Comparing Z3801A Receivers

I managed to borrow a few Z3801A GPS-discplined frequency standards to temporarily add to my own two units, and ran a comparison of their performance. This is the first part of a two-step process to characterize the Z3801As; the second step will measure the short term stability of the oscillator output. I'm still assembling the equipment to do that part with the precision I'm looking for.

This page shows a chart of the internal stats for each of the five receivers.

Test Configuration

The test period began at 20:32:30 UTC on 9 February 2003 and ended, due to technical difficulties, at 07:01:02 UTC on 7 March 2003.

The first two receivers, N8UR #1 and N8UR #2, had been running without interruption since 9 January 2003, while the other three units were all started on the afternoon on 9 February 2003 and therefore had been running for a couple of hours when the data capture began (of the three, N8ZM #1 had the most time to stabilize before the test began).

All five units shared the same antenna. Because I didn't have a proper antenna splitter that could feed all five units, there's a bit of a feedline haywire (I used Radio Shack DSS satellite splitters and pre-amps to kludge together a four-way splitter to add to the two HP 2-way splitters I already had). Because of this, N8UR #1 and N8UR #2 probably had somewhat more signal to work with than the other three units.

The two N8UR units shared a power supply, as did MVUS #1 and N8ZM #1. WB8GXB #1 was on its own supply.

N8UR #1 and N8UR #2 were physically adjacent to each other in one location (on top of an equipment rack), while the other three units sat next to each other on a table about 30 inches away, and 36 inches lower. The N8UR units were closer to the ceiling and to the outside wall of the basement; they were probably exposed to a bit more temperature gradient than the others.

All data was logged using custom software on a Linux machine (one that has many serial ports!). There's info about the software at the bottom of my main GPS page.

There were two charts generated for each receiver. The top chart shows the time interval between the 1pps signal derived from the unit's 10MHz oscillator and the "filtered GPS 1pps" time signal (The HP documentation I have doesn't describe what this filtering is).

The second chart has two graphs. The one in blue, with its Y axis labels on the left, shows the EFC correction applied to keep the oscillator on frequency. This is a relative value (using the minimum recorded value as zero). The actual EFC value isn't too meaningful as long as it's within the unit's correction range; all these units were operating with EFC values between 700,000 and 800,000, well within the safety zone. Information from the web indicates that each EFC step equals a correction of about 1x10-12.

The graph in red, with its Y axis labels on the right, shows the "Predicted Uncertainty" which is the receiver's calculation of how much time it will gain or lose during the first 24 hours after the GPS reference is removed.

Note that the Y axes of the time interval and EFC graphs are autoscaled, so you can't directly compare the plots between units as the scales are probably different. Also, the Y axis of the predicted uncertainty graph is logarithmic rather than a linear scale; this reflects the fact that while the maximum PU is 432 microseconds, a reading much above ten microseconds indicates a problem.

Note 1: If you're having trouble displaying these graphs in Internet Explorer, here's info on why, and how to fix the problem.

Note 2: On the evening of 20/21 February, a glitch caused data to be lost for several hours for MVUS-1, N8ZM-1, and WB8GXB-1. On the afternoon of 25 February, there is another short data loss (about 20 minutes) for those three units while some housekeeping was done.


The Results

Last 36 hours -- Click on a graph to see the full record...

... or click here to see the full record of all charts.