Pulses out of SNM 18-1 (at last)

This is bound to get mixed up with things in Electronics, check both. Physics-specific stuff here, mostly.

Re: Pulses out of SNM 18-1 (at last)

Postby Joe Jarski » Fri Aug 26, 2011 8:44 am

Doug Coulter wrote:You'd see it fine with either scope. Say the sample rate is 25 mhz. That gives you 125 samples during a 5uS pulse -- which if displayed direct, would make the pulse about half the screen wide with a lot of detail on the pulse shape.


Yup, you're right... that's what I get for thinking about this stuff at 3am - only off by a factor of 1000. :oops: I had it in my mind samples would be taken at 4us missing most of a pulse, but it'd be 4ns. The good thing is that the lower end scope would do the trick! Whew... :D
User avatar
Joe Jarski
 
Posts: 231
Joined: Thu Sep 16, 2010 8:37 pm
Location: SouthEast Michigan

Re: Pulses out of SNM 18-1 (at last)

Postby Doug Coulter » Fri Aug 26, 2011 9:30 am

I've been real happy with the low end guy. In a perfect world and if it was my only one, I might have gone for the 4 channel version though - I note they have a new version that is a little quicker (60 mhz) for about the same price I paid for the slower one. These have their own a/d per channel, so even when looking at things like relative delays between channels, they do a real nice job. They're just not quite fast enough for total troubleshooting of fast RF or some logic race conditions. But the buncha channels thing is sweet down on the fusor if you hook up a bunch of detectors and can watch them all at once, time-aligned, or in some circuits like the HV driver bridge where you need to see when this fet turns on vs when that one turns off, and maybe the drive signals too so as to see the delay on both of those.

If I'm up at 3 am, I'm usually doing the "don't stumble over the beer bottles on the way to bed" dance. I think then...but...I don't usually write it down. If it was good, I'll remember it the next day.
Posting as just me, not as the forum owner. Everything I say is "in my opinion" and YMMV -- which should go for everyone without saying.
User avatar
Doug Coulter
 
Posts: 3515
Joined: Wed Jul 14, 2010 7:05 pm
Location: Floyd county, VA, USA

Re: Pulses out of SNM 18-1 (at last)

Postby chrismb » Fri Aug 26, 2011 6:49 pm

Mine is the OWON 5022, suspiciously similar nomenclature to Doug's "X X022" it is probably all the same Chinese innards, 25MHz spec too.

I was debating buying one of the 'new generation' of budget digital scopes for a long while, and when this model came out with a FFT function (thereby doubling up as a rough-and-ready spectrum analyser - yes, you can plug an antenna straigh in and see all the main AM/FM frequency bands!) and with a price tag of £197, it was a 'no-brainer'.

I don't think mine has a peak-hold, though.
chrismb
 
Posts: 620
Joined: Thu Aug 05, 2010 6:32 pm

Re: Pulses out of SNM 18-1 (at last)

Postby chrismb » Fri Aug 26, 2011 7:24 pm

I did say I'd wait for different components. But call it an obsessive-compulsive thing; if I can spend time trying to tie up a loose end, I usually do. So another day was 'spent' fiddling.

And... success! First off I picked higher input resistors than Doug's circuit as I felt this was a problem before. I replaced with 200k and 1M off the pnp base. This suited the BC557 I was using perfectly at 5V, which is where I will be running it (through a regulator, so as to keep a 'known' voltage setup). I think the BC557 has much the same voltage characteristics as Doug's choice and even slightly higher gain, albeit worse noise/speed no doubt.

After various fiddling, I got a nice copy-cat trace giving me a 200-500mV response to a 10-20mV input off the tube. Joy! That's a signal I can now do something with!

It then dawned on me (a 'doh' moment) that Doug may've picked the 15k with a view to having the tube output come straight through the coupling cap, as per the tube spec sheet with the resistor tying it to ground. So I decided to ease off the high resistances to see if that gave a bigger response. It didn't. But what I did find later on (during various experimentation) that the circuit sometimes went a bit potty after turning it off/on. I'm no whizz with such things, but it seemed to me that the circuit was self-exciting itself on my breadboard. Not altogether surprising, but I went back to the higher resistances anyway, 200k and 1M, and the problem went away.

