Page 1 of 1

Detector noise problem

Posted: Thu Apr 19, 2018 9:31 pm
by Dan Knapp
We are having a problem with discriminating neutron peaks using a He-3 detector tube and seek ideas on how we might solve the problem. The detector is a one inch 4 atm. Saint-Gobain “ruggedized” He-3 detector tube operated at the recommended 1150 volts. The tube is inserted into a central bore in a 6.5” diameter cylinder of HDPE as the moderator. We are using a Cremat CR-110 charge sensitive preamplifier feeding an Ortec 490 Amplifier/Single Channel Analyzer and then an Ortec 871 Counter Module. We are monitoring the preamp output and the SCA output with a dual channel oscilloscope. This counting arrangement works fine with a Californium neutron source where we can discriminate the neutron peaks from the gamma peaks with the SCA, but in the presence of a fusor plasma, there are electrical noise peaks which are higher than the neutron peaks, and thus cannot be discriminated by the SCA. The photos below show scope traces for a noise peak and a neutron peak. Evidence that the second type of peak is a neutron peak is that these peaks disappear when the He-3 tube is removed from the moderator.
First a little background on the experiment. Dan Barnes and I have been working for some time now on extending the Los Alamos work on the Penning trap as a small fusion reactor (https://nucleus.iaea.org/sites/fusionpo ... /Knapp.pdf). We have recently begun attempts to actually do D-D fusion in the Penning trap. We decided as a positive control to implement a spherical fusor in our vacuum system. The fusor has an 8 inch diameter grounded spherical anode (8 longitudes; 5 latitudes) made from 1/16 inch 316SS welding rod. The cathode is a 2 inch sphere of 0.022 inch 304SS wire (6 longitudes; 1 latitude). The cathode is connected to a Universal Voltronics -60kV/60mA X-ray power supply via a 120K ballast resistor. Pressure control is accomplished by closing the gate valve to the turbo pumped chamber and pumping through a one inch flexible metal hose and bellows valve. Deuterium is admitted from a one liter buffer volume via a Granville-Phillips 203 Variable Leak. A stable plasma without excessive cathode heating is obtained by adjusting the deuterium pressure to about five millitorr at a cathode potential of -25 kV, with about 6 mA cathode current.
Under these conditions, a large number of peaks are observed consisting of spikes about 200 nsec wide and lower amplitude broad peaks about 200 usec wide. Detail of the two types of peaks are shown in the attached photos. The broad peaks disappear when the detector tube is removed from the moderator, and on this basis are assumed to be neutron peaks. The noise spikes (as well as the neutron peaks) are only present when the plasma is lit. The detector tube is attached to the preamplifier via a 3 inch coax line with BNC connectors. The spikes disappear when the coax and tube are detached from the preamp, but reappear when the coax without the tube attached is connected to the preamp input. Thus it appears that the noise is entering the system via the preamp input. Both types of peaks disappear when the plasma extinguishes.
We have searched for possible ground loops by trying many different ground connection schemes (normally the entire system is grounded in a star pattern to a large copper buss connected to earth ground). We have also tried a large number of different arrangements of additional shielding. The only thing that made any difference was that adding a grounded foil shield around the preamp power cable attenuated the noise spikes a bit. We tried a new Cremat preamp module, but it made no difference. Putting an isolated grounded foil shield around the outside of the moderator and extending over the coax to the preamp made no difference as did putting an insulated shield on the tube inside the moderator. Moving the detector/moderator/preamp off the main system chassis reduced the number of neutron peaks (moved the detector farther away from the source) but made no difference in the noise.
We are seeking advice on how else we might attack this noise peak problem. Hopefully, the shape of the noise peaks with the ringing tail might be a clue to the source. Any suggestions would be appreciated.

Re: Detector noise problem

Posted: Fri Apr 20, 2018 12:43 am
by Richard Hull
If you look at the fusor IV's neutron detector you will see there is no cable between the preamp and the Rueter & Stokes 4 atm 3He tube. This is the ideal. I do run about 6 feet of coax from the preamp to the pulse shaper and SCA in my NIM bin.

The first scope trace is a damped wave and might be indicative of an arcing within the system. A common wave form from an arc discharge in a line transmitted or picked up elsewhere.

