'Slightly' reversed biased tantalum.

Trinitite?

It has always amused me that one of the Rosenbergs' co-conspirators was named Greenglass ;-)

The ones in Tek 7XXX 'scopes were usually silent but deadly. The first thing you knew was when the lab filled with smoke.

Hide quoted text -

Guess I'm still reacting to the thread a week or so ago where someone had designed a circuit that deliberately used polarized caps in a way that reversed them.

The caps are sitting on my bench, it took a

The devil is always in the details. A more interesting test would have been to reverse the caps in the product and run a few through the production pipe to see if the problem showed up in the standard QC procedure.

If it does then we can hope to

Still think your test is inconclusive. If the parts catch fire, you've learned that you shouldn't put parts in backwards. If the parts don't catch fire, you've learned that five parts in a different configuration didn't catch fire. You already knew that you shouldn't put parts in backwards. And you still don't know if the QC process catches it. That five parts on the bench didn't catch fire is not license to ship product with backwards caps.

Production errors happen no matter what you do. If a peek at the production and field repair logs suggests that the incidence of this cap being put in backwards is higher-than-normal, you need to determine why and fix that.

If the consequences of the inevitable random failure are great, an addition to the QC procedure to "check C1 polarity" can help.

This is the kind of statement that frustrates me. Reversed tantalum caps are NOT OK CAPACITORS!!!! Never have been. Never will be. Testing five samples does not make it so. Anybody who thinks so needs to spend more time at gamblers anonymous meetings.

text -

etting flaky

a competent

(4.7uF

it. =A0All

e it was put

he pulled

ears. =A0But it

t=92s hard to

any idea

ndering how

my bench for

e leakage

ing

istor.

versed

sent me

suring

=A0 =A0 =A0 =A0 =A0 ...Jim Thompson

=A0 =A0 =A0 =A0| =A0 =A0mens =A0 =A0 |

=A0 =A0 =A0 | =A0 =A0 et =A0 =A0 =A0|

nus =A0 =A0|

=A0 =A0 =A0 =A0 =A0 |

Rat =A0|

=A0 |

food.- Hide quoted text -

ture

d be one

go

their

of

sky :-)

=A0At

e,

ent

Hi Mike, Well we are a very small company. I get to do the final testing on a bunch of instruments. (six and counting) On a few insturments I've demanded to visually inspect the pcb before they are installed. (so I can catch errors like this.) But I'd much prefer if this could be done by someone in production.

I think you work(ed) in a different world, so to speak. But my goal this week is to get all pcb's visually inspected for reversed tantalums. This seems to be an error that slips right past any test I can invent.

OK as long as it won't get your blood pressure too high I'm willing to argue the other side. So first please take this as an academic discussion. I'm not going to build any instruments with tants in backwards, nor do I advocate that anyone else do it. After all there are plenty of other non-polarized caps out there.

That said, If I run the tantalums way below their rated voltage, say

George H.

talum

quoted text -

So I could put one in a low voltage, low impedance circuit

Couldn't agree more. If it quacks like a part used in a manner contrary to it's defining physics/chemistry...swims like a flawed design...

There is zero, none, not any justification for reverse biasing a polarized cap...none, not any. We've already got way too many unintentional design flaws. Putting them in on purpose is unacceptable. The notion that an engineer would even consider the concept boggles my mind.

If you don't care how it behaves, just leave out the part.

I'm not emotional about this, just unyielding... Did I mention zero, none, not any justification???

"Slightly pregnant" comes to mind...

it behave as a capacitor. =A0If it quacks like a duck and swims

..

nd.

Oh sorry, I'm a Physicist, not an engineer.

Well BS in EE and more alphabet soup in Physics.

I was thinking you could use the nonlinear impedance as some sort of voltage control. Maybe a real low frequency Wien bridge oscillator.

But there's lots better ways to do that. .

Yeah sure... If I ever think of something useful, I won't tell you.

(Not that it's very likely, there's lots of people way smarter than I am.)

George H.

=A0 =A0(Richard Feynman)

I fixed a backplane in one of my beloved 7xxx Tek scopes once.

_Dead_shorted_ across the power planes, so I was pleased to desolder a tant and find it shorted. But, that didn't fix it.

Then I found another. And another. And...another!!! They were all shorted. Every single one. Nasty. Worked like a charm once fixed, though.

-- Cheers, James Arthur

pg. 6

I've seen back-to-back tantalums used to filter analog inputs, esp. thermocouples.

But, the explanation above doesn't fly. On a transient, the reverse- biased device will see half the transient voltage--not negligible at all.

-- Cheers, James Arthur

flaky

All

put

it

to

idea

for

leakage

Hide quoted text -

:-)

He he, just do a reverse bias test: 30V reversed supply on all your boards... The tants that do not explode were mounted backwards.

Folks also sometimes put a big resistor from the junction to the appropriate power supply, to avoid reliance on leakage to bias the caps correctly. The other problem is the huge tempcos of electrolytics--as the C value changes, current has to flow to make V stay the same, resulting in really putrid low frequency noise. (This happens even in the leakage-biased case, if there's any significant voltage across the terminals.)