I then put the signal into a LM311 so that I could threshold it against a voltage set by a 220k resistor and 10k pot, to give ~0-200mV setpoints to threshold with. Then I fed the 'pull-up' [low] side of the LM311 into a HEF40106 to clean up the signal.

Initially I got a lot of 'bounce' from the noise being carried through in the signal, so I changed it over to a 'pull-down' output with a 10nF cap to low, which removed the bounce completely (I used 1k pull up and 100 pulldown, so I suppose that's around a 1us RC 'anti-bounce time', much less than the pulse period).

The results are very satisfying:
P8260703_sample_trace.gif


Here is 3 days work!:
bread_board_18-1_detector.gif


I did go back to 50Mohm to try to keep the pulse length short, but it didn't seem to make much difference. As Doug says, it looks to be a capacitance thing, and this isn't a tiny tube! But it looks like it is going to top out at around 1,000 counts/s, which appears to be the price you pay for higher sensitivity. It does seem pretty sensitive, though. Some interesting 'day-time dependency' I have noticed already - come the afternoon there are so few background neutrons about that I couldn't make consistent progress, then in the evening there are bags of them falling from the skies!

So, now all that's left to do (!!) is solder all the bits up and get used to how it works with some trial background measurements.
chrismb
 
Posts: 620
Joined: Thu Aug 05, 2010 6:32 pm

Re: Pulses out of SNM 18-1 (at last)

Postby Doug Coulter » Sat Aug 27, 2011 10:20 am

Good show! Looks like so much delay in the final output sq pulse that it could have been looking at the post pulse overshoot...but what works, works. We have this advantage over this non-living stuff -- it has to sit there and take it while you beat on it till it works. But there's usually some beating involved...

The gain of that circuit is set by the ratio of the output collector resistor to the input emitter resistor. The two transistor chain (with the bypassed output emitter) has gain ~ Hfe1 * Hfe2 -- huge.
You can certainly make a circuit with ~60 mhz bandwidth oscillate (bonus if you can get 3 or more frequencies at a time), particularly with all the parasitic 5 pf caps to ground and other signals you take on board with a kludge board. For your final build, you might want to look at the pix of mine -- that one is very carefully laid out, even on perfboard, to limit those sorts of things.

Yes, we just put the coupling cap from the tube right onto the input. I biased it the way I did (and the way I bet you'll wind up with) because the noise injected from a bench supply will definitely create tons of false counts in the presence of a fusor with its little arcs and corona. Hence the use of a battery, inside the box with the preamp, all shielded, and zero ground loop issues where it matters. Even then, there are issues when our coffee can shield is a little loose or all the screws aren't tight.

This required the rather large emitter resistor in the output transistor. The first transistor tends to bias itself to a fixed voltage, which is whatever is 1 Vbe with the bias divider. If no or a very small emitter R was used in the second stage, the DC gain becomes so large that you can't get a stable bias, especially when you're working with 6-9v variability in the supply voltage - areal wide range. There is of course a little DC NFB due to the output collector resistor tying to the input emitter resistor, but that's still too much DC gain to use over a wide supply voltage. And of course, in our case, we were trying for direct drive for the counter, and a few mv isn't enough for that, so I went for gain to get 3-4v pulses directly out of the preamp which at least can go down a cable without additional noise pickup there being an issue. The main trick there was having that much available across the output transistor, and not already all used up in the emitter and collector resistors. If you're only getting a couple hundred mv...it might not be all right just yet. Should be more like 2-3v. Could be you need more HV supply volts? You want to go up into the range where nothing changes with a change of 50v or so on the HV.

If you are looking to get minimal transistor noise, there's often a curve in any low noise transistor data sheet that shows what collector current to bias it to to minimize total noise ref the input for a given input impedance. This is trading off voltage input noise with current input noise in the transistor. I used levels similar to what we used to use in phono preamps, as the impedances are about the same here (the input R). Somewhere in the 50-100 ua region is usually about right (and then you select the input transistor collector resistor to just turn on the output transistor at the resulting voltage drop).