I assume the second scope image is not a simultaneous capture on the DSO but a noise pulse on one channel overlayed with the neutron pulse as a time ordered example for us to see.

Wherever there is coax, I clamp on a ferrite block externally. I also attach ferrite blocks on the HV line right as it enters the fusor. I have about 8 of these blocks on my coax and 1 on the HV line. These blocks help prevent coax ground bounce bumps and squelch resonant "antenna" line noise on the shield. One can never establish a "good enough ground" around a fusor. Ferrite blocks are usually contained in a plastic clampable housing designed to be snapped together around a shielded cable near the cable's connector ends. Long cables in a noisy environment can have several along its length.

It is great that you have a scope for analysis. I use my DSO for such noise hunts regularly. Noise introduced after a preamp can usually be easily dealt with as it can be attenuated and the clean premap neutron pulses can be amplified over the noise and the noise discriminated out. However if noise gets into the preamp and is greater than the neutron pulses, your are sunk! I run my 23 inch long, 1 inch diameter, 3He tube at about 1600 volts. You might try boosting your voltage a bit to increase the neutron pulse amplitude over that of the noise pulses. Hopefully, doing this will increase the pulse amplitude without increasing the noise pulses. First try a few ferrite blocks on that short BNC cable between the tube and preamp.

This sounds line a ground issue, arcing or shield bounce noise. Use a little portable AM radio to search for noise sources. The fusor, itself, is typically noisy.

Richard Hull

Re: Detector noise problem

Posted: Fri Apr 20, 2018 7:02 am
by Dan Knapp
Sorry for not explaining the two traces. The blue trace is the preamp output, and the yellow trace is the count output from the SCA. On the noise trace, there is no count pulse because the threshold was set to reject the noise peak (which also blocked the neutron peaks). The neutron peak trace was captured with the SCA threshold set lower and hence had an accompanying SCA output peak on the yellow trace.

Re: Detector noise problem

Posted: Fri Apr 20, 2018 2:41 pm
by Peter Schmelcher
Looking for EMI is a witch hunt and your coupling could be M or E field or simply conducted.

If it were me I would float the preamp with a 9V battery and only connect the scope to begin the hunt.
If I could float the scope on batteries I would and also position it on a plastic shelf.
Then start introducing various shielding and grounding ideas and observe the results.

Don't overlook that all BNC RF connectors are not created equal. Nickle plating has bit me in the past. I always use silver or gold for both inner and outer contacts when I do low signal work.

Best of Luck
-Peter

Re: Detector noise problem

Posted: Fri Apr 20, 2018 4:13 pm
by Bruce Meagher
A few years ago I had noise issues with a large 2” x 36” He3 tube connected to Cermat 110 preamp (mounted on a Cermat 150 board and in their box). I had a 6’ coax cable between the preamp and the detector that was a significant factor. I switched to an HP 5554A preamp and a shorter cable and got much better results. I didn’t really fully investigate the issue because I was more interested in the neutrons.

A couple things to potentially consider:

Try directly connecting the preamp to the detector with just a high quality connector. Switch out the input connector on the Cermat 150 box to a better quality one (assuming you are using their preamp board and box). Trying a different preamp brand. Run the NIM bin, HV power supply, preamp, and scope on a UPS power supply disconnected from the mains. Also have you tried capping (not terminated) the end of the BNC cable to see if the noise is still present?

Good luck!

Bruce

Re: Detector noise problem

Posted: Fri Apr 20, 2018 10:29 pm
by Dan Knapp
We’ve tried running the preamp with battery power, attaching the preamp to the tube with a double male bnc (quality of connector unknown, however), and capping the preamp input with a terminator (which eliminates the noise as does leaving the preamp input BNC open.
We also tried ferrites on the preamp to tube short coax and on the preamp output cable per Richard’s suggestions. We already had ferrites on the HV line.
We have another preamp on the way, and also plan to build one per an extensive discussion here several years ago. I’m wondering if we should try some different R values on the ballast to try to reduce the noise at its source.
A possible solution is to select the “fatter” neutron peaks with a pulse shape analyzer/single channel analyzer/time to amplitude converter/single channel analyzer, but we haven’t found any TAC NIM modules on eBay.
If and when we cure the problem, I’ll post the solution. Thanks again for the input, and we welcome any other ideas.

