Needing more than a spark test?

[graham-xrf]

Thanks for the update. I watch with interest. Here, I have soldered down the AD7622 A/D converter on 0.5mm pitch LFQP. I managed OK, but this is not something I would expect many others to do. Easily the most awkward was getting the package into position under assembly microscope, for which mine has too much magnification to see the whole package. I used the sharp point on re-shaped paper clip, with base end held down under a handy nearby Vee-block. The ease with which even a slight touch to a pin can bend it sideways I knew about, and tried to avoid. It's now down, and I will be adding the rest of the components soon.

Re: The AD8646. I get it that you chose it because of noise figure. The only thing I see there is the 24MHz Gain-Bandwidth product, and 74°phase margin. To get the Signal/Noise ratio locked in from the first stage gain will require about 30dB. Getting as much gain in at the first stage as possible to lock in the noise performance is traded against the ability to reproduce and measure the area under a pulse that has 100kHz to 400kHz components, and needs a slew rate and settling time not to have overshoot when returning to zero.

There is, of course, a wide disparity in impedance for power ratios, but loosely, trying for a first stage gain of 50,000 we would have a bandwidth of only 480Hz. That is why I used the LTC6269, regardless the size for the first stage. Are we sure we cannot find a bigger package op-amp with a GBW product (say) about, or more than 200MHz?
- - - - - - - - - -
At home, things are damn difficult. I happen to be in the South-East of England, and in a unprecedented total legal lock-down. Where we are tonight is 20,000 new infections, and 1000+ deaths/day from the new highly contagious strain. For the 60M size of our population, its about as bad or worse than most other countries. About 1 in 50 are infectious, and in some situations, more than that. Pretty much unless you have been in shielding quarantine with someone else for days, you have to act as if anyone you encounter has already got it.

I had thought to settle in playing with the XRF circuit board, which is code-speak for finding all, and somewhat sorting my stash of SMD components, but fate intervened. A water leak at a dishwasher inlet had been doing it's damage for some days unknown before we discovered it. I had to move out all the kitchen white goods, take up the flooring, and found the substrate floorboards soaked through. Once I found and fixed the leak (replaced fitting), I sort of put together the kitchen again in camping mode without top flooring. My wife has bought a de-humidifier to help dry out the place, but it's a stop-gap. Replacing the floor and tiling it to fit up effectively a new kitchen is now planned.

 

[homebrewed]

Sorry to hear about your situation regarding Covid. Hopefully the vaccine will begin to contain its spread. Also sorry to hear about your plumbing problem, those can cause a lot of damage. We've had a couple of water related problems at our place too, and, like you, one of them was caused by an appliance (a front-load washing machine). Fortunately I caught on to that one pretty quickly so we didn't get any damage to the floor or subfloor. The floor in the utility room is slate, a decision we've been thanking ourselves for.

Regarding my use of the AD8646, I did worry about its lower speed and how that might affect the pulse response. Simulations looked OK (and converged quickly, so it didn't appear that there was a stability problem), so I went ahead and ordered a few. I made the mistake of replacing the 9V LDO with the 5V version before I discovered the problem with the LTC6269, so I also was looking for an amplifier that can run with a 5V supply. That limited my choices right there.

Regarding an adapter board for the LTC6269, it looked like it was going to be difficult to reliably solder it down -- it can't be any larger than the 8 SOIC footprint. And I wasn't particularly excited about the idea of soldering 8 teeny wires to the IC footprints on the board. Hence the search for something with decent noise specs that comes in the 8 SOIC package.

 

[graham-xrf]

For me, the adapter board(s) are purely a temporary hack to get a prototype lash-up working. My apologies for the crappy out of focus picture, but this is the AD7622 ADC mounted on one of those "universal" quad flatpack mounting boards. Maybe it's a good thing the focus is not great, because the soldering, while absolutely fine as joints, does not have the re-flow uniformity. I did it with a soldering iron.



I think one can get similar things for op-amps. I have in the back of my mind that a purpose-designed PCB will be needed anyway, but for now, I have a way. The whole subject is having me re-think, and perhaps try again, looking for suitable large-package parts that electronics experimenters can get their fingers around, and maybe even folk who have never done it before can attempt (with enough instructions and pictures). We have come quite far simplifying the thing (think back to PMTs).

 

[homebrewed]

Update:

I have started working on an enclosure for my XRF setup. The main reason is that I need to remove most of the copper shielding foil in order to replace some of the ceramic capacitors with better film capacitors -- so I will need an enclosure to eliminate EMI. One of the capacitors is the 1uF cap in the final bias voltage RC filter so the capacitor has to have >25V operating voltage. This makes for a rather large capacitor, but what the heck.

The other cap I want to replace is the .1uf interstage coupling capacitor in order to extend the low frequency response. Pulses coming out of the detector are pretty slow anyway so I don't want to reduce pulse amplitude any more than necessary. That one will be 1uF, too.