I didn't go for an onboard comparator as that would kill expensive 9v batteries very quick (and I didn't want to force people into even more expensive multiple Li cells -- which would have made all the bias issues trivial). The original design lets them live "forever" if you don't leave the power switch on when not in use. I'd have used a reminder led in series with power, but the basic circuit doesn't draw enough to light one. I get very tired of changing batteries in a complex setup -- you only have to have one bad to ruin a run (that thing with multiplying probabilities for a bunch of individually reliable parts, till you get to the point where somethings always broken somewhere).

With some tubes, we see a bunch of random pulses at power up, sometimes lasting 15 min or so. It's in the tube in our case, not the preamp, we see it on the scope with no preamp attached too.

We have a diurnal cosmic profile here, but not huge -- maybe 10-20% tops swing. I don't see it every time I look either. You could well be picking up some noise there -- we really did have to put ours into a very good shield to stop it. This is one reason I advocate an analog output and audio monitoring -- makes that obvious so you go fix it instead of reporting bogus numbers that lead to embarrassment later.

Peak mode in my scope is accessed through the "acquisition" menu. At the very least, yours has different software for the display - it looks different. One perk you get with the higher models is that you can take the right side menu off the screen and get more for the signal -- both ways are useful if you remember to have the right stuff up there when you take a picture (or grab a screenshot with a PC). For better or worse, the only *good* software for these is linux stuff and it would be nearly impossible to port to windows. The GW Instek stuff comes with a very flakey .net app that kind of translates the scope over to the pc screen, with some dropped frames and sometimes the settings from the PC actually work on the scope...it crashes a lot, requiring a windows reboot as it leaves the USB driver hung. I can post the linux stuff in our library if anyone wants it -it's no longer out there on the web because Tektronix complained about them using "trade secret" stuff (it works with their scopes too). Basically, it renders the bash shell scope-aware and you can write scripts there. It also hooks up the scope as a data input to "octave" which is the linux version of matlab more or less (with nice plotting and every signal processing trick there is). Once nice thing about having the pc stuff going is that the scope actually captures a lot more than shows on the screen -- a few k samples, and you can get them out via the pc interface, and see pre and post trigger data.
It can make for a quick and dirty data aq system off the pc, useful for lots of things.
gds2000tools-0.14-source.tar.gz
Scope software for linux
(241.81 KiB) Downloaded 549 times
Posting as just me, not as the forum owner. Everything I say is "in my opinion" and YMMV -- which should go for everyone without saying.
User avatar
Doug Coulter
 
Posts: 3515
Joined: Wed Jul 14, 2010 7:05 pm
Location: Floyd county, VA, USA

Re: Pulses out of SNM 18-1 (at last)

Postby chrismb » Sat Aug 27, 2011 11:02 am

I think the delay is partly my anti-bounce scheme. But what I get in return is a pulse length proportional to the charge in the signal.

I'm also thresholding the pulse quite aggressively. I'm missing out maybe 30-50% of the pulses, but as you mention, what we have in the tubes is a random direction of ionising projectiles, so it should (supposedly) all remain in counting proportion and I can afford to do that with the sensitivity of this tube. A higher threshold also shortens the pulse time a little (gets to the threshold a bit later, and back again earlier).

Also to comment, I do get a bounce after the bigger pusles, suspect that is just the current layout, but, all the same, by aggresively culling the threshold I eliminate the double-count.

I would have liked to do 'no' processing, but there are some benefits to this scheme. Once I collect the pulses I can reject them if they are too long or short. Genuine pulses are in the 60 to 120us range, and I can reject the rest as noise, which is usually 'spikes' (<<10us) or long periods of fuzz that appear in my pulses as a >~1ms long pulse. Easy to reject during collection by audio jack with the computer (also another 'additional benefit' of the longer pulses this tube makes; we're talking 'AF' here rather than 'RF'!)

