Cease and Desist Order!
Regrettably I am stopping all work on the Phasing Transmitter Rig. After finishing the KK7B Rig following the 2Q4 sojourn, I am left with a pile of crap. I have been unsuccessful in making a suitable and working hardware based phasing transmitter.
The KK7B Board has a mysterious problem that I have been unable to resolve. So one option is to stop work today, as my efforts, so far, have failed to crack the code. Spending some time thinking about the why is more productive that just removing parts and soldering in new ones. So following a tip from N2CQR --put it aside and think, research and come up with possible resolutions.
I know this project has stirred some interest no more so than my own; but there comes a time to fold them and play another day.
73's
Pete N6QW
The QuadNet Software from Tonne Systems
May 10, 2020 ~ A story from WWII.
Failed to mention that May 8th, was the 75th Anniversary of VE Day. There was such great celebration across this nation on that day in 1945. Perhaps we all will live to see the same for Covid19.
There are some interesting stories from WWII. As the country geared up for war, there was a huge infrastructure construction build up to support the war effort. While the work was done by civilian contractors, the US Army oversaw the security for these sites. Everything going in or out of the facility was searched by the Army personnel.
So there was the contractor person that everyday would go through the gate pushing an empty wheelbarrow from the inside to the outside. The Security personnel would check over the wheelbarrow and seeing as it was empty would wave him on through the gate. What the security forces did not realize was that the worker was stealing wheelbarrows.
The posting from May 9th asked why were those frequencies chosen as it was not clear to me. Thanks to Joe, I have been enlightened. And what I am about to share is more than I really know.
Think about capacitors and their response to frequencies. If you impress DC on a Capacitor (think of it as "O" frequency) what do you have? It is like an open circuit to DC -- essentially DC is blocked. So OK now if you impress a Very High Frequency on a Capacitor it acts like a short circuit to ground. Capacitors have names where they are referred to generically as "Blocking Capacitors" and/or "By-Pass Capacitors"
Now since the non-inverting input has the capacitor affixed to it we have the input to Pin 3 being affected such that a phase shift may occur when the cap is blocking/bypassing.
I think this comes from an internet source (Wikipedia) regarding all pass networks...
This
implementation uses a low-pass filter at the non-inverting input to generate
the phase shift and negative feedback.
In fact, the phase shift of the all-pass filter is double the phase shift of
the low-pass filter at its non-inverting input.
At high frequencies, the capacitor is a short circuit, creating an inverting
amplifier (i.e., 180° phase shift) with unity gain.
At low frequencies and DC, the capacitor is an open circuit, creating a
unity-gain voltage buffer (i.e., no phase shift).
At the corner frequency ω=1/RC of the low-pass filter (i.e., when input
frequency is 1/(2Ï€RC)), the circuit introduces a 90° shift (i.e., output is in
quadrature with input; the output appears to be delayed by a quarter period
from the input).
Another revelation from Joe, The filter elements can be in any order and not necessarily from low to high or high to low.
More parts on the board yesterday and maybe a finish today. Keep in mind this is NOT plain old perforated vector board but Single Sided Copper Vector Board so the top is a copper ground plane with all point to point wiring done on the reverse insulated side.
I even wear nitrile gloves when I work with the board as the acids in your hands will tarnish the copper surface of the board. If you look at the T2 -- it has a resemblance to KK7B's layout. How is that for guessing a bunch? I will take some photos of the underside as another objective is short direct connections (just like a PCB) and to minimize any cross overs.
Every penetration through the board and not to ground has the copper removed from around the hole to prevent shorting to ground. I use a 1/8 inch drill bit fitted with a handle. All ground connections are simply passed through the hole and soldered to the top. If you look closely you will see some of the ground connections.
This takes some effort but makes for an excellent alternative if you lack a CNC machine. The reason I chose this approach versus using my CNC is to achieve a very compact layout.
73's
Pete N6QW
May 9, 2020 ~ Does Science Meet Data or does Data Meet Science?
I have been pondering that question and suggest it depends. It depends if you want to implore science to verify and/or validate data. Or is it a that you have weird data and are looking to science to explain the why.
This leads me to a point made yesterday about the "odd ball" capacitor in the KK7B T2 schematic. It was pointed out to me (thanks Joe) that in the schematic for the individual op-amps in the network the only variable is the RC component values for the individual op-amp. You know the old time constant T = RC (in seconds) for discharging a capacitor.
