its discharge overnight. It held up substantially and could work.

I discharged it for installation. And then I remembered about dielectric absorption and recovery voltage.

It's going to operate in-circuit at 5V, as measured. I can't find any information on the net about terminal recovery voltage. Obviously recovery to more than 5V is a problem.

Right now after a full discharge via 100R for a while then a 30 minutes wait it's back up to 0.14V from 0V.

Rubycon states "Recovery voltage peaks between one and three weeks after the terminals are disconnected,and then gradually decreases." But no details on the magnitude.

Do I need to worry that the capacitor will recover to > 5 V in circuit? Can I ensure that won't happen with a long enough prior discharge time? (of what duration?) I searched the eevblog forum and didn't see any relevant discussion on this.

Thank you for your ideas.

Brett

[Gyro]

Welcome to the forum Brett 

I very much doubt that the capacitor would recover to anywhere near 5V especially with the load (albeit minimal) of the SRAM. That would be more than half the previously applied voltage. Leave it unconnected overnight, ie. about the same period that it had approx 9V across it, and measure again. I'm sure what you measure then will give you reasonable reassurance.

You are wise to be cautious about such things though. 'Dielectric recovered' capacitors have taken out more than a few low cost component testers, and in the case of high voltage ones, caused some very nasty surprises!

[mariush]

You could just replace the supercap with 2 cr2032 in series with 1-2 diodes  to drop some voltage and get total voltage below 5v.
ex 2 x cr2032 gives you 6v, 2 1n4148 or 1n581x will drop around 0.5-0.8v per diode at that low current.

It may work even with just one battery and its 3v voltage.

You can find cr2032 in shops in packs of 2-4 or more for a dollar or so.

[george.b]

You could always do an experiment. Considering a data retention current of 1µA (typical value for a HM628128), at 5 volts, that works out to a resistance of 5 M . Charge your capacitor then let it discharge through a 4M7 resistor. Measure every now and then, or, if you know your DMM's impedance, work out the necessary resistance to go in parallel with the DMM (for an equivalent of about 5M ) and just leave it hooked, monitoring the voltage.

I very much doubt there will be any noticeable recovery under load, though.

You could just replace the supercap with 2 cr2032 in series with 1-2 diodes  to drop some voltage and get total voltage below 5v.
ex 2 x cr2032 gives you 6v, 2 1n4148 or 1n581x will drop around 0.5-0.8v per diode at that low current.

It may work even with just one battery and its 3v voltage.

You can find cr2032 in shops in packs of 2-4 or more for a dollar or so.

SRAMs will typically retain data down to 2V. Using two in series is, in all likelihood, unnecessary. Also, two diodes in series would result in a drop of about 0.5V in total. A single 1N4148 will drop less than 0.5V at 1µA. See:


One CR2032 with a series diode (to avoid damage to the primary cell) should work.

[mariush]

Yeah, I meant to say 1n400x but was thinking of adding 1n581x which are common but schotky (probably not correct spelling) and low voltage drop) and i wrote 1n4148 instead. 

[BrettD]

Welcome to the forum Brett 

Thanks Gyro! I've been a lurker for a while. EEVblog is a wonderful community full of fun, good vibes and opportunities to learn, but this is my first post. I was licensed as a radio amateur in the mid 70s and enjoyed working as a embedded SW developer in small companies where I was very close to the HW development.

I very much doubt that the capacitor would recover to anywhere near 5V especially with the load (albeit minimal) of the SRAM. That would be more than half the previously applied voltage. Leave it unconnected overnight, ie. about the same period that it had approx 9V across it, and measure again. I'm sure what you measure then will give you reasonable reassurance.

I shall do that then.

You are wise to be cautious about such things though. 'Dielectric recovered' capacitors have taken out more than a few low cost component testers, and in the case of high voltage ones, caused some very nasty surprises!

Thanks. I was aware of the risks around high voltage caps but it never occurred to me before that this property could blow up a component tester. Makes sense.

I did figure that I can easily take more electrical abuse without dying than the SRAM and uP in my receiver.  :-)

The capacitor is in my Yamaha RX-V3000 A/V receiver I've had for about 20 years. It was a wonderful receiver when I bought it and is still lovely now. It was a pleasure to see Yamaha's construction when I opened it up. I removed one board to unsolder the dead supercap then soldered some wire to the PCB so I can connect whatever cap I use without removing the PCB again.

I was impressed with Yamahas service manual too, with complete schematics, PCB layouts, exploded diagram - where do these four 'extra' screws go? , block diagrams, complete parts list :-), etc. And hooray for them: the vast majority of screws are identical. (I must have removed 70 of them.)

That's why I'd hate to blow this receiver up with my repair.

[BrettD]

You could always do an experiment.
...
One CR2032 with a series diode (to avoid damage to the primary cell) should work.

Wow! I am blown away with your in-depth response, building on mariush's response, following one and a warm welcome from Gyro. A big thank you to all of you for turning a basic repair question to an engaging afternoon and educational experience.

I was actually surprised to find the capacitor power backup instead of a lithium cell, when I dug into this repair. It actually took me a while to find the capacitor when I got the service manual because I was expecting and looking for a cell.

As it turns out, I do have CR2032s, diodes, and even a cell holder in stock, so I will likely go that way, as I won't have to open up the receiver for many years, hopefully. But first I will do the recovery voltage measurment experiment because you've engaged my curiosity!

[george.b]

Wow! I am blown away with your in-depth response, building on mariush's response, following one and a warm welcome from Gyro. A big thank you to all of you for turning a basic repair question to an engaging afternoon and educational experience.

You're very welcome!

I was actually surprised to find the capacitor power backup instead of a lithium cell, when I dug into this repair. It actually took me a while to find the capacitor when I got the service manual because I was expecting and looking for a cell.

It's understandable that they chose to use a supercap instead of a lithium cell, I mean, the supercap you can recharge, so it should theoretically outlast the expected useful lifetime of the receiver itself. At least they didn't go for a NiMH battery - those famously end up destroying whole boards when they leak. Speaking of which, it might be wise to clean the area around the failed capacitor, as they sometimes do leak as well. It may not be as conspicuous as a battery leak, but you might be able to notice a thin layer of electrolyte coating the area around the capacitor, if it did indeed leak.

As it turns out, I do have CR2032s, diodes, and even a cell holder in stock, so I will likely go that way, as I won't have to open up the receiver for many years, hopefully. But first I will do the recovery voltage measurment experiment because you've engaged my curiosity!

Again using the HM628128 as reference for guesstimating - it typically uses 10µW in standby. A CR2032 cell has a typical capacity of 645mWh (according to a datasheet from Varta), so that works out to an endurance of 64500 hours, or about 7 years and 4 months, not accounting for self-discharge (~1%/yr). In practice, if your receiver is plugged in most of the time, the battery should last a lot longer than that.

Let us know how the experiment goes