SoftwareSRAM SetupSRAM I/ODATA XferDevice SelectMulti-Freq PSKSimple  PSKPerformanceCommand Line Call

Empirical Measurement of the PMG-M-130 Spectral Purity

This chapter is dedicated to showing the test results of the PMG-M-130. The test setup used for each of the tests will be shown specifically for the test data displayed.  We have constructed a simple spectrum analyzer using the PMG-M-130 to cover a 1 MHz band.  The basic test setup is shown below for the two analog output channels from 1Hz to 130MHz.  The PLL test is conducted with two PMG-M-130 devices in the same system using a similar test setup as the two analog channels.




This test is a simple baseline test with the analog channels 1, 2 and 4 off, the clock channel 3 on at 100KHz and the output sitting at about 1 volt DC.  The channel 1 analog output is then activated and set to 2.5 KHz with a four volt peak-to-peak sinewave output.  Since this OEM design has the options of selecting a broad range of filters they are generally selected based on the OEM application.  For this application we are using a simple 4 pole 7 KHz low-pass active filter on board. There are connectors for separate modules that may be tunable from 1Hz to 130 MHz on a separate daughter board for two channels and a separate VCO connection for the PLL channel. The Phase Frequency Detector (PFD) is an on-board function incorporating a 1Hz to 80MHz integer divider. For higher frequencies this would be part of the external VCO circuitry and may be programmable via the user configurable IBRs. These are prototype test results from the PMG-M-130 and we are expecting much better baseline results from the Production OEM unit with off board filtering. The graphs below are of the actual raw data using a Success Approximation type ADC analog input card. This is used to obtain the absolute noise spectrum of the channel. Shown also below is a similar graph that a analog input card using a moving average filtering algorithm.  With just averaging 2 points per point the spectral noise baseline drops an additional 20 db average. This is very misleading when publishing specifications. BASIL Networks will be publishing other waveforms of data collected from other analog input cards as time permits to see the effects of various ADC technologies.

1.00 MHz Sampling Rate 256K points

Click on image to enlarge

4 Volt Pk/Pk 2.5KHz sinewave Channel 1
50KHz sampling Rate,  12.5KHz Marker 256K points
Click on image to enlarge


Typical moving average filtering for the 256K point array changes spectral dynamic range measurements.

This is a typical moving average smoothing from MathCad to show what happens when different data collection schemes are used. X and Y are the data arrays, indexed from X = 0 to N-1 points. the variable w is the number of moving average points, in this case it is set w=2 points.

As you will see below the spectral noise has dropped an additional 15 to 20 dB average.  When testing true performance it is very important to choose the right test methods and well as the proper data collection methodology. In a system this would generally be overlooked and may show that each sub-system functions within specifications, however the total system performance may not.

BASELINE NOISE with Moving Average
1MHz Sampling Rate 256K Points

Click on image to enlarge

Moving Average A/D Conversion
Moving average every 2 points, 100KHz Sampling Rate

Click on image to enlarge

Copyright 1990-2018 BASIL Networks, PLLC. All rights reserved