Needing more than a spark test?

On a whim I fired up my de-shielded pocket geiger. Unsurprisingly, lots of AC pickup. I haven't replaced the LMC662 with a better device yet, but I have gotten a 5V LDO to replace the 9V version that came on the board. The original is in an SOT23-5 package so it's best to replace the original with one from the same family. Simulations don't reveal a problem running the LMC662 LTC6269 well beyond its abs-max supply voltage, but that's not surprising. Many SPICE device models aren't designed to catch that.

On a slightly different subject, I've been looking into making my own fluxgate magnetometer (for a completely different project) and came across a product that could be very useful if we find it necessary to fall back to a scintillator/PMT approach. It's a magnetic shielding tape that is claimed to not suffer from degradation of permeability if it's bent or formed. Not cheap, but it could be useful if you have some odd mechanical configuration you need to shield: This. Unlike permalloy, the product contains a lot of cobalt rather than nickel. Maybe it also could be used as a standard to aid in the XRF spectrometer analysis. It wouldn't be useful as an energy filter because it (likely) would only be useful to help separate nickel and copper peaks.

I've been wanting to make the lead shield to do some experiments but right now my mill (which I would use to make the aperture hole for the detector) is need of repair. That's due to a mistake I made in a modification to the mill's Y axis; but I should have the items necessary to address that pretty soon.

Edit some time later: struck out "LMC662" and corrected.
 
Last edited:
I look forward to messing about more with the XRF, Right now, most of anything I would be doing as interest projects are temporarily boxed up. The throughput on building related stuff has made the garage more or less unusable. The final straw was dragging in the crate with the mill drill!
I spent yesterday exploiting the let-up in the rain to get out there with the laser level (handy thing), and with my stakes and profile planks, setting out the foundations and surrounds for the outbuilding.

As I recall, in simulations, I did get significant performance differences between using the LTC662, and the LTC6269.
Beyond that now. The 662 has been lifted off the PCB. It yielded to the Polish coil wire trick. This has to be a lash-up. It would be great if in a final design, the diode could be mounted at right angles to the PCB, so all could go down a metal tube, with a little lead insert at the end. The shielding would be inherent.

TIA Amplifier Design
If a low bandwidth TIA is built having extreme high Rf, then most FET input op-amps end up performing a bit similar in simulation. In my simulations of "event" pulse current injections, I deliberately spread the gain, trading lower first stage gain for bandwidth, so I could capture the pulse area integral with sampling. I went against some of the aims of directly putting current electrons into a summing input of an op-amp with hundreds of MegOhm Rf. All this, while trying to keep the noise down.

In conventional TIA design, there is a noise factor benefit in simply making the feedback resistor as high as possible. The transimpedance gain (Rf) increases faster than the Johnson-Nyquist thermal noise.

Thermal Noise = √(4*k*T*B*Rf). Gain = Rf.
Signal/Noise improves by √Rf

This, I think, is why we get TIA circuits with 400MegOhm, and 1GigOhm feedback resistors. The 66MegOhm in the Geiger 5 seems relatively low by comparison. Here is where we get to what we want it to do. For detection and measuring of low level light, this instrumentation is OK. When it comes to detecting pulses, where we care also about the real amplitude and duration of the pulse, we depart from photovoltaic. We change technology to PIN. We use a bias to lower capacitance. We trade off noise contributions. We exploit the extremely low offset voltages and extremely low bias currents in the modern op-amps.

I was finally getting consistency, and feeling better about the simulations, but to get there, I had to first test LTSpice in artificial, simplistic setups, and set the tolerances, and convergence, to make it credible when counting femto-amps. LTSpice had been tweaked to speed simulation of switcher power supply chips.

The shunt resistance of our PIN diode is 40MegOhm. That is "low" compared to common photovoltaic photodiodes. After going all around the houses with simulations, like you, I got to the point where I needed to try it out. I will push the TIA first stage gain as far as I can, consistent with not losing the pulse integration area information, because that is the one property that discriminates the material that caused it. The amplifiers need to be oriented to keeping some of the nature of the pulse, rather than just counting the fact they happened.

We can put up with the pulse being distorted from passing through bandwidth limited circuits, so long as different sized pulses can remain recognizably different in calibration. We cannot have local circuit energy storage artifacts, such as waveform response overshoots, contaminating this information. That is why the pulse circuits I was happy with do not have these overshoots.

If you haven't come across it already, I attached a 1994 application note from Burr-Brown before it joined with Texas Instruments. Figure 2 amplifier has Rf = 1GigOhm. The photovoltaic circuit Figure 3 uses 400MegOhm. The "wider bandwidth" 100kHz circuit has to use Rf = 1MegOhm.
In the circuits I was interested in, I used 510K up to about 1MegOhm.

