开云体育

You may recall that I have an S-1 sampling head with blown diodes, which I replaced with M/Acom MA4E1339 Schottky diodes in SOT-323 packages. I used both diodes in the package in series, max 1.2 pf each, to keep the capacitance down. They are mounted on small pieces of copper-clad and the package spans a gap in the copper. Hopefully low capacitance.

Anyhow, it works fine on fast pulses, and at DC too. But I ran a 1 MHz square wave through it (something I would not normally use a sampling scope for) and found that the transient response is waaaaay undercompensated. There's a bad overshoot, then undershoot, finally recovery. Per the S-1 manual that is the input to use while adjusting the transient response. But there isn't enough range on the pot (R13) to get it near flat, although with it all the way "up" it's better. I'll post display pics tomorrow.

I conclude that either my mount has excessive capacitance causing blowby in that frequency range, or possibly Q13 is bad. Although the trigger pickoff does function (Q13 drives Q17 which sends the trigger to the 3S2). It will be difficult to make a mount with much less C... the diodes themselves, at 0.6 pf, should not be a problem.

So I could use some suggestions. If I have to modify the circuit to allow for more out-of-phase signal to bypass the sampling bridge, what would be easiest to change with minimal side effects? Thanks.

Using double diodes changes a lot of the conditions. Try going back to single diodes instead of series doubles, and see how it works. You can temporarily short one of the diodes in each pair for experimenting. There may not be enough strobe drive - it may be adjustable and fixed that way. Also, check the other adjustments and study their purposes and effects.

Ed

The strobe is adjustable in several ways (avalanche volts, snap-off current, bridge volts). But wouldn't insufficient drive show up as an incorrect sample time, lack of sensitivity, or both?

As I mentioned, it works properly with my pulser (4 ns flattop pulse with 650 ps risetime). The sampler risetime is at least in the ballpark too (would expect to display sqrt(650^2+350^2) = 738 and that looked about right when I checked).

The problem I'm seeing is exactly consistent with the photos in the manual (setting transient response at 1 MHz). There is too much feedthrough aka blowby and the inverted signal through Q13 does not have enough adjustment range to null it out. I thought that double diodes would help this. How would going back to single diodes help, which will double the feedthrough capacitance? Please explain further, I'm learning as fast as I can.

Also it will be mechanically difficult for me to jumper one diode without damaging or destroying anything ;)

Charles, the reverse-biased capacitance of two diodes in series is not necessarily one-half of the nominal C (from the spec) of either one. The capacitance is nonlinear, and is maximum near zero bias. So, putting two in series not only changes the DC bridge voltage and balance and strobe requirement, but may also (counter-intuitively) actually increase the capacitance. There's not all that much reverse bias on the bridge in the first place, so the diodes operate possibly in the steep part of the C-V curve. Since you are using a known part that hopefully has specs available, check the C-V curves in the data sheet, and estimate the C you would get for one diode only, from the actual bias conditions on the bridge. Now consider how much C each diode would have at half the voltage, assuming they each share the total reverse bias equally, if you have two in series. Then divide that C by two for the series C estimate for the pair. Depending on the curves, you may find one diode is better than two. Ideally, you want the bias to land it in the flatter part of the C-V curve, where the C is low, and fairly constant. This is optimized at the maximum possible reverse, but of course, that means more strobe is needed. It's all in the trade-offs. Also note that all the adjustments tend to interact, and the diode C-V characteristic is a big part.

The easiest thing conceptually, is to just go back to single diodes and see what happens, where the system was intended to operate. Yes, I understand it may be physically difficult to make the mod, but consider that you may end up doing it anyway, after studying the situation.

Ed

Thanks! Feel free to keep sharing your knowledge ;) I have never played with varactors so that C-V curve didn't even occur to me... so what happens to the junction capacitance under forward bias?

But the data sheet does not show a curve- the only entry is:
Total Capacitance (Ct ) Vr = 0 V, F = 1 MHz, 1.2 pF max.

That leads me to believe that my 0.6 pF conjecture is not so far off, if the capacitance is maximum at 0 (as you said, and where the data sheet figure was measured). I'll take a closer look when I get a chance, and revisit the strobe adjustments too. Possibly I adjusted it into a corner where everything is working except the transient response?

I looked up the part and see there is no C-V curve, only the max C at 0 bias. So, that's the max C to expect. What you don't know then is what the actual C will be when biased up properly - it must be less than that, but who knows how much. So, you won't be able to paper-compare the various combinations without more spec info, or possibly testing the parts, which may be difficult without the right facilities. It may be possible to find some app notes or such on these diode types or families, that include more info.

