Below is a simulation of the band pass filter used at the input of the BitX20 receiver designed by Ashhar Farhan.
I assumed the input and output impedance is 50 ohm. I tried various capacitance values for best coverage from 14.0MHz to 14.35MHz (20m band). The best values I found and the filter response are shown below:
It can be noticed that at 14.0MHz the power loss is -16dB, while at 14.35 the power loss is -12dB.
Extending the simulation range, we can see there is another frequency where the filter shows low rejection.
After posting this subject on BitX20 yahoo group, I was expecting a bit more interest from the other people, after all, the topic of the group is BitX20. And the input filter is also essential for the parameters of the radio.
Anyway, I got some answers by email and I found out the filter used by Farhan for the input stage is called a “top C coupled band pass filter” or a “Capacitor coupled resonator BFP”. One of the posters explained to me the mysterios spike that appears around 15MHz (see http://hobbytronix.net/images/Electronics/Bitx20_Band_Pass_Filter.htm). It was purely because of my lack of experience and limited amount of time spent to “tune” the three trimmer capacitors during the simulation. I was pointed in the right direction that the resonator formed by capacitor C6 and inductor L2 is not tuned properly.
Next day I had a closer look at the freeware filter design software from AADE (Almost All Digital Electronics) and I found these types of filters as well (named Coupled Resonator – Butterworth, Chebyshev, etc). A filter designer software is an extremely powerful and useful tool. I decided to design such a filter using the middle of the band (14.175MHz) and set a 3dB bandwidth of 500kHz, ripple 0.5dB. My reason for a larger bandwidth (should have been 350kHz) was to have better band-pass linearity within the band. I’ve also chosen to use 0.5uH inductors because it is a good value for an air core inductor. By doing some simulations with other values of inductors, I found out that if I use larger inductances, the capacitors in dipole 5 and 7 must have lower values (for inductors of 2uH as in the original design and the above settings for the filter, these capacitors should have 1.4pF).
Using the components from the diagram and doing fine adjustments (in simulation) of the trimmer capacitors, I got the following shape. It is possible get even better results by adjusting the capacitors in dipoles 4, 5, 6. However, the filter design software I am using does not allow me to view the shape in real time. Ideally, if I could adjust a component and see the filter shape changing in real time, it would be much easier to optimise a filter for some preset values for some of the components. You can see that, at the moment the filter minimum loss is -6.72dB at the center of the band. I assumed that the quality factor of the coils built on toroids is 130. I will correct this value in the future, when I will have more information.
This is how the filter I decided to build looks like:
I was glad that all the capacitors were close in value to some values I had in my drawers.
I made quickly 3 coils, which measured 0.52uH. I will provide later details of how to make them.
Instead of 209pF, I had to silver mica caps, with a value of 215pF, 1% (really good accuracy!).
For the rest of the caps, I used ceramic NPO capacitors.
I will make a comment about capacitors. A few months ago I purchased 1000 caps of various values from Ebay, at very good price. I was happy to have a large number of common values. And they were physically very small as well, convenient to place in tiny places. But then, I decided to measure the caps while I was heating them with the soldering iron. I was not so pleased anymore with them. For example a 1000pF capacitor was going down in value to about 500pF when warmed up. Clearly, they are not to be used in any RF circuitry where some stability is to be desired. With the smaller values, I can imagine I may use some to compensate the frequency drift of VFOs, since coils usually (to my knowledge) increase their inductance when temperature rises. Later I found there are capacitors with an almost zero temperature coefficient and they are marked NP0 (negative positive zero coefficient, the last symbol is a zero, not letter O), while a different standard lists stable capacitors as C0G. Usually, they are larger in size than other ceramic capacitors.
In Australia, NP0 capacitors can still be found at reasonable prices at Jaycar Electronics. I purchased a healthy stock of NP0 capacitors. Also, in my wish list I set my eyes on a big box of SMD capacitors size 1210, with many of the values being of type NP0.
My first attempt to build the filter had this result:
We can see the center frequency was higher than 14.175MHz. Apart for this, the shape was not too bad, the left side had slightly larger loss.
What concerned me quite a lot was the loss of about 6.5dB. Compared to the simulation, my filter had a significant loss.
Because I gave my best shot measuring the components and making the coils, I really didn’t know what went wrong. So, back to the simulation table, to figure out what was wrong with my filter.
It didn’t take long to start asking myself questions if the quality factor could play a role in the large loss of the filter. Luckily, the AADE software which I am using is able to take into consideration
any Q factor for the coils. So I identified that my coils, no matter how shiny and nice, are not perfect. It hurts a bit to know that you tried your best and your best is not good enough 🙂
Using the simulation software, I came to the conclusion that a 6.5dB loss is caused by a coil with a quality factor of about 100.
Knowing this, I redesiged the filter, taking into consideration Q. I got these results:
Now, I decided to do some changes to the filter, to bring the shape symmetrically within the 20m band.
All I had to do is to rebuild the coils, to allow a slightly higher inductance than theoretical. Since the only adjustable elements in my filter are the coils,
I must make them with a higher inductance than required, so I can stretch them a bit, to bring them exactly to the required inductance.
Adjustment was not exactly a piece of cake. There are three resonators and each of them is changing the shape of the plot.
During the adjustments, I used a ferrite bead next to each coil to see what changes on the spectrum analyzer’s screen if I adjust that particular coil.
Ideally I should also have a piece of brass, which reduces the inductance of a coil, but I didn’t have any.
Instead, I carefully stretched the coils and if I stretched them too much, I was compressing them back.
For a perfect shape of the filter, the coils need to be stretched to the order of tenths of millimeter.
Once I stretched the coils, some of them moved back toward the original position, showing some elasticity. In the end I got it right.
I am not sure if tomorrow the filter will look as good as today, but the filter shape is only a stretch away 🙂
I will do later some experiments, heating the filter with a hair drier, to see how badly gets missaligned.
You can notice the minimum loss is a bit over 6dB, which is an indication for a quality factor of around 100 for the coils.
When I will receive the toroids I ordered, I will build some coils on toroids and will build another similar filter, to see if the inductors built on toroids have higher Q.
Here are some photos of the actual circuit:
A more detailed image of the coils. You can see they are stretched just a bit. I built the coils on a Philips screwdriver with a diameter of 4.8mm.
Each coil has 17 turns of enamelled copper wire 0.6mm (probably AWG22 = 0.64mm).
The base for the filter is a high quality PCB on berylium oxyde substrate, working in frequency range up to X-ray. Nah, just pulling your leg…
I found what seems some interesting pages about Q factors of coils and a Q meter, plus a method of measuring/calculating the Q factor of an oscillant circuit:
Now the filter is ready for the next experiment – connecting it to a one transistor preamplifier.
I am really interested in learning how the frequency response will look like at the output of the transistor and what the gain will be.
We should get some valuable indication about the input impedance of the preamplifier. A perfect matching (S11 = very high) will leave this shape unchanged, except it will have larger level.