Needing more than a spark test?

[homebrewed]

I've been working on improving my XRF results, based on some simulations I did to see how much the output pulse shape changes in the presence of pulse overlap -- in terms of the front-end bandwidth. The result showed that I needed to reduce the time constant of the final RC low-pass filter, which was a 1Kohm resistor and a 68nF capacitor. I reduced the capacitor value to 10nF, and accordingly changed the expected risetime/falltime values, which are used as part of the pulse qualification process. One beneficial side-effect of this was that the pulse amplitude increased.

The result is the following spectrum (background subtracted), using my cadmium elemental sample.


I also am running my ADC at its max rate, 1MSPS. The Theremino approach uses pretty heavy-handed filtering due to the fact that they are using an audio ADC with a much lower sampling rate, so the ability to use a much higher sampling rate can be helpful when it comes to detecting pile-up. The resultant higher count rate will also cut down on the time needed to analyze a sample.

This result shows that there are some strong relationships between the system bandwidth and sampling rate (given my current approach toward pulse-qualification). So there still could be some improvements to be had.

 

[homebrewed]

FYI, I'm using the same americium-241 setup that I made for the X100, so the count rate likely is quite a bit lower than it would be with a larger aperture in the lead shielding plates. The apertures I'm using are a little more than 10mm in diameter, while the scintillator is quite a bit larger (38mm). The small-sized aperture was sized to accommodate the X100's 10x10mm PIN diode.

The Theremino-style filtering scheme WILL limit the maximum count rate because it increases the odds of pulse overlap occurring.

 

[homebrewed]

I just came across a web site that describes a different kind of DIY x-ray detector that has some advantages regarding XRF analysis of steel alloys. It's a gas-based proportional detector, which WAS briefly discussed very early in this thread -- but at the time all I found was versions that required a specific operating pressure and somewhat-exotic gasses (like a flammable argon-methane blend called P10). But the detector (described here) uses a continuously-flowing gas mix of argon and CO2, basically the same stuff used for MIG welding. And the detector's operating pressure is atmospheric pressure! No vacuum pump needed. No mixing valves, mass flow controllers, none of that spendy stuff. A simple tapered-tube flowmeter would work just fine.

The detector referenced in the above link does use a thin aluminum window but I also have found there is a class of windowless gas proportional detectors, which would have the absolute least amount of x-ray absorption. A window is always going to be a problem when it comes to the low-energy x-rays we're interested in. For this kind of detector, the main problem likely will be stray 60Hz noise. It is NOT sensitive to light, so it doesn't need a light-tight enclosure, either.

The other nice thing about this type of detector is that it can be built from raw materials using tools that most of us already own. It DOES require a high voltage power supply capable of outputting something on the order of +2,000 volts DC, but the current requirements are very low. That's good from a safety standpoint.

I don't have a MIG welding setup so I can't really go much further with this scheme. At least, not yet. I'm still hoping that I can get one of those silicon PIN detectors to work.

Apparently you can get small amounts of MIG welding gas in disposable containers, but the only sources I found were in Europe and they won't ship to the US.

 

[homebrewed]

I messaged the fellow who made the gas proportional counters, asking if the sealed version lasted very long. He indicated that it only lasted a couple of months due to outgassing from the 3D printed pieces, and he didn't bake it to drive out the water. The buildup of contaminants "killed" the tube.

The continuous-flow approach is much more robust regarding these kinds of issues. 'Course, its drawback is that it consumes gas all the time it's on. But the gas flow doesn't need to be all that high, once the detector tube is completely filled with welding gas. Considering all the requirements needed to construct a reliable sealed-tube counter, I think the tradeoffs strongly favor a continuous flow approach.

 

[homebrewed]

An atmospheric-pressure proportional counter requires a higher voltage than a PMT to operate, something on the order of 2KV. Due to my choice of HV diodes and capacitors the HV supply I made can't go that high so I've been thinking about that. I have a commercial HV supply that can go up to 3KV, made by Ortec but its noise+ripply specification is on the high side. The supply was originally meant for a phosphor + PMT detector combo for a scanning electron microscope so its noise specification didn't have to be all that great, compared to an XRF application. On the other hand, it appears to be pretty good in terms of stability.

So I was thinking of using a simple capacitor multiplier RC filter -- basically it's an emitter follower driven by an RC filter, to clean up that HV supply's output. The transistor used for the emitter follower doesn't have to have a very high breakdown voltage because it won't much of a voltage drop across it. But its performance is only so-so, as can be seen below:

At 60Hz the filter's output is about 24 dB down from a 0 dB "ripple" signal.

Then I started thinking about so-called "ripple eater" circuits I've seen, but the simulations I'd done on the 1-transistor versions didn't perform all that well. So I came up with an opamp-based circuit, and it performs MUCH better -- see below.


This circuit attenuates 60Hz ripple by about 70 dB. It's stable because the circuit gain is mostly determined by the ratio of two capacitors, C2 and C3. The 20 meg resistor is there primarily to provide some bias current for the opamp's inverting input, it doesn't play much of a role in the circuit's AC performance.

This version of a ripple eater is pretty easy to understand if you know anything about opamps. Basically the inverting input of the opamp is a summing node. The feedback acts to force it to equal the non-inverting input, which is at ground -- so the ripple voltage on the other side of C2 is greatly attenuated.

The AC sweep also shows that the circuit is nice and stable, no wild oscillation issues to worry about.

In a real-world application I would add some diodes to protect the transistor and opamp input. If something happens to the input HV that causes it to abruptly drop Q1 probably would be zapped due to the stored charge in C2. And in turn C2 would pull current out of the opamp's input pin. Cheap insurance.

 

[homebrewed]

Need an excuse to consume your favorite beverage?

