Sound Card Spectrometry

[Steven Sesselmann]

Hi Guys,

Sound card spectrometry is moving forward in leaps and bounds, it's almost becoming a competition between the software authors, who can write the best software, and we are the winners...

Marek Dolleiser has just released PRA version 6.0.0, now improved with simpler energy calibration and built in oscilloscope screen.

The Theremino team have just released their Theremino MCA version 2.11, which can now match the resolution of PRA and now has energy calibration with isotope identification both for Gamma and XRF.

Can't wait to see where this will end...

Steven

Software download links on my site..
http://www.gammaspectacular.com/software-downloads

[aka47]

Steve

Yes indeed, I was rummaging through the 3 identified on your site last night in fact. (As well as your site, nice work)

For my self and for my other life I am looking for similar but on Linux. As neither I nor the SCADA I am working on in my other life use windows at all.

Death before dishonour...

At the moment though there does not appear to be much that is linux based and as targeted/focused to the task.

I am sort of torn between writing a Linux equiv (targeted to the Raspberry Pi initially) or offering to contribute effort towards porting one of the leading 3.

As you suggested Theremino is looking pretty good at the moment.

For porting any of the leading 3 though it means hacking my way through windows code and maintaining portability. See above re dishonour.

If going any of the above ways I could do much worse than to cherry pick some of the code base.

Another option I have been considering is looking into what I can make out of CERN's ROOT analysis package. Which fortunately is Linux ready.

Given that I need to tackle ROOT in my other life anyway, this is looking promising as well. If only to reduce how thinly I am already spread.

Whichever way, anything I produce, like all else I do, will be published Open Source.

The Raspberry Pi also has various bits of exposed IO, so if I need to I can add on some signal processing and custom DtoA instead of using the sound subsystem. Pi's are also very cheap for a full Linux system to be dedicated to the task.

[Steven Sesselmann]

Andy,

Linux is no problem, both PRA and Theremino MCA 2.11 will run under Wine on Linux.

Steven

[aka47]

See previous re Death & Dishonour. <|;)

I wouldn’t waste precious processing cycles on windows much less emulate it.

I made a conscious decision a long time ago to actively use, contribute to and promote Open Source Hardware and Software.

It is pretty much a personal decision and everyone else is welcome to pay their M$ taxes if they so wish.

Wine on the Raspberry Pi is a bit of a challenge due to lack of resources.

[Doug Coulter]

I am, among other things, a Linux guy too - I made my fortune fixing windows, then ditched it as soon as I didn't have to have it around.

I've been looking at this, but feel the serious limitations of soundcard bandwidth are going to limit its usefulness. You really need an analog/digital hardware front end "peak picker".

For example, my NaI "gallon jjug" sensor typically produces mhz count rates when a medium hot cal sample is put near it. No way you're doing that with a soundcard, and if I wanted to measure my fusor spectrum - 100 times as hot anyway...no go for certain.

If you have a hardware front end that can:
Peak detect above a threshold.
Hold that peak, ignoring further input till that one is digitized.
Only turn on again when signal below threshold (so you don't catch the "peak" of the back half of some random pulse)

Then you can do something with a DC coupled a/d (no soundcard is), and it doesn't have to be super fast - it can just throw away some of the peaks (as long as that is random, more or less, and here it is) and still build up a good histogram.

At that point, you can use most small dev boards - a pi might be overkill, and you could use an arduino, a pic, you name it, and spit histogram data back to the pc at a sane data rate.

I'm holding out for that - something that can be actually useful with strong sources. If I get my round-tuit, I'll make such a device, open source of course.

Right now, I 'm doing other data aq with pics and usb interfaces. Output data is human-readable ascii, so it works with any PC opsys. My fancy graphics at present work mainly in linux, but if you are willing to put in perl, some CPAN modules, gnuplot etc, it'll run anyplace.

I would use the same plan for something like this. Every so often the device spits out a histogram, which could be accumulated in the PC for higher dynamic range (say, spit it out everytime you're about to overflow the highest bin).

My money is already where my mouth is. If you go to my site, http://www.coultersmithing.com/forums/index.php and search for things like sdaq or data aq...you'll find some freebies I've done already, from standard calibrated geiger counters to general data aq.