The enclosure will be a box with a partition in it, with a hole drilled/bored in it for the XRF photons. One side of the box will contain the detector and signal conditioning board, held in place by vertical rails; and the other side will contain the Am241 sources and sample. That end of the box will be removable, probably a lead sheet captured by two grooves on either side. This way the entire source/detector volume is shielded so no super-delicate aluminum guilding foil is needed to shield the detector from ambient electrical noise. I have enough material for the base but need to order more for the sides and top of the box. I will use .25" thick plate, extreme overkill for shielding purposes; but that will make it easier to drill/tap holes to assemble everything. Hey, the skin depth of 60Hz in aluminum isn't that much so it will cut down on magnetic coupling, too.....

I will line the box with lead sheet to keep those 60Kev x-rays in their place. I think I have enough. .25" of aluminum will also absorb a substantial portion of the x-rays.

Also, one small update regarding my SPICE model for the detector. I had noticed that my simulations predicted a much shorter pulse width than what I was really seeing. Additional investigation revealed that I had misread the schematic, placing a 1pF capacitor in the second amp's feedback loop -- but it really is 10pF. Only an order of magnitude error . Simulations look more realistic now.

The section on mechanicals make this more appropriate for a machining forum. Wasn't it about time?

 

[rwm]

Definitely do NOT understand all the technical details of the circuit design. Once it is complete, I hope you will 'splain the circuit design in a broader respect so that those of us without training can understand some of it?
Still watching.
Robert

 

[graham-xrf]

I also have some "mechanicals in the brew.
I had thought about the upset of unwanted AC fields getting into the (very) high gain low noise amplifier, but it seems that if one plonks the open end against the metal to be tested, that sufficiently stops it. The "optics" circular array of sources, slightly recessed into their lead drilled pockets, all set at an angle pointing inwards (like a ER32 collet end angle) is such that no emission from the lead can "see" around corners so as to reach the sensor.
Samples smaller than the circular aperture can be tested by placing them on some aluminum cooking foil, and placing the sensor head over it. There is some (low) count from aluminum 1.48keV and 1.55keV that is barely possible, but if the sensor can see it at all, we still know where it came from. If the idea fails, I will adopt Mark's split box.

The conical lead shield shape is there ostensibly to stop the Am241 gamma from getting out, but it occurs to me that one could turn the thing out of aluminum, and settle for some thin discs of lead cut or punched out of 2mm lead sheet, and a piece of sheet "rolled up" to butt on the split, pushed up the aluminum outer tube.

I have got 3 parts to this project part not quite done. I have made a daughter board fitted to the Raspberry Pi, carrying the A/D converter, and a display scheme that plots onto a web page on the Pi. You can see it from a PC, or a direct connected monitor, or a smartphone browser, or any screen that can be on it's network. In theory, if the Raspberry Pi is also running it's server connected to an ISP domain (like mine), you can see it from anywhere, but the limited usefulness of such a stunt is obvious!

The next part is the low noise amplifier electronics, with low impedance buffer. This part is right on the back of the sensor, and drives a screened cable pair to computer box.

The last part is the tube sensor head mechanical assembly. While I am trying to keep this kit as simple as possible, I also want it to be a serious performer. HM members wanting a kit would have the plans for the tube, lead bits, etc. and would need to purchase a Raspberry Pi. Then, plug a pre-assembled PCB kit of A/D converter daughter board. For getting the tiny electronics soldered, it is not a thing one would expect most users to do. We are open to suggestions as to a small casing and power supply for the Pi. Mine is running off a USB charger. Voltages used for the low noise parts are conditioned by in-circuit regulators and filter components.

The little amplifier electronics PCB is similarly a preassembled part of the kit, connected to the cut off end of a Geiger 5 board containing the diode.
- - - - - - - - - - - - -
Right now, I am coming up to my "big day" when I am to be shot full of AstraZenica's best, (stage 1). I am also madly making (wood) chips, and the occasional smell of table saw and scorched router end wafts about. The sound of hammer drill Makita doing 8mm fixings into concrete as well. My fire door, with it's exact specification frame clearances and intumescent strips is all happening. I can pick up and move a normal door, but a fire door is a very different story! Oof - but they are really heavy!

I will post some more visual stuff as soon as I can.

 

[homebrewed]

You're making good progress if you already have your ADC going. Nice!

 

[homebrewed]

I got the aluminum in for making my enclosure, and have started cutting it down into the rough sizes I need. As part of the enclosure, I need to drill/bore an aperture in one end for the x-rays to enter.

One thing I (belatedly) thought about was the fact that I need to place the Am241 disks around the aperture, to "illuminate" the sample with 60Kev gammas. The detector is 10x10mm so it needs a ~.56" diameter aperture to avoid any vignetting, which would reduce the count rate. My preliminary enclosure design uses 2.5" wide end plates so that's not a problem. However, can I fit up to 8 disks around that size of aperture?

