So after studying some designs of multiple feedback band-pass filters for CW by other designers (see my previous post), I’ve decided to try my hand at designing and building one myself. Actually, I’ve made two different versions. The first one is designed by myself and is a two stage filter with an fc of 750 Hz. The second one is the filter designed by Tom Hall GM3HBT and published in Ham Radio Today 1987. I selected this filter because the gain ( ~1x), the center frequency ( ~750 Hz) and the Q / bandwidth ( ~5x / 100Hz).
Design | Stages | R1 [ kΩ ] | R2 [ kΩ ] | R3 [ kΩ ] | C1=C2 [ nF ] | Gain [ ] | fc [ Hz ] | Q [ ] | BW [ Hz ] |
13. GM3HBT | 2-3 | 390 | 12 | 820 | 2.2 | 1.1 | 740 | 4.20 | 176 |
PA3COR | 2 | 110 | 2.1 | 220 | 10 | 1.0 | 747 | 5.17 | 145 |
Tom Hall GM3HBT
The design by OM GM3HBT is a 2 or 3 stage filter. Tom claims a center frequency of fc of 740 Hz and a filter bandwidth (presumably 2-stage) of 120 Hz. The filter was build on a 9 x 6 cm of Perfboard. Input on the left, output on the right. The power supply connection is at the left down and should be somewhere between 9V and 18V and well stabilized. The capacitors were not selected but randomly grabbed from stock.
The filter was measured with my soundcard interface and the excellent audiotester V3.0 software. The results are show below:
The results are actually pretty impressive! The measured maximum obtained attenuation is around 94 dB! The second ( unconnected ) channel shows a gradual increase above 1.5 kHz. and so does the actual measured channel. So the maximum measured attenuation is most likely limited by the measurement set-up and not so much by the filter. The peak is around 760 Hz and the -3 dB around 100 Hz. So all pretty nice!
After this I discovered the audiotester software also does also have the ability to send a sine burst and test the impulse response. The software was set-up to send a burst of 35 sine waves with a frequency of 750 Hz at a level of 0dB
The builtin ‘scope’ was set to 10ms/div horizontal and 200mV/div (vertical). I have my doubts about the last spec, for I have never tested it or calibrated my soundcard. However, for this test this is not important. In the graph, you can see 35 sine waves before the output start to decline again. It takes slightly more than 1 division before the output has stopped ringing, so about 11-12 msec.
PA3COR
This is a 2-stage MFBF filter with a gain of 0 dB (1x). The gain was chosen deliberately so that it can be inserted in an existing receiver. The center frequency fc was set to circa 750 Hz. It was build on a piece of single-sided copper-clad board. Designs build in this way tend to be a bit larger than as build on Perfboard. It is more easy to swap out components and to experiment with values, however.
Below you can see the measured filter response. The sides are less steep as the GM3HBT design. This expected as this is only a 2-stage filter vs the 3-stages from GM3HBT.
As you can see the gain matches pretty nicely and is close to -0.5 dB. The center frequency is a bit higher. This might be because the capacitors selected (see below) were 9.48 nF and the 10 nF the filter was designed for. In practice this will make little to no difference.
Matching components
To get the best results (or at least results that are best matching with theory) it is sometimes necessary to use parts with close matching values. To get best matches, I use the following three step approach:
- Measuring
- Sorting
- Grouping
1. Measuring
This is actually the most labor intensive part of the process but still it only need to take about 30 mins or so. I would recommend to take group of maybe 10x the desired number of matched components. For a two stage filter with each 2 matched capacitors, I would go for a batch of about 40 capacitors. Then just measure the value with your desired instrument, this can be a multimeter with capacitor function and RLC tweezer set of some sort or whatever you have on hand. Measure the values and write them down (my preference is to write them down and later transfer the data to a spreadsheet, whatever works best for you)
2. Sorting
If not already done at 1) above, transfer the data to your favorite spreadsheet program on your computer. I prefer OpenOffice Calc because of the truly open file format. Not absolutely necessary but I prefer to add an index number in front of the column with measurement values. This will help to find back the part later on more easily. You should end up with something like the data in the left. Now sort the data off the measured capacitors. You should end up with something like the sheet on the right.
You can clearly see that the measured capacitors are now listed in ascending order.
3. Grouping
We now want to select a group of 4 capacitors (2 for each filter section), that are as close together as possible. Of course, we can manually browse through the table and the fact that the data is now ordered will help us greatly. A little bit of spreadsheet math will help us greatly. We calculate the standard deviation of 4 consecutive values of column C (Values). We now, only need to pick the lowest value to get the 4 best matching capacitors of the lot. In cell D12 we can see an SD value of 0.000 meaning there are 4 capacitors with the same value (see cell C12-C15).
So there you have it! A simple 3 step approach to finding matching capacitors from a lot.
Recent Comments