EMRFD Message Archive 14983

Message Date From Subject
14983 2018-08-08 08:35:52 Michael Hadley Dishal filter calculator
Hi .

I am having problems trying to get an eight pole 9.0 MHz 600 Hz filter to work correctly using the Dishal calculation program. Despite several attempts the filter always comes out low in frequency by about 1.5 KHz. Other bandwidths (2.5 and 3.1 KHz) using the same program and crystals work out correct with centre frequency almost dead on 9.0 MHz.

The filter requirements are 600 Hz BW, 0.1 ripple.

I have measured all necessary parameters of the crystal using the G3UUR method. These are Cm - 7.28fF, FS1 - 9.0011 MHz, FS2 - 8.9994 MHz, FS3 - 8.9982 MHz, FP - 9.0125 MHz, Crystal holder 3.0pf

--
Mike Hadley G4JXX
14984 2018-08-10 11:14:15 rcbuckiii Re: Dishal filter calculator

Mike,

I think you got lucky on the SSB filters. I have designed a couple of filters over the last year or so and mine did not come out at 9.0 MHz using Dishal. I'm not sure if that is actually how the program is supposed to work or not. The input field just asks for the Series Frequency of the crystals which is never the marked frequency. I always put in the average of the crystals I am using. Maybe I should enter 9000 in that box??

I built 3 filters about 6 months ago for a 40/20/30 receiver. The center frequency of the 3 filters came out as follows:
SSB    8.998000
CW    8.999370
AM    9.002480 (this is a 5 kHz wide filter)

Since I am using a Si5351A for the LO and BFO, I just handle the offsets in code so that the LCD shows the actual frequency of the signal. It does present a problem if I am copying a CW signal in SSB mode and switch to CW. I have to mentally add 570 to the SSB frequency to arrive at the actual frequency of the signal. This is due to the difference in center frequencies of the SSB and CW filters minus the 800 Hz BFO offset that I am using for CW.

If someone else has managed to get their filters on the same center frequency as what the crystals are marked, I would like to know how that is done.

Ray,
AB7HE

14985 2018-08-10 12:48:20 Bill Carver Re: Dishal filter calculator
Ray, a few times, to make up a "set", I've put an inductance in series with the xtal at 4434 KHz to move the frequency down a little. I used a small u=20 ferrite core (-67 or -68, don't remember now) to get 50uH and high Q at 4434 KHz. The external coil needs have a high Q: it will be lower than the hundreds of thousand for the crystal itself, so the net effect is to move BOTH the Q and the resonant frequency down. But keeping the Q of the external coil high minimizes the Q degradation. It worked to move an xtal about 100 Hz to make a set for a 200 Hz wide CW filter.

If you wrote, or have the source code for the Si5351A, you can move the LO and BFO to adapt to different filter center frequencies. Move both by the same amount (the LO might be going down, and the BFO up, depending on which side of the signal frequency the LO is) until the CW filter and pitch is where you want them to be when you switch mode. Think about how heterodyning works....adding/subtracting frequencies, and with a pencil and paper you can figure out which way they have to move.

Bill W7AAZ


14986 2018-08-10 12:59:57 Mike Hadley Re: Dishal filter calculator

Hi Ray.

I am using crystals supplied by Mini Kits Australia (in their own kits) and matched them all to within 20 Hz.

I found that changes to the (supposed) holder capacitance has a marked effect on filter capacitors.

In my application I cannot change the vfo offset only the product detector cio frequency. 1.5KHz down from 9.0MHz though is a bit too much to ask.

Have you tried the mesh capacitors calc program on EMRFD cd? If so how was it.

The three filter board, (0.5, 2.5 & 3.1KHz) will be made available as a kit, primarily to compliment the HyCas kit already available from me.

Mike
G4JXX

Sent from my Xperia™ by Sony smartphone



---- rcbuckiii@yahoo.com [emrfd] wrote ----

 


Mike,

I think you got lucky on the SSB filters. I have designed a couple of filters over the last year or so and mine did not come out at 9.0 MHz using Dishal. I'm not sure if that is actually how the program is supposed to work or not. The input field just asks for the Series Frequency of the crystals which is never the marked frequency. I always put in the average of the crystals I am using. Maybe I should enter 9000 in that box??

I built 3 filters about 6 months ago for a 40/20/30 receiver. The center frequency of the 3 filters came out as follows:
SSB    8.998000
CW    8.999370
AM    9.002480 (this is a 5 kHz wide filter)

Since I am using a Si5351A for the LO and BFO, I just handle the offsets in code so that the LCD shows the actual frequency of the signal. It does present a problem if I am copying a CW signal in SSB mode and switch to CW. I have to mentally add 570 to the SSB frequency to arrive at the actual frequency of the signal. This is due to the difference in center frequencies of the SSB and CW filters minus the 800 Hz BFO offset that I am using for CW.

