I sent a copy of your text examples back to Dave, and he replied with the following:
[paste] He did not indicate whether he validated his calibration of the VNA, but the results suggst that the calibration was performed correctly. The SA/TG equipment should be discarded; if there's a disagreement between VNA results and SA/TG results, one would always discard the SA/TG results, because an SA/TG is not vector corrected. His statement that the precision 50 ohm load is the most important piece is not correct. The most important bit of the calibration kit is that it's been characterized in a lab that's traceable to NIST, and the measured data for that calibration kit has been loaded into the network analyzer. There is an erroneous assumption that when one connects the precision 50 ohm load and pushes "cal", the analyzer measures the load, assumes it's 50 ohms, and applies a mathematical correction to itself. That's not really how it works. Instead, you first enter the calibration kit's coefficients into the VNA (usually by loading a file from a floppy disk that came with the kit), then you attach the 50 ohm load and press "cal". The analyzer measures the load, then refers back to the calibration file and corrects itself so that it will measure that load as it was measured by a traceable analyzer at a lab. The load could actually be 60, 70, 80 or whatever ohms, as long as it has been accurately characterized, and the data entered into the VNA. However, the 85032 loads are good enough that at the low frequencies we're discussing, it's close enough to 50 ohms that if he doesn't bother loading the cal coefficients, he's probably not measurably in error. As I said, his measurements suggest that his cal routine was adequate. Getting fussy about having the exact cal coefficients in the instrument is only worth bothering with if you want to make return loss measurements of devices better than 40dB or above several GHz. The values he measured are consistent with some old and some poorly-assembled connectors that I have measured. RF connectors are not intended to be used and re-used more than a few hundred times; in the instrumentation industry we think of them as "dispoables" and keep a budget item for their replacement. The characteristics of RF connectors varies a lot by how well they're manufactured. I have seen no correlation whatsoever between price and measured performance. The stack of adapters that I measured some years ago was simply gathered. I cleaned them before measurement, and I discarded the type N connectors which had "fingered" ground connections that did not appear tight, as those are a well-known trouble spot. I'm not sure why anybody built type N male connectors like that, as it's a lousy way to do things. Adapters in particular, take a lot of abuse. It's good to measure these things. If you don't, then you have no idea whether you're using a good connector or a bad one. D > "nj902" <[EMAIL PROTECTED]> wrote: > > I hope nobody is buying into this BS. > > Connector and adapter mis-match and loss issues are hardly "amateur > radio lore" > > RF manufacturers and instrumentation companies have stressed this for > years - justifiably. > > It's easy to verify if you have access to instrumentation. Just out > of curiosity I decided to run a couple of tests. > > When you measure a device for SWR you are actually measuring return > loss. You can do this on a network analyzer or with a spectrum > analyzer and tracking generator plus a signal separation device such > as a return loss bridge or directional coupler. > > I decided to try both instruments and compare the results. You first > calibrate the measurement setup with precision components from your > cal kit [e.g. Agilent 85032B]. The critical component is a precision > 50 ohm load. With a good bridge or coupler you should be able to > measure a return loss of 40 dB or greater [equivalent to an SWR or > 1.02:1] > > Normally your instrument test ports will be type N although some > instruments have APC-7 test ports. Your cal kit should match the test > ports. > > I started with a full 2-port cal of the VNA and a normalization of the > Spectrum analyzer. My reference load did measure as expected. > > The next step is to "insert" the device under test. In this case we > are interested in looking at a UHF connector. Obviously, since the > instrument is type N, we have to use adapters. Just out of curiosity, > I decided to compare two pairs of adapters. First an NF-NF mated to > an NM-NM. [UG-57B/U and UG-29B/U] This created an "insertable" device. > These are "standard" adapters as opposed to "precision" components > and as expected, there was some degradation of the return loss > measurement. > > I then compared this to two other adapters, NM to UHF-F and UHF-M to > NF [UG-146/U and UG-83/U], again creating an "insertable device" This > pair has just a single UHF male to female junction and results in a > significant decrease in return loss [i.e. increase in SWR.] > > I also tried inserting an 90 degree UHF adapter [UG-646/U] in line > between the UG146/U and the UG83/U which resulted in further > [significant] return loss degradation. > > I measured several sets of these parts and the results were not only > consistent from part to part but generally match within a few tenths > of a dB between the VNA and the spectrum analyzer. I always like to > see comparable results from two significantly different instrument > setups - it confirms your thought process and results. > > Results at: 150 MHz, 450 MHz, 900 MHz > > NM-NF adapters: > [VNA]: 36 dB, 35 dB, 31 dB > [SA/TG] 37 dB, 35.7 dB, 30 dB > > N-UHF + UHF-N adapters: > [VNA]: 26 dB, 16.5 dB, 11.4 dB > [SA/TG]: 23.6 dB, 17.2 dB, 11.5 dB > > N-UHF + UHF 90 + UHF-N > [VNA]: 20.2 dB, 12.03 dB, 9.3 dB > [SA/TG]: 21.6 DB, 11.9 dB, 9.1 dB > > Now just for reference, a return loss of 11 dB would be equivalent to > an SWR of 1.785 : 1 [ONE UHF junction @ 900 MHz] and a return loss of > 17 dB would be an SWR of 1.329 : 1 [@ 450 MHz] > > UHF connectors "invisible" below a gigahertz??? Hardly. > > > Also - someone asked about the 90 degree N fitting at 900 MHz so I > tested a few of those. That's easy because being M on one side and F > on the other, it is inherently "insertable" > > The round kind [UG-27A/U] averaged return loss of 22 dB. The square > ones[UG-27D/U] are better at 27 dB [1.094 : 1 SWR] > Yahoo! Groups Links <*> To visit your group on the web, go to: http://groups.yahoo.com/group/Repeater-Builder/ <*> To unsubscribe from this group, send an email to: [EMAIL PROTECTED] <*> Your use of Yahoo! Groups is subject to: http://docs.yahoo.com/info/terms/