Noise source design question aka 'How Agilent does the magic trick?'

[alex.m]

Hello everyone,

I am currently building a RF white noise source to measure simple filters with a RTL-SDR dongle. I am trying to design a noise source that could provide a fairly flat response up to 1 GHz.
While trying to understand the basic principles of a RF noise source and how they can be built (normally built around Avalanche diode) I stumbled across Agilent 346A/B/C Operating and Service Manual [1]. In page 29 (36 of the pdf file) a simplified diagram is shown.
A particular detail looks puzzling to me. Why the +28VDC that comes from the input BNC supplies a 5.6 kHz oscillator that then is rectified to supply a constant current source that bias the diode? Why not use the DC input to directly supply the constant current source?

[1] http://cp.literature.agilent.com/litweb/pdf/00346-90148.pdf

[T3sl4co1l]

Beats me.

Is it isolated?

Tim

[rfeecs]

The magic of just Google it.

Have a look at this teardown with schematic:
http://www.simonsdialogs.com/?p=1969

[alex.m]

Thank you. 
It seems that the oscillator is used to generate a negative voltage for the current sink that bias the noise diode inside the can. Not really magic. In the block diagram and description it looks way more enigmatic.

[LaserSteve]

For microwave ham radio these days, we use a current limited  avalanche B_E breakdown of an  NPN  RF transistor...
This makes it easier to match to 50 ohms, as well as have wide bandwidth... If you choose one with a "winged" package with 50 ohm leads, life is even easier. Usually these are directly coupled to a MMIC amp.. This avoids sourcing an expensive avalanche diode.

If you'd like a specialized diode with a known calibration, Noise-Com has been known to be friendly for small orders if you have a ham license.

I was not interested in "perfect" so I bought one of these, besides making my own.  All I can say is It works reasonably well as my  SA is an old Avcom and is not known for its gain flatness.  I have not had a chance to take it up to Cleveland and run on the known good HP 22 Ghz SA at my friends office....

http://www.ebay.com/itm/SMA-noise-source-Simple-spectrum-external-tracking-source-DC-12V-/171924968530?hash=item280786d452

Steve

[edavid]

The magic of just Google it.

Have a look at this teardown with schematic:
http://www.simonsdialogs.com/?p=1969

Maybe I'm dense, but I don't see anything there that explains the "why".  Why did they want to drive the diode with a negative voltage?  And if that's important, why didn't they just reverse the drive polarity?

[SeanB]

They need to bias the buffer transistor correctly, so need a negative voltage ( and a positive voltage as well to bias) the transistor so that the output will be around 0V DC without needing any DC isolation capacitors. So they use a negative voltage so the collector of the RF buffer can be connected via a terminating resistor to 0V, and the base biased so it will stay at around that level but amplify the noise nicely. That way there is no loading other than the coax output cable, and no loss in capacitors and terminating resistors. They likely select the actual resistor values based on the gain of the actual RF transistor used, selecting them for a specific gain in a test jig and then further selecting the lowest noise devices for this so there is no added noise over the diode. Then the final adjustment is done during assembly to get the response flat within limits.

[edavid]

So why didn't they just drive it with a split supply?  It wouldn't have been hard to design it so only one side needed to be switched, or they could have had a separate switch lead.

[edavid]

So why didn't they just drive it with a split supply?  It wouldn't have been hard to design it so only one side needed to be switched, or they could have had a separate switch lead.

It looks like this is designed to operate from the accessory power jack of the HP noise analyzer, which may only have the +28V available, so they had top do what they did...

The noise source and noise figure meter were designed together, so they could have done anything.

[SeanB]

True, but they probably wanted a power source for an external antenna amplifier, and many of those run nicely on 28VDC with good gain, so they optimised for that so the major use would work best. Powering an external antenna up the power cable is common, and there are only 2 standards, the old one of 60VAC and the more common one of 28VDC.

