Comparing three versions of Noise Source

 Below is the set of comparative results for the Nooelec, secondly my first version using a 5.1 volt zener diode and thirdly my improved version with a 7.5 volt zener selected from the same series as the 5.1 volt device. The Rigol DSA815TG was set to 0dB attenuation and the noise sources each had a 20dB attenuator fitted. That was to reduce the total noise to a safe level within the spec of the Rigol. That means the actual noise measured at 150MHz for example from my 7.5 volt version was not -58.23dBm but -38.23dBm which roughly equates to the recognised standard of S9 +35dB (subject of course to the bandwidth in which the signal is measured).

 

 Scans are included below which show the various comparative results for the latest 7.5 volt version. To smooth the traces to provide accurate readings the lowest practical VBW/RBW setting was used which means the various indicated baseline and noise levels will vary. As the resolution bandwidth is decreased the trace will lose roughness, become smoothed and drop down in indicated level. For example 10KHz gives 85dBm, 3KHz gives 92dBm, 300Hz gives 103dBm, or 10dB change for each factor of 10 decrease in bandwidth. A quick check on the 20dB attenuator accuracy showed that removing it when testing 150MHz proved the noise jumped from -59dBm to -39dBm but with the Rigol complaining about something... probably total power output too high. My guess is the crackle produced by the on/off switch being turned on tipped the noise output to a larger figure for a brief instant.

Results are not as I expected... all the pictures are from the 7.5 volt zener version.

 NOISE SOURCE

 Measurement Bandwidth

 Adjustment dB

 NOOELEC

 G3PIY 5.1

 G3PIY 7.5

 FREQUENCY CENTRE Hz

 RBW/VBW

 Relative to 100Hz

NOISE dBm Note 1

NOISE dBm Note 1

NOISE dBm Note 1

 150M

 10KHz

 -20.0

-51.34

-58.77

 -59.23

 15M

 3KHz

 -14.8

-63.14

-65.59

 -66.20

 1M

 300Hz

 -4.8

-62.55

-78.21

 -66.79

 250K

 300Hz

 -4.8

-62.83

-81.29

 -60.52

 100K

 100Hz

 +0

-66.76

-86.12

 -64.83

 25K

 100Hz

 +0

-53.23 **

-88.77

 -70.25

 750M

 300KHz

 -34.8
   

-52.37 *

 Note 1. The figures for noise are dependent on the bandwidth setting and also the true noise output would be 20dB higher once the attenuator has been taken into account. For example the 150MHz noise signal of -59.23dBm shown above is really -39.23dBm but relative to the 15MHz reading which is measured in a narrower bandwidth of 3KHz subtract 5.2dB making -64.43dBm (= -44.43dBm true roughly S9+30dB). See scans below for parameters used during each test. Of course the lower numbers mean stronger RF.

 

* I added the scan for 1.5GHz as this shows the remaining oscillation proving difficult to eliminate and no doubt messing up noise output in that frequency range, however all is not quite as it seems because this seemingly strong 750MHz noise level related to the 15MHz noise can be translated as -52.37-(34.8-14.8)=-72.37dBm in a similar bandwidth of 3KHz.

** The Nooelec LF/VLF figures shot up after I'd added extra 12 volt smoothing to their circuit board to eliminate serious interference from the 12 volt booster chip.

 

 

 

 

 

 

 

 

 

 

 

 
   

 

 Latest circuit above, possible final circuit below

 

 I carried out a further test once I'd repaired the sticky pointer on my HP Wattmeter. Previously I'd measured total noise power from the initial circuit as +1.4dBm but checking this measurement with the 7.5 volt zener gave me areading of -1dBm**. I then inserted my 30MHz low pass filter to see the level within the range seen by a standard shortwave communications receiver and measured -13.6dBm. What do these numbers mean in practical terms? This is not straightforward because of the nature of the noise output but let's say to give a reading on the power meter of -13.6dBm at 7MHz would require a discrete CW signal of about 47mV or roughly S9 +60dB. The total power of -1dBm is similarly equivalent to 200mV and the previous value of +1.4dBm about 260mV. Comparing with the Nooelec gave readings of +12.8dBm total and -5dBm filtered. All these figures are shown below for clarity. At first sight (by comparing total power and filtered power), the Nooelec is producing more power at frequencies beyond the capability of my noise source, in the region 2 to 10GHz.

How do these numbers relate to the spectrum analyser scans above? Looking at the shortwave scan the signal strength of the noise at 7MHz is -62dBm and, taking account of the 20dB attenuator, this represents about -42dBm which is about 1.8mV or roughly S9+30dB. This represents a reduction of about five S-points. In terms then of using the homebrew noise source for shortwave receiver alignment the equivalent noise signal is about the same as that from a signal generator producing 1.8mV. Therefore, to reduce this level to that of a weak signal requires 40dB to 60dB of attenuation.

** Note that this discrepancy was probably due to the effects of instability and oscillation at 1.2GHz

 

 Noise Source

 Total Power

 Total Power

 Volts equiv

0-30MHz power

 0-30MHz volts 

 0-30MHz S Meter

 G3PIY

 -1dBm

 0.8mW

 200mV

 -13.6dBm

 47mV

 S9+60dB

 Nooelec

 +12.8dBm

 20mW

 1.0V

 -5dBm

 126mV

 S9+68dB

Return to Reception