an electronish puzzle

Actually, if you just dump the collector current into a 1M scope input, there'd be numbers like a quarter of a volt in this case, plenty of signal. Tau might be in the 10s of microseconds, so millisecond thermals (should there be thermals!) should be obvious.

Once the b-e junction is zenered, all the transients should be over pretty quick.

John

Typical queen response....

No he's not- he's a little dodging slanderer and bullsh_t artist puke- not unlike yourself. You think you are sophisticated- but that passes only with the little dog-manure class of people you're used to dealing with- you are in reality a low-life and second rate.

Oh reeeeeeally? And do you flip your wrist while you write that? You're another dodging wimp who gets his ass kicked when there's no room for you to play your queen games. Like most scum who habituate usenet, you are 100% mouth... that's all you know how to do- a mouthy punk and fraud.

"Ken Smith" wrote in message news:d1ultn$8gp$ snipped-for-privacy@blue.rahul.net...

Quite so. I plan to use single frequency excitation. That makes extraction of signal from noisy sample data a little easier too.

Well, they are nonlinear, but the voltages across both junctions are or can be held pretty constant. (I plan to use a conventional, low Z-in photoamp circuit to collect the collector output current.)

If I read that correctly, you mean to permit a DC measurement of the b-e zenering effect, separated from the temperature effect, by holding some temperature within the device constant under DC conditions. (If it was transient or AC conditions, the cure would be harder than the problem.)

One problem with that approach is that there are temperature gradients within the device when you dissipate power at/around/in a junction, but the "bez" effects are not necessarily occuring at the same places as the dissipation. For example, the temperature effect on c-b leakage itself occurs throughout the base region and in the c-b depletion region. You would need to hold those temperatures constant.

With the AC excitation at a suitable frequency, both the capacitance effect (due to junction voltages not being constant) and the thermal effect have a fortunate phase relationship to the "bez" effect, which has no way to be slow enough to be anything but in phase with the excitation. (That's just "theory", of course; the lab may well play some of its usual pranks.)

In my work on a precision square wave generator one time, I found that junction scale thermal tails occur with about half mS timescales for small RF transistors. So I'm thinking 10 to 30 KHz will be a suitable frequency for getting the thermal effects (which are slower because not at one junction) to appear with a 90 degree phase shift.

I will (and do) appreciate any predictions as to problems with the AC excitation approach.

Once, in the drunk tank, some guy said I looked like Chuck Norris. ;-)

Cheers! Rich

Do you mean in "the dunk tank" with your shirt off or wet? Or do you mean "jail on public drunkeness charges"? (I ask only because you do not seem the sort to be found in the gutter.)

(And if you were gaffing, do take a point. ;-)

Nah, Chuck Norris isn't as good looking as you are.

In article , John Larkin wrote: [....]

You betcha

Charging the junction capacitances is going to cause a current spike that may be bigger than the value to be measured. This makes for some instumentation problems. The fact that the capcitances are not linear makes life even more interesting.

OR:

Figure out some way to make an electric heater keep the junction temperature constant. The semiconductor to case resistance is a linear effect. The effective location of the heat source within that semiconductor may move a little when the current changes, but I think that that is going to be too small of an effect to worry about.

Jail for getting caught driving while having over 0.10% blood alcohol content. (I don't get "drunk" until about 0.25%.) Actually, it was only "setting off the cop's breath box". And I wouldn't have got caught if I'd checked the taillights. )-;

Cheers! Rich

But the transition from zero emitter current to lots of current will happen in 0.1 volts delta Vbe or less. And the base is an ideal shield between the emitter and the collector, which is why people like cascodes (hell, you could *do* it as a cascode.)

So you could easily make a pulse that goes from Ie = 0 to Ie = 10s of mA, with almost no capacitive coupling to the collector/scope. And, as I noted, the recovery time constant will be fast. There's plenty of signal, so shunt the scope with a 100K resistor (or a 10K if you have something exotic like a 7A22 handy) and really bash the tau.

