Page 1 of 7

Not Quite As Simple CSA

Posted: Tue Mar 02, 2010 6:18 am
by Richard Hester
Attached is a schematic for a charge sensitive preamp that runs off a 9V battery, and should be adequate for the output of He3 and B10 proportional tubes. The output pulses will be around 100-200 mV peak, ~2 usec wide with the 0.1-0.2 pC pulses from that sort of detector tube. Circuit hasn't been built yet, but it's pretty straightforward and simulates very cleanly in ORCAD. Those with an appropriate detector tube (I don't have one) may benefit.

Re: Not Quite As Simple CSA

Posted: Tue Mar 02, 2010 11:44 am
by Starfire
Excellent Richard - do you have a printed circuit layout for this?

Re: Not Quite As Simple CSA

Posted: Tue Mar 02, 2010 8:13 pm
by DaveC
Richard -

Clean and simple. Very nice.

Thanks for posting.

Dave Cooper

Re: Not Quite As Simple CSA

Posted: Wed Mar 03, 2010 3:11 am
by Richard Hester
For John - sorry, no layout. If I build one for myself, I'll use perf board or "dead bug" technique and have it built faster than I could lay it out. Dead bug style has the advantage of leaving the summing node up in the air, helping with leakage. A good wash-down with alcohol after all is said and done (hang it up to dry and don't touch it with your grubby fingers -all our fingers are grubby) helps, too.

Re: Not Quite As Simple CSA

Posted: Wed Mar 03, 2010 3:12 pm
by Richard Hull
Great preamp, Richard! I gotta diddle this one into a reality.

For those not highly electronically inclined or not used to electronic assembly, the first two caps C1 and C2 both need to be rated for ~3kv or more for a decent safety margin. (not seen on diagram)

For my purposes, I like to wash or clean the bodies of all components in contact with the gate lead of the input FET with absolute ethyl alcohol and not board mount them, but instead, keep them in air over the board. Use a good ground plane based board if you go that far and surround the whole set up in a faraday box. for the ultimate set up, try and work a male HN connector directly to the box and cable nothing to the input. This would allow the preamp to directly mount to a stock 3He tube, reducing noise tremendously. All this is worth the effort, especially in a noisey or uncontrolled home environment

The far side of the components would be allowed to be board inserted. This construction is seen in the best charge sensitive preamps and electrometers.

I mount my Princeton Gamma Tech preamp in this manner.

Re: Not Quite As Simple CSA

Posted: Wed Mar 03, 2010 4:51 pm
by Doug Coulter
Air wiring is good, but gilding the lilly here. At 2.2 meg input impedance (not counting the fact that it's really much less, due to negative feedback), the leakage of a decent board is far into the noise, relatively speaking. Even the rather high leakage of a 1n4148 would be larger (25 nA on the spec sheet). Typical PCB leakage is far above the 100 meg+ range, particularly if the board is post-build coated with something (solder mask or something you add at home) to keep the conductors insulated from any surface contamination (humidity, dust, electro-migration).

This doesn't apply to perf board, which can be pretty bad (forget phenolic), due to the fact you can't really clean them as well after building (all the extra holes collect gunk that seems to re-spread after cleaning attempts). I've only had to resort to air wiring (which also can collect floating dust etc) in cases where I was working with impedances in the e-14 to e-16 ohm range and worried about DC, vs AC errors.

Quote from the National semi apps book:

"Board leakage becomes more of a factor as impedances are raised. It may not be advisable to take advantage of the full potential of the LM11's 5 pA bias current in all cases, when hostile environments are involved. Anyone designing high reliability equipment that is going to be in trouble if combined leakages are greater than 10pa at 125 C had best know what he is about." Which kind of sets where the numbers were back when this was written -- 1980 -- things have gotten better since. Lessee, 10 pa into 10 megs is.... e-11 * e7 -- e-4 volts offset-- that's not all noise, not mostly noise, just DC, varying with humidity and temperature. Tenth of a mv DC error due to that -- and this circuit is AC coupled internally -- C2 and L1 take care of that (and some normal variations in the fet IDss).
So, don't paste peanut butter between the FET leads, and do clean the board. But these impedance levels, no sweat on a PCB if the rosin is cleaned off after soldering. I've never had to clean any but the oldest most corroded badly stored parts for things in this impedance range -- the trashcan is a better choice for those anyway. You're going to have more noise from the bypass electrolytics specified here.

