Message	Date	From	Subject
8257	2013-02-12 12:01:31	Wes	Double Tuned Circuit Thoughts
Hi all, A recent series of posts, starting with #8174, dealt with double tuned circuits built from EMRFD Table 3A, page 3.14. The filter of interest was one with a 0.4 MHz bandwidth at 14.2 MHz. Note that there was a typo in the table that is now reported in the errata. The BW is 0.4 MHz and not 0.2 MHz. Bob Kopski expanded the experiment of the original post by building, measuring, and reporting on a no-tune filter at 14.2 MHz based upon the table entry. See posts #8183, 8210, and 8226. As Bob demonstrated, it is indeed possible to build no-tune filters at this 3% fractional bandwidth. A no-tune filter is one without trimmer capacitors or variable inductors. It is a circuit where the components are selected, measured, and assembled with little or no additional tweaking. Most of the filters that we might build from Table 3A, or from the accompanying design procedures do not use the no tune approach. Rather, the filter schematic is modified by the addition of trimmer capacitors or inductors. I built a version of this filter. It was designed, simulated, built, measured, and studied at both HF and VHF. The results are presented at http://w7zoi.net/dtc_in_emrfd.pdf Thanks to all for the related posts, especially those from Bob, K3NHI. 73, Wes w7zoi
8263	2013-02-13 08:14:21	KK7B	Re: Double Tuned Circuit Thoughts
Really nice work Bob, Wes and all. Regarding no-tune techniques... As Wes mentioned in his excellent article on filters, I've been fascinated with this topic since my student days, long ago. My early work was at microwaves, using both hand-cut tubing resonators and printed resonators. All of my audio filters in EMRFD are no-tune, including the all-pass phase shift networks covered in chapter 9. My approach has been careful and deliberate, with measured component variations and construction tolerances included in the design. In the early days those were random variables in design equations, but now they are more often swept component values in a simulator. An example of my no-tune approach in EMRFD is illustrated in the four plots in figures 9.58, 9.59, 9.60 and 9.61 showing the effect of using 1% 0.5% 0.2% and 0.1% tolerance resistors and capacitors in a third-order all-pass network. To design a circuit for duplication using no-tune techniques, I usually start by looking at designs that work well with perfect components, and then either write design equations and take derivatives over each component value to find the sensitivity to each component, or else do a simulation with each component value as a variable that may be swept over a small range. Some designs are more tolerant to fuzzy component values, and those are good candidates for no-tune duplication. Then I do a statistical study at the bench to find out how much variation I actually get in the individual components and construction methods. For my first no-tune microwave filters using hand-cut copper tubing resonators, I cut the 12 resonators I'd need for 4 filters using an exacto knife and magnifying glass, and then measured them all with a dial caliper. Then I calculated the average length, standard deviation, etc. for the 12 resonators so I'd have a measurement of the expected variability in resonator length. I plugged that variation into the design equations and found that I could expect to build no-tune filters with 5% bandwidth. I also plugged in all the other mechanical variables I could identify. I built 30 or 40 filters using that procedure, and all of them worked as expected with no tuning, when measured on a microwave network analyzer. It took several hours to build each filter, so I naturally looked for a way to reproduce them photolithographically. Microwave engineers have built printed filters for many decades, but the standard procedure is to use a precision substrate--often ceramic in those days--so the design could be exactly duplicated with micron tolerances. Jim Davey's work with precision photolithography on Teflon substrates was a real inspiration in that era. I wanted to use cheap FR-4, which is notorious for having variable dielectric and mechanical tolerances. I measured many resonators printed on FR-4, and concluded that 10% bandwidth was about as narrow as I could reliably reproduce. That put some constraints on IF frequencies, but since it is easy to build printed filters, I used more of them in the system design, which resulted in clean signals. The key to successful no-tune systems is to include the component tolerances in the design. FR-4 tends to have bias errors rather than a random distribution, so that was included in the design decisions as well. For HF filters using toroid inductors, I'd wind ten of the same value, measure them all and squeeze the turns to tweak them to the same value, and then measure them again the next day to find out how much they change as the winding stress relaxes overnight. I might solder them into a circuit that puts a little stress on the leads and measure them all that way too. After I had some good measurements of the expected inductor variations I'd proceed with the simulations. For an individual application it is often easier to include tunable elements and optimize the circuit after it is built, using test equipment. That also permits narrower bandwidths, as Wes has noted, since component and mechanical variations may be tuned out after construction is completed. i often use that approach. But no-tune filters have two advantages: since component variation is included in the original design, no-tune circuits tend to be more tolerant of aging, mechanical stress, vibration, and the shocks and bumps of portable operation; they also tend to be more tolerant of external variations such as drive and load impedances. When carefully designed and constructed, they tend to work the first time, and from then on. I have no-tune transverters in my shack that go back 30 years, and they still work the same as when they were first fired up on the bench. Again--nice work Wes and Bob--this sure brings back fond old memories of mailing printed filters back and forth with Jim Davey in the early no-tune microwave days. Also, note that I was not the first fool to build printed microwave circuitry on FR-4. Paul Wade W1GHZ published a 1296 LO system and mixer

