The diagram to the right shows the Variable Frequency Oscillator part of the circuit for testing. The 74HC4060 is a 14-stage binary counter, with an internal oscillator driver. When an external tuned circuit is connected, a buffered oscillator output is available at pin 9, and various divide-by-2 outputs at the other pins. I used a small air-spaced variable capacitor, which I estimate to be 0-50pf. The toroid has 15 turns of thick copper wire. This toroid came from my junk box and I am unsure of its specifications.
CMOS based oscilators aren't particularly stable, but they are simple to build and this does at least provide a good test of the Huff & Puff stabiliser. I found that the whole thing was very sensitive to lengths of wire to the relays etc. The only solution to this is to keep everything as tidy as possible, and when in a receiver etc., apply generous shielding.
When a toroid is placed in a magnetic field, its permiability decreases, causing a decrease in inductance. A varying magnetic field can be generated by applying a varying voltage to a relay coil. In this way a correction can be applied to the VFO by the stabiliser circuit, with the relay coil and magnetic field replacing the conventional varicap diode which you see in all the usual Huff & Puff circuits.
I used a 12V open-frame relay, from which I removed the spring and moving contact arm. The relay must be placed in close proximity to the toroid, the exact position determined by trial and error for maximum effectiveness. The photograph below left shows the orientation which I found best. I did a test to determine how the frequency shift obtained varied with applied coil voltage. I decided not to apply more than 9 volts to the relay coil, since although it is rated for 12V, it seems to get alarmingly warm when run at 12V! This test is somewhat difficult and inaccurate to perform, because while the measurements are being taken the VFO can drift all over the place. It is best to be as quick as possible. For the frequency measurements, I used my Radio-Frequency Counter project.
The results (click above right) show a nice linear variation of frequency shift with applied coil voltage. According to my understanding of the theory, increased magnetic field reduces the permeability of the toroid, decreasing the inductance and hence increasing the VFO frequency. However, in my case I found the opposite to be true! I do not know why this is, if anyone knows please tell me.
27-Aug-08: Cor Eiff PE1GTV (Analog RF engineer) wrote to me with his theory for why the permeability of the toroid appears to be opposite to the expectation:
I think I have an answer why the tuning direction of the permeability steered toroidal core seems to go the wrong way ie frequency down instead of up.
The initial permeability of most ferrites is lowish, so the range in which the VCO oscillates [small amplitude] does use only the part of the curve around zero. As you force an external magnetic field on the core the permeability goes up, so the inductance goes up and the frequency of the VCO down.
Only with a very strong magnetic field [for most ferrites around 300mT] saturation sets is in and inductance goes down forcing the frequency up.
Also using the core in an oscillator with an higher amplitude will skip over the lower starting permeability, but using too much power is not so good for stability for which all was started..
Relay Coil amplifier and Driver
David White simply used a CA3140 to drive his relay coils, but they can only source 10mA. I found that my coil has a resistance of only 100 ohms so needs more current than this. Therefore I purchased a power op-amp type L2726. When it arrived I realised it was a surface mount device arrggh. I managed to fix it to a piece of PCB material, and it worked Ok. However, I decided I wanted to get maximum frequency shift, so drive the relay coil at more than the 0-5V of the Huff & Puff integrator. It is a 12V coil, after all. So I configured the driver op-amp as a x2 amplifier.
Unfortunately within a few seconds there was a puff of smoke, a blue spark, and a moderately loud bang that made me jump, making a fairly reasonable job of separating my bone structure from my skin. That L2726 had overheated. Which meant all my efforts to surface mount it were wasted. How are you supposed to heatsink a tiny surface mount chip anyway, it's hard enough just soldering to it.
Lesson 1: Check that you know what kind of package an IC comes in before you buy; Lesson 2: When making a circuit change apply power for a short time first and check the temperatures...
This left me without a driver circuit for the coil. I built a new one using a NE5534 op-amp followed by a BD131 power transistor. Unfortunately the input resistance of the NE5534 is quite low (100K), which places too much loading on the integrator circuit. You need to use an op-amp with an FET input. So I used a CA3140 I found in an old unfinished project (really a CA3240 which is a dual CA3140, and I used only one half). My final amplifier/driver circuit is pictured to the right here, it is built on a small piece of matrix board which is fixed to the main board by 4 wire legs.
To test that this circuit works correctly, I monitored the voltage across the relay coil, and applied a varying input voltage between 0 and +5 V, using a 1K potentiometer. Below left is the circuit of the coil driver, and on the right the voltage across the relay coil plotted against the input voltage. Beautifully linear...
