Replace tantalum caps with aluminum?

tantalums

or

should

They still make wet slug tantalums that use sulfuric acid as the dielectric. Not pretty when they explode.

dies

much.

$0.01

You didn't go far enough. Hardware manufacturers are actively working with part manufacturers (not that either one will admit it publically) to find ways to make parts less reliable, even if it costs a little more. The goal is to not have any failures before 3 years (so they don't have extended warranty costs), but as many failures as they can before 5 years.

goal

what's your basis for that assertion? Not that I'm challenging it, but I wouldn't take it on faith, either...

Bullshit

--

    Boris Mohar

to

In round numbers, the military/space/avionics market consumes about

1% of the integrated circuits produced, the automotive market about 4%, and various flavors of telecom and consumer electronics the other 95%. In general terms (and therefore open to much debate), the parts made for the different markets 30 years ago were quite different, but the parts made over the last 10 years or are quite similar. The reason for this is the fantastic strides made in reliability of even the cheapest parts made them "good enough" to use in a lot of applications where previously, "high reliability" parts were required. But that caused another problem for OEMs; commercial electronics have become incredibly reliable. The OEMs initial response was brilliant, they pushed "extended warranties" on consumers who remembered problems they had had with previous generations of hardware. Over time, consumers have become much more leary of buying extended warranties that their new experience showed they would never collect on.

The big problem for OEMs is how to get consumers to replace hardware that doesn't fail. The computer and cell phone OEMs have done very well in providing increased functionality to make this a mute point, but other product sectors (e.g. TVs) haven't (don't get me started on the "digital TV" shell game). The obvious solution was to literally build in timers that would FORCE the hardware to fail after a certain date, but the companies legal departments didn't like that idea. The second solution was to design in parts that would wear out. The beauty of this solution was that in most cases the parts that will fail early are SLIGHTLY less expensive, and often have slightly better performance. So now, an OEM can win on both ends, a cheaper product up front, and selling a replacement at an earlier date than they otherwise would. The problem is that the automotive market needed the long term reliability because the automotive environment ages parts a lot faster than sitting on a desk does. The problem for the part manufactures is that 95% of their customers want a product that the other 5% can't tolerate.

What is happening today is that the differences between the automotive grade parts and the military grade parts is becoming almost nil, but that they are becoming more and more different from consumer grade parts. Little things like passivation, metalization thickness, current density, bond wire materials and thickness, and ionic impurities in the plastic molding compounds are all becoming different whereas they used to be the same. The company line always is something like "to reduce costs/to increase market share/to provide more performance/etc" and in most cases those things are true, but they aren't the whole truth. Also, since this is real engineering, there are always tradeoffs. In many cases you CAN'T get the increased performance without decreasing the reliability or radically increase the cost. And none of this is evil, it is capitalism, pure and simple.

I attended a IEEE Reliability Group meeting where this was shown to be a fact. Most manufacturer's goal is to acheive a 3 yr MTTF. The is the expected lifetime.

Al

--
There's never enough time to do it right the first time.......

to

But there's a far cry between "trying to achieve a 3 yr MTTF" and doing that "even if it costs a little more."

It is well known that manufacturers aim for a certain system-wide MTTF. For instance, there are stories of Henry Ford studying junked cars to see what parts were still good, because those parts were by definition overengineered: if some of the car had already worn out, but some of it hadn't, he was wasting money on the good junked parts.

But that was with the goal of saving money. Then we get to "planned obsolescence," that is, the effort to make sure something will fail in a certain time so that customers will buy more. This still isn't quite the same as paying more to have something fail; it's more a matter of intentionally choosing the less reliable (but still presumed cheaper) parts, across the board, to create an object that is disposable rather than lasting. For instance, my HP654A test oscillator, the progenitor of this thread, was built to last. My Neutrik Minirator, though a convenient, powerful, and highly performant instrument, is made of plastic and is completely nonmaintainable. It'll die before the HP does, even though it's

35 years younger.

Finally we get to Ken's premise, that manufacturers of equipment are actually willing to work with parts makers to design, and pay extra for, parts that fail sooner. I can't reject that claim, but I've never seen any evidence of it from the designers I've spoken with (a small and biased sample), and I'm skeptical.

Ken's response to my query didn't quite defend the premise; if I may paraphrase his response, it was that for consumer parts, the prioritization is cost, then performance, then reliability; so reliability is suffering. That is different from saying that they are actively searching for ways to reduce reliability even at higher cost.