So for a constant "T" the values of R and C can float all over the place just as long as their product remains a constant "T". There is of course a relationship between the Time and Frequency Domain which is why the time constant is so critical. Thus f = 1/T or 1/(RC). OK 1 millisecond is 1000 Hz and 1 microsecond = 1 Megahertz.
So did KK7B simply run out of 0.001 uf caps and then made a quick calculation and said Ahh this size 1% resistor will keep the T constant and boom there we go. I simply don't know!
If we look at that last capacitor resistor combination we have 0.0027uF times 649K and that product in seconds is 0.001753 seconds and thus 570.678 hertz. The other value of 0.001uF and 511K comes out to 1956.947 hertz.
Some of the other values have an interesting reveal. One combination is 0.001uF and 15K. Doing the math ~ 001uF * 15K = 0.000015 seconds which translates to 66.667 kHz. Yet another of 0.001uF * 232K which translates to 4.310 kHz. For the same value capacitor smaller resistors result in higher frequencies.
Keep in mind the T2 was developed in 1993,which is ancient technology (almost 30 years ago) so I am uncertain what tools were available then. Even the QuadNet is 15 years old; but certainly more interactive than would be available in 1993. The KK7B Values result in the following frequencies.
- U2C = 66.67 kHz
- U3D = 8.85 kHz
- U3C = 1.956 kHz
- U2B = 19.12 kHz
- U3A = 4.31 kHz
- U3B = 0.5707 kHz
At this point I have no clue as to the why these are the frequencies chosen by KK7B. Now lets look at the 6 pole filter used in the QuadNet where all capacitor values are the same at 0.01uf. I will use the same designations as in the KK7B.
- U2C = 47.62 kHz
- U3D = 5.24 kHz
- U3C = 0.9524 kHz
- U2B = 12.406 kHz
- U3A = 2.262 kHz
- U3B = 0.2551 kHz
Again not obvious the "why" of the values. The first three are in cascade on one output as are the second three in cascade for the second output. I am just not that versed on progressive filtering and the impact of values and what passes through the chain. [More learning opportunity.]
But the data does show the 1st grouping is passing a higher band of frequencies as in 66 kHz to 1.9 kHz and the second grouping 19 kHz to 0.5 kHz. That same relationship is evident in the Quadnet values.
Certainly it has to do with passing the same band of audio frequencies but out of phase by 90 degrees. Just doesn't jump right out at me and grabs me by the stacking swivel.
I was made aware of the stage to stage coupling in both the KK7B and the Quadnet model as being "AC" coupled via the 0.001uF in the KK7B and the 0.01uF in the Quadnet to Pin 3.
It was suggested that I swap the capacitor and resistor locations (just the R &C) so that the stages are now "DC" coupled to avert noise pick up issues. Initially I sort of gulped over switching the circuit location of R and C. But if you do a LT Spice Simulation the results are no different; but as we know --noise and op amps are like oil and water.
Thus for my first build I am switching the R's and C's. Keep in mind T still = R*C. I understand that W6JL in his magnificent build used this same approach. (Thanks Joe)
Below is the beginning work on the KK7B T2 Phasing Transmitter. I am using single sided copper vector board where the top of the perforated board is covered in copper (read expensive). At each location where there is not a ground connection using a 1/8 inch drill bit with a special handle I ream out the holes so the component or socket does not ground itself to the top surface. All hardwired connections are made on the underside. If a connection goes to ground don't ream the hole and solder right to the top of the board. Using the photo of how Campbell laid out his PC Board, I more or less followed his layout and away we go
As I well know from the 2Q4 this could go bust at any moment but at the very least there are now some analytical tools that let you have the very best of the Science and Data worlds.
SITS and HUAH.
73's
Pete N6QW
********************************************
Noteworthy I also do not see a direct mathematical relationship between the sets of values for the two filters.
There is much to be said about someone who discovers a new religion and suddenly is imbued with a terrible missionary zeal. Thanks to two hams (Joe and Dave) who suggested I look at the Tonne Systems QuadNet software as a design tool for op-amp audio phase shift networks. They were right!
First some important observations:
- Unlike many tools this one is user friendly and intuitive. If you are a button pusher this will work for you.
- It has some pretty nifty built in tools like the Monte Carlo Analysis to rapidly evaluate a network's response to an array of stimuli in graphical form.