The full circuit diagram of my VFO and magnetically coupled Huff & Puff stabiliser can be viewed by clicking here (1096 x 608, 57K). I included a +5V regulator on the board, with plenty of large electrolytic capacitors. The supply to the CMOS oscillator chip is additionally smoothed with a choke and capacitor. The output of the oscillator is buffered by an unused XOR gate to try to minimise any nasty effects arriving from the direction of the frequency counter.
I initially used an 80 MHz canned oscillator as my reference oscillator, but this poor device died at some point so I constructed a replacement 5 MHz oscillator using a crystal and 74LS04 chip (see below). Later I procured a 64 MHz canned oscillator, and used it to replace the 5 MHz one.
With the values shown, the VFO oscillates in the 40m band. The tuning range was about 6980 KHz to 7120 KHz. I haven't yet tried to see if I can acheive a stable output on higher frequencies. The step size of the output can be varied by choosing a different division ratio in the 74HC4060, by connecting a different Q output to my Q8. My calculated step size is 47 Hz but I think I have calculated something wrong because the observed step size is half that.
I varied Chas Fletcher's circuit slightly by using one half of the 74ACT74 dual D-type flip flop as a divide by 2 to optain equal mark/space ratio from the canned crystal oscillator. I think otherwise the performance of the stabiliser will suffer. The datasheet for the crystal oscillator specifies 60%/40% as the mark/space ratio. I also found that I could obtain the HEF4517 128-bit shift register easily (from RS components, rather than just use an 8-bit shift register.
I have included a switch to close the stabilisation loop. When the loop is open, the VFO drifts massively. When closed it locks almost instantly to the nearest lock point and stays there indefinitely. This switch is mainly included for the satisfaction obtained from seeing this happen and knowing the stabiliser is functioning so well.
The results obtained with this stabiliser are quite impressive. Even with such an unstable CMOS VFO, once the stabilser is switched on, it locks immediately and does not move more than 1 Hz in either direction. Long term after several hours, I have seen it move by only 4 Hz or so, and even that is most likely due to drift in my frequency counter rather than the stabiliser. (Ahem, the frequency counter does get rather warm due to its extreme miniaturisation...).
A moving coil meter displays the integrator voltage. I opened it up and drew a new scale, calibrated 0-5V. I also drilled some holes and inserted two 5mm high-brightness green LED's to illuminate the display. This looks VERY cool particularly at night with the lights out. These two LED's are driven from the +12V supply via 330-ohm resistors (top left of circuit diagram). The other LED (560-ohm series resistor) is a 3mm red LED mounted on the board to indicate power is on. This helps avoid soldering the circuit while inadvertently leaving the power switched on. That is something that can lead to nasty surprises considering the iron tip is earthed, and so is the 0V output from the power supply.
The integrator voltage can be set to just below its half-way point by a push button. Originally I noticed the VFO had more of a tendency to drift upwards, which is why I chose a reset point just below half way. However I think this was because the low input resistance of the NE5534 op-amp I was using at that time was dragging the integrator voltage upwards the whole time. After replacing it with a CA3140 this effect seems to have dissappeared. I obtained best results when I used a 39K integrator resistor, much lower than the 2.2M-ohms Chas used. I'm not sure why this is.
I built the circuit on a piece of single sided copper board. Some amateur's call this "ugly" construction, but I prefer the term used by the Huff & Puff guru, the late Klass Spaargaren PA0KSB: "Amateur Surface Mount Technology". Where necessary small fixing islands were made by glueing square copper board fragments. I fixed the tuning capacitor, integrator voltage meter and reset button to the board. In practice, when this stabiliser is used in a receiver, I will of course move these controls to the front panel. Additionally I'm sure plenty of screening would be highly beneficial.
Unstabilised VFO drift experiment
To test how much the 7 MHz VFO would drift when not stabilised, I opened the stabiliser loop and left it running for an hour, taking measurements at frequent intervals. The results are plotted on this chart, which shows drift of some 900 Hz in an hour Click here for a larger view (838 x 603, 107K). There is also a huge amount of short term variation that isn't shown on this chart. Every reading on the frequency counter is different by many 10's of Hz. As I said before, this kind of CMOS oscillator isn't necessarily inherently stable (I don't want to be rude to it).
A small Huff & Puff theory experiment
To test the Huff & Puff theory, I decided to see how the integrator output varies with the incoming frequency, without closing the stabilisation loop. Ok, that's not strictly true, in reality nothing was working and I didn't know why so I needed to find a way to debug it. It's very hard to test if the stabiliser is working properly when you have a VFO which is moving around all over the place. If the stabiliser is working the VFO stays dead on frequency, but otherwise there's little indication of what's wrong. This was when I was using the 5 MHz crystal as the reference oscillator.
