Home built mass spectrometer, continued

[Richard Hull]

Nice work! The dedication to the effort paid off. You have the right to be quite proud of your efforts. It has been a joy to follow your build and commentary all along the way.

Richard Hull

[Chuck Sherwood]

This is a trace with some nitrogen tracer gas on the dual sector to show increased resolution over the single sector shown in previous traces.

Interstage slot is 0.165. System pressure is 2E-5 Torr and emission current is 150uA. One volt is 10 pA so each division is 20 pA on the first trace and 10pA on the second trace.
The ramp shows the acceleration voltage.

Large N2 peak at the second division. Very large H2O peak just after the center line and a small N1 peak on the right hand side.

Of particular interest is the H2O peak and the aberrations associated with it. Here is only the H2O.

Notice the shelf on the right side that was all blurred together in the single deflection unit. Amazing, this shelf is actually AMU 17 created by knocking a hydrogen atom off the water molecule. This is explained at jeoluse.com. Full link below.

https://www.jeolusa.com/RESOURCES/Analy ... try-Basics

Note my scans are increasing voltage which results in heavy ions on the left and light ions on the right so things are reversed from the Jeol graph. You can even see the small response to oxygen O1, AMU 16, on the second division after the centerline. You can also see the AMU 19 and 20 response distorting the beginning of the H2O peak. Personally I find this quite amazing and it would be even better with stronger magnets. Why, the stronger magnets require more acceleration voltage increasing the voltage difference between ions.

[Chuck Sherwood]

I wanted to test the single sector at high mass units. One eventual goal was to look for radon at AMU 222. I decided to try mercury vapor at AMU 200. There are some concerns using mercury but I went ahead anyway. More on this later. Initial calculations indicate my electromagnet can make 12-13 thousand Gauss at 3 amps. I had no data at higher currents.

I cut a 1 inch glass test tube so I could connect it to the vacuum system with a compression fitting as show here. The test tube is vertical even if it shows up otherwise.

A trace developed quickly as the pressure went down showing that Hg truly does produce vapor at low pressure and room temperature.
Here is a trace with the magnet current at 3 amps and another at 5 amps. Also a simple graph showing the calculated magnetic field based on the atomic mass at 200 and the acceleration voltage.

The calculated field strength at 5 amps is 30K gauss which is quite extraordinary. Once again this is only at the center of the core.

Some interesting observations. One is the peak is very broad. This is explained by the fact that Hg has many isotopes ranging from 196 to 204 AMU where the most abundant one is 202 at 29.7%.

Another observation was that after removing the mercury drop and cleaning the glass test tube, I still saw a response of 5 pA compared to the 50 pA response shown on the traces. This indicated there was obviously some Hg left behind in the plumbing. I disassembled and cleaned the magnetic sector and the associated vacuum plumbing to get rid of the residue. I do not think I will use Hg again.

For the next test I put five old watch hands in the test tube that contain radium. My hope was the decay would include radon and I could detect it. There was no significant response.

This has been interesting and educational, but I'm running out of idea. Unless I can find a way to measure the magnetic field and set it to a known value the electromagnet is of limited use because hysteresis makes the exact value unknown and therefore its impossible to know exactly what is present.

[Richard Hull]

The introduction of mercury into a vacuum system is the end of that system for many uses.
Along time friend use to do all manner of glow discharge systems from neon to special apps tubes.
He told me he had a separate vacuum system for tubes that used mercury in them. Once used in a system that system retains traces nearly forever.
He had a separate system for deep pumping in another room. He kept his mercury out doors in a shed well stoppered. He was paranoid about mercury vapor in his home getting into his super clean vacuum system.

Mercury places a hex on a vacuum system unless that system will forever be used to work with it. Thankfully such systems never almost never contain a secondary pump and are very simple to construct and use. Technical vacuums are good enough.

Richard Hull

[Chuck Sherwood]

All previous scans have been done sweeping the acceleration voltage while holding the magnetic field constant. This test does the opposite. Keeping the acceleration voltage high improves resolution. There are issues associated with this, primarily magnet hysteresis which makes it hard to determine the exact magnetic field. Regardless, it is very repeatable.

Here is a trace with the acceleration voltage held at 200 volts and the magnet current sweep from 0-3 amps which changes the field from 0-12K Gauss. Hysteresis adds 10% uncertainty. The emission current is 200 uA and the system pressure is low 6s.

I think the tiny peak on the left side is H2. I am certain the center saturated peak is H2O. The double peak is H2, O2. I think the next peak is argon and the last peak is CO2. This fits the graphs I previously made but I there is no reason argon should be present. I need to run some tracer gases to verify things. The bottom sloping line is the input voltage to the power supply that drives the electromagnet which gives an estimate for the magnetic field.

Here is the lab setup. The HP power supply by the turbo pump has been "strapped" to use an external voltage source to control the output. This power supply powers the electromagnet. The very old Wavetek function generator outputs a ramp into the HP power supply. It has been adjusted for a very slow ramp taking 18 seconds from min to max. I love old equipment because it can usually be configured for odd lab setups like this.