To answer that, I did some internet searching and found a formula that looked close to what I need -- but it was cast in terms of the distance from the center of the hole to the centers of the surrounding disks (so the disks would intrude into the aperture hole). Changing the formula by adding the disk radius to the desired aperture hole radius, I came up with a formula that I could use to test my enclosure design. I measured the radius of the little Am241 disks to be about .238". From there, I used the formula to determine where they needed to be placed in order to fit up to 8 around the aperture, and that came to R=.384". This is the distance from the center of the aperture to the nearest edge of the disk. This is .768", larger than .56" -- so I can fit 8 around the aperture without any of them extending into the aperture hole. Also, the aperture center-to-disk-outer-radius comes to a little less than .86", so the disks will fit quite comfortably on the 2.5" wide end plate.

So to sum up: The aperture hole is .560" in diameter. 8 of my Am241 disks should fit around the aperture without intruding into the aperture hole.

However, I also am going to draw this up to verify the result. I want to find the same result using two different paths to get there. If there's a significant discrepancy, it's time to reexamine my assumptions!

 

[homebrewed]

I wrote a simple openSCAD program to test the emitter/aperture geometry and it looks like the numbers I got work OK -- see below:


If anyone wants the openSCAD code, just let me know. I wrote it so it is easy to play with the geometry.

 

[graham-xrf]

The Single-Ended Illumination Optics Concept
Using a aluminium ring just tall enough to block sideways radiation, and exploiting the fact the radiation can't go around corners.
Here we get an idea of some bits users will need to machine up out of aluminium (OK - spell check - aluminum).

The optics principle
Unlike light rays, the 60KeV coming out of Am241 sprays anywhere, all directions at once, and only some of it is useful. What happens here is all 60KeV hits only aluminum, or lead-backed aluminum. The 55° tilt angle sets the height above the sample, and the cone is blocked by the iris. Only XRF from aluminum, or the test sample can be seen, and the responses from aluminum are too low to be significant.



This scheme is a radiation source carrier, and a sensor/amplifier mounting all in one.
It requires a aluminium machined shape like this..



The Am241 radiation sources (from smoke detectors) are interference fit pressed into the little pockets. They do not have to be pushed in deep. They can have a small lead disc inserted first. Any X-Rays from their steel carrier, and indeed anything the 60kV photons hit, other than the test sample, cannot get back to the photodiode.


Here is a part structure illustration, showing the sensor.

There are all sorts of ways to mount the amplifier PCB. In my circuit, the little white PCB from the PocketGeiger 5 is not useful beyond being the photodiode carrier, so the idea is to leave it soldered to the PCB, but cut away the part not under the photodiode, and mount it at right angles to the low noise high gain electronics PCB. It's a bit home-made, but the hot melt glue was a cheap and quick solution.



There are more tubes needed. I was looking to find sizes that could readily be cut from standard aluminium pipe, but failing.
Also - more 2mm aluminium, that being the 100mm square piece the tube is put up against. Aluminium will only glow at two energies. One is at 1.487KeV and the other is at 1.557KeV. Both of these would be at only 1% to 2% efficiency, but if their count is seen, it will be a known artifact, and counts at those energies could be subtracted from the plots by software anyway.


The thought occurs that the entire plot without sample present could be counted, then subtracted from the plot when the real material is present. in the same way that dark current noise is removed from CCD images.



The rear tube "handle" cover cab be reduced in diameter - perhaps carved from a piece of aluminium scaffold pole, or trade display stand, or sartorial lounge lighting stand.

Lead bits
A disc of lead under each Am241 source, to stop gammas making it back into the electronics. To stop cosmic incoming, perhaps a lead ring layer around the optical shield, to stop direct sideways gamma making it to the sensor. A lead bucket lining in the machined pocket would do the same thing. Those determined to stop even incoming cosmic stuff can wrap a thin lead sheet around the Outer Stand-off Tube

Magnetics
Not shown is the Mu-Metal shield box around the amplifier PCB. We keep this small, so it is affordable

In use, it is stood over the 100mm square aluminum plate, which would have the test sample on it. Alternatively, on bigger stuff, the probe is simply put up against it.

Dev items
The whole computing end can be done with the $35 Raspberry Pi Model 3B. It has 1GB of RAM, and is a 1.4GHz 64-Bit little thing, used here because the Pi4J Java library works with it. This one is apparently a current best seller!

Pi4J software library is still being updated for Raspberry Pi4 models. The development hack-up of matrix board was to get the ADC connected. On that Raspberry Pi, under piece of buff perf board, is a little fan and heatsink. Final add-on GPIO daughter board with the sampling electronics would not have all that stuff under it.

Keep in mind that the display is any phone or PC that has a web page browser - because this one plots it's display in real time to a web page in it's own memory space. If you can network to it, you can see it, and interact with it. This part, at least, already works. It can also run a web server (nginX) if need be. It has WiFi and BlueTooth built-in, and has a EtherNet connector. It also has a monitor display and USB for mouse and keyboard, if you want direct dedicated operation.



The project kit box has accumulated quite a little pile of stuff - all development aids..



I have left out quite a lot. There may well be good suggestions. Please do!