Cheers

Phil Hobbs

I've done the same thing with aluminum electro's mostly for AC coupling where I don't really care about the exact value of the capacitance. But now I wonder what the real capacitance is/was. At low voltage it looks like two caps in series, (1/2 C) but then at high voltage when the reversed one starts to leak... does the capacitance increase?

The other thing that bothers me is why do EE=92s always talk about an exponential as having a =91knee=92 voltage. They never heard of a log scale?

Here=92s the reverse current vs voltage for the 4.7 uF 35V tant and a

V rev. I tant I alum. (V) (A) (A) --------------------------------- 1 2E-7 2E-7 2 1.3E-6 2.7E-6 4 1.2E-5 3.2E-5 5 3.2E-5 2E-4 6 8.9E-5 1.1E-3 7 1.7E-4 2.8E-3 10 5.7E-4 2.7E-2

and a pic

George H.

The 7xxx backplane/interface assembly is a bitch to change out components, and a bigger bitch to take out, Capacitors on there have been responsible for many problems wrongly attributed to the power supply. I've had aluminum electrolytics, as well as tantalum, go short on there. I once had a tech waste nearly a week trying to troubleshoot a 7104 PSU, before giving up, and reporting a faulty transformer. Turned out to be a shorted aluminum electrolytic on the 50V rail on the interface board, didn't show up under DMM ohms range voltage. PSU totally blameless as usual.

BTW, if you want to run a 7xxx PSU in isolation, on dummy loads, disconnect the feedback cable, there are 100 ohm resistors built into the PSU to allow you to run in isolation. They burn up if you leave the external feedback connected with a shorted rail, and the PSU comes up.

I still have a 7904A on my bench, and a 7104 hovering around on a Scopemobile. Wouldn't be without them.

I do have "modern" digital scopes as well. Don't like 'em much.

ed,

The 7104 PSU has a really cool bjt-as-synchronous rectifier trick: a buck, where a tap off the flyback inductor drives a bjt catch switch. Slick.

I have one (for sure) or two 7104s in the stable (can't remember...I once had four, then divested), with sampling plug-ins good to 12GHz. I interfaced one of the samplers to be swept and digitized by the PC...neat stuff.

Yeah, I've got a few digital 'scopes, including a 150Mhz Yokogawa/HP, a 4-ch plug-in for the 7xxx, and others.

Digital's handy for web-stuff and documentation. I usually prefer analog for troubleshooting. Digital rocks for one-shot and low-duty cycle.

-- Cheers, James Arthur

Using a 7104 with samplers is a waste of a good mainframe. A 7704 is just as good for this purpose and a lot cheaper (those microchannel tubes were

Which was one of the big selling points of the 7104. I much prefer M/DSOs, now. I suppose it's all the way you look at debugging.

e:

rst

lic

all

fixed,

nts,

ble

had

ted

how

the

ust as

I mis-"spoke." It is indeed a 7704A.

I don't have every plug-in ever made for it, but a pretty darn close-- counters, DMM, DSO, event trigger, samplers, low-level analog, differential input, etc.

I also have a 7603 (3-bay, 100MHz) in the dungeon.

-- Cheers, James Arthur

That makes more sense. ;-) The 7104 was an amazing beast but it did have some severe limitations and cost about 3x a 7904.

Nice stuff. I remember my first day in IBM (1974(, when I sat down in front of a 7904 these beasts (and couldn't find the power switch - a big bat handle in front of my face). In college the best I'd worked with was a 485, but usually an HP140. ;-)

Haven't used anything that crude in almost 40 years. ;-)

They just delivered a pair of Agilent MSO-X3104s (I ordered 3054s - no idea who changed them to 3014s) this week, one for me and one for my partner. So far I've looked at the calibrator. ;-)

I just got a Tek DPO2024, with a gorgeous hi-res panoramic screen. Once it boots up (which takes ~~forever) it's wonderful.

Rob, down the hall, topped my by getting the Rigol 4-channel 1 GHz scope, whose screen would work for home theater. It's serious. Rigol doesn't just make toy 60 MHz scopes any more.

We still have one analog scope that we use, a 7104 on a cart.

A fabulous "classic" is the 11801 sampler with an SD24 20 GHz TDR/sampling head. It's beautifully quantitative. You can get the combo now for under $2K. I must have a dozen, and a lot of heads, pushing a million bucks at the original price.

I just bought a spare SD24 for $103. (We'll see if it works.) I'm just setting up to use one to measure the step response of a bunch of photodiodes, to see how much it changes with position on the active area.

I don't have a good handle on the effects of the epi layer and recombination at the oxide. You can sometimes deplete the epi right at the oxide surface, which gets rid of the surface recombination and improves the QE some. I expect it'll also improve the spatial uniformity and 1/f noise, for the same reason.

Cheers

Phil Hobbs