But this problem needs more than even a pro audio fast-sampling card can deliver, if it's to be useful to me - I'm less interested in counting things at near-background than looking at activations from the fusor and such - far too high count rates for any audio card to work with. Too much data is lost before you can apply PC smarts to it with one at high count rates.

[Carl Willis]

The Raspberry Pi has no audio input, no ADCs on board at all, and you would have to come up with your own hardware to accomplish the major task of the spectrometer. It would seem to be an odd choice for a spectroscopy platform based on what I know about it, unless of course the entire point of the exercise is the ol' linux purity crusade. Some MCA tools have been developed for small platforms, e.g. DSO Nano and the iPhone, that do have ADCs on board.

The most comprehensive discussions about these soundcard-based spectrometers seem to be on George Dowell's Yahoo! group at

http://tech.groups.yahoo.com/group/GammaSpectrometry

Well worth joining.

-Carl

[aka47]

Doug

I am pretty simlar, must admit though Coprorate IT Consulting for M$ products never made me rich. I can't say the same for my ex employers though. But it did keep me alive.

On Detectoring and Spectrometry, your thinking is pretty much the same as my own.

The raspberry Pi has a small GPIO port with some Digital IO and a SPI port.

It does also have, SD Card, USB, HDMI graphics, sound and Ethernet already on board.

So network/internet keyboard, mouse, monitor and Linux is a done deal.

It also has a port on it already for an LCD touch screen so moving towards a handheld spectrometer later is doable.

Alternatively if the SPI is'nt fast enough there is a serial web cam interface as well that comes under control of the GPU.

Altogether though I am heading the route of a signal processing card to connect via the GPIO and do feature detection, as you sugested to reduce the badnwidth requirment.

Quite what all the feature detection will entail is as yet up in the air.

It certainly will have a buffer amp, bipolar signal route, unipolar signal route, zero crossing detection on the the bipolar route and peak detection on the unipolar route.

There is also likely to be some thresholding (Nuclear folk call it discrimination) and gating on one or both routes.

How I combine this to get useful feature extraction is up in the air also. There is certainly room for programmable logic (CPLD) not sure if I will need FPGA though. Probably a reasonably fast SPI ADC with good resolution. DC coupled as you sugest.

PIC's are pretty good, I ditched them a few years back for the ease of arduino and AVR. The Xmegas have some interesting event driven circuitry for their peripherals, but I dont think anything realy fast is needed if as you pointed out the front end does the quick stuff.

Purely as an aside the latest Arduino is using an ARM cored AVR one of the SAM9 series.

Detector dead time combined with pulse pile-up puts an upper limit on the rate of single detector counting anyway.

For detectors at the moment I am principaly looking at Scinitilation detectors as it has a certain amount of crossover with my other life.

[aka47]

Carl

Thanks for that I will have a rumage through.

As in last post I don't need a Pi to have analogue IO or sound input.

It's interesting to note the speed with which Linux is dismissed as being elitist yet exclusive use of M$ and Apple is not.

[Steven Sesselmann]

Doug,

You have been a constant adversory against soundcard spectrometry, but have you even tried it?

Yes, it won't do Mhz count rates, but why can't you just back off the source a bit if the source is too hot?

Attached: Cs-137 spectrum taken with a sound card in 300 seconds, showing a 6.3% resolution.

Detector: Bicron NaI(Ti) 1.5" x 2.5"
Driver: GS-1100A
MCA: Theremino MCA 2.10

Steven

[Edward Miller]

We should be able to build off a xylinx platform a much higher audio sampler. What is the target performance?

[aka47]

As this is more my area of research than Fusors per ce.

I can honestly say I have an interest in both. I also recognose the differences and in some cases limitations of each.

My personnal take on both of these methods is that they are to a degree complimentary, so glib comparisons are likely to be flawed. It is a case of horses and courses.

F1 Cars and SUV's are very differrent animals but they are both cars, have 4 wheels and take you somewhere faster than walking. Claiming one is better than the other is very much missing the point. It depends on exactly what you want to do with it. F1 off road is as useful as a chocolate teapot, SUV on a race track will never win anything. etc etc etc.