[rfeecs]

They need to bias the buffer transistor correctly, so need a negative voltage ( and a positive voltage as well to bias) the transistor so that the output will be around 0V DC without needing any DC isolation capacitors. So they use a negative voltage so the collector of the RF buffer can be connected via a terminating resistor to 0V, and the base biased so it will stay at around that level but amplify the noise nicely. That way there is no loading other than the coax output cable, and no loss in capacitors and terminating resistors. They likely select the actual resistor values based on the gain of the actual RF transistor used, selecting them for a specific gain in a test jig and then further selecting the lowest noise devices for this so there is no added noise over the diode. Then the final adjustment is done during assembly to get the response flat within limits.

Not exactly.  There is no RF amplification taking place and no RF transistor.  The transistor in the current regulator is for DC current only.  There is a blocking capacitor and attenuator after the diode.

The reason for the negative voltage is due to the diode die layout so the die could be directly attached to the housing ground.  A more detailed description of the RF part of the noise source is shown here on page 4 (page 28 of the article) "A noise source for noise figure measurements":

http://www.g8wrb.org/data///HP/8970A_HPJ.pdf

The +28 volt supply has been a standard for many decades for HP noise sources.

[rfeecs]

So now we have the schematic for both the DC and RF section of the 346 noise sources.  No magic, just a diode driven by a current source.

On the +28V, it looks like that was already a standard before the 346 and the 8970 were designed.  For example, here is an Eaton noise from 1980 that also uses +28V:
http://exodus.poly.edu/~kurt/manuals/manuals/Other/EATON%207600%20Series%20Operations%20Manual.pdf

[Fraser]

Interesting thread. I have some commercial noise sources and all seem to require 28V. I have not looked inside mine to see what resides within. I shall have to address that situation 

Fraser

[alex.m]

So now we have the schematic for both the DC and RF section of the 346 noise sources.

Thank you rfeecs. You provide all the pieces of the puzzle. 
It was a quite interesting read.

As reference for my noise source design I am following the examples provide in the ARRL Handbook (a really nice book btw).
Both designs presented uses a NOISE-COM NC302L diode as noise source. Also in rf-microwave.com shop they are several noise diodes available to buy in 1 qt. [1].
In one of their pages Franco Rota (I2FHW) describes how to build a 10 MHz to 3.5 GHz noise source using a NS303 [2].

For microwave ham radio these days, we use a current limited  avalanche C_E breakdown or Diode Wiring of an NPN  RF transistor...
This makes it easier to match to 50 ohms, as well as have wide bandwidth... If you choose one with a "winged" package with 50 ohm leads, life is even easier. Usually these are directly coupled to a MMIC amp.. This avoids sourcing an expensive avalanche diode.

Also in Paul Wade (N1BWT) article, it is stated that an emitter-base junction of a small microwave transistor can be used to make a good noise source (pdf pg. 7) [3].
What are the design parameters I should be looking for to select a RF transistor suitable for the task?
Surely they are more low-cost than a 30€ diode. I am sure that for my applications almost anything will do the job but it is always good to know how the design can be optimized if needed.

[1] http://www.rf-microwave.com/en/shop/0/446-noise-generator-diodes.html
[2] http://www.rf-microwave.com/en/shop/0/446-noise-generator-diodes/4953-NS303.html
[3] http://www.w1ghz.org/QEX/Noise_Measurement_and_Generation.pdf

[LaserSteve]

I'll ask Paul Wade  indirectly, he and I are on a ham radio mailing list.

Edit, question sent  on W1MBA Microwave Mailing list...

Steve

[LaserSteve]

First reply from Jari in Finland was Quote:

" BFR92/BFR96 is quite common noise source. Amplified by a MMIC."

So I do some Googling and there are a few BFR92 Noise source pages:

http://vk1bl.net/html/noise.html

Steve

[LaserSteve]

Kent, WA5VJB  Came back with Quote:


" Up to 1 GHz the MRF901 and NE021 work well.
 For up to 3 or 4 GHz, and reduced Noise up to10 GHz the NEC64535 works well."

Off topic: A bit of a plug for WA5VJBs antenna products:

Wa5vjb.com

Steve

[LaserSteve]

Worth a read:

http://www.simonsdialogs.com/?cat=435

Steve

[LaserSteve]

NL310s are readily available....

http://www.geraldj.networkiowa.com/papers/2012/predictablenoiseA.pdf

Steve