Needing more than a spark test?

[rabler]

Clough42 just released a youtube video using a Scanning electron microscope to analyze a drill bit failure. Gives some good high level summary of what is trying to be achieved here, obviously without the imaging.

 

[homebrewed]

I've been attempting to use a higher-order polynomial fitting scheme to see if that will improve the spectrums I'm getting, in terms of the FWHM distribution for photon energy. The idea is that a higher-order polynomial will permit the use of more samples around the peak (in addition to a higher sampling rate with a higher-resolution ADC), which should improve the effective SNR. Turns out there are some interesting things one has to address when using higher-order polynomials. The most significant is that they can have multiple zeroes, so there's a possibility that the peak-finding routine will converge on the wrong value....or just head off toward positive or negative infinity. I ran into this as a grad student so it's not totally unfamiliar (but the previous work was a LONG time ago).

I started by using a 4th-order polynomial fit, so ithere could be three solutions for the peak (defined as where the first derivative is zero, signifying a local maximum value). Since I thought it might be necessary to go to an even higher-order polynomial, I decided to use an interative scheme to find the zeroes, using the Newton-Raphson method where Xn+1 = Xn - F(Xn)/F'(Xn). Turns out there were all kinds of "interesting" programming issues to address, but I think I have the approach mostly tamped down. The main thing was to specify an initial guess based on a simple find-the-maximum-value routine that scans the actual pulse data.

Evaluation results aren't done yet. I'm getting some spectra now to see how they look.

 

[homebrewed]

The Scionix PMT + scintillator I got has been a bit of a puzzle to figure out. There isn't any information on the Scionix web site regarding the connections. And speaking of connections, there are wires coming from both ends of the thing. It's pretty clear which set comprise the pulse output and HV, because it includes a coax for the pulse output. I'm not sure just what the connections at the other end are for. I'm speculating that they are for monitoring the state of the NaI(Tl) scintillator.

But I DID find some very useful information at the very end of the Theremino "PmtAdapters_ENG.pdf" document. Their drawing is consistent with the results I got when ohming-out the connectors. It doesn't include the wires at the other end so I have to assume that they really aren't necessary for proper operation of the PMT. I measure ~50megohms between two of the wires, maybe leakage current through the crystal. If so, that doesn't bode well for the health of the scintillator. I may end up replacing it with a different scintillator. I have a couple of LYSO crystals but their form factor isn't all that good for a 1.5" PMT.

The Ortec HV power supply I scrounged from an old SEM doesn't have very good ripple specifications, according to the Theremino folks, so I'm thinking that I need to build something similar to what they are using. The problem is that both transistors they specify for the flyback circuit are obsolete. So instead of that, I'm going to use a CCFL power supply module as the starting point. I found a "universal" CCFL power supply for a bit under $8, but will need to add some high voltage diodes and capacitors -- and a control circuit to stabilize the voltage. I found at least one example of a PMT supply built this way so I can use that as a starting point.

 

[homebrewed]

I've been doing some simulations to test some ideas I have for a PMT power supply. I have no idea how to include a CCFL module so I'm just using a VCVS (voltage controlled voltage source). Results are shown below:



The top plot shows the output of the power supply over a 0-50C range, and the bottom plot shows the output of the VCVS over the same temperature range. I had thought that the phase shift due to the three RC filters would guarantee oscillations but that isn't the case. Probably due to the _really_ serious rolloff imposed by the integrator.

The pulsed current source is there to test the circuit's response to an impulse load. I also performed an AC sweep from 10Hz to 100KHz and didn't observe a problem.

The original circuit I found on the internet doesn't use the transistor "capacitance multipliers" shown here. It uses an opamp-based capacitance multiplier. Interestingly, with a 1V ripple voltage applied in series with the HV input, the output of Q1 shows a fairly respectable amount of ripple: but it isn't caused by ripple on the base input. It's due to collector-emitter impedance (or perhaps Early Voltage effects?). That's why I have two in series -- as a result the output of Q2 has just a few microvolts of ripple. Even with the 10 meg base resistors the voltage drop across each transistor is pretty small, because their base current is just 25nA.

One oddity I noticed with the simulations is that transient simulations indicated that the apparent beta for Q1 is very low, around 8 (!). Q2 looked fine. Drove me nuts until I did a DC setpoint simulation and found that both transistors' beta was close to 300. Not sure why the transient simulation is so far off in that regard. Perhaps LTspice thinks Q1 is in saturation when it does the transient simulation? Guessing here.

Since the transistors only have a few volts' drop across them it isn't necessary to use high voltage transistors. My simulation uses 2N3904 transistors, and would likely work just fine in this case. Since their emitters are looking into fairly high impedances I don't think they would go into parasitic oscillation. A real-world design would include a clamp diode between the collector of Q1 and the emitter of Q2....just in case of an arc....

 

[graham-xrf]

Hi Mark
Just to catch me up here - I take it you are making up a HV supply for the PMT you recently bought?
Is it that it is a devoted "ripple remover", like a very low pass filter?