Yes, I would have expected a bit more amplification here. I don't know which way you are suggesting for the resistor-calc gain, but if I am using 1M feedback then presumably it should've been more gain? May be it isn't because of something to do with the particular output from this tube. Also, I am using a 2n5551 as the npn, which has a slightly different 'on' voltage - do I need to find one with the same 'on' voltage as the pnp?
chrismb
 
Posts: 620
Joined: Thu Aug 05, 2010 6:32 pm

Re: Pulses out of SNM 18-1 (at last)

Postby Doug Coulter » Sat Aug 27, 2011 12:42 pm

The main issue with thresholding here is that everything has to be real stable, or the sensitivity varies. You might not care for short term tuning uses, which is where these things shine anyway.

You'll always have some post pulse overshoot as the coupling cap charge changes a little whilst driving the current into the preamp input, and it's gotta go someplace after that. It's just the usual AC coupling artifact. It's one reason there's a range of values that work right for that. The other is current limiting the tube during the pulse. It's expected and OK to have some overshoot time there, as the tube takes longer to recover than the circuit anyway.

For audio (other than PC input) even short pulses are hear-able. And you get additional info from the width/height as far as the ear is concerned - it gets more "bassy". I just use the signal straight for that, no processing at all. On my big-fast 3He tube (1" by 30") I can hear when it gets a double hit from a burst of neutrons, nice.

The gain is set by the negative feedback from the output, through the load R, to the emitter of the input. The circuit is an opamp with open look gain >>50k, so that ratio sets the actual gain. The bias network has no effect if the emitter of the NPN is bypassed as shown -- there is no AC signal on it at that point (the .1 uf being huge in the context of the speed of a pulse). There is, however, some DC feedback there, to help it stay biased in range as the supply voltage changes over 50% or so with a battery. This keeps the bias resistors from having any effect on the AC gain as a nice by product.

(I'm going to go find the schematic so I can see what I put in for the pnp emitter resistor, be back soon) OK, it's 10k/2k, which in the standard non-inverting opamp formula gives gain of 51 (with a hell of a lot of loop gain to make it frequency-flat). You can check that with an audio input and just scope both input and output. Should be dead on at 51 and not vary with supply voltage at all. I used the rather large 10k value as the only bad effect of the high impedance is to slow down the pulse a little bit when driving capacitive loads like coax, and then only on the transistor turn-off side of the pulse. During the onset of the pulse, the transistor drives as hard as need be (that negative feedback loop) so this impedance is different on the rising and falling edges -- this doesn't matter in this case, as the onset of the pulse from the tube is fast, and so is the preamp. During the back side, the pulse decays slowly, and so does the preamp. If you want to use tweaky words, it's an ideal "matched filter" for the actual signal. In this case, it lets the design draw lots less power, which is what I was after. http://en.wikipedia.org/wiki/Weiner_filter

Theres no need (and no desire) to match the Vbe's of the transistors, just bias the pair up right it all that's needed. Here, I specced a fast, low noise, low current "audio" transistor for the front end, and a regular old switching NPN for the output, because noise doesn't matter there, but output drive peak current does -- again, even a tiny cap, like a couple feet of coax, draws infinite current with 1/infinite risetime. I did it pnp input, to match the signal characteristics. The pulse turns that one on (well, both of them), which is quicker in transistors -- on time is faster than off time, which also matches the signal here. While turning on, full current from the tube is hitting that base, during the off time, only the bias R is turning it off, so the speed is asymmetric, again, just like the signal. Gives a little extra noise rejection without actually slowing anything down. The fact that the pnp collector is driving an AC short circuit makes it faster, as it doesn't have to charge any capacitor (internal to the xistor, or external) during the pulse. The input to the second stage looks like a short circuit to ground for AC.

As you can see, there can be a lot more to even a simple circuit (or would that be especially a simple circuit) than meets the eye at first. This was all pretty well thought out.
I've been using variations on this circuit since about 1972, when I stole it from a Marantz tone control buffer... It's pretty good for many things in the mid low-level domain, as changing the various values can set it up for fast or slow, low noise for a given impedance and so on. If this guy doesn't do it, you can go to a more "heroic effort" but usually you don't need that.

Edit: Added a copy of the preamp schiz so you don't have to flip around to see what we are talking about here:
NewDPre.jpg
edit: modified diagram with battery symbol reversed
NewDPre.jpg (61.32 KiB) Viewed 10292 times

The protection diodes should be 1n4148 FWIW, unless you have low leakage versions of the germaniums shown.
Posting as just me, not as the forum owner. Everything I say is "in my opinion" and YMMV -- which should go for everyone without saying.
User avatar
Doug Coulter
 
Posts: 3515
Joined: Wed Jul 14, 2010 7:05 pm
Location: Floyd county, VA, USA

Re: Pulses out of SNM 18-1 (at last)

Postby chrismb » Sat Aug 27, 2011 3:26 pm

Yeah, that diagram confused me at first, before I went 'doh' after wiring the neg to 'earth'. The battery symbol should be upside down - it's the earth being pushed up so should be attached to the 'long' line.
chrismb
 
Posts: 620
Joined: Thu Aug 05, 2010 6:32 pm

Re: Pulses out of SNM 18-1 (at last)

Postby chrismb » Sat Aug 27, 2011 3:52 pm

I had a poke around on my 'scope and, cool!, it has peak detect also! Well, it should have, seeing as it is only a digital function. I think it is absolutely a great bit of kit for £200. Impossible to do this stuff without it (and, indeed, it was impossible for me to do it when I first got these tubes... 'cos I didn't have it!!)

OK, so the longest scan is 200 s per division, but you'd not see all the background signals if I used that timebase, so here is one at 25 s per division (i.e. ~5 mins total horizontal) on peak detect:
P8270716_5_min_scan.gif


I set the threshold to ignore anything below 0.2uA ( which is 20mA looking at this plot as I have a 100k to ground). I'm not really missing many potential counts, whilst not getting any false ones. It even correctly missed out on a few 'noise' spikes (such as where it matches the spikes above the top peak) - I have an additional 150pF in series and stuff much slower than a few 100us doesn't get through. (The only one that did register wrongly was when I switched off the light which was next to the bread board, at ~220 s, just before I took the shot of the 'scope screen in the dark!)

I also connected a 220k resistor and a small cap back into the feedline off of that crate I posted on last week (which is providing the 1900V I am running the tube at). This gives a 'kick' back down the line for the electronics to register, which it does perfectly without any multiple counts (providing there's no noise, of course). So, potentially I might make up a box to connect this with the crate, then it'd be a fully sealed box with just the screened power coax running into it from the crate. And I think I can also run this just off an internal battery for a long while - the power supply wasn't registering any current at all to run it, when it wasn't counting, so I ran it powered up through a 10 ohm and got 20mV drop. I think if I assume 5mA, conservative, then that's probably about the current draw, which'd mean a regular 9V battery should power that for a few hundred hours.
chrismb
 
Posts: 620
Joined: Thu Aug 05, 2010 6:32 pm

Re: Pulses out of SNM 18-1 (at last)

Postby Doug Coulter » Sat Aug 27, 2011 4:58 pm

Looking good! I looked up a standard 9v alkaline on digikey and the data sheet shows 100 hours (to 5.4v) at 5ma -- in other words, 5 ma is a real battery-eater. 4 days for a few bucks. Now the original 150ua I designed that to draw doesn't look so dumb. Those specs are always quite optimistic in my experience. There's a lot of air and packaging inside a 9v battery, not so much is actual battery.

Yeah, those scopes are really a good thing to have, eh? Does yours have provision to connect to a PC or put things on an SD card? I find that really useful here, and you get much better pix.

What you might do is build the thing inside the existing crate, and just separate the HV from the signal input inside it to get it going right and all tight and shielded. With a quiet regulated supply, you can run this off either polarity -- just hook the appropriate power rail to ground and the other to your supply - it's ac coupled in and out after all. The way it's shown is slightly better for a non-perfect supply (running off negative supply).
Posting as just me, not as the forum owner. Everything I say is "in my opinion" and YMMV -- which should go for everyone without saying.
User avatar
Doug Coulter
 
Posts: 3515
Joined: Wed Jul 14, 2010 7:05 pm
Location: Floyd county, VA, USA

PreviousNext

Return to Metrology

Who is online

Users browsing this forum: No registered users and 6 guests

cron