Needing more than a spark test?

[homebrewed]

As a follow-up to the discussion regarding the Radiacode, I need to say that it is really inexpensive for what it does. The next-closest I found, an offering from Gamma Spectacular, is about $1300. It uses a PC or Mac so not as portable as the Radiacode.

One little detail that is especially relevant for our effort is that the energy resolution specifications are based on a much-higher energy reference -- Cs-137, which emits gammas @660Kev. A good NaI scintillator detector can achieve around 7.5% at that energy. At lower energies the resolution goes down. At 6Kev it's substantially worse than that. Other scintillator materials can perform much better than that, but they cost MUCH more.

Silicon PIN diode detectors can perform much better at lower energies, so that's why our efforts have centered around them. Unfortunately, the large 10x10 detector that was sold as part of a ~$100 geiger counter is no longer available so I've been exploring alternatives. The best results, in terms of energy resolution, that I have seen for PIN detectors, would be great for our needs but costwise are out of the reach of hobbyists. I found some smaller ( 3x3mm) detectors that are pretty reasonable, pricewise, and one of them already has been used in a simple "demo" system -- see here. Improved electronics, along with (likely) a Peltier cooler, could get us where we want to be.

The PIN detector used in the demo is a Hamamatsu S1223. Newark sells them for $11 or ~$19, depending on the detector area. The more-expensive one has about twice the area, at 13mm^2, and is one of the ones I bought to play with (the other is an even cheaper, smaller-area PIN diode sold by Osram). So definitely in an affordable range for folks like us. Some fairly good electronics will be needed as well, but I think there are a number of folks on this forum who are well-qualified to contribute in this regard.

There is one big unknown regarding the use of PIN detectors that weren't specifically designed to function as x ray detectors -- and that is, of course, how well they will work in that application. The link I included shows that the S1223 can at least detect x rays but it's not clear (yet) if its energy resolution and its sensitivity will be sufficient. There's only one way to find out......and I'm working on that!

HB

 

[rwm]

It would be nice to get away without the Peltier cooler. What was the cost on the 10x10 detector?
I am also a little confused about the detector. In the link they indicate "In our tests a S1223 Hamamatsu photodiode was used: it is a “general purpose” photodiode, sensitive from 320 nm to 1100 nm." Isn't that wavelength a lot longer than the 6KeV photons we are trying to detect? What did I miss here?

 

[homebrewed]

It would be nice to get away without the Peltier cooler. What was the cost on the 10x10 detector?
I am also a little confused about the detector. In the link they indicate "In our tests a S1223 Hamamatsu photodiode was used: it is a “general purpose” photodiode, sensitive from 320 nm to 1100 nm." Isn't that wavelength a lot longer than the 6KeV photons we are trying to detect? What did I miss here?

The 10x10 detector plus the rest of the electronics and PCB for the "pocket geiger" cost about $100 from Sparkfun. That was a killer price because the detector alone, when it _was_ available from Digikey and Mouser, cost about $130. Now, it's possible that the ones Sparkfun was using were devices with relatively poor performance -- the PocketGeiger never was touted as an x ray spectrometer, it was sold as a cheap geiger counter that could be connected to a cell phone. Its output signals were pulses generated by a couple of on-board comparators. For our x ray spectrometer development I hacked into the signal path to gain access to the TIA output. I also replaced the original opamp with a much better one, in terms of noise performance but it didn't seem to make much difference. The dark current was the main noise source and there wasn't much that could be done about that (other than cooling IT, which I did consider).

Even if cooling the 10x10 detector had worked out OK, it now is sort of a moot point in terms of getting something into the hands of more machining-type hobbyists.

The deal with using a PIN detector is that it has a much wider depletion region so it has a higher probability of detecting x rays. The 10x10 detector was a PIN diode as well. I ran some calculations to determine the depletion width of the 10x10, using the capacitance specified by the manufacturer and that turned out to be pretty comparable to the Hamamatsu (and Osram) PIN detectors. Another way to compare the diodes on a per-unit-area is to look at the dark current. It is mostly due to thermal generation so should be (approximately) constant, and proportional to the depletion width. Again, the diodes all have about the same dark current per mm^2. I looked at a number of PIN diodes and most were NOT comparable, so I think I may be on to a way to identify potentially useful PIN diodes.

Regarding cooling the detector, my main reason for considering it is the fact that commercially-offered PIN detectors with good resolution at low x ray energies are all cooled. The dark current is going to contribute noise, and detector resolution is strongly affected by the detector's SNR.