If someone else has managed to get their filters on the same center frequency as what the crystals are marked, I would like to know how that is done.

Ray,
AB7HE

14987 2018-08-10 16:42:27 rcbuckiii Re: Dishal filter calculator
Mike,
No, I have not tried the mesh capacitors calc program on the EMRFD cd. I believe the Xtal Tuning program (under Xtal on toolbar) is doing something similar. Using the calculated values more or less worked. I still had to make slight adjustments in capacitor values while watching the shape on the spectrum analyzer. I used the L-C Match program in Dishal to match my filters to 50 ohms.

W0QE has a couple of good YouTube videos that show how to use SimSmith to fine tune crystal filter parameters. Maybe some of his suggestions will help you move the CW filter up in frequency.

Bill,
I wrote the code and I accounted for the different center frequencies in the code. I wrote some test code to automatically account for the differences in the center frequencies when changing modes. I kept track of which SSB mode I was in and applied the differences and it worked OK. I had to handle switch debouncing and delayed changing modes for about a second (in case switch was turned slowly) after the mode switch was changed. Everything worked well as long as I  paused for more than one second between modes. I finally decided to not worry about it and put it in my "todo list" to be handled at a later date.

I will try your idea with the inductors the next time I design a filter. I could pick crystals that are all above the desired frequency and try using the inductors to shift them all to a common frequency.

Ray,
AB7HE
14988 2018-08-10 19:02:35 Bill Carver Re: Dishal filter calculator
Ray, maybe I shouldn't have mentioned the inductor tuning, because it adds another level to the trickery you have to pull out of the hat to get a filter tuned perfectly!  A series inductor will move the effective xtal frequency down by one half of its inductance divided by the xtal "motional" inductance. So if you have xtals whose motional inductance is, say, 100 millihenries, then a 20 microhenry coil is 0.02% of the xtal's inductance. When put in series it will move the resonant frequency down by .01%. So a 4434 KHz xtal will move down by 443 Hz.  I only moved two xtals in a narrow 10-pole CW filter, not to move the center frequency but to get all the meshes on the same frequency.  With only two coils at opposite ends of the filter I still had to put a shield around the coils! Make sure the added coils can't "see" any other coil, or the input transformer (if you use one at the input/output for matching).

Getting the right shape and width are tricky enough without also trying to hit a specific center frequency.  But if you're determined to move the filter CF with a set of matched xtals, a (better?)  approach IMHO is to adjust the filter impedance. That is, lets say you've designed a narrow 50 ohm ladder filter but you find it's center frequency is lower than the SSB filter (the usual case). You can raise the center frequency of the CW filter by raising the design impedance. A good first shot might be to redesign the filter (hopefully all this is "on paper"!) for 200 ohms. This higher impedance will reduce the value of all the coupling capacitors: this has no first-order effect on the shape of the filter, but the smaller coupling capacitor values will move the center frequency higher.  This 200 ohms as "first shot" is simply because you can put a relatively simple bifilar 1:2 turns ratio transformers (1:4 impedance ratio) at both ends of the filter to again obtain 50 ohms to the external circuitry.

Another approach would be move the SSB filter down, by reducing it's design impedance. A good first shot being the same 1:4 ratio. Redo it for 12.5 ohms and see if that moves the SSB filter center frequency down enough that the CW filter's relative position is acceptable. Again, a bifilar transformer with 1:2 turns ratio would match it to 50 ohms, but the transformers would be turned around, with the 2 turn side facing the 50 ohm source/load, and the 1 turn side facing the SSB filter. I've moved filter CF's a few times this way.

You can get yourself in trouble, so look carefully at the frequency response of the filter before and after you change the impedance. Remember that because of the "holder capacitance" shunting the series RLC of the crystal the filter shape is not symmetrical. The smaller the coupling capacitors are the larger the asymmetry becomes. So raising the impedance of an SSB filter, which is wider and has smaller coupling capacitors, will produce more of a change in shape than moving a CW filter up.

Something that improves the symmetry of SSB-width filters, especially ones with 3-4 poles, is to neutralize the holder capacitance of the input and output xtals. This uses the same idea as neutralizing the grid-plate capacitance of triode tubes in the olden days. You use a center tapped transformer at input and output. Ground the center tap, drive the filter off one end of the winding, and connect a capacitor equal to the "holder" capacitance of the bare xtal ( maybe 1-4 pF) to the other side of the first xtal. And another transformer with the same design, it's capacitor going to the other side of the last xtal. Now you can "cancel" the "holder" capacitance....and at least those two stages will be symmetrical. If you're feeling your oats, you can even over-cancel it, and put a notch below the LOW side of the filter the same amount as the other xtals put their notches on the high side.