The OPA128 was an electrometer grade amplifier. Very good in it's day, but only 1MHz GBW. The specifications cannot match things like LTC6269.
- - - - - - - - - - - - - - -
We should do OK messing with butchered Geiger 5 boards for development.
 

Attachments

Playing with the Pi4
On first lash-up, it's only the Pi4 hooked to the A/D converter via SPI.
The general scheme is the Pi4 is much like the one right next to me running the web server with NGINX.
We already know that a whole lot of phone "apps" are just links to a web page. This is a similar idea.

Basically, the Pi4 does everything, and is with the TIA amp kit. It also hosts a web page with the plots in real time.
It's on Wi-Fi. It's on BlueTooth. It can be on a LAN with an IP address. It can even be the other side of the planet.
With zero extra software, any smartphone with a browser can see and interact with the plots.

I am hooking up Pi4 Model B to a skeleton lash-up AD7622BSTZ A-to-D converter to shift at least 20 chunks of 16-bit samples in a 10uS period.

I very much wanted to have it as a sensor with a bunch of electronics, and a phone, as an available choice for interested folks. There is nothing at all wrong with alternative displays, like a WiFi connected laptop, or using one of the Pi4-B HDMI display connectors to a monitor, and mouse + keyboard. Direct to Pi4 would likely be the way I do it. For those using a phone to display, the web page has buttons and inputs to control it.

Maybe a full Pi4-B with maxed out memory is overkill, but the ease with which it can be a web server is what convinced me.
 
graham-xrf said:
Playing with the Pi4
On first lash-up, it's only the Pi4 hooked to the A/D converter via SPI.
The general scheme is the Pi4 is much like the one right next to me running the web server with NGINX.
We already know that a whole lot of phone "apps" are just links to a web page. This is a similar idea.

Basically, the Pi4 does everything, and is with the TIA amp kit. It also hosts a web page with the plots in real time.
It's on Wi-Fi. It's on BlueTooth. It can be on a LAN with an IP address. It can even be the other side of the planet.
With zero extra software, any smartphone with a browser can see and interact with the plots.

I am hooking up Pi4 Model B to a skeleton lash-up AD7622BSTZ A-to-D converter to shift at least 20 chunks of 16-bit samples in a 10uS period.

I very much wanted to have it as a sensor with a bunch of electronics, and a phone, as an available choice for interested folks. There is nothing at all wrong with alternative displays, like a WiFi connected laptop, or using one of the Pi4-B HDMI display connectors to a monitor, and mouse + keyboard. Direct to Pi4 would likely be the way I do it. For those using a phone to display, the web page has buttons and inputs to control it.

Maybe a full Pi4-B with maxed out memory is overkill, but the ease with which it can be a web server is what convinced me.
Click to expand...
That sounds pretty sweet, especially for the price -- and you have a number of options on how to implement the user interface.

I think my first attempts will use my Teensy for the front end ADC and peak detector. It will send the data via its USB to an old Dell laptop running a pared-down version of linux. On that end, a bit of code written in plain-vanilla C will partition the peak data into bins -- that's the MCA core -- and output a csv file I can post-process and display with a spreadsheet. In addition to the MCA function, the C program will handle the setup stuff for Teensy -- start/stop acquisition, sample rate, etc. Not in the least portable....not in a physical way, anyhow. You will be ahead of the curve in that regard.

If the massively-filtered pulse approach doesn't work out, the same basic setup still can be used, but the front end -- TIA and ADC -- will be different. Since the Teensy will just output the peak voltage values, the data rate going to the back end will be the same regardless -- it's just the number of clean pulses culled out of the raw ones coming from the detector.
 
homebrewed said:
That sounds pretty sweet, especially for the price -- and you have a number of options on how to implement the user interface.

I think my first attempts will use my Teensy for the front end ADC and peak detector. It will send the data via its USB to an old Dell laptop running a pared-down version of linux. On that end, a bit of code written in plain-vanilla C will partition the peak data into bins -- that's the MCA core -- and output a csv file I can post-process and display with a spreadsheet. In addition to the MCA function, the C program will handle the setup stuff for Teensy -- start/stop acquisition, sample rate, etc. Not in the least portable....not in a physical way, anyhow. You will be ahead of the curve in that regard.

If the massively-filtered pulse approach doesn't work out, the same basic setup still can be used, but the front end -- TIA and ADC -- will be different. Since the Teensy will just output the peak voltage values, the data rate going to the back end will be the same regardless -- it's just the number of clean pulses culled out of the raw ones coming from the detector.
Click to expand...
The Teensy is fine, and the slowed up distorted pulse can also work OK, so long as the pulse varies a bit with the energy that caused it, and can be "recognized" in a calibration. At this stage, goodies like phone display are not important. We gather that stuff up at the end.