Re: Detector noise problem

Posted: Sat Apr 21, 2018 1:34 am
by Rich Feldman
Looks like there's a huge difference between the frequency content of the bad pulses and the good ones.
The bad ones ring at about 10 MHz, and the big spike is of comparable frequency.
The good ones would not change much if viewed with a probe and 'scope bandwidth of 100 kHz (3.5 us risetime).
Does your 'scope vertical channel menu have any bandwidth limit (digitally filtered) options?

Can you put a real lowpass filter in the path?
Not just some ferrite. Those are usually specified for loss at 100 MHz. Often chosen by designers in a trial and error process, until functional problems stop, and EMI regulatory compliance test sweeps pass with reasonable margins.

Details would depend on where the signal path is accessible, and the associated voltages and impedances. My personal experience with radiation detection instrumentation is rudimentary.

Here's one homebrew passive LPF: http://www.k5td.com/500kHz_Low_Pass_Filter.html
Here are some you can buy: https://kiwa-electronics.com/100-khz-lo ... ilter.html https://www.thorlabs.com/newgrouppage9. ... up_id=8613
I bet they're all much fancier than you need, in terms of flat passband and sharp cutoff. DC voltage ratings might be an issue.

Re: Detector noise problem

Posted: Sat Apr 21, 2018 5:11 am
by Peter Schmelcher
Try disconnecting the HV tube bias supply and cap or short the connection.

Re: Detector noise problem

Posted: Sat Apr 21, 2018 6:53 am
by Dan Knapp
The noise is still there with the detector HV bias turned off.
The low pass filter suggestion sounds like a possible solution. The neutron peaks have a fast rise time, so I’m not sure how they will be affected; but it’s an easy thing to try.
Thanks again for the comments.

Re: Detector noise problem

Posted: Sat Apr 21, 2018 5:11 pm
by John Futter
dont forget that input capacitance is directly proportional to noise
each input pf adds noise and limits bandwidth in a charge amplifier.
this seen in some ORTEC preamps where they try to match input capacity to the feedback capacitor ie they change the feedback cap with a switch.

the best match is when input cap = feedback cap

Re: Detector noise problem

Posted: Tue May 29, 2018 8:58 pm
by Dan Knapp
I had indicated I would post the solution when this problem was solved. It is now at least partially solved, so here it is.

Richard suggested operating the He-3 tube at higher voltage to increase the size of the neutron peaks versus the noise peaks. It turns out there is a published paper on this issue from the cold fusion era (perhaps the only useful product of cold fusion work?): T. Aoyama, et al., Radioisotopes 40:188-192, 1991 (copy attached). Cold fusion people were apparently claiming detection of spark noise peaks as detected neutrons. In this paper they generated sparks with an induction coil and showed the peaks heights did not change with He-3 tube bias whereas the neutron peaks increased, and that neutron peaks could be raised above the spark noise with higher bias voltage, as Richard suggested.

Rich Feldman suggested a low pass filter. We built a filter (circuit attached) designed to pass the neutron peaks and block the ringing noise peaks shown in the original post. It blocks the big noise peaks, but doesn't completely eliminate noise.

Peter Schmelcher suggested looking at the quality of the connectors being used. I replaced the signal input BNC on the preamp box with a silver plated panel mount HN connector that attaches directly to the He-3 tube eliminating two BNC connections. I also replaced the BNC bias voltage connector with an SHV connector. I replaced the Cremat supplied banana jacks for the preamp power connections with a DB-9 connector and made a shielded power cable to connect to the power jack on the back of the NIM Amp module.

Not directly related to the noise problem, but as another change from the original post, I made a new 200K / 40W ballast resistor by connecting eight 100K / 5W ceramic film resistors in a series parallel configuration mounted on a sheet of Lexan inside a piece of 1-1/2 inch PVC pipe.

Using all of these changes, I am now able to count neutron peaks that go away when the tube is removed from the moderator. I found that one problem I was having in trying to observe neutron peaks disappearing upon removal of the tube from the moderator cylinder is that I was seeing thermalized neutrons being reflected from my stack of plastic containers filled with boric acid solution (some may recall my earlier post on this shielding arrangement). Since it is obvious that I'm not producing enough neutrons to need shielding, I rolled the shield stack to the other side of the room, and this problem went away.