Another bit of trickery you can use if you do make a transformer: typically, "they" say, make the self-inductance of an impedance matching transformer at least 10 times the impedance of the filter being matched. FB, but that's a rule developed for wideband circuits. Since you're making a very narrowband filter an alternative that really works better is to resonate (tune) the transformer to the frequency of the filter. Voila, it becomes almost transparent. So if you're at 4434 KHz, and make a 1:2 transformer whose larger turns are 40 uH, the put a 32 pF capacitor across that winding, resonating it to 4434 KHz.  This relieves you from having to worry quite so much about that "10 times the filter impedance" in your winding, and letting you concentrate on making the transformer have tight coupling between windings (balun core, or at least a toroidal core). Resonating can even enhance performance more by letting you use -61 balun cores which have slightly lower signal loss, rather than of -43 cores. In addition, if you find the ideal impedance is 2.5 times 50 ohms, requiring a 1:1.58 turns ratio, you can get pretty close with turns ratio of 2:3 and obtain close coupling between non-bifilar windings by resontating  with a capacitor from the last turn (end of winding) to ground (ground = start of winding) As I said, "trickery".

But I'd rather see you solve your center frequency problem by simultaneously changing the LO and BFO frequencies. If you are familiar with "PASSBAND TUNING" (Collins 75A4) or "IF SHIFT" (Japanese translation of PASSBAND TUNING), having a knob to tune both oscillators the same amount gives you adjustable positioning of the narrow CW filter on the signals without changing their beatnotes.  With that one stroke you've solved your filter CF problem and greatly enhanced the utility of your receiver on CW.

Regards - Bill - W7AAZ


14993 2018-08-12 12:52:27 rcbuckiii Re: Dishal filter calculator
Bill,

Interesting comments. Designing and building crystal filters is always an interesting undertaking. Fortunately we now have software programs to help with the design process. The building and adjusting process is much easier with today's test equipment. My DSA815-TG allows me to make component changes and see the effect in real time.

I agree with you, changing the LO and BFO frequencies is the easiest way to address the different CF problem. Unfortunately that won't work in Mike's case.

Ray,
AB7HE
14994 2018-08-12 14:30:57 Mike Hadley Re: Dishal filter calculator

Hello Ray.

Interesting comments that are well noted.

As a different approach I have ordered some more crystals made by the same company (ECS) with slightly different characteristics. Again, these crystals are already being used by Mini Kits for the same application using Dishal. I will post my findings after the swap out.

Mike

Sent from my Xperia™ by Sony smartphone



---- rcbuckiii@yahoo.com [emrfd] wrote ----

 

Bill,

Interesting comments. Designing and building crystal filters is always an interesting undertaking. Fortunately we now have software programs to help with the design process. The building and adjusting process is much easier with today's test equipment. My DSA815-TG allows me to make component changes and see the effect in real time.

I agree with you, changing the LO and BFO frequencies is the easiest way to address the different CF problem. Unfortunately that won't work in Mike's case.

Ray,
AB7HE
14995 2018-08-12 16:14:58 Bill Carver Re: Dishal filter calculator
Agree on all, Chuck. Having the ability to accurately model without having to build allows you to try the impedance change trick to move the CW filter upward. It's hard to move it 700 Hz up, unless your filter is higher than 4.434 MHz. But you can get it partway....and try it with only some time at the computer.
Bill

15005 2018-08-15 08:36:06 mhadley157@gmail.... Re: Dishal filter calculator
Changing the crystals
15007 2018-08-15 11:40:33 Bill Carver Re: Dishal filter calculator
Ah yes, that was one other "trick" I forgot to mention. The crystals are really the same except their series resonant frequency is moved just enough so parallel resonance with the specified capacitance is at the nameplate frequency. GOOD SHOW!
W7AAZ

15008 2018-08-15 12:16:00 Norman Dennett Re: Dishal filter calculator

 

 

15009 2018-08-15 16:52:47 Tayloe, Dan (Noki... Re: Dishal filter calculator

When I matched crystals, I had to use a specific test osc that used the crystals in the series resonant mode.  Most crystal oscillators seem to use the parallel resonate mode.  Thus matching the parallel resonate frequency does not guarantee that the series resonance is also with 20 Hz.

 

I found that out the hard way a long time ago.

 

The test oscillator that I built used a series resonate L/C in series with the crystal.  I would short out the crystal socket, tweak the L/C to “free run” on the expected crystal frequency, then remove the crystal short and insert a real crystal and measure the resulting frequency.

 

It takes a specific oscillator configuration to work in this mode.

 

  • Dan, N7VE