 

[homebrewed]

Hi Mark
Just to catch me up here - I take it you are making up a HV supply for the PMT you recently bought?
Is it that it is a devoted "ripple remover", like a very low pass filter?

Yes, and yes . I ordered a couple of these, which can output up to 900VRMS, so the peak voltage should be around 1200V (a guess, since I'm sure that the output waveform is not a sine wave).

In the actual circuit, the inverter's Vin will be driven by the buffered output of the integrator. Another option is that the inverter has a 0-5V "brightness" control, but that may just change the duty cycle of the waveform. I will need to make myself a high voltage probe so I can safely examine the waveforms on my 'scope.

The inverter frequency is between 40 and 70KHz so that will be easy to filter out. I'm using much more aggressive filtering because I want to make certain there isn't much low frequency noise present, which will modulate the gain of the PMT.

The inverter can output up to 8mA so I definitely will need to be careful when using it! The 1 meg resistor on the output is mostly there to reduce the maximum output current. And the actual supply won't have a 47nf capacitor out there, either. It can be much smaller, since the current pulses coming out of the PMT are very short.

 

[graham-xrf]

Yes, and yes . I ordered a couple of these, which can output up to 900VRMS, so the peak voltage should be around 1200V (a guess, since I'm sure that the output waveform is not a sine wave).

In the actual circuit, the inverter's Vin will be driven by the buffered output of the integrator. Another option is that the inverter has a 0-5V "brightness" control, but that may just change the duty cycle of the waveform. I will need to make myself a high voltage probe so I can safely examine the waveforms on my 'scope.

The inverter frequency is between 40 and 70KHz so that will be easy to filter out. I'm using much more aggressive filtering because I want to make certain there isn't much low frequency noise present, which will modulate the gain of the PMT.

The inverter can output up to 8mA so I definitely will need to be careful when using it! The 1 meg resistor on the output is mostly there to reduce the maximum output current. And the actual supply won't have a 47nf capacitor out there, either. It can be much smaller, since the current pulses coming out of the PMT are very short.

For less than ten bucks, I think you made a smart choice. One can add some electronics to dress up the output to get it regulated and clean. "Regulated" is the key word. Keeping a fluorescent backlight display lit is not like keeping a PMT gain nicely constant.

I use two approaches to aggressive filtering. Low pass filter type, with lots of energy storage as in "smoothing capacitor" works if the storage is so massive that it can supply current transients without significant ripple. Using a less storage higher impedance very filtered derived voltage can be used as a smooth regulation reference, to a fast, high bandwidth active device like regulator or integrator, able "regulate away" the ripples, and noises, just by being faster. High voltage regulation in the usual format is a bit more tricky.

My PMT tube does not have any circuit in there at all. Mine just has a socket and a bunch or wires.I have to make the HV supply, and also the set of electrode voltages.

-------------------------

At home, things are just getting more chaotic! Yesterday, I had (my first ever) tooth extraction. It was from upper right, and now, I am nursing the wound. It's the kind of thing that can put a damper on the whole day.

 

[homebrewed]

For less than ten bucks, I think you made a smart choice. One can add some electronics to dress up the output to get it regulated and clean. "Regulated" is the key word. Keeping a fluorescent backlight display lit is not like keeping a PMT gain nicely constant.

I use two approaches to aggressive filtering. Low pass filter type, with lots of energy storage as in "smoothing capacitor" works if the storage is so massive that it can supply current transients without significant ripple. Using a less storage higher impedance very filtered derived voltage can be used as a smooth regulation reference, to a fast, high bandwidth active device like regulator or integrator, able "regulate away" the ripples, and noises, just by being faster. High voltage regulation in the usual format is a bit more tricky.

My PMT tube does not have any circuit in there at all. Mine just has a socket and a bunch or wires.I have to make the HV supply, and also the set of electrode voltages.

-------------------------

At home, things are just getting more chaotic! Yesterday, I had (my first ever) tooth extraction. It was from upper right, and now, I am nursing the wound. It's the kind of thing that can put a damper on the whole day.

Sorry to hear about your tooth. Dental problems can impact your entire immune system so even if temporarily painful you did the right thing.

 

[MikeWi]

I've actually followed this complete thread and it's a fascinating read, but SO far over my head!

 

[graham-xrf]

I've actually followed this complete thread and it's a fascinating read, but SO far over my head!

Hi Mike - welcome to xrf thread.
The thread is outrageous, and are an exception! The first thing to know is that much of this stuff was also unknown to me and others here at the beginning! In wringing out the science on the way, I now know a bit more about how this stuff works. HM is actually full of knowledgeable folk who know, no matter how obscure the subject.

We do have the goal of getting something together that can be acquired and used by HM folk without spending sums that would buy a couple of nice machines, and without having to be expert at nuclear physics. The hard stuff will be figured out here. There is room for mechanical creativity. Certain bits have to be machined, or fabricated. If the idea can work at all, this design effort should pretty much wring it out so we can finally know.