OTOH, the idea of making this into a box that screws directly onto the sensor is a very good one.
Looking at millivolts right near a high voltage supply that can easily make 10's of volts corona noise on a scope probe placed where the detector would be, even when things are "quiet" is hard to do -- at that point, you've got to have 10k to 1 noise rejection just to get a 1::1 SNR! Much cheapo coax isn't 100% shield coverage (if you were going to have the preamp remote), and of course doesn't address ground loop troubles, so having the preamp right there and shielded *tight* is wise indeed. Even then, the key word is "tight" here, as just a couple screws holding a Bud minibox together won't make it tight enough to reject loud pulses on the HV (including the AC magnetic fields those make with most HV wiring schemes that look like a big transformer primary turn).

I'm in fact using something simpler here (just bipolars) and doing fine with it, but have had troubles with EMI getting in through a sloppy aluminum box at the "joints" that swamp all these leakage effects, including the noisy leakage from the protection diodes that "see" X rays to an extent when reverse biased, as well -- they are pretty good "photodiodes" for those energy photons.

Re: Not Quite As Simple CSA

Posted: Wed Mar 03, 2010 6:31 pm
by Richard Hester
I beg to differ with you Doug, but I saw differences in the pulse response even at these impedance levels between air mounted and board mounted components, mostly in the fall time of the pulse. I did a lot of work with these things a few years ago when researching preamp designs. With the small value of cap used in this design, leakage will start to have an effect. Good construction practice never hurts.

If you insist on using perf board, putting the input components on teflon standoffs is not a bad idea, follwed by an alcohol wash. It also makes mods easier, in case you want a different charge storage cap or decay constant.

Re: Not Quite As Simple CSA

Posted: Wed Mar 03, 2010 7:17 pm
by Doug Coulter
You can differ with *me* all you want (why should I care, I'm right!), but not with physics and EE math and get away with it. 3dB point of 10 nf and 2 megs is order of 7 hz. It goes up a lot faster if the other end of the 2.2 megs isn't grounded, as here (because it no longer looks like 2.2 megs to ground from negative feedback), but hey -- this still implies something else, not normal clean-board leakage, was going on for you to see any difference whatever on fairly fast pulses. Else virtually all the analog stuff I designed for 30 years would not have worked, but instead it did and made me able to retire early.
I'm way -way past guessing on issues like this one.

Try it in your simulator if you don't believe me, just add another 2 megs straight to ground on the fet input, or to a bypass to ground so as not to mess up DC levels, and see what difference it makes. And board leakage R is hundreds of times higher unless there's visible corrosion there.

Re: Not Quite As Simple CSA

Posted: Wed Mar 03, 2010 8:29 pm
by DaveC
I think, both of you are correct about different aspects.

Getting the pulse into the circuit is a characteristic impedance issue, which input geometry and the board material will affect. It may take teflon PCB to get the permitivity low enough.

Inbound, the circuit impedance needs to match the tube characteristic impedance... which from geometry has to look something like 75 ohms or therabouts. Air wiring, while asking for trouble in the RF pickup mode (if un-enclosed, that is) reduces parasitic capacitances..

But Doug is correct also, that for the short retention time needed for digitizing, not much leakage will occur with almost any type of decent construction.


Dave Cooper

Re: Not Quite As Simple CSA

Posted: Wed Mar 03, 2010 10:28 pm
by Doug Coulter
Still not right, sorry. And this from some of the smarter folks here, re electronics. Sigh. This is not a personal attack of any kind, just an attempt to educate ;~}

From the geometry, the tube "impedance" at RF is in the few hundreds ohms, the center conductor is tiny indeed in these -- to make it 75 ohms with over an inch ID of the outer conductor, the inner one would have to be well over 1/2" diameter (I'd have to go look it up in a handbook, but after awhile the eyes are close enough for things like this). All the tubes I've looked in use a tiny center conductor to get gas gain where the field gets high near the small radius. Tiny wire in big tube, dielectric constant basically one, high impedance compared to most coax. (the old stuff they used for car AM radios is the exception - air dielectric mostly, and tiny wire, same idea) With dielectric == 1, the impedance of a coaxial affair is a function of D/d - a larger dielectric constant makes it lower for a given size ratio.