 

[WobblyHand]

Any progress to report? I have some mystery steel that I'd like to identify. I'd like to be able to know it's properties so I could do an FEA simulation on a part. There's enough variation on the yield strength to swing the results from safe to potential failure. I messed up a part - there, I said it. So I will have to scrap it. If I knew the material, I'd be able to determine the safety margin. Right now I can't, so the part needs to be scrapped.

Still, it would be useful to know what material it was, so I could bound the design. I was attempting to modify a stock part. Typically manufacturers don't inform consumers about materials.

 

[homebrewed]

Any progress to report? I have some mystery steel that I'd like to identify. I'd like to be able to know it's properties so I could do an FEA simulation on a part. There's enough variation on the yield strength to swing the results from safe to potential failure. I messed up a part - there, I said it. So I will have to scrap it. If I knew the material, I'd be able to determine the safety margin. Right now I can't, so the part needs to be scrapped.

Still, it would be useful to know what material it was, so I could bound the design. I was attempting to modify a stock part. Typically manufacturers don't inform consumers about materials.

Progress has been slow at the moment due to summer-related distractions (a 2000 ft^2 vegetable garden, kayaking, etc.). I have a front-end design for one of those low-leakage ~3mm^2 PIN detectors I found. It uses a JFE150 JFET, designed by Burr-Brown (now owned by TI). It simulates pretty well, so there is a PCB design that's being designed. I'm doing it with the option of adding a Peltier cooler to further improve the SNR. This is where I'm hung-up at the moment, mulling over the best way to achieve a reasonably hermetic package w/o impacting the circuit performance due to parasitic capacitance. I purchased some activated-alumina desiccant beads that should be able to dry the internal volume down to where the dew point will be less than -40C. This will be particularly important if I decide I want to try using the CsI scintillator crystal I got on ebay. I also have thought about putting the detector in a leak-tight enclosure and pulling a vacuum around it, but this puts additional constraints on the window. It would have to be strong enough to withstand the pressure difference, i.e., thicker & therefore would absorb more of the incoming xrays.

My experience with the scintillator/PMT combo was very useful w/regard to finding and fixing some problems in my MCA code, but it appears to crap out below about 20Kev -- "nothing to see here" -- probably mostly due to the relatively thick aluminum around the NaI scintillator crystal. That's why I'm back to some sort of direct-conversion scheme. The Hamamatsu PIN diode I got comes in a metal TO-8 style package so it should be possible to remove the lid if that is needed to extend its sensitivity down to the ~6Kev range of interest.

Just to keep things interesting, I also have continued my investigation into the possibility of using an ambient-pressure gas proportional counter. The best performance is achieved with CO2 percentages in about the 10% range, on the low end of TIG welding gas blends. But it might be good enough. And on this side of things I found a relatively inexpensive line of flow sensors that could be used to set the gas flow low enough to where the cost per analysis wouldn't hurt too much, $-wise. This is in a continuous-flow situation. It also may be possible to fill the detector, close a valve and use it awhile with no flow.

One of the really nice aspects of a gas proportional counter is that it's not affected by light, unlike all the scintillator or PIN diode approaches. The fact that a TIG gas blend may work OK might be attractive to those who already own a TIG setup. Heck, it could be yet another reason to get MYSELF a TIG setup .

So the answer is a qualified "yes", but unfortunately nothing's in good enough shape to help you with your current problem. Now that gardening season is coming to a close progress should pick up again.

 

[homebrewed]

Here's a screenshot showing the current state of my latest TIA, designed to use one of the smaller PIN diodes I recently found. It includes a small temperature/RH sensor so I can determine if the internal RH is too high to operate a Peltier cooler. I want to avoid condensation & the attendant surface leakage that would mess up the circuit operation.



I added some copper on the back side of the board, underneath the regions that would be most-sensitive to 60Hz noise pickup. It's connected to the GND net. I AM concerned about its impact on the parasitic capacitance it will introduce, so it will be interesting to see how it behaves. Initial testing will be done at room temperature, no need to get fancy until I know if the design is going to be stable or not.

Adding the TEC will be "interesting" and will likely reveal some mechanical design shortcomings on my part. That's definitely TBD.

While designing this PCB I discovered yet another EasyEDA bug. I created several copper regions on the board using separate rectangular blocks and found that the S/W had problems figuring out the connectivity, despite the fact that I defined them all to be on the same net. The only way to satisfy the DRC was to add vias that "connected" the rectangles together, despite the fact that they were overlapping on the same layer of the board! That was a bug that took me awhile to figure out because I thought I was doing something wrong.

 

[homebrewed]

A couple of notes, one regarding the PCB design. One application note I found suggested putting a ground trace underneath the TIA feedback resistor, to shunt-out some of the parasitic capacitance that's across the resistor. This is meant to improve the circuit bandwidth. There wasn't enough room to do that between the 0805 SMT resistor pads so I placed traces on the two inner layers. I _could_ go to a larger resistor like a 1206 but they have higher parasitic capacitance anyway. And there's nothing that can be done regarding parasitic capacitance across the top of the resistor, anyway. The feedback resistor is R5 on the layout. A thru-hole resistor that has a wrap of thin wire around its middle would _really_ knock down that parasitic capacitance, but do I actually want to go there??

The second note is that I found a surprisingly inexpensive flow sensor whose range matches the low flow rates I'd want for a continuous-flow gas proportional counter. It could be used to make a relatively cheap mass flow controller in the 0-30 SCCM range. Digikey has them for $45 but unfortunately they drop-ship directly from the German manufacturer so shipping is a bit high. I also have been looking into DIY flow sensors that use cheap NTC thermistors. Thing is, I'd still need a "real" flow sensor to calibrate the home-made one.