Richard Wilkinson

I have had some success casting various objects from aluminium and now I am trying to cast a flywheel using the “lost foam” technique.

Showing the router with 6mm cutter which includes a top bearing. Below is a 3d printed pattern, from telford3dprinting.com, attached to a 10mm block of expanded polystyrene. The idea is the ball race bearing follows the pattern and cuts out the ‘foam’ object. The block was made from 3 layers of ceiling tiles bonded together.

The cutter makes quite a good job but there are a few holes from the ceiling tile pattern which are filled with hot wax.

Left showing lost foam patterns buried in sand, right Feeders and Risers before and after removal.

Showing my backyard casting setup. The Feeders and Risers are made from Plaster of Paris an idea from the “vegoilguy”

The generator was invented by Robert Van De Graaff in 1930. As can be seen it is made from a 568ml lager can and some plastic pipe fittings. The motor is from an old tape recorder and runs with an 8 volt PSU.

The top electrode is made from a 4mm brass rod and a glass tube from an old fuse. The lower electrode (right) is made from a 4mm brass rod which fits on to the motor and slides into a short length of PVC rod, the static charge is picked up using a short length of flex. The charge is transferred between the 2 electrodes using a 6″ half inch rubber band. It is a bit fiddly to get the band to run straight, so a few turns of PVC tape are added to the centre of the PVC rod.

The capacitor (*condenser) is made from 4 cans, 2 inside and 2 outside with a PVC tube 360mm long and 65mm diameter which gives a capacitance of 425pF. The spark gap is about 19mm so the voltage is around 20kV and takes about 25s to charge depending on the humidity.

• * I can see why they called it a condenser, the spark is a lovely thick blue colour but without it the spark is very thin! They were orginally called Leyden jars, and I think Marconi used the word ‘jar’ as a unit of capacitace.

More about the electrodes: Although the generator looks simple the theory is quite complicated. It works by making use of the triboelectric effect. Materials that gain -ve charges are Teflon, Silicon and PVC, materials that gain a +ve charge (ie tendency to give up electrons) are Air, Rabbit fur and Glass. Because the lower roller is removing electrons to ground the top roller becomes +ve and so does the “dome”.

In the future I would like to increase the capacitace and make some more ‘condensers’ but of course that would entail drinking more Lager!

I have been looking at the Oxford Electric Bell which has been running since 1840 using Zamboni piles*. There are 2 stacks each having a voltage of 2kV. These piles(cells) have a very high impedance and a voltage around one volt, so 4000 piles would be required. I have done some work trying to create these piles from zinc paper, MnO2 and wall paper paste, but so far I have not been too sucessfull! So as proof of concept I created a 4kV / 5uA supply from a Hartley Oscillator and Cockroft Walton multplier * 10 which only uses 5vdc at 10mA .

The ball of the pendulum is made from masking tape rolled into a ball and covered with aluminium tape, it is suspended from a cotton thread. The 2 contacts/plates are 5cm apart, this was calculated by trial and error… The principle is when the ball hits say the positive plate it picks up the 4kV charge and being of the same charge, it is repelled, the opposite happens at the negative plate.

The proof of concept seems to be good, so I now need to start making the Zamboni piles!

*There is a high probability that they are Zamboni but as yet nobody has opened them up to investigate.

Tests on the basic NRF24L01, 2.4GHz

• 

After buying a pair of these devices I had the usual problems with the PSU and shielding.

I decided to power the module from 5 volts which also powered the Arduino Nano.

The 5v was regulated down to 3.3v and 100uF tantalum caps were added for decoupling.

A standard sketch which sent a message between the 2 modules, was downloaded on each Nano. I made a slight change by using one of the nano’s pins to identify whether it was a transmitter or receiver. Also I connected an LED that indicated if the message had been received without error. They worked fine at low powers (-18,-16,-8dBm)  but when I tried 0dBm they stopped working. I made an RF shield from a piece of aluminium foil sandwiched between 2 pieces of card. This went under the nano, NRF and 3v regulator, this worked well.

Range tests:

The transmitter was placed about 20m above the ground and I walked away with the receiver until I lost the signal. With the NRF set to 0dBm I was able to get about 450m.

This was further than I had expected and I started to think how far I could get with the NRF24L01 + PA + LNA + rubber duck antenna.  This time I used Arduino Pro Minis to drive the NRFs.

• 

I had the same problems with RF shielding but this time I had to wrap the NRF in insulated aluminium foil, this worked up to +15dBm but +20dBm failed! In the end I made 2 aluminium boxes and connected a piece of braid from the box to the NRF, finally this worked.

To actually see the message I connected a FTDI serial interface to the ProMini then to a tablet via an OTG cable. Our highest hill in Shropshire is the Wrekin at 407m. I spent some time using various apps to see if I could get line of sight to the hill. I was looking for about 5km but found one of our friends had line of sight at a distance of 7.02km.  