This is an easy (if unnecessary) measurement to make. With a scope, a pulse generator, a power supply, and a resistor or two, it could be done in minutes.

John

John

It is still less than the ideal situation. My main concern would be:

If you do it this way, the spike from charging the C-E capacitance will be something like 8V. This means the full scale of the scope must be about

With a 12 bit converter this would not be a show stopper but it is still not nice to lose those bits.

Considering the time constants:

The tail after this has lets say a 10uS TC. If we want to measure to 1%, we need to wait about:

This number is much less of a problem since the juction is not likely to change temperature hugely in that much time.

Yes, I agree with that.

How are you going to deal with the non-linearities of the capacitances and the fact that they don't have a zero damping factor? Using a single frequency seems like a bad idea because it allows confounding variables into the issue.

What will happen to the capacitance of the E-B juntion as the zener current changes?

I'll rem Newsgroups: sci.electronics.design Subject: Re: an electronish puzzle In article , Ken Smith wrote: [...]

I still think this is true.

The capacitances interacting with each other form capacitive voltage dividers. The E-B junction has a low and real impedance when it is zenered. This makes a path for the capacitance to carry an in phase AC component to the collector lead.

I assume you are running above 100MHz. In general cores become near useless at those frequencies. You may still be able to find some with cores. It would be interesting to know if their tempco could be counted on.

"Ken Smith" wrote in message news:d201na$ag6$ snipped-for-privacy@blue.rahul.net...

....

Can you elaborate on that? (maybe after considering the discussion on non-linearity as developed below) The "damping factor" issue completely escapes me.

It seems to me that it allows better separation of the confounding variables, at least the ones I know of. Which variables do you believe are more confounding with single frequency excitation? (Mind you, I am not saying only one frequency is to be used in the experiment -- just that they are used one at a time.)

I don't think it is sensitive to current per-se. To get that current change, the voltage must change. And that modulates the depletion region width, which of course is as non-linear as the doping profile is non- constant (away from the junction itself).

I agree that non-linearity can be a real complication. But its impact can be reduced by using the AC excitation approach, with signals made small enough.

....

I think so too. I did not mean to contravene that, but to bring up a larger difficulty with trying to hold some relevant temperature constant. I will try to state the point less obliquely.

The heat generated by the b-e zenering flows from the b-e junction to either { mainly toward the collector } or { about equally toward the collector and the other way }, depending on the package type. That heat will flow thru thermal resistance, leading to a temperature difference between the b-e junction and the regions where the temperature effect mainly occurs. I imagine you could hold the b-e junction temperature constant, but that would mean the regions where the temperature effect occurs would not be constant.

If you are claiming the temperature drop between the b-e junction and { the base region and c-b junction } is too small to justify worry about its variation, then I can only say that depends on the magnitude of the temperature effect relative to the sum of the others. I don't know how to avoid the worry based on what is known about that.

I cannot see that. The b-e capacitance and the b-e voltage change required to modify the tunnelling probability (as real an impedance as can be) are in parallel. If the base is AC grounded, the current from those two components of the emitter admittance can be separated as the in phase and quadrature phase components.

Here is what I think the setup will look like: ___ .------|___|------. | | | * || | o-------------||--o | || | | |\| | o--------|-\ | | | >-----' | +3V-|+/ GND DUT | |/| DC source | |/ / \ GND-| ( ~ ) |> * = HF stabilization only \_/ | GND | ___ | | '----|___|---o / \ | ( ~ ) --- \_/ Sig gen --- | ___ | '----------|___|---'

Constructive comments are welcome.

Are you going to measure it, or talk it to death? I could have done it in 10 minutes, about twice what it took to do the DC version.

John

"John Larkin" wrote in message news: snipped-for-privacy@4ax.com...

I'm going to get my sig-gen back from a friend, get a current (and paying) mini-project cleared off my limited workspace, and measure it as I've shown. It could be a few days or a week. Is that going to cause any real problems here?

Since I don't expect that you will find the thermal effect, and because the point is entirely academic anyhow, no problems.

John