 

[rwm]

Thanks. That makes some sense. So the PIN detectors will sense these photon energies even though they are out of the published response range? Or am I misinterpreting the manufacturer data?

 

[homebrewed]

Thanks. That makes some sense. So the PIN detectors will sense these photon energies even though they are out of the published response range? Or am I misinterpreting the manufacturer data?

You are not misinterpreting the manufacturer's data. In taking this approach I'm doing something I normally advise strongly against, which is using a part in a manner that doesn't appear on the data sheet.

This is risky for several reasons. As shown by that physicsopenlab web page I linked to in #2171, there's at least one example of successful x ray detection using the S1223-1 --- but Hamamatsu _could_ change their IC fabrication process (or the foundry they use might), and, poof, it's no longer useful for what we want.

The other risk is that the device's performance as an x ray detector depends on some uncontrolled process variable that doesn't affect its visible-light performance -- so there could be batch to batch variations.

However, due to the cost of re-qualifying a product (in terms of reliability testing, characterization etc.), IC manufacturers try to avoid changing the IC process: and the device characteristics, in terms of the DS dark current and capacitance specifications, would be REALLY hard to change w/o affecting the intended use. Lot to lot variations are more likely, but I think that variations large enough to affect performance in terms of detecting x rays would be rejected by the mfr's ATE because they wouldn't meet their DS specifications, either. I think I'm just being a Nervous Nellie about this. Or hoping so .

As an aside, I like the S1223 because it's packaged in a metal package, and it would be a piece of cake (for someone with a lathe) to remove the lid without damaging the diode. This would remove the optical window, which would otherwise absorb some of the low-energy x rays we're interested in. It _would_ render it more susceptible to damage, but, since it's light sensitive, it would be behind a thin aluminum window no matter what. I'd like to use a thin beryllium window because it would be physically more robust but the dollar signs suddenly show up in large quantities!

 

[rwm]

Thanks for explaining all that. It makes perfect sense.
Now we are machining semiconductors on a lathe! I love it.

 

[epanzella]

After a spark test if I'm still lost I check to see how it machines and how it welds. Then I go to the speedy metals website and try to match up those properties with something. This method has helped me but it fails as much as it succeeds.

 

[homebrewed]

After a spark test if I'm still lost I check to see how it machines and how it welds. Then I go to the speedy metals website and try to match up those properties with something. This method has helped me but it fails as much as it succeeds.

Yes, the test results would just be one of the steps needed to get a good result. For instance, the alloy composition won't tell you if it's been hardened, tempered or annealed.

 

[homebrewed]

Another update....

I finally got around to assembling my silicon PIN diode-based detector circuit. Just to refresh folks' memories, this design uses an inexpensive PIN diode, made by Osram (the SFH2202). It's about 3mm^2, far smaller than the X100. I chose it because its capacitance and leakage per unit area are very close to the X100. Before testing it I had to correct an assembly error on my part -- I soldered the wrong parts into the second-stage's active filter block. That's what I get for waiting long enough to get a little hazy on the layout!

Anyway, the first thing I did was to test the TIA gain. The design has a string of three series-connected 1pf capacitors in the feedback loop (.33pf total) so the TIA should have a gain of about 1/3.3*10-13 = 3E12 volts/coulomb. To test that I connected a 10pf capacitor to the input and connected my signal generator to the capacitor. It was outputting a square wave with an amplitude of 40 mV, and the TIA was outputting a 1V pulse. Going through the math and back-calculating the TIA's effective feedback capacitance yielded 0.4pF -- not bad! Particularly since the capacitor is just a cheap 10% tolerance component.

Just for fun, I put together a simple lashup to see if I could detect any x rays with this thing. I have to say: MAYBE. There's just too much electrical noise and, even in my darkened basement, stray light that also injects noise via the PIN diode itself. After all, I was using my oscilloscope to test the setup and it most certainly emits light!

One concern I have regarding the photodiode is that it has a fairly thick layer of transparent silicone on top of the die. That will absorb some of the low-energy x rays I'm interested in. The stuff is actually pretty soft and rubbery so it could be scraped off the die -- but I have my doubts that the bond wire would survive the procedure (there are chemical ways to do it but the chemicals are nasty _and_ expensive) . So it's quite possible I will need to move on to using the Hamamatsu S1223 photodiode, which is packaged in a metal can that can be turned off by using a custom-made collet on my lathe. But that's for a future post .

Either way, I need to fabricate a light-tight/EMI-tight enclosure before I can go any further. I'm thinking that I should go ahead and design it so I can drop in my Peltier coolers as well.

 

[rwm]

This it great! Thanks for the update.