However, remember that a mismatched transmission line has it's *lowest* resonant frequency (for ringing) at the point where it is 1/4 wavelength -- so let's work that out and see if that's what matters here.

The tube is a little over a foot long, so the place where that kind of thing happens is far higher in frequency than these couple uS pulses (a full wavelength is ~300/F in meters and mhz, which is why the 2 meter ham band is 145 mhz more or less), and the tube looks like a capacitor for all intents and purposes (with a relatively tiny series L), which is why (partly) someone uses the moniker "charge sensitive" preamp for a circuit designed to capture total pulse energy without looking too hard at the shape of the pulse. There's not much useful info in the shape from a gas tube, that depends on things like where the reaction took place inside the tube, which could be anyplace, (and along it vs across it), and the waveform will vary accordingly in all but saturation (geiger) conditions, which aren't relevant here -- we don't run the tube volts up that high to get into that operating regime where we get constant energy pulse output. We're only interested in knowing how many ions happened in the event.

In this case, the ringing frequency of the mismatched line impedance of the tube itself would be of the order ~100mhz (5 ns per half cycle) so....doesn't matter here, the rest of the circuit will just lowpass all that out and give the correct answer anyway -- Ft on the pnp is lower than that, and the output darlington even lower I think (didn't check, but darlingtons are slow in general). With another piece of differently mismatched coax between the tube and preamp, the number changes, of course, and a long enough cable *could* get you in trouble there.

Basic theory. Known correct. Well tested.

An event creates a certain amount of charge, which we want to collect and amplify. This amounts to a current multiplied by a time, and an integrator is what is most often used for that, though you could usefully put a preamp in front of all of it and integrate later on -- and get better S/N by allowing some voltage drop to occur so that a *power* is developed with that current. The reason most put the integrator first is it's more elegant that way, and cuts the HF noise (from other things) early on in the chain. And the integrator is rarely a perfect one (else the pulse would never return to baseline among other things!) -- so it works out, even though not technically ideal -- an impedance matched preamp would be better for S/N, but luckily we don't need the best we can possibly get here. You'd couple more signal power out of the tube if you didn't load it with so low an impedance, allowing a bit of voltage drop to occur so the old currnent times volts times time = joules thing is happening and the rest has some actual power to work with to swamp its own noises.

Technically speaking, this isn't even a charge sensitive preamp. Even the miller capacity of the FET is suppressed via the low input impedance cascode pnp here! That's where the integration usually is in a CSA, or around the entire amp loop, and there is no explicit C is across the feedback 2.2 meg R here (always some actual C, of course and at 2.2 megs, the parasitic C is about right, actually). Does that actually matter? Only to a pedant, as the circuit will work fine anyway -- it's a current to voltage converter with current gain and both high and low passing, not all explicitly drawn, but there none the less.

The real time constant of the input circuit can't be known without also knowing the R in series with the HV supply for the tube, btw -- the tube is a current source when there's an event, and open circuit when not -- essentially infinite impedance at all times, just one that draws some current when a neutron hits it, for a little while.

The real impedance the input coupling cap sees is that resistor (not shown or specified) in series with the amp effective input impedance (zero by comparison). Talk about a mismatch -- we have about a meg or more at the tube end (with a tiny C across it, intrinsic to the tube and connectors), and basically zero at the amp. But this is an easy signal processing job, and that pedant stuff doesn't matter as much as you may think in this case -- remember the active tech in use when these tubes were really in use was thermionic triodes.... This is, though, why the HV power has to be quiet, as any noise on it will get in here via the series resistor to the supply, and show up as an inpu current (the supply R is a voltage to current converter for noise on the HV).

Depending on the open loop voltage gain of the amp, which I've not looked at hard, but it's high, essentially the transconductance of the fet driving 82 k ohms via the cascode pnp with elimination of the DC bias current of the fet via the clever inductor. Fets are all over the place on IDss, which is why the source R is also a good idea. We have a high pass filter or differentiator due to the inductor.