Really nice work Bob, Wes and all.

Regarding no-tune techniques... As Wes mentioned in his excellent article on filters, I've been fascinated with this topic since my student days, long ago. My early work was at microwaves, using both hand-cut tubing resonators and printed resonators. All of my audio filters in EMRFD are no-tune, including the all-pass phase shift networks covered in chapter 9. My approach has been careful and deliberate, with measured component variations and construction tolerances included in the design. In the early days those were random variables in design equations, but now they are more often swept component values in a simulator.

An example of my no-tune approach in EMRFD is illustrated in the four plots in figures 9.58, 9.59, 9.60 and 9.61 showing the effect of using 1% 0.5% 0.2% and 0.1% tolerance resistors and capacitors in a third-order all-pass network.

To design a circuit for duplication using no-tune techniques, I usually start by looking at designs that work well with perfect components, and then either write design equations and take derivatives over each component value to find the sensitivity to each component, or else do a simulation with each component value as a variable that may be swept over a small range. Some designs are more tolerant to fuzzy component values, and those are good candidates for no-tune duplication.

Then I do a statistical study at the bench to find out how much variation I actually get in the individual components and construction methods.

For my first no-tune microwave filters using hand-cut copper tubing resonators, I cut the 12 resonators I'd need for 4 filters using an exacto knife and magnifying glass, and then measured them all with a dial caliper. Then I calculated the average length, standard deviation, etc. for the 12 resonators so I'd have a measurement of the expected variability in resonator length. I plugged that variation into the design equations and found that I could expect to build no-tune filters with 5% bandwidth. I also plugged in all the other mechanical variables I could identify. I built 30 or 40 filters using that procedure, and all of them worked as expected with no tuning, when measured on a microwave network analyzer.

It took several hours to build each filter, so I naturally looked for a way to reproduce them photolithographically. Microwave engineers have built printed filters for many decades, but the standard procedure is to use a precision substrate--often ceramic in those days--so the design could be exactly duplicated with micron tolerances. Jim Davey's work with precision photolithography on Teflon substrates was a real inspiration in that era. I wanted to use cheap FR-4, which is notorious for having variable dielectric and mechanical tolerances. I measured many resonators printed on FR-4, and concluded that 10% bandwidth was about as narrow as I could reliably reproduce. That put some constraints on IF frequencies, but since it is easy to build printed filters, I used more of them in the system design, which resulted in clean signals. The key to successful no-tune systems is to include the component tolerances in the design. FR-4 tends to have bias errors rather than a random distribution, so that was included in the design decisions as well.

For HF filters using toroid inductors, I'd wind ten of the same value, measure them all and squeeze the turns to tweak them to the same value, and then measure them again the next day to find out how much they change as the winding stress relaxes overnight. I might solder them into a circuit that puts a little stress on the leads and measure them all that way too. After I had some good measurements of the expected inductor variations I'd proceed with the simulations.

For an individual application it is often easier to include tunable elements and optimize the circuit after it is built, using test equipment. That also permits narrower bandwidths, as Wes has noted, since component and mechanical variations may be tuned out after construction is completed. i often use that approach. But no-tune filters have two advantages: since component variation is included in the original design, no-tune circuits tend to be more tolerant of aging, mechanical stress, vibration, and the shocks and bumps of portable operation; they also tend to be more tolerant of external variations such as drive and load impedances. When carefully designed and constructed, they tend to work the first time, and from then on. I have no-tune transverters in my shack that go back 30 years, and they still work the same as when they were first fired up on the bench.

Again--nice work Wes and Bob--this sure brings back fond old memories of mailing printed filters back and forth with Jim Davey in the early no-tune microwave days.

Also, note that I was not the first fool to build printed microwave circuitry on FR-4. Paul Wade W1GHZ published a 1296 LO system and mixer