• 

I set the transmitter up at our friends on a 2m pole and powered it with a small 5v PSU.

• 

• 

I set off to the Wrekin and when I got to the top I put the receiver on a 2m pole and connected it to the tablet using ‘Serial USB Terminal’ to my amazement it worked first time and I could see the string of characters being sent. The data was 100% and I did not notice any retransmissions. The path loss at 2.4GHz between the points is about 116dB.

The only failure was taking a screen shot of the data on the tablet, fortunately I did get a pic on my phone but not of very good quality. So I wonder how far they will go…

• 

• 

Early days

I hope to be able to ‘see’ the Hydrogen line at 1420MHz. I have a helical antenna pointing NE with an Elevation angle of about 80° and when the Milky Way comes over my location(I am using Stellarium) there should be an increase in noise, just how much I am not sure. I have written a Python app that samples the data from an RTLSDR dongle and converts it to RMS then to dBm. I do this every 10s then after say 1hour, plot a graph using pyplot.

The helical antenna feeds the 1st LNA,32dB(with perferations) which includes a SAW filter which I bought from gpio Labs. This connects to the BPF which feeds the 2nd LNA (SPF5189Z, 20dB) then to the RTL dongle. The dongle is set to maximum gain so the total gain is about 100dB. I chose a linear PSU to cut down on any RF hash from a SMPSU.

The RTL dongle uses a USB link which is connected via a 35m CAT5 cable to my laptop in the house. I use a StarTech USB extender at each end which works very well, if a little pricey! I did buy a £20 unit which did not work at all even on a short length of CAT5 cable, so I think on this occasion ‘cheap and cheerful’ does not work.

Antenna Gain (G) = 10.8+(10*log(nTurns * Space b/w coils))

This gives a calculated gain of about 30dBi, I think that might be a bit optimistic so closer to 25dBi, but this compares well with a 1m dish.

After some more researching I came across Michel Klaassen excelent site http://parac.eu/index.htm. I now understand the importance of averaging and a typical number would be between 50k to 100k. From what I can see he is ploting the doppler frequency shift of the Milky way as it passes ‘over head’

Each line represents a 5 minute 2MHz scan using the RTLSDR dongle, 288 lines make up the whole picture which takes 24 hours to complete.
From the plot it looks like about a -260kHz red shift from 1420.4 so using the formula v=df * c/f that gives 260,000 * 3*10^8/1420.40^6 = 55km/s

LNA problems

I now have 2 working LNAs, Nooelec and GPIOlabs. After some discussion I decided to shelve the helical antennas and try a dish with a coffee can feed horn. I think there were 2 (or more) things wrong:

I was feeding the 2 LNAs from the antenna, with 2m of coax, I now have the GPIOlabs LNA connected to the feed horn via a bias tee. Originally I bought it without the tee but noticed there was a surface mount pad connected to the output. After discussing it with GPIOLabs I added a 6uH inductor and that worked fine.

The feed horn is made from a 200mm * Ø 150mm coffee can with a ¼ wave probe that is ¼ wave from the base. It is connected to an SMA connector which connects to the coax cable. For the probe I used a narrow brass tube and inserted a piece of thick copper wire that I could slide in and out, trombone like, so as to tune it using my VNA-F V2.

This image was created with the dish pointing NW at an elevation of about 55 degrees. It correlates well with Stelarium as the arm went passed, also although I am not sure if it is relevant, the constalation of Cassiopeia was in view.

I have added another LNA, the GPIOlabs LNA and the Nooelec LANA next to the RTL, are connected directly to the 5 volt PSU; the second Noolelec is powered via a BiasTee module. (nb I am now not using the BiasTee on the GPIOlabs LNA)

The image above was created from the same data that produced the ‘wiggly line’ image but shows more detail. Also I am only displaying 1024 bins, y axis, so as to focus on the part of interest. I have swopped the axes so x now shows time in 5 minute intervals, so again this has taken 24hrs to produce. The time difference between the 2 blobs is about 8hrs.

I have now witten my own Python data creation program, it works in the same way as Michel Klaassen’s software but can be run without the use of SDR# and that allows me to create data automatically at any time. The 2048*128 sample from the SDR dongle is passed through matplotlib ‘power spectral density’ funtion 350 times, averaged by column, ‘convolved’ across the 2048 bins and written to file. Each file takes about 90 seconds to create.

Help from Megan Argo

• 

• 

• 

As mentioned in the diagram, getting the holes for the inlet port and outlet port in the right position was the most difficult part of the project. The flywheel and crank shaft wheel were made by melting aluminium cans then pouring them into green sand molds.