Listen to Latest SolderSmoke Podcast

Thursday, May 7, 2015

IF Selection -- Which Filter for the New Rig?

With some prodding from Pete Juliano, I am moving forward on my next transceiver.  Same wood box enclosure (with copper flashing), but this time I am greatly relaxing some of the radical fundamentalist restrictions:  Chips will be allowed. VFOs and VXOs will be replaced by an Si5351. Filters will not have to be homebrew.  Pete has been putting his CNC machine to use and making me some nice boards with isolation pads already milled in.  Oh, the luxury!

I am going to use the Termination Insensitive Amplifiers designed by Wes Hayward and Bob Kopski back in 2009.  These are especially useful in bilateral type transceivers because they allow you to nail down the termination impedances on the crystal filter IN BOTH DIRECTIONS.   That's is important if you want the same filter shape on both transmit and receive.

But now, with the trauma of my unfortunate IF selection on the BITX 20/40 (now just 20, sniff...) in mind,  what filter should I use on this rig?  The three main candidates appear above.   The 9 MHz Yaesu filter was given to me some time ago by Steve "Snort Rosin" Smith.   The Heath filter (3.395 MHz) and the larger silver one (2.215 MHz) were given to me by Armand Hamel. (Thanks Guys!)

My main band of interest for this rig is 40.   But if possible, I'd like to be able to use it on 15 and 12 meters,  and maybe even 20 and 17,  hopefully without having to change filters. 

 So what say the gurus?  Which one should I use?   Or should I put two of them in there, with provisions that would make it easy for me to move from one to the other? 

Right now my inclination is to go with the 9 MHz filter, perhaps with the 3.395 MHz filter also available.

Our book: "SolderSmoke -- Global Adventures in Wireless Electronics" Our coffee mugs, T-Shirts, bumper stickers: Our Book Store:


  1. Go with your 9 MHz inclination, though with caveats. 2 MHz will produce image frequencies within 20% of the desired frequency (at 20 MHz) so will be harder to filter.

    Another thing about 9 MHz is that it's at the point above which hams use USB and below LSB. If you have your filter set up so that only the lower carrier crystal is used (to produce USB) you can have the DDS local oscillator above the band desired for LSB (eg 16 MHz - 9 MHz USB for 7 MHz LSB) and below the band desired for USB (eg 5 MHz + 9 MHz USB = 14 MHz USB). That way you could use a DDS that only goes up to 20 MHz yet still cover up to 29 MHz. You also get the 'normal' sideband selection automatically without having two carrier oscillator xtals that are switched.

    But it's not all plain sailing, as some combinations might cause problems. Articles and nomograms have been published showing the troublesome ones. Eg 18 MHz with a 9 MHz crystal filter and low side LO is 9 MHz. Mixers are not linear and you might get some odd products if there's products like double both the local osc and carrier osc signals that will be hard to filter out since all, like the desired signal, are near 18 MHz.

    It may be easier, if your DDS goes high enough, to always have the local oscillator on the high side of the xtal filter/IF and accept that you'll need to switch the sideband (noting you're inverting on all bands). High side use means that the image is always higher in frequency, which is easier to suppress if you're using a good transmitter LPF (and have it switched in on rx).

  2. .I agree with Peter, for a single conversion transceiver the 9 Mhz. filter would be the best choice. The main band that would be a problem from IF harmonics is 18 M.mhz. With the second harmonic falling near the band edge. The Heathkit filter was used in a dual conversion system with a first IF around 8.4 to 8.9 Mhz. , and.a 3.395
    Mhz. second IF. Using the filter directly in a single conversion would put harmonics of the IF near several bands Harmonics at 6.79, 13.58, 16.975, 20.37, 23.765 could be problematic and hard to filter. That is not even taking into consideration mixer products with the LO frequency. Using the 5351,and high side injection, .changing sideband is just a couple of lines of code. With the upper frequency limit of the 5351 you could also generate the LO and BFO signals at 4 x the.frequency and use a 74ac74 to divide back down to the actual frequency. This should reduce the phase noise and crosstalk between clocks that seem to be the issues that are brought up with the 5351.

  3. One thing about wide range DDS synthesizers is that you can be flexible.

    If the IF strip is broadband, then changing filters only means changing the LO frequency into the mixer. Presumably in this day the frequency readout can show the operating frequency.

    In the old days it was too much trouble to shift this way, so there'd be multiple conversions to make use of different filters on different centre frequencies. (Though, that method did cascade the filters.)

    That flexibility isn't so needed with an SSB only rig, but when building a general coverage receiver being able to switch to an AM or FM or SSB filter could be useful, especially when scrounging isn't likely to turn up such filters on the same frequency.

    For that matter, if building general coverage, shifting an IF or even the LO may get rid of a birdy. When you can easily generate frequencies, you aren't stuck in the old model where you try for optimization before building, and then live with unwanted responses. Popping in a scrounged filter for some ranges may simplify overall design.


  4. Hi Bill,

    Ok I prodded you this far along --so here is the next round.

    Switch to the Mega 2560 Arduino so you have more pins to play with and more room to program.

    You have received some excellent input from Peter and Duwayne about the why of 9 MHz with of course the caveat about 17M. So why not use two filters the 9.0 MHz for all bands except 17M and a homebrew filter for 17M at 4.9152 MHz.

    You can keep the single conversion scheme and with a flip of a switch when going to 17M a different filter is selected as is different BFO frequencies. The extra pins are needed so you can switch the filters using Omron G5V1 relays and to again relay switch Band Pass and Low Pass Filters.

    The Mega has plenty of program room to do this --you could even have a keypad to select the band and tune the radio from the keypad. The Mega additional pins enables having a color display too!

    In for a penny might as well be 5 pounds!

    Pete N6QW

  5. Or you could go phasing or third method and not need a filter at all. Hans Summers is one of the few who has built a Weaver receiver.

  6. Considering that each of those filters are likely to be at least 20 years old - and that they can go bad just from sitting due to atmospheric contamination, mechanical shock, excess energy or who knows what - I'd build a test fixture using (lossy!) resistive matching to its appropriate impedance for your chosen filter before building a rig around it and verify that its passband shape and ripple is what is expected.

    If it is not, either the info the matching impedance is not correct, you goofed up on the resistive matching and/or the filter has gone bad. In general, if the passband looks good, so will the stop-band - which is much harder to check with a simple test fixture since that is less likely to be able to "see" down 50-60dB, either due to the dynamic range of your Rigol or if you are using one of those log amp chips, the added loss of the resistive match+limited RF drive, "blow-by" (if you are lucky!) or a combination of the above..


Designer: Douglas Bowman | Dimodifikasi oleh Abdul Munir Original Posting Rounders 3 Column