Operating the fusor at 25 kV, 8 ma, and 3 mTorr of deuterium, with the He-3 tube at 1500 V and 30 cm. from the center of the fusor, I see a count of 10.5 neutrons per second that goes to less than one per second upon removal of the tube from the moderator. There are still some noise peaks even with the low pass filter, which I block by raising the lower limit control on the SCA to eliminate all counts with the tube out of the moderator. It is apparent from the scope traces that this setting is also blocking smaller neutron peaks, so I still don't have an optimal solution. However, by counting peaks that go away upon removal of the tube from the moderator, I can be confident that the peaks I am counting are neutrons. Based upon a previous calibration of the He-3 tube located 30 cm from a Californium-252 neutron source, this count corresponds to an isotropic neutron production rate of 8500 per second. This calibration was done with a lower bias (mfr. recommendd 1100 volts) and a much lower limit on the SCA (just enough to block gammas). We need to do another calibration in situ with the same settings. It is expected that the calibration source will give fewer counts with the higher SCA lower limit, and thus we were actually producing more than 8500 n/s.

This neutron production rate is nothing to write home about, but at least we now have a positive control for neutron detection. With a power supply that goes to 60 kV / 60 ma, we should be able to generate a much larger neutron flux; but the system is very difficult to stabilize. The x-ray power supply is being controlled by the vendor's LabView interface, which is written to allow voltage adjustment only in 1 kV increments (OK for an X-ray supply but not for a fusor!). We only have the standalone LabView software interface without the source code, so I can't modify it for finer control. There is clearly a need for fine analog control of fusor voltage. With this crude control of voltage, the system has a tendency to go into runaway mode with grid overheating and current spiking to overload shutdown of the supply. At present we’ve not been able to get above 25 kV without the system going into runaway and overheating the cathode.

This experience of trying to quickly cobble together a positive control for neutron detection has given me much appreciation of the fine art of tuning a fusor, and the need for fine control of voltage. I have a couple of Spellman -50 kV / 4 ma supplies that I can parallel to get 8 ma with fine analog control of voltage, but this current is at the edge of adequate. With some modifications, I should be able to near double this current output, so it looks like this will be the next step.

Hopefully, this description will be of use to others using He-3 detectors. Clearly, accurate counting of neutrons with He-3 tubes is not a trivial undertaking (which may have been the root of cold fusion fiasco). Thanks to all who offered suggestions.

Re: Detector noise problem

Posted: Tue May 29, 2018 10:12 pm
by Richard Hull
Dan,

I am glad you found a workable solution and that the many comments helped. I printed out the nice paper on the 3He detector setup and noise elimination. I have harped for some time the need to set the discriminator level high enough to take care of local noise. The fusor kicks out a very strong and consistent neutron that is obvious on a scope to be rivaled only by cosmics which can't be removed. In general operating a 3He tube can go a good bit beyond its ideal voltage in spec sheets which is often a broad range, anyway, provided you have the original spec sheet.

You also begin to see, as so many other folks who stuck with fusion in a fusor, that amazing results require amazing control skills. Manual skills in control that lead to better and better results with a "feel" of operation of each fusor are often a learned function by a stable hand at the "wheel" with better than average feedback to the human by multiple measuring devices. These devices guide control.

A computer controlled system for a fusor might best be implemented by the person who has first mastered his fusor's "fine control" hysteresis characteristics, manually.

Great work and thanks again for the paper...

Richard Hull

Re: Detector noise problem

Posted: Wed May 30, 2018 6:23 am
by Dennis P Brown
In your low band pass filter circuit, what are the values of those resistors: Ohms, kilo-ohms or mega-ohms?

Re: Detector noise problem

Posted: Wed May 30, 2018 7:11 am
by Dan Knapp
Resistors are 100 ohms. Purpose is to make input and output impedances 50 ohms to match preamp output impedance and cables (amp module has high input impedance).

In rereading my post, I realized I failed to thank all who had made suggestions. Thanks for your input on this.