Wenzel 5 and 10 MHz Ultra Low Noise OscillatorsWenzel makes some of the lowest phase noise crystal oscillators you can buy. Their noise floor (out 10kHz or more from the carrier) is 10 to 20 dB better than most frequency standards.
Thanks to a loan from TVB, I was recently able to put four of these oscillators on the bench and test them against each other, as well as an Oscilloquartz 8607-008 BVA, which has lower close-on noise, but can't match the Wenzels' noise floor.
This page shows the results of those measurements. I've also created another page that discusses the Wenzel ULN's sensitivity to whatever is on the power supply pin. Because of this sensitivity, all these plots were taken with the oscillators running from a very low noise voltage regulator fed from batteries that were isolated from the AC power system.
The four oscillators tested were one each of:
- 5 MHz 501-04534F
- 5 MHz 501-16833
- 5 MHz 500-12610 (proprietary, but similar to 501-7175E
- 10 MHz 501-16835
The 501-04534F and 501-16833 are electrically identical, but packaged differently. Both are "premium" 5 MHz SC cut oscillators spec'd at -120dBc/Hz at 1 Hertz offset. The 501-16835 is the 10 MHz version of the same oscillator, spec'd at -105dBc/Hz at 1 Hertz. The 500-12610 is a proprietary part number for a 5 MHz "Blue Top" oscillator which is quite a bit larger and heavier than the standard ULN, and is spec'd at -115dBc/Hz at 1 Hertz.
First, here are phase noise and Allan Deviation plots of the four
Other tests have shown that the two "normal" 5 MHz ULNs had almost identical performance, so I'm not including the full set of possible comparisons in this plot. There are three important things to note:
- The noise floor measurements are limited by the TSC-5120A phase noise
set, which is specified at less than -169dBc/Hz for 5 MHz inputs, and less
than -175dBc/Hz for 10 MHz inputs. The 5 and 10 MHz ULNs are all spec'd
to have a noise floor of -176dBc/Hz or better, so beyond a few hundred Hertz
they are better than the TSC.
- When comparing two oscillators of similar performance, such as "TVB 5 MHz"
and "N8UR 5 MHz #1", it's appropriate to subtract 3dB from the results. I
have not done that in these plots. The 1 Hertz offset of the pair
of 5 MHz units shows about -125dBc/Hz here; subtracting 3 dB would yield
-128dBc/Hz at 1 Hertz offset, which ain't bad at all!
- When using input and reference signals of differing frequencies, the TSC requires that the results be scaled. I have scaled the 10 MHz results in this (and all subsequent phase noise plots) by subtracting 6dB.
Here are the Allan Deviation plots gathered from the same data runs:
At extremely close offsets (less than about 10 Hertz), the Oscilloquartz 8607-008 BVA is better than the Wenzels; the one here has measured noise of -130dBc/Hz at a Hertz, though its noise floor is only around -160dBc/Hz. The following plots show the four ULNs measured against this BVA; data beyond about 10 Hertz for the 5 MHz units, and 30 Hertz for the 10 MHz unit, are limited by the BVA's noise floor. At 1 Hertz offset, the approximate -127.5 reading of both 5 MHz standard units lines up nicely with the results above.
The improved performance of N8UR #1 over TVB 5 MHz at longer tau is probably due to #1's much longer cumulative running time and associated reduced aging. And, N8UR #2's better still long term performance may be due both to running time, and to better thermal stability because of its much greater mass and increased heater current (I believe that the Blue Top oscillators have a double oven).
Finally, here's a direct comparison of the 5 MHz versus 10 MHz ULNs, both measured against the third premium ULN: