Needing more than a spark test?

[homebrewed]

Second the strangeness, on this side of the pond. I live on the outskirts of Portland, Oregon, and have seen lots of crazy stuff going on downtown. The Portland cop who lives next door says what we've seen on TV is about 1/10 of what it really was like .

The issues w/regard to transistor selection may be overblown. It is all to easy to dive into the weeds to get some hypothetical optimal solution -- I've gone there and had to haul myself back when I figured out I was worrying about absurdly small improvements. As you say, at some point we just need to get 'er done and see how it works out. Who should ever believe that their first attempt at something will be the last? It's all a learning experience when doing this kind of thing

On a slightly different subject, the Americium capsules I ordered from China (via ebay) arrived today. Curiously enough, the package had been opened (probably by US Customs). Things have been too busy to fire up my detector circuit yet, hopefully that will happen soon.

Mark (A.K.A. Homebrewed)

 

[homebrewed]

I've been thinking more about the problem of accurately determining the peak voltage of pulses coming out of the Pocket Geiger circuit. I'm planning on extracting the peak value from the digitized data (rather than building a peak-hold circuit) so this is relevant.

I decided to analyze a (simulated) pulse coming out of the circuit, by looking at the first through fourth differences (a simple form of numerical differentiation). As expected, the first difference switched from positive to negative near where the peak voltage occurs. The second difference was almost constant for values around the peak, suggesting that a second-order polynomial fit might be good enough. So I found an online polynomial regression calculator and entered some data. It looks pretty darn good:


If you look at the residual error, it is about 4.5E-13. Not bad, not bad at all. But this is for noise-free data so it's a best-case result.

There are many code examples on the web that can be used for a starting point if you want to write your own code; or, of course, numpy/scipy will do it for you with relative ease. Using more than 4 data points will improve the performance in the presence of noise.

The reason I'm looking at this approach is that the PC-hosted MCA programs I've found all expect the data to come from a sound card outputting data at one of the standard audio sampling rates (which are slower than what I want). I want lots of samples around the peak value so I can reduce the impact of noise as much as possible. My Teensy 4 processor chip running at 600MHz has more than enough processing power to do the polynomial regression (it has a built-in FPU) so I likely will roll my own second-order polynomial fitting code and just send peak values to the host PC.

 

[graham-xrf]

According to the folks at Angel Guilding, the thickness of their chemically-deposited silver is around 50 nm. I assume this is somewhat dependent on deposition time, but if it's like electroless nickle, once the surface is coated the dep rate goes down by a large factor.

The silver layer would be pretty delicate. A thin coating of clear-coat (no pigments please) wouldn't absorb x-rays to any significant degree. And it might help the exterior stay a nice shiny color, too!

Re: the electroless nickel reaction. I was startled at how it initially just plates withing seconds, but very thin, the a slowdown, but only for about 15 seconds, and then it really gets going!

The reaction is "self-catalyzing". Palladium, platinum and various other metals will catalyze the nickel deposit - including nickel. I don't fully understand how nickel can be it's own catalyst, nor exactly what happens with the reaction, but no question, it gets going with some fury if hot enough. Seeing hydrogen come streaming off, (without any electrolysis power supply connected) is completely convincing!

[BTW - Do NOT let your significant other read the label on the nickel sulfate reagent bottle]!

Silver is used as a "strike" metal because it can be reduced from silver nitrate to plate into the microstructure of almost anything, like glass or plastics. If you leach out the butyl in ABS by wiping with solvent, or wipe over the vinyl with acetone, (dulling it's surface permanently), then deposit the silver, you can then plate over that a sufficient thickness of copper, which will be mechanically bonded into the spongy surface microstructure. Buff that up, and follow with nickel, and then chrome, whatever if you like. The thin layer, if protected with paint, is good for mirrors, though not for first surface mirrors. Optics are usually "silvered" with aluminum, deposited in vacuum, and covers over with deposited other substances (zirconium)?

Electroless nickel for plating on iron and steel has the advantage that it deposits an even thickness everywhere, not depending on the siting of electrodes, nor current density. Initially, a "Woods Nickel Strike", started with the help of a current, for a minute or so, grabs on in the same way. (1:1 nickel chloride and hydrochloric acid). Then copper, if wanted, laid on thick enough to allow the pits and imperfections to be buffed away. Then back to the nickel.

I have the various recipes for plating, some dating back a century or more. I could post them, but I thought that if I could find them, so could anybody.

I have been well distracted by a peculiar mix of lathe fix-ups, house fixups, XRF PMT circuits design, Python software that is out of my depth, and now, adventures with electroplating nickel. So far, it is the nickel that has had the most success. It could be because the little cellar, where the chemistry is done, is also a haven of cool in a humid UK heatwave!

 

[graham-xrf]

I've been thinking more about the problem of accurately determining the peak voltage of pulses coming out of the Pocket Geiger circuit. I'm planning on extracting the peak value from the digitized data (rather than building a peak-hold circuit) so this is relevant.

I decided to analyze a (simulated) pulse coming out of the circuit, by looking at the first through fourth differences (a simple form of numerical differentiation). As expected, the first difference switched from positive to negative near where the peak voltage occurs. The second difference was almost constant for values around the peak, suggesting that a second-order polynomial fit might be good enough. So I found an online polynomial regression calculator and entered some data. It looks pretty darn good:
View attachment 333753

If you look at the residual error, it is about 4.5E-13. Not bad, not bad at all. But this is for noise-free data so it's a best-case result.

There are many code examples on the web that can be used for a starting point if you want to write your own code; or, of course, numpy/scipy will do it for you with relative ease. Using more than 4 data points will improve the performance in the presence of noise.

The reason I'm looking at this approach is that the PC-hosted MCA programs I've found all expect the data to come from a sound card outputting data at one of the standard audio sampling rates (which are slower than what I want). I want lots of samples around the peak value so I can reduce the impact of noise as much as possible. My Teensy 4 processor chip running at 600MHz has more than enough processing power to do the polynomial regression (it has a built-in FPU) so I likely will roll my own second-order polynomial fitting code and just send peak values to the host PC.

In the end, however good is the sample quality, and however great is the resolution of the A/D converter, however careful one is to prevent the common mode current from the roll-over of digital counters delivering into the analogue part common connection point, the number you get depends on an analogue thing in the front.

There is something sweetly efficient about using an analogue circuit to capture and preserve a buffered peak value, noise free as best any circuit could have it, to be sampled at relative leisure. By that I mean, sample rates from Hertz to some MHz, if you like. The whole business of getting the number, with fewer of the least significant bits jangling about, only to be steadied by software averaging, Kalman filtering, and the rest, ending up delivering at a the same rate you could have had with digital filtering, but without all the software struggle.

A analog contrivance, including filters, integrators, peak-hold, etc. delivers at the speed of electrons, and the "program" is inherent. Even now, the very fastest digital processing, can never deliver faster than the analogues that were up front anyway. I don't say we need it fast. I am saying that even modest analogues, put in the right spot, can hugely ease the digital post sampling processing load that follows.

 

[homebrewed]

Re: the electroless nickel reaction. I was startled at how it initially just plates withing seconds, but very thin, the a slowdown, but only for about 15 seconds, and then it really gets going!

The reaction is "self-catalyzing". Palladium, platinum and various other metals will catalyze the nickel deposit - including nickel. I don't fully understand how nickel can be it's own catalyst, nor exactly what happens with the reaction, but no question, it gets going with some fury if hot enough. Seeing hydrogen come streaming off, (without any electrolysis power supply connected) is completely convincing!

[BTW - Do NOT let your significant other read the label on the nickel sulfate reagent bottle]!

Silver is used as a "strike" metal because it can be reduced from silver nitrate to plate into the microstructure of almost anything, like glass or plastics. If you leach out the butyl in ABS by wiping with solvent, or wipe over the vinyl with acetone, (dulling it's surface permanently), then deposit the silver, you can then plate over that a sufficient thickness of copper, which will be mechanically bonded into the spongy surface microstructure. Buff that up, and follow with nickel, and then chrome, whatever if you like. The thin layer, if protected with paint, is good for mirrors, though not for first surface mirrors. Optics are usually "silvered" with aluminum, deposited in vacuum, and covers over with deposited other substances (zirconium)?

Electroless nickel for plating on iron and steel has the advantage that it deposits an even thickness everywhere, not depending on the siting of electrodes, nor current density. Initially, a "Woods Nickel Strike", started with the help of a current, for a minute or so, grabs on in the same way. (1:1 nickel chloride and hydrochloric acid). Then copper, if wanted, laid on thick enough to allow the pits and imperfections to be buffed away. Then back to the nickel.

I have the various recipes for plating, some dating back a century or more. I could post them, but I thought that if I could find them, so could anybody.

I have been well distracted by a peculiar mix of lathe fix-ups, house fixups, XRF PMT circuits design, Python software that is out of my depth, and now, adventures with electroplating nickel. So far, it is the nickel that has had the most success. It could be because the little cellar, where the chemistry is done, is also a haven of cool in a humid UK heatwave!

I've used electroplated and electroless nickel for various applications at work. I mixed up the electroplating solution pretty much as you describe, but bought the electroless nickel solutions premixed -- one solution contained the nickel salt and other magic chemicals to do things like get the pH right, and the other contained the reducing agent, either phosphorous or boron-based (it probably was sodium hypophosphite). The reducing agent didn't last all that long once the bottle was opened because the oxygen in the air destroys it. I was able to extend its lifetime by backfilling the bottle with nitrogen, which was readily available to us -- the whole building was rigged as a fab and our lab occupied a small part of it. Thankfully, I didn't have to wear a bunny suit, though!

IIRC, electroless nickel is used to tack down diamond grit as a first step to making "diamond plated" grinding and cutting tools. Not to tempt you down yet another rat hole .

 

[homebrewed]

In the end, however good is the sample quality, and however great is the resolution of the A/D converter, however careful one is to prevent the common mode current from the roll-over of digital counters delivering into the analogue part common connection point, the number you get depends on an analogue thing in the front.