I find magnetic sweep quite promising for better resolution.
Cheers
Chuck

[Jan Ohlsson]

I have followed Chuck`s achievements, which are great, but have not posted anything on my system for a long time.
I have cleaned up my electronics with proper PCB:s, and made a new analog ramp generator for less noise in the plots. The old DDS waveform generator was very noisy.

Bottom left is the arduino wit shields for analog in/out and relays. Bottom right is the analog ramp generator. Top is the PCB for the DC/DC converters for the high voltages for anode and sweep, with adjustments for gain and offset. The electronics is more dependable now, and easier to adjust and service.

Here are a few spectrograms:

Left is hydrogen atoms, then hydrogen molecules and to the right is injected helium. These peaks separate very well.

This is after just turning on the ion source filament. Oxygen atoms are visible at 390 on the X-axis, it is a fragment of water molecules. Oxygen molecules are also visible at 590. Nitrogen is the almost buried peak at 530. The other hardly visible peak at 430 is also a water fragment, OH.

The peak at 730 is argon leaked into the system. The peak at 590 is oxygen molecules, and the badly resolved peaks to the left of this is water fragments.

The last spectrogram is propane. There are three clusters, the last very weak. The first is fragments with one carbon atom, then two, then three. Inside the clusters there should be a lot of peaks with different numbers of hydrogen atoms, but as those are only separated by one mass unit, my spectrometer is not able to resolve them. So the spectrometer is good for seeing whats inside my system with respect to the air gases, but not for analysis of any organic substances. That would require a larger magnetic sector. And the AMU limit for my system is at 44 AMU, that is the peak for carbon dioxide.

[Richard Hull]

This project is on the order of difficulty with a fusor! Great work.

Richard Hull

[Dennis P Brown]

Your spectrometer seems to work better - less noise, cleaner peaks - than the commercial unit I use to use at work. Really a great build on your part.

[Chuck Sherwood]

In the beginning of this project I used a varying high voltage to accelerate ions through the magnetic sector. Plotting the ion current vs acceleration voltage produce a mass spectrum. This is the simplest approach but it has limited range and resolution. The sensitivity also varies and is much higher at higher voltages. So I embarked on a design to scan using a changing magnetic field. The beginnings of this endeavor are shown in some previous posts. The next several posts will show how I proceeded with the design and build.

One of the most annoying artifacts of a continuous running slow sweep is waiting for a trace to start after you changed some parameters. This might require waiting many seconds for the previous sweep to complete and the scope to trigger. Therefore a single sweep option that started on demand was a "must have". The circuit shown here uses some basic concepts and components to build a sweep generator that can run continuously or in single mode.

The heart of a ramp generator is a current source to charge a capacitor. The voltage across the capacitor is buffered and provides a linear ramp which is the control voltage. A cascode current source is created by Q1 and Q2 which charges C1. The voltage across C1 is buffered by U1, which becomes the linear ramp. When the voltage across C1 reaches 10 volts it triggers a comparator (U2) to reset the circuit. In continuous cycle mode, U4 discharges the timing cap, C1, via Q3 and R5, and holds it discharged for a short time. In single sweep mode, U3 is used to hold the capacitor discharged until a new start signal is received. The schematic below shows all the details. This is all pretty basic stuff so I won't get too deep into the explanation. The current source is covered very well in "The Art of Electronics".

The range and offset is controlled by the following circuit. A key point is the caps across the feedback resistors creating low pass filters. They have no affect on the slow increasing ramp but slow the fall time so the magnetic field does not collapse rapidly creating a large voltage spike.

There are two desirable modifications to this design. One is to invert the magnetic field output and scale it to match the calculated output. The reason for this will be shown later. This can be achieved with a single op-amp with adjustable gain of approximately 3/4. Second is to make the time between sweeps somewhat adustable. This ensures the scope is ready and waiting for the next trigger pulse if the sweep time is slightly longer than the scope timebase. Adding a pot in series with R11 would accomplish this.
Continued in the next post:

[Chuck Sherwood]

The problem with using a simple signal generator and a DC power supply to to control an electromagnet is hysteresis. You just never really know the exact magnetic field. My solution was to measure the magnetic field with a hall sensor. I found some very good surplus ones on ebay (SS94A2D) with a sensitivity of 1mV per Gauss. They have a linearity spec of 2% over the range of -2500 to +2500G when operated at 8 volts. I determined they can operate up to about 4000 gauss in the positive direction when operated at 10 volts.

The hall output is buffered by U1 and level shifted from mid power supply to 0 volts. U2 is used to bring the 0-4 volt range (4k gauss) up to 10 volts to match the ramp generator. Both amps have caps across the feedback resistors to form low pass filters for noise reduction.

The last stage is the error amplifier / integrator. It ensures the magnetic field matches the control signal and compensates for magnet hysteresis. It is very slow because of the cardinal rule; never apply error correction faster than the system can respond. Originally the output of U3 was fed directly to the power amp. When I started using a commercial Bipolar power supplly, I added R26 to reduce the gain and insure U3 was not overloaded.