The soundcard method clearly lowers the bar to entry ($$$) and if carefully matched to detector type and the limitaitons of it's capabilities it is certainly a very useful tool. I guess it is also a realy good teaching tool for gamma spectrometry. For well matched tasks it is perfectly good enough

From a technical signal processing point of view it has it's limitations (which if you make sure you are using it within, will not cause you any problems) There's plenty of volumes been writen and plenty of practical experience and engineered examples that back up the the volumes that have been writen on the same subject. If you want a quick hit, Nyquists sampling theorems etc are a good place to start. Being AC coupled the input is effectively diferentiated only, there is no chance of measuring levels only rate of change.

Having said the above if I am a detection experimentalist wanting to evaluate a wide range of detectors and as yet unidentified and unquantified sources under a wide range of conditions. Some of which, I won't know what they are until I have got there, because that is research and experimentation. Then the limitations of the sound card method are something I might quite rightly seek to avoid. I am going to select a method that for the effort I am prepared to put in will give me maximum flexilbility.

This perhaps is why NIM Bins are so enduringly popular as instrumentation methods as they give that degree of freedom to reconfigure and deal with unknowns. They are'nt exactly portable though and there is an argument that they encompas a lot of technology which is now well past it's prime (But still useful if you dont need to find components to fix it).

The DC coupled fast front end route then, offers greater flexibility and will give quality results across a wider range of sources and detectors. But takes much more effort to do and will be potentialy more expensive. Being DC coupled it will measure rate of change and constant levels. It will also do this at higher frequencies than the bandwidth limitations imposed by the Soundcard Input. Which incidentaly is preprocesed analogue style ( filtering etc ) as that is it's intended purpose.

In comparing the front ends from an electronics and signal processing point of view, the DC method (if implemented well) will give rise and fall times that are a more faithful reproduction of the input signal. The sound card input physicaly cannot see these at all. This is before getting to the discusion of pulse repetition rates, noise pulse supresion and pulse-pileup reduction strategys.

Re- fusors they are a great (and slightly cool, OK realy cool) source of detectables and are capable of generating more detectables than they put out in their own right.

If I am only building a fusor then what I want to detect is already fairly well defined, Being a known spec helps me match this to Sound Card MCA techniques and the correct detection method. Giving good results and being an indispensible tool.

If I am building a fusor as a source of detectables and detectables generator then I am off that specification and out of that teritory completely.

Guys your all clever, and doing clever things. Arguing whether an F1 car is better than an SUV is a bit pointless.

Personaly I want to hear about the technology that goes into both and cherry pick it for my own projects.

[Steven Sesselmann]

Andy I agree with your comments...

The sound card spectrometry is proving very useful in Japan, where people are identifying trace amounts of Cs-134 and Cs137 in their food. Using a good NaI detector with lead shielding, the count rate is low and the system when used with PRA or the japanese equivalent Bekumoni can easily detect trace amounts of Caesium.

On the other hand, I know a scientist working on a pulsed ground penetrating neutron source, who needs a spectrometry system capable of mega counts per second, to identify minerals in the ground, clearly one system can't do the same thing.

Steven

[aka47]

Steve

> The sound card spectrometry is proving very useful in Japan, where people are identifying trace amounts of Cs-134 and Cs137 in their food. Using a good NaI detector with lead shielding, the count rate is low and the system when used with PRA or the japanese equivalent Bekemoni can easily detect trace amounts of Caesium.

Yes I have been researching into detector technology for a quite a bit now. I think there is a pressing need for Opensource detectors (That will work well with Opensource MCA's) and use commodity (Hardware store+ availability) material in their construction. They should be implemented in such a way as to be largely patent free. I don't have much time for ambulance chasers on the make.

The Bust then Boom of the GM Tube and Detector markets courtesy of Fukushima reinforces the point. Crap tubes and antique instrumentation of dubious functionality are currently selling for ridiculous prices.

NaI(Tl) Crystals are good but relatively expensive and heading slowly towards unobtanium. Similarly 3H. I thin this will get more pronounced as semiconductor detectors gain acceptance.

They really can not be described as cheap (even those that are less expensive than their rivals) I just attended a demo by Kromec and their little CZT room temp offering, nice unit, can't afford it on my pocket money.

Whilst Fukushima has eclipsed the fall out from Chernobyl it is still spread all over northern europe in sufficient quantities to turn up in farmed and foraged foods.