There is something sweetly efficient about using an analogue circuit to capture and preserve a buffered peak value, noise free as best any circuit could have it, to be sampled at relative leisure. By that I mean, sample rates from Hertz to some MHz, if you like. The whole business of getting the number, with fewer of the least significant bits jangling about, only to be steadied by software averaging, Kalman filtering, and the rest, ending up delivering at a the same rate you could have had with digital filtering, but without all the software struggle.

A analog contrivance, including filters, integrators, peak-hold, etc. delivers at the speed of electrons, and the "program" is inherent. Even now, the very fastest digital processing, can never deliver faster than the analogues that were up front anyway. I don't say we need it fast. I am saying that even modest analogues, put in the right spot, can hugely ease the digital post sampling processing load that follows.

I understand what you're saying, and usually take more of an analog route when the path is well defined. In this case I'm trying to keep the hardware to an absolute minimum and move the smarts into the digital realm ASAP....because I don't know precisely what will be needed.

I've built more than my share of rats-nest circuits, adding this block and then that block to address various problems; and past a certain point I end up spending more time troubleshooting my circuit tangle than solving the original problem. And if I quit and come back a few months later....well, it's just a tangle of parts and wires looking back at me. With well-commented code I can usually figure out what I was up to.

 

[graham-xrf]

I've used electroplated and electroless nickel for various applications at work. I mixed up the electroplating solution pretty much as you describe, but bought the electroless nickel solutions premixed -- one solution contained the nickel salt and other magic chemicals to do things like get the pH right, and the other contained the reducing agent, either phosphorous or boron-based (it probably was sodium hypophosphite). The reducing agent didn't last all that long once the bottle was opened because the oxygen in the air destroys it. I was able to extend its lifetime by backfilling the bottle with nitrogen, which was readily available to us -- the whole building was rigged as a fab and our lab occupied a small part of it. Thankfully, I didn't have to wear a bunny suit, though!

IIRC, electroless nickel is used to tack down diamond grit as a first step to making "diamond plated" grinding and cutting tools. Not to tempt you down yet another rat hole .

Nickel with a small percentage of phosphorous, like a "alloy", if you can call it that, is a standard, desired smooth hard bright type. It is something to do with the crystalline matrix. Steel is like that too. It has a matrix of all sorts of non-metallic additives. The mix of nickel with phosphorous can be made directly when plating, instead of it being a furnace operation.

Funny you should mention the locked-in diamonds. Only days ago, I finally purchased another double-sided credit-card sized diamond metal hone. Two grades of diamond grit 320 and 600 I think. It replaces one that I have had for 17 years, and I won't throw it yet, because it still has life in it. So far, the most use has been in rubbing up tungsten carbide stuff like router tools.

 

[homebrewed]

Nickel with a small percentage of phosphorous, like a "alloy", if you can call it that, is a standard, desired smooth hard bright type. It is something to do with the crystalline matrix. Steel is like that too. It has a matrix of all sorts of non-metallic additives. The mix of nickel with phosphorous can be made directly when plating, instead of it being a furnace operation.

Funny you should mention the locked-in diamonds. Only days ago, I finally purchased another double-sided credit-card sized diamond metal hone. Two grades of diamond grit 320 and 600 I think. It replaces one that I have had for 17 years, and I won't throw it yet, because it still has life in it. So far, the most use has been in rubbing up tungsten carbide stuff like router tools.

The son of one of my co-workers had a project where he electroplated a special ferromagnetic alloy composed of iron and nickel. I helped him iron out a few equipment related issues to improve film uniformity. It's an interesting approach for putting down thin alloy layers, but has a lot of variables you need to control in order to get a specific composition. Bath temperature, iron/nickel salt concentration and ratio, pH, complexing agent type and concentration, cell voltage, agitation..... lots of knobs to turn. It was a good thing that his dad knew how to design experiments!

If you need an even finer-grit diamond hone, take a look at lapidary diamond lapping disks. I have 300, 600, 1200 grit and 3,000 grit ones for shaping and honing carbide scrapers. Not too expensive if you shop around.

 

[homebrewed]

For a first crack at using the pocket geiger as the detector for an MCA, I was going to use the Teensy4.0's 12-bit A/D and just tap into the output of the filter, an analog signal. But my circuit simulations indicate that the filter's DC level is about 4.5V, far above the Teensy's 3.3V supply. A voltage divider can drop the voltage to below this, but at the expense of reducing the already-small pulse height. So I have decided to make a simple signal-conditioning circuit using some low-noise opamps. One for adjusting the signal offset, and another to boost the signal level to take better advantage of the A/D resolution.

It's a simple enough circuit to hand-wire, and I already have the active components I need (and most of the passives, too). Time to get out the soldering iron....

 

[graham-xrf]

Just a note: I am still here, and wanting to press on with some construction. For the present, a whole lot of other (good) stuff is getting in the way. I want to also get the same diode detector as @homebrewed . Mark may be further along that route by now.

While I am going to make the PMT tube work, if only because I already have one, (and it's fun ), in the longer term, the attractions of small, lightweight, low voltage, and possibly not needing a scintillator count big-time with me.