I also looked at the S-1 manual, and the sampler diodes are described as a GaAs matched pair, with no C or other info to go by. The new diodes are Si Schottkys, and likely already have higher Vf than the original parts, so using series pairs would push the bias conditions even further from the original design. Do you have any more info about the original parts, or still have them?

So, I still think the most straightforward approach boils down to figuring a way to hook it up with single diodes. From the series pair SOT picture, it looks like a thin wire tacked over to jump out one diode in each should be doable, if the parts are accessible. Also, which end of R13 limits out for sure - the hot end or the ground end? It could be that the diode C is too small rather than too large compared to the original parts. It may be worth a close look at the circuit layout to be sure which direction the pot actually goes, and whether the blow-by compensation needs more or less gain. Is the tilt rising or falling at the limit?

Ed

Hi Charles,

I have had multiple S-1 and S-2 heads with transient responses that could
not be leveled. In all but one case, the problem was caused by R18 that
had drifted high by 100% - 150%. It also caused increased signal noise,
especially with the sampling head case removed. (All my sampling heads
still have original diodes, though.)

I thought I'd mention this, as it is quick and easy to check, and fairly easy to fix.


Good luck,
dan

Thanks, Ed & dan. I will check R18 in particular. But I apparently have overthought the problem. It's mechanical, not electrical! I posted three pics in "Sampling with 3S2" (and set the order to most recent first, to save you thumbing through in alphabetical order).

I carefully adjusted the pots (ALL of which interact, what a PITA) as in the manual. Since I don't have a really fast edge, I used my ~600-650 ps pulser edge and set it to display ~750 through the S-1, which should be close to 350 ps risetime. Going back to the 1 MHz square wave, it looks like crap as I initially described. Not only that, suddenly the transient response pot wouldn't do anything either! But flexing the preamp board the right way caused it to pop into flatness (second pic). Now I just have to find the bad solder joint (or carbon comp resistor)...

The slight ripples in the S-2 are probably because I haven't adjusted it either - or possibly from the 114, cables and adapters, and too fast to see on my 60 MHz scope mainframe)... I'd try my delicate S-4 but it has a new hybrid and hasn't been readjusted either! Guess it's time to buy a Leo Bodnar pulser :)

You can find the likely part numbers for the original diodes in the "Tek schottky Diodes memo" in the reference materials files on Tek Wiki.
Specs for many of the part numbers are in section 12 of the Semiconductor common design parts catalog also on tek wiki.
Craig

Thanks... but I broke the nylon holder, and those diodes are rather prone to destruction (which is how I got the S-1 in the first place). Also they are REALLY tiny and I just don't have the mechanical skill to work with them - in fact I broke a lead, on the remaining good one of course, just removing them from the holder. I decided that more recent diodes could be substituted, and SOT-23 packages are about as small as I'm comfortable handling... and don't forget that they are available from electronic suppliers like Digikey, instead of (50+ year old) dwindling Tek parts stashes.

Anyhow, I finally found the problem... the push-together connectors carrying the trigger from the sampler board pickoff to the preamp board were just barely mating. So when they parted, the center pin(s) were so close that there was enough capacitative coupling to send a fast pulse to the trigger amp, but way too small for the blowby compensation circuit. Verified with scope. On further investigation, the case is supposed to have four "wiggly" long spring strips to make contact with the ground planes on the outside boards, and presumably to also keep a tiny bit of pressure on the connectors at issue! There were only three, and the missing one was on the side where the problem is. I moved one strip to that slot, reassembled and everything's working fine again :)

p.s. to dan: the notorious R18 (36 ohm) was 37.7 so I left it alone :)

Congratulations on getting it all working, Charles. Another victory for persistence!

-- Cheers,
Tom

--
Prof. Thomas H. Lee
Allen Ctr., Rm. 205
350 Jane Stanford Way
Stanford University
Stanford, CA 94305-4070

Thanks :)

While comparing the S-1 to the S-2, I discovered another reason to prefer the diodes I used.. I was swapping cables around and plugged the avalanche pulser (at least 10 volts) into the S-1 without the 10x attenuator! Oh my, I said, and yanked the power lead to the avalanche supply as soon as I saw the trace shoot off the screen...

So I put the attenuator on the CORRECT side of the divider tee and the S-1 still works. I am sure the original diodes would have bit the big one instantly. Also glad I didn't do that to the S-2!