Putting aside the above though and getting a little more back on topic. A Detection tool-kit for Fusor constructors that covers the well specified minimum for safety and successful building, may have some mileage in it too. If only as an exemplar. I see the sound card spectrometry as a definite component of that. From my own point of view the detectors will be home constructable and not need anything too exotic (Nuclear folk are a bit too kinky for exotic materials, perhaps with reason).

> On the other hand, I know a scientist working on a pulsed ground penetrating neutron source, who needs a spectrometry system capable of mega counts per second, to identify minerals in the ground, clearly one system can't do the same thing.

Agreed.

Pulsed Neutron work is something else again, some of the demands are readily deducible and there are a bunch more to discover. The relative low average flux and the need to pickup and derive meaning at high speeds brings lots of interesting challenges. Depending on what your field of application is and how short your pulses are has a big impact on this.

Back onto fusor related topics. having had a look at the PSU designs in use I am curious as to how much the fusor builds are pulsing due to ripple. (If at all) I would be interested to hear from folk who have measurements of the PSU ripple and can instrument to the level of seeing if the neutron flux has a matching ripple. Although it may be greater or maybe lesser...

[Richard Hull]

The power supply on neutron counters, whether Nim or home built is not pushing ripple into the counts for the simple reason that the discriminator level setting will not allow it. Most Nim and other supplies like PMT supplies are really well regulated and filtered.

On neutron counters, the discriminator level is easily set to where no detections occur at all and then back set to ~5 to 20 cpm which is a normal range of cosmics which depends on the NV of the detector used. Ripple busting through from the supply just can supply that low of a count. Thus, ripple is aced out, upfront.

The PS draw from even a flooded tube would be tiny, perhaps 200ua. That much current ruipple filtering is easy to eliminate and already accomplished in NIM and other PMT supplies used by amateurs.

Those with real gear in hand and using it, understand the above.

Richard Hull

[aka47]

Richard I fear we are talking at cross purposes.

"Back onto fusor related topics. having had a look at the PSU designs in use I am curious as to how much the fusor builds are pulsing due to ripple"

I think makes it pretty clear I was refering to the Fusor power supply. Not the PMT or Detector supply.

The last statement :-

"Those with real gear in hand and using it, understand the above."

Tells me that probably based on the above we are also jumpint o incorrect conclusions.

Is there a sub text here I need to be aware of ??

[Carl Willis]

The sound card spectrometry thread began as a nice little self-contained thread, but the discussion is off on one or more tangents now. That probably explains the confusion.

As regards the neutron source rate varying with voltage ripple on the fusor supply, there is a good thread started by Jon Rosenstiel in which he shows his data on that subject. At one time, unfiltered, full-wave-rectified x-ray power supplies were the dominant source of high voltage for neutron-generating fusors, and so there was considerable ripple in the neutron yield. It's probably three or four years old now, but it that thread is here and it is the place to learn about that issue.

-Carl

[aka47]

Carl Willis wrote:
> The sound card spectrometry thread began as a nice little self-contained thread, but the discussion is off on one or more tangents now. That probably explains the confusion.

Agreed

> As regards the neutron source rate varying with voltage ripple on the fusor supply, there is a good thread started by Jon Rosenstiel in which he shows his data on that subject. At one time, unfiltered, full-wave-rectified x-ray power supplies were the dominant source of high voltage for neutron-generating fusors, and so there was considerable ripple in the neutron yield. It's probably three or four years old now, but it that thread is here and it is the place to learn about that issue.
>
> -Carl

Carl, thanks for the pointer. I will have a nosey. I like a bunch of what Jon has done. (That I have read about so far)

[Peter Schmelcher]

Slightly off topic. I have been considering the $59 USB starter kit version of the DI-149 for voltage, current, and pressure logging http://www.dataq.com/products/startkit/di149.htm, however, this device might also be a reasonable alternative input device for spectrometry. The free but rate restricted windows software limits the A to D sampling to 240Hz, however, the device is capable of 11,000 samples per second if you write your own windows application or purchase the $195 unlock code. Still the A to D is rather slow for a multi channel analyzer so a pulse stretcher is needed. A secondary charge amp with a long time constant would probably do the trick.

Alternatively, if the detected pulse is converted into a purely digital pulse using a comparator, the device also has a total pulse count input pin and a rate input pin that would likely be good for counting thermal neutrons.

Peter