The effective value of the 2.2 meg resistor for frequencies the amp can pass (see that inductor - it's not a DC amp), yup, or pretty close! (check this in the simulator) -- it would be 2.2 meg divided by the amp open loop voltage gain which is a fairly large number (100?) so call it 20k. And even though I spit out that figure above as 7 hz -- no one caught it (heh;-), so I now know that no one knows basic opamp closed loop theory (at least who has yet posted, other than myself). Or didn't realize that it applies here. So these words need to be read and understood.

Just because you didn't draw the opamp symbol doesn't mean you didn't make one -- and this is pretty close to some current opamp (and transimpedance amp) designs in most of the ways that matter. It just lacks a non inverting input (as did many of the older designs from Philbrick). And it has several things in it doing high pass too -- all those bypass caps but mainly the inductor which probably sets the main corner high pass frequency and the rest don't really matter for skinny pulses in the few uS range.

Since we've made a virtual ground (for AC) with the 2.2 meg feedback from the output, the effective AC *input* impedance there is basically zero, or low anyway -- and hooking even a big leaker across that makes no difference (unless the leaker is noisy, like a diode being hit by X rays! Or one with shot noise or 1/F noise in the leakage), any more than it does in an inverting opamp circuit when you hook a resistor from the - input to ground at the virtual ground point at the minus input. when you do this, you lose the advantages of low noise at high impedance a FET has, but it really doesn't matter that much here. The signal is, gratefully, loud anyway.

Guys, I'm not making this up! Sorry if I sound negative here -- not intended -- but I just can't let factual error pass completely unmolested, as we're not the only people listening here -- some people are just learning electronics, and big misconceptions will lead them astray in later efforts. This is dead center of my expertise. Other things, sure, there are people who know more or have more experience. This stuff, nope.

If you like, I can get a "real" analog chip designer on my team (Joe Sousa who works for LTC in that capacity) to comment further...this is a fine design, but perhaps not as well understood here as it could be. You probably own some of his (working!) designs now and use them every day. He maintains an antique opamp forum too, at
BTW. Way not dumb on this stuff. Designs A/D's for a living for LTC, mainly.

I was simply pointing out that going for certain things (like insulation) in extremis was indicative of a misunderstanding and superstition -- neither are good for being able to repeat results like we demand in science.

Sure, no harm keeping things clean, but this is like thinking you need an atomically sharp screwdriver tip to turn a 10-32 screw. You don't! And as I pointed out, shielding may make far more difference, since it has to have 10k to one goodness to even get to 0 dB signal to noise(!), and it's not trivial to get a couple hundred times that to really ignore noises around a fusor. That's one of two reasons to put the preamp right on the tube. The other is that not only is coax not a perfect shield, but it's also a capacitor in parallel to ground, which will "eat" some of the charge signal from the tube.

Remember, these are not working in a range where the resonance of a mismatched cable of reasonable length makes *any* real difference. Off by orders magnitude unless you've got a whole spool of wire in there.

Changing the leakage currents from a ten thousandth of what matters to a millionth really doesn't make much difference. Anything you think that did, there's something else going on -- where the rosin spattered or peanut butter smeared while you were breadboarding or similar could do things like that for example. Of course, you keep things clean no matter how you do it, that's just basic prudence and hygiene, and some things like rosin can be hygroscopic and become quite conductive with time due to that and other effects, like electro-migration (electroplating by another name, only not desired here).

But I've had air wiring fail too -- amazing how a cat hair can enter into the equation....They seem to seek out things like that to bridge, collect dust and dampness, and so on downhill we go.

For a reality check on what I'm saying, take an ohmmeter and measure track to track leakage R on a clean PCB. I want to see the ohmmeter that will do that -- they exist (and I have one), but not the Radio Shack, bench-grade that only go up to 20 megs. The number is going to be giga-ohms in nearly all cases. The number is higher yet if the board has solder mask or a good coating. And a cat hair laying on a coating may not affect that -- it can actually be better than air in that regard, given the dust issues.