- It offers options on the precision of the network such as what happens if you use 5% resistors versus 0.1% resistors (It is dramatic!). If you really want 90 Degrees across an audio range at suppression levels >50 dB then 0.1% or better is required!
- The default for the capacitors is all of the same value of 0.01uF . But you can readily change that to any value say 0.001uF or even a mixed set of values.
- One of the outputs is a link to LT Spice so that the QuadNet Design automatically appears as a LT Spice design. Now that is a nice feature.
- It also can be used in conjunction with LT Spice as an evaluation tool to compare networks. More on this.
You have options for the numbers of poles such as every integer up to 10. I must admit a 7 pole filter looks strange as we tend to think in an even number of poles.
Before being introduced to QuadNet (dragged is more like it) I had decided to build the KK7B phasing transmitter and ordered the parts which have since arrived. Since I now have a new found religion, I decided to use QuadNet to design a 6 pole filter which is what Campbell uses. The next step was to compare the QuadNet with the KK7B using LT Spice.
Campbell uses the NE5514 which is difficult or impossible to find let alone purchase. Yet the QuadNet uses a generic op-amp. I suspect you really cannot use just any op-amp; but one that is low noise, has good isolation between the sections and since this is audio having really wideband characteristics like 200 MHz is not so critical. Several candidates look promising as substitutes like the N35534, NE5532 and LM324. I happen to have all of these in the bins. Linear Technology has a slew of op-amps --but I just am not as familiar with those as I am with the three mentioned.
Noteworthy in the KK7B design is that five of the six capacitors are of the same value (0.001uF) and one is not (0.0027uF).
In one analysis I use with LT spice is to designate the I and Q outputs as "A" and "B" and tag them as such Then setting up the AC Analysis for the range of 100 Hz to 3kHz I look at A & B with reference to the V (audio input). I get two plots which shows the phase difference. Boom two parallel lines separated by 90 degrees over the audio range.
Now here is the interesting part. If you use the KK7B values (with the one odd cap value) for the most part the output curves are parallel. But if you make that odd capacitor the same as the other five --the output curves are skewed and not parallel. So that odd capacitor value is there for a reason (A big one).
Now here is the interesting part. If you use the KK7B values (with the one odd cap value) for the most part the output curves are parallel. But if you make that odd capacitor the same as the other five --the output curves are skewed and not parallel. So that odd capacitor value is there for a reason (A big one).
Above the plot with the same capacitor values
Above is the plot with the one odd ball capacitor.
The lines are for the most part parallel and 90 Degrees apart. But observe the skewing at the very low and high end of the range
Now if you use the QuadNet Values where all caps are the same at 0.01uF, then you get the two parallel lines BUT the lines are really appear parallel over the entire range whereas in the KK7B design appear a bit skewed at the extreme ends.
Note above that the very low end that the curves still remain parallel, whereas below (KK7B) there is a skewing. How much of an issue that is remains to be decided whether just a plot point or real issue. I suspect the 40M SDR police running their FLEX 6600's with 72 Inch LCD will tell you something about your signal at 100 Hertz when running the KK7B.
Bottom line is the use of the tools. I have parts on order for the 8 pole QuadNet and likely will build three versions. The two six poles for the transmitter and the 8 pole for a receiver.
BTW the 1% resistors can be had for 7 cents a piece in lots of 10 from DigiKey. Thus 60 piece of 1% resistors is less than a $5 bill. Of course the 0.1% cost more and are more readily found in the surface mount versus thru hole.
QuadNet is your best friend and a new found religion.
73's
Pete, N6QW
The QuadNet 6 pole all same cap value.
Note above that the very low end that the curves still remain parallel, whereas below (KK7B) there is a skewing. How much of an issue that is remains to be decided whether just a plot point or real issue. I suspect the 40M SDR police running their FLEX 6600's with 72 Inch LCD will tell you something about your signal at 100 Hertz when running the KK7B.
The KK7B plot, note the very low end skewing.
Bottom line is the use of the tools. I have parts on order for the 8 pole QuadNet and likely will build three versions. The two six poles for the transmitter and the 8 pole for a receiver.
BTW the 1% resistors can be had for 7 cents a piece in lots of 10 from DigiKey. Thus 60 piece of 1% resistors is less than a $5 bill. Of course the 0.1% cost more and are more readily found in the surface mount versus thru hole.
QuadNet is your best friend and a new found religion.
73's
Pete, N6QW