AnsweredAssumed Answered

Noise of 10+ year old HP 8720D vs a new Agilent N9923A

Question asked by drkirkby on Oct 13, 2012
Latest reply on Oct 16, 2012 by daras

Dr. Joel Dunsmore said on another thread about VNA noise figure
Noise figure of a VNA


*There are three or four common receiver types, for worst to best:*

*1) Sampler based: very high noise figure - 8753 and 8720 use samplers operting on 200th to 300th harmonic of the VCO*
*2) High harmonic mixer (or low harmonic samplers) - many mm-wave systems use this; operation on harmonics up to 10 or 12 or more*
*3) Low harmonic mixer - PNA uses third harmonic mixing in analyzers above 26.5GHz, PNA-L uses it above 13 GHz, I think.*
*4) Fundamental mixing - Lowest noise figure.*

when I asked what the N9923A FieldFox uses, he said "fundamental mixing".

I deceided to to compare noise levels of two quite diferent VNAs I own: 

1) HP 8720D
At least10 years old, 50 MHz-20 GHz, lab instrument, which uses the type of receiver with the highest noise figure. 

2) Agilent FieldFox N9923A
Current model, 2 MHz - 6 GHz, portable instrument, which uses the type of receiver with the lowest noise figure.

Needless to say, I was expecting the noise to be lower on the N9923A. But at least according to the method I used, this was not the case. Perhaps I'm doing something wrong. 

The following settings were used on both instruments. 

Averaging = 1
IF bandwidth = 1 kHz
Span = 50 MHz to 6 GHz
Points per scan = 1601

I then measured S12 with nothing on the test ports of either machine. 

Because of the very nature of noise, the peaks on the traces can occasionally be very high. With 1601 points collected, I adjusted the reference level so the noise exceeded the reference only once or twice per sweep. This was done by eye, rather than collecting data down a GP-IB, USB or LAN. So this is near the maximum of the noise level, rather than any sort of average. Only the occasional data point shows a noise higher than this. 

The results were: 
* HP 8720D = -87 dB
* Agilent N9923A = -75 dB

Adjusting the reference level so about half the points are above the reference and half below gives lower noise levels of course. 

* HP 8720D = -100 dB
* Agilent N9923A = - 90 dB

So it appears to me the noise level on the older model is about 10-12 dB *less* than on the new model. 

The noise on the N9923A appears to be about 5 dB lower at 50 MHz than 6 GHz, whereas on the 8720D the noise was flat with frequency to 6 GHz - the highest frequency at which I have an N calibration kit. (I could have switched to 3.5 mm and calibrated to 9 GHz, but I did not do so. In any case, the N9923A's maximum frequency is 6 GHz). 

The fact the N9923A is newer, it almost certainly has a higher resolution screen, so when sweeping 1601 points, an occasional peak is perhaps more likely to be seen on the higher resolution screen than on the lower one of the old HP 8720D. But there is no way such an effect could explain a 10-12 dB difference. 

Any thoughts why I'm getting these results?