The Hall sensors are ratiometric to supply voltage and are therefore powered by a dedicated 10 volt voltage power supply built from a MIC2951 IC to minimize drift and noise. Note that the zero adustment on U1 is powered by the same 10 volt reference.
The Magnetic field output was overlooked at initial design and I would have preferred to invert it. Presently I just invert the scope channel.

10v2.JPG (15.42 KiB) Viewed 452 times

There is one problem with this design discovered during testing on the system. The indicated magnetic field does not match the calculated magnetic field required to deflect the ions. See graph below obtained by measuring ion currents of known gasses at measured magnetic fields. For this test the acceleration voltage was set to 250. I believe the discrepancy is due to the fact the hall sensor is right next to the steel pole of the magnet and the ions are in the center of the gap between the poles. The gap is about 0.57 inches. The only correction idea I have at this point is to attenuate the hall output to the scope so they are closer.

The calculated magnetic field is derived from this formula. Derivations can be found in a number of text books.
B = 144/Rn * square root (mE)
R is radius which is 5 cm, n = charge on the ion, which is hopefully 1.
m = mass of ions in AMU, E is acceleration volt which is set to 250

Continues in next post:

[Chuck Sherwood]

Initially I designed my own power amp to drive the magnet and this was the final circuit. It is designed around a high voltage op-amp and the application notes than came with it. It works fine, but I never got around to buying or building power supplies to drive it and my goal was to make two for the dual deflection unit. This was used for a lot of my initial testing, but was replaced with a commercial bipolar power supply called a BOP.

Many of the magnetic field calculations shown in previous posts didn't make sense and I sat out to figure out why. (I discovered I made some stupid algebra errors) Using this setup, I mapped the magnet current to the measured field. The magnet makes about 1000 gauss at 3/4 amp and this seems to scale well up to 4000 gauss which is where the sensor saturates. Initially I was going to attempt to operate the magnet at higher levels but presently the sensor sets against the south face and the brass sector fills the rest of the gap.

Here is a trace of the control circuits in action. Yellow is the control signal and purple is the magnetic field from the hall sensor.
Blue is the error signal and light blue is the magnet voltage. Of particular interest is the peak on the light blue signal as the magnetic field collapses and the energy is discharged into the filter caps via the protection diodes. I lost a couple parts without the diodes.

In order to control the collapsing magnetic field, I started experimenting with shorted turns, much like is done with some relays. My results indicate a few shorted turns really limit the collapse of the field so put a copper band around each coil.

Below is a picture of the coil voltage when driven by a relay with the shorted turn in place.
Yellow is the coil voltage and purple is the magnetic field measured by the hall sensor. The shorted turn is very good at controlling the kickback at high flux levels but eventually decouples and the magnet kicks still producing a significant voltage spike but most of the energy is gone and this smallish spike is easily absorbed by the catch diodes and passed into the power supply filter capacitors.

Here is a picture of the test setup of all the various boards and the power amp.

I bought several books on mass specs that focused on magnetic sectors. All discussions and equations show the magnet shaped to match the deflection angle of the sector. There are a few discussions on changing the angles to achieve secondary focusing. Most of this is beyond my knowledge base so I decided to machine the pole pieces to match the deflection angle as shown below. The original permanent magnet also had 60 degree angle. I believe this is an improvement over the round pole pieces. I think a little more needs to come off so that cores can fit deeper into the magnetic sector so the center of the core aligns with the center of the ion beam. For perspective, this round core measures 1.75 inches of 1010 steel.

This is the commercial bipolar power supply I am using to drive the magnet. It can drive plus or minus 36 volts at 5 amps with a bandwidth of 13KHz. I bought 4. Only one works. If anyone has a manual, for a Kepco BOP 36-5, I would love to get a copy. Apparently is was made in the mid 1970s.

The only significant system change associated with using the BOP is my original amp did not invert and the BOP is an inverting amplifier. This requires reversing the leads to the magnet. A cool feature of the BOP is you can disconnect the input and engage the reference input shown on the left upper corner and use it like a conventional power supply.

There are a number of mass scans in the "making hydrogen with a PEM cell thread" so I won't repeat them here. I can post more if people are interested.
cs

[Dennis P Brown]

Extremely good post and so informative. Keep posting, please.

[richnormand]

This is an impressive thread. Very well documented indeed.
Please keep active.
Cheers

[richnormand]

" If anyone has a manual, for a Kepco BOP 36-5, I would love to get a copy. "


There is a listing for the manual for the BOP 36-5 on ebay if you are interested:
https://www.ebay.com/itm/363645512201

Many other are listed on the Kepco website but not that one.

[Chuck Sherwood]

I have been in contact with Kepco and they provided a few pictures of key pages. But finding a complete manual is a huge advantage.

And on top of that they are in the next suburb, probably less than 10 miles away so I can swing by and pick it up and check out their inventory.

MANY MANY thanks
cs