[SOLVED] Capacitor failure --- handling reverse biases.

I've designed a heartbeat monitoring circuit similar to the one below, before finding this one through googling.

|500x301

My circuit uses a pair of electrolytic capacitors back-to-back to act as a non-polar capacitor, due to unavoidable reverse voltages.

On the page about his circuits, he says he uses tantalum capacitors instead, because "they can withstand some reverse bias".

I tried googling for info about tantalum caps and reverse voltage, but papers I found were both vague and in disagreement with each other about what (if any) reverse voltages are acceptable without damage to tantalum capacitors, regarding both the voltage level and the duration of reversal.

I just want the units I sell to not come back to me due to capacitor failure.

What is the safest solution to be sure the circuits will continue to operate year after year?

I have use back to back caps in many projects. There are however, lots of np caps you can choose from. http://www.ebay.com/bhp/non-polarized-capacitor

FYI Use a LM339 cct. See page 33: http://www.ti.com/lit/an/snoa654a/snoa654a.pdf

non-polarized hold up best if they have no average
DC level, such as AC going to a speaker.
If you are really worried, you can bias the center lead with
a high ohms resistor.
Such as, if the + leads together place a 1 meg resistor to
the + power lead.
The capacitors will always remain properly biased.
You can also find 5uf ceramic capacitors. They are available.
Dwight

Why use polarized caps in the first place? 4.7 uf ceramics are readily available and dirt cheap.

And, uh, I hadn't heard anything about tants being okay with reverse bias...

The posted circuit is poorly designed. The diodes do not create a very good pseudo-ground and the bias level will fluctuate with the signal.

I suggest to look for another, with proper biasing. You shouldn't need unpolarized caps.

Edit: take a look at this (randomly chosen) input stage, which is an AC-coupled amp, biased at the midpoint of the power supply, and does not require a polarized cap. http://www.learningelectronics.net/circuits/images/heart-rate-monitor-circuit-diagram-2.jpg

jremington:
The posted circuit is poorly designed. The diodes do not create a very good pseudo-ground and the bias level will fluctuate with the signal.
I suggest to look for another, with proper biasing. You shouldn’t need unpolarized caps.

+1

DrAzzy: Why use polarized caps in the first place? 4.7 uf ceramics are readily available and dirt cheap.

And, uh, I hadn't heard anything about tants being okay with reverse bias...

High value ceramics are often unsuitable for analog coupling capacitors as they are microphonic and non-linear, depends on the application where this matters (for audio the microphonic behaviour is a killer). For analog filters they are no good because of the wide variation between devices and high tempco.

Perhaps rework the circuit component values so you can use plastic film - reduce the capacitor values and increase the resistances? Plastic film don't age like electrolytics.

Asking this question has really been a helpful lesson for me, teaching new aspects of capacitors I need to know. Thanks to all of you who answered.

LarryD: I have use back to back caps in many projects.

Glad you told me. Now I know I can trust the "back-to-back" technique, knowing the person who previously told me about it wasn't making an unfounded guess.

LarryD: Use a LM339

That's an excellent idea, by using an LM399 comparator instead of an op amp, a feedback resistor from output to input could set the range between positive going trigger and negative-going trigger, and then send just a "1" or a "0" to the software; instead of the way I have it now, sending the software analog data in a string of bytes for the sketch to accomplish the same tasks as the LM399 could have already done.

dwightthinker: bias the center lead with a high ohms resistor. Such as, if the + leads together place a 1 meg resistor to the + power lead. The capacitors will always remain properly biased.

Now that's an extremely smart idea I'd never thought of. If I were staying with the electrolytics, I'd use it!

DrAzzy: Why use polarized caps in the first place? 4.7 uf ceramics are readily available and dirt cheap.

D'OH! I feel embarrassingly dumb, not to have discovered this on my own. I saw the many polarized versions online and assumed that's all that existed. I would not have needed to post this question. But thanks to you, I've now ordered ceramics in both 4.7uf and 10uf values. (And they really are cheap.)

jremington: I suggest to look for another [circuit] with proper biasing. You shouldn't need unpolarized caps.

take a look at http://www.learningelectronics.net/circuits/images/heart-rate-monitor-circuit-diagram-2.jpg..., which is an AC-coupled amp [for heart monitoring], biased at the midpoint of the power supply, and does not require a polarized cap.

Thanks. I'll build and test yours example, hoping it really will work better for me. Also the website it's from, LEARNING ELECTRONICS, seems an outstandingly interesting place to explore.

MarkT: High value ceramics are often unsuitable for analog coupling capacitors as they are... microphonic... non-linear, wide variation between devices... high tempco.

Perhaps ... use plastic film...? Plastic film don't age like electrolytics.

All really good points. I'd forgotten about "microphonics". I first discovered that when I brushed my finger across a ceramic cap, and heard it in the speaker almost like I'd brushed across the actual microphone!

I've now looked up info on "plastic film" capacitors, and they are just too big! This project is a tiny item to be worn.

I'm glad you confirmed electrolytics don't last. I worked in a computer repair shop where it seemed we were having to replaice a lot of blown electolytics from motherboards. So, they won't be my choice (where I have a choice).


All said and done, I think ceramics are the best choice for this circuit, as I only need it to see the heartbeat frequency, and report when a beat occurs.

Below is the circuit I'm actually using. It is better than some mentioned above because it filters out noise very well, which is a must for Arduino analog input sampling. In designing it, I varried part values until I got the best output; highly sensitive and clean.

|500x266