-w

working

publically)

more.

before

that

For

parts,

it's

any

prioritization

I did say something that was a little wrong. I know of OEMs that have approached parts manufacturers and said the 3 year/5 year thing. I also know part manufacturers who have said "well, we could do this", and that "this" did cost "more", but the "more" was on the order of thousanths of a penny per part or less, so it is really in the noise. Given the time value of money, I wouldn't believe anyone either that said that OEMs were willing to spend X% more to obtain parts that fail sooner.

Thank heavens! A little bit of good news is welcome amidst all the rest of the hoo-hah on this newsgroup. I'm glad to hear that my faith in humanity is at least partly justified. Thanks :-)

-w

I would extremely doubt that any of the tantalums are bad so why would you replace them? As well as low ESR they exibit much lower "leakage currents" and generally better "dielectric absorbtion" compared to aluminum caps. Even with todays technology. There are some good aluminum caps, but they will cost as much or more than the tantalums. Not knowing where these caps are used in the circuit, but these features could be important.

In a lot of years experience, I have never seen a bad tantalum, when use properly in a circuit. They are not good as coupling caps and really hate reverse polarity, but in most other instances they are very relilable.

Take care.......Gary

But there is no need for the manufacturers to do that. Simply design the thing (whatever) for lower reliability. e.g. If a resistor has to dissipate .475 watts, good practice would be to use a 1 watt (or more) unit. To lower the reliability, use a 1/2 watt unit. Specify a 16 V electrolytic in a 12 volt circuit instead of a 25, 35 or 50 volt cap. Don't use ESR caps. Use smaller heat sinks. Essentially, use the minimum spec'ed parts to do the job, and leave out parts that are for reliability only.

I think the economics automatically drives this, and there is no need to purposely lower the designed in reliability. The dictate is to make it cheaper. That pressures the engineer toward cheaper parts and leaving out stuff that is not absolutely necessary. Protective diodes - we don't need no stinkin' protective diodes. And so forth.

From the parts manufacturer's viewpoint, the same thing happens. The clients are clamoring for a lower priced xyz component. If you can make the component with 10% less material, you can reduce the cost, retain the client, and maybe increase the profit margin. So what if the life expectancy of the component is reduced as a result, as long as it still lasts for ~3 years.

What makes them not good for coupling?

The always need a forward bias. You have to ride the AC on a DC bias. This is not always acheivable.

Al

--
There's never enough time to do it right the first time.......

Are you saying that tantalum electrolytics have more severe bias requirements in coupling applications than aluminum electrolytics? (My experience is that they are less severe). Granted-- neither will be as tolerant as a true nonpolar capacitor.

-frank

Tantalums may tolerate about 0.5 V reverse bia whereas an aluminum electrolytic may tolerate up to 1.5 V reverse bias. In audio coupling applications this can make the diff.

Al

--
There's never enough time to do it right the first time.......

I believe that this is correct. The manufacturers' literature will back this up.

OTOH, if there was no bias in the circuit, a coupling capacitor wouldn't be necessary. We could just use direct coupling.

My impression has always been that capacitive coupling was always used precisely BECAUSE there was a DC bias present.

I've used lots of tantalum capacitors as coupling capacitors to replace coupling caps that were originally alum electrolytics. The reliability of the devices I've done this in has always gone up as a result of this.

-

----------------------------------------------- Jim Adney snipped-for-privacy@vwtype3.org Madison, WI 53711 USA

-----------------------------------------------

Sometimes you don't know what the bias will be; you just presume there could be one. For instance, at the input to an amplifier. Could be some DC riding on the input; you don't want to amplify it; so you block it. Or output; FWIW, the HP654A test oscillator that started this thread has two

4000uF, 3V nonpolar caps between the output amp and the attenuators.

Amusing factoid: many "balanced outputs" in audio devices have blocking caps. Because the presumed load impedance is 600 ohms, the caps need to be high value, so they're electrolytics, usually rated at somewhere near the supply rail voltage. But it's common to plug balanced outputs into mixer mic inputs, which often have phantom power present. Phantom power is usually 48V behind a 6.81k resistor. Not all 16v caps will hold up well, if you connect them to that. Bang.

That's a good point. I had only been thinking about coupling caps between stages WITHIN a device. Your examples are for blocking caps at the interfaces. I'd think such caps would need to be non-polar so that they could handle I/O offsets of either polarity.

-

----------------------------------------------- Jim Adney snipped-for-privacy@vwtype3.org Madison, WI 53711 USA

-----------------------------------------------