Progress with my PHSNA calibration
anglerhamming
Hi everyone,
Latest update. I zeroed out the polynomial coefficients in the Parameters.txt file. Next, took new data with the homebrew VE7BPO 10 dBm sine wave calibrator (checked against the K3NHI CMOS 10 dBm square wave calibrator) and my homebrew step attenuator inputting to the LPM with the SNA running. I got data from the SNA at the 10 dbm points from 10 dBm to 70 dBm. The PHSNA program calculated the slope (0.09083) and intercept (82.93 dBm), which I manually entered into a copy of the Parameters.txt file. Then I connected the SNA output to the LPM, and I took data of actual dBm from 1 MHz to 30 MHz in 1 MHz increments. I entered the dBm data points into an Excel 2003 spreadsheet, and I got the scatter plot with the 5th order polynomial trend line. I couldn't get the actual equation at that point, so I brought the .xls file upstairs and imported it into my wife's business PC that has a much newer version of Excel. Tomorrow, I will enter the coefficients of the equation into the Parameters.txt file and then proceed with setting up for the return loss bridge (a homebrew version) and the crystal test fixture. This stuff is really interesting! Ted, KX4OM


Nick Kennedy
Glad it's coming together. I use the same system  the CMOS square wave generator, which then calibrates the VE7BPO sinewave oscillator. I was just thinking I should offer to run your Excel data file through my program to calculate the coefficients, but I'm glad you found a better way. Watch out for it defaulting to a small number of significant digits on the coefficients. There are plenty there, but you may have to tell it to show them. 73 Nick, WA5BDU
On Wed, Nov 16, 2016 at 5:59 PM, anglerhamming@... [PHSNA] <PHSNA@...> wrote:


anglerhamming
Nick,
Thanks for the heads up on the coefficients. They seem to be in line with Jim's example in his PDF on "adding calculations to compensate..." Here is what I have from my plot, in the format of the Parameters.txt file: float C0 = 7.0624; float C1 = 0.2112; float C2 = 0.0544; float C3 = 0.004; float C4 = 0.0001; float C5 = 2E06; I'll create a new folder in the Files section to hold some stuff from my build, and I'll upload the plot with the equation tonight. Ted, KX4OM


anglerhamming
Nick,
I loaded the .xslx file in LibreOffice Calc version 5.1.6.2 on my laptop, and it displays the equation with a *lot* more decimal places than it did on the Windows 10 machine where I did the plot. For example: C0 = 7.0623883289, and C5 = 1.51876617279347E06 I also inadvertently created my new folder under someone else's, and I had to redo it, hence 2 upload messages. Apologies to K5ESS. Ted, KX4OM


Nick Kennedy
I'm glad to learn that the LibreOffice spreadsheet does calculate the coefficients. I don't know how many significant digits we need, but I remember an instance of just having two or three and that wasn't enough. The proof will be in the pudding when you do a frequency compensated plot and see it nice and flat. 73 Nick, WA5BDU
On Wed, Nov 16, 2016 at 10:51 PM, anglerhamming@... [PHSNA] <PHSNA@...> wrote:


anglerhamming
I entered the coefficients in long form (all they appeared in the plot) in Parameters.txt, I ran a few sweeps, and I've uploaded the plot of one one to my folder in the Files. It is 1 to 30 MHz in 10 kHz. I also uploaded the raw data as a text file. The program normalized the compensated plot to 0 dBm.
I ran the program to measure my homebrew K8IQY return loss bridge, which checked out at 18.something dBm. The crystal test board has a problem. About 60 dBm output with the board jumpers set per the program, also verified with my 20 dBm test source. I probably mishandled the ERA3 during the buildup of the board. I got 3 free samples, so I can fix it once I've eliminated all of the other possible causes. Ted, KX4OM