It's just that I have, somewhat by accident, and partly by rabbit-hole suck-in, managed to maintain the undistorted theoretical pulse as a high resolution sampling. I am tempted to wind the resolution discrimination up as far as it will go, but that is all - just tempted, not acting on it! The A/D could allow 40 samples in a 10uS shot, and the number of totalizer buckets can be as many as we can find statistically separated sets of.

There is, of course, all sorts of smarts filtering to have it do sensible actions when confronted with pulses very short, pulses very long, three pulses in 15uS, a huge energy pulse that did not come from an X-Ray, 50/60Hz, a set of numbers so untypical, it "knows" the sample is not there, etc.
I am in two sections here. One is all about the A/D sampling via SDR, and making that bulletproof. The other is about what is upstream of the ADC. Polish coil-wire trick may have worked just fine in removing the LTC662. Getting the LTC6269 to stay put long enough get soldered in it's place is a bit more fraught. Maybe i had too much coffee. I think I have to break out the binocular illuminated inspection/assembly (low power microscope) thingy, currently in storage.
 
graham-xrf said:
Getting the LTC6269 to stay put long enough get soldered in it's place is a bit more fraught. Maybe i had too much coffee. I think I have to break out the binocular illuminated inspection/assembly (low power microscope) thingy, currently in storage.
Click to expand...

My stereomicroscope never goes into storage, just because it makes working on stuff like that SO much easier. Not to mention pulling nasty metal swarf splinters out of fingers.

Correct me if I'm wrong but it sounds like you're initially going to repurpose the pocketgeiger by simply replacing the LTC662 and Rf? I'd thought you had removed the X100-7 for your own board but perhaps not.
 
homebrewed said:
My stereomicroscope never goes into storage, just because it makes working on stuff like that SO much easier. Not to mention pulling nasty metal swarf splinters out of fingers.

Correct me if I'm wrong but it sounds like you're initially going to repurpose the pocketgeiger by simply replacing the LTC662 and Rf? I'd thought you had removed the X100-7 for your own board but perhaps not.
Click to expand...
I had thought my final TIA circuit is just too different to keep the pocketgeiger board. An arrangement with sensor mounted at right angles to a PCB, at it's narrow end, would allow the whole thing to fit in a metal shielding tube, (so no metal tape), and the radiation illuminated region being circular, so having a turned lead shield that mounts the Am241 sources in little drilled recesses, arranged all around. These can be angled to shroud signal returns from stuff made of lead. They squirt all directions at once, but only controlled apertures count. It seems the geometry may have such a shape, where the photons that get loose, and hit lead in various places do not have a lead target x-ray glow path into the diode.

I use the pocketgeiger board as a lazy temporary convenient breadboad lash-up. It mounts the diode, and and a first stage. The value of finding a way to keep a re-worked version of it, if it can be done, is not lost on me. Keep in mind that you and I may be able to bring out the stereomicroscope, and modify PCBs, but a lot of folks who would love to have one, just can't do it. That is where an dedicated board may make sense. I also thought that most folk wanting one, would love to machine up the tube and shields, and so get invested it it being partly their own work.

A final physical arrangement could have the sensor mounted on the PCB, as like pocketgeiger, but it forces the circuit board to stick out sideways through the lead shield, and ruins the symmetry of a circular array of sources blasting away at the test surface. I think my choice end product would be a diode, with the piece of board under it, cut free of the pocketgeiger PCB, and fixed at right angles onto a new amplifier PCB. If it proves to be possible to build a decent amplifier on a somewhat modified pocketgeiger PCB, I suppose that coud be a big advantage. No need to make more PCBs.

It is much easier to carry a amplified buffered pulse signal up to a ADC mounted at the Pi4, instead of trying to run up to 70MHz SPI signals up a cable. This time, having a Pi4 more or less in the same box shield, and relying on network solves most of the 50/60Hz interference coupling.

On question that comes to mind is, can there be (useful) signature energy groups from stuff other than metal alloys?

[Edit: Thanks for the swarf splinters suggestion. It would not have worked on my last encounter with showers of broken off wiggly wires flung free of a rotating wire brush in my Makita drill to end up somewhere on the garage floor. Later, heading through side door to fetch something out of garage, but only wearing sock on the feet. The wire went into my foot pad roughly between big toe and the toe next to it. probably about 9mm in, and the end disappeared. No amount of magnifier will let you see under the foot. The fix involved help from others, and pointy scalpel blades, and made my eyes water! ]
 
Last edited:
Yay - getting help from No.1 Son (satellite engineering software developer) working up the AD7622 ADC into life. :)
 
You must log in or register to reply here.
Back
Top