I built a VXO (variable crystal oscillator) using a 6 MHz crystal and 74LS04, all from the junk box of course. The 5 Mhz crystal oscillator that I used before I got the 64 MHz canned one (see above) also used this same type of circuit, but without the 350pF tuning capacitor in series with the crystal to pull the frequency. This circuit is shown in the diagram below left. In this way I found I could vary the frequency from 5,997,386 Hz to 6,000,293 Hz, yet it was very stable so it was possible to find out what was happening in the stabiliser circuit. Or, what wasn't happening. Of course, the measurements below were taken when everything which should happen was happening.
I disconnected the VFO components from the 74HC4060's internal oscillator and connected the 6 MHz directly to pin 11, the clock input. I measured the integrator voltage using a digital multimeter. During this experiment I used a 3K9 integrator resistor, so that the level settled very fast, I took a large number of measurements across the tuning range. These are plotted above right.
You can see the classic variation in integrator voltage as expected (read the Huff & Puff articles!), which is the average level of the digital signal coming out of the XOR gate, whose mark/space ratio depends on the relationship between the VFO and reference oscillator. The stable points are the midpoints, i.e. 2.5V. The measurements become more widely spaced to the right of the plot, this is because 1) the tuning of the VXO was highly non-linear so it was difficult to make small adjustments at this end and 2) my patience was starting to run out.
Top View of the Huff & Puff Stabiliser board The VFO is at the bottom left. Top left is the 64 MHz canned crystal oscillator. Along the top is the power supply smoothing capacitors and +5V regulator. In the centre is the digital stabiliser circuit, and the relay coil driver amplifier. At the front of the board is the tuning capacitor and integrator voltage meter. Connections to the board for +12V, 0V and the Output frequency are via pads at the top left.
Another view from above In this view the crystal oscillator can isn't obscured by the flash reflection. I never claimed to be any kind of expert photographer.
Stabliser with Frequency counter Here you can see the stabiliser with my frequency counter. The counter is resting on top of a breadboard which I use extensively for prototyping parts of the circuit before soldering them properly on the board. At the top right you can see an ex-PC power supply which I am using. The light behind the frequency counter is a 12V car brakelight bulb. I have to connect it to the power supply otherwise the supply is insufficiently loaded and its protection circuit shuts it down.
From the front Here's a nice photo showing the integrator voltage meter, with my hand drawn calibrated dial for 0 to +5 volts. Notice the two green high brightness LED's which fit inside the case. Left of the meter is the tuning capacitor. Not exactly easy to tune to a specific frequency: in a receiver I'll add a reduction drive and a larger knob. To the right of the meter is the integrator reset button. You need to push it for a couple of seconds for the integrator discharge to occur and watch the needle return to the centre.
A darkened view showing the meter illumination With a little less lighting and by switching off the flash on the camera, I was able to obtain this photo which shows the integrator voltage meter nicely illuminated, with the frequency counter begind. It still looks a lot nicer in real life, but this is a little more like it. They'll make a photographer of me yet...
Stabiliser at Night! Here's my attempt with no light at all. So, it's not great but it's better than those "London at Night" postcards. I'm still trying to capture the magic of the illuminated meter... (and, in case you're wondering why the frequency readout is different in every picture: it's because the lack of shielding means that every time you touch the board to move it a little, the frequency moves miles. When you take your hand off, it locks to a different lock point).
At work on the Stabiliser A view of the workbench. Various chips and non-used components and my multimeter litter the area. Note the orange tubes of instant araldite epoxy resin glue, which I used to attach the board fragments to make connection islands in my amateur surface mount contruction of the stabiliser board. At the rear is a medium wave/long wave broadcast radio which belonged to my Father, he purchased it in 1976. On top is a lable saying "Repaired by Hans, 1988". I listen to it while working, and it's powered by the +5V output of the PC power supply. Certain oscillator frequencies in the stabiliser cause a lot of whistling in the radio.
At work on the next receiver module Here you see a view of the bench looking the other way. In the foreground is my polyphase network. This is the next part of the receiver under construction. Note again the PC power supply and brakelight in the background. The copydex glue was used to fix the new scale in the meter. In front of that, a fragment of board from my Fifth-method Huff & Puff stabiliser project, which I started years ago but never finished. I'll never finish it now with this superior version. So I didn't mind stealing from that board the CA3240 op-amp, and a 74HC4060 when I blew the one in this stabiliser by pushing +12V into one of its output pins...
The polyphase network A sneak preview of the 90-degree audio polyphase network mentioned above. The network has 8 stages, with resistance and capacitance values calculated to give zero insertion loss. It uses 0.1% tolerance resistors, and capacitors also matched to better than 0.1% by adding small capacitors from the junk box in parallel. This photo shouldn't really be in here, of course. Hopefully eventually if there is ever any acceleration in the slow pace of construction at G0UPL, the polyphase shift network will be completed and part of a receiver, which will deserve its own page!