Diy Noise Source up to 3GHz

[Yansi]

Hello!

I am just in need of a noise source, that would work up to 3GHz with some significant measurable output.  I have tried myself to make some circuit for that, but failed. Can you help me please what I have done wrong?

The schematic is here.  As the noisy element, I thought that a reverse biased BE junction of an UHF transistor will produce enough RF garbage. But, nope. With the 2SC5773 which I have plenty of, the thing just sits and does nothing. (Output monitored on spectrum analyzer)

Then I have changed the transistor to some old soviet KT382AM, which is a ~2GHz something NPN. Bingo! This one does something around 4mA, but the production of noise is weak and over 400MHz it is no longer detectable on the SA. (way below -80dBm).

How critical is the current setting? (I think it should be fine adjusted for any device). Is my circuit approach correct, or have I screwed?

Currently I do not have those MMIC amps connected, just testing the noisy element itself. The noise have to be reasonably measurable level even without the gain stages.

Thank you for help 

[Yansi]

I could buy whatever, even the chinese noise sources (they are just about usable at the range of my interest), however I needed this for an experiment and didn't want to wait for parts to come. Hence why I am building it.

Yes, the small 2SC5773 does pass significant current even at very low reverse BE voltages. It almost seems it does not have any kind of a zener like behavior.

So do you think the schematic approach is correct? (current source into the noisy junction, terminated to 50ohm, amplified)

[Wolfgang]

Hi,

if you want a flat response its better to use a diode instead of a transistor. A Zener at 6.3V should work.

[German_EE]

I am just in need of a noise source, that would work up to 3GHz with some significant measurable output

It's called a 'wife' or SWMBO, they're expensive and not really worth the trouble 

[Wolfgang]

How many dB is "significant".

Most (broadband, well matched) noise sources provide between 7 and 20dB.
Is that what you need ?

[EVS]

its better to use a diode instead of a transistor.

This is not quite true. Even with a simple BFR93A you can get a very decent result:

[RoGeorge]

with a simple BFR93A you can get a very decent result

That spectrum looks very good. 
Can you post the schematic, or a link to it, please?

[EVS]

This project is a cheapest alternative for Micronetics NMA-2414 series noise modules.
This is my own design, but it's not my property, sorry  .
And if you have any technical questions - I'll try to answer them.

[Yansi]

How is the BFR93 supplied current and how is it matched to the output? The performance seems to be too god to be true. 

[EVS]

How is the BFR93 supplied current and how is it matched to the output? The performance seems to be too god to be true. 

- Series resistor without a current source
- No special measures for matching, just a base-emitter juction decoupled by cap
- I don't care what you believe.

[Wolfgang]

Very good is relative. When you look at the vertical scale, this noise source is +/-2dB up and down at least.
The commercial ones are usually within +/-1dB, the good ones within +/-0.5dB.
I am not surprised by this one, its OK.
Could you make a measurement with better vertical resolution ?

[RoGeorge]

Very good is relative. When you look at the vertical scale, this noise source is +/-2dB up and down at least.
The commercial ones are usually within +/-1dB, the good ones within +/-0.5dB.
I am not surprised by this one, its OK.

Indeed, everything is relative. Please share if you know other 6GHz wide noise source, with at least the same flatness than the exemplified BFR93A, and that can be approached at a DYI level (also, a BFR93A is about $0.4 sold in 1pcs quantity).

Do you have any examples?

[rhb]

The EVS design is so much better than an Ebay BG7TBL there is no comparison.

[Wolfgang]

Plus or minus 2dB should be makeable. When you need to be more flat, it gets difficult for a very wide range. For a smaller range (a few 100MHz), sub 0.5dB flatness can be achieved by a DIY solution, like here:

https://electronicprojectsforfun.wordpress.com/making-noise/

There is an old HP article how they constructed their commercial noise sources (I dont find it right now), but IIRC it was a coax or stripline design with SMD parts and a stripline equalizer. The point is not if you can homebrew this, but how stable and repeatable your design are  The money is not in the parts, but in the time spent to make it work.

On top of that, commercial manufacturers age and temp cycle their noise sources.

[Yansi]

The EVS design is so much better than an Ebay BG7TBL there is no comparison.

Not sure what BG7TBL means, apart that I have seen this designation with a lot of chinese RF junk, but have had a hands on experience with some extremely hot running chinese noise source board with I think a couple of ERA-5 amplifiers, that produced "something" (not anywhere close to flat by any means) up to some hundreds MHz, then steeply and wildly rolling off.

I should have image somewhere...

What current do you run through the EB junction of the BFR93? (or, respectively, what voltage and resistance). Guessing higher resistance with more volts is better.
It seems promissing, so I will try to repeat your result. A broadband noise source even within 2 dB is good enough for a homegamer, if it can do at least those 3GHz, even better those 6!

//EDIT: Found it! The upper trace is directly the output of the chinsy RF noise source. Well, lol. Not flat anywhere it seems. (Ignore the lower trace)

[Wolfgang]

The BG7TBL ones are so wobbly that they are not suitable for a reasonable filter align job 

[Wolfgang]

Just to show others hacking with same approach and what they found out:

http://www.simonsdialogs.com/2015/08/tws-15n-noise-source-some-rf-transistors-as-noise-generating-devices/

Also using BFR93. No rocket science anywhere. Its a common approach.

[rhb]

The EVS design is so much better than an Ebay BG7TBL there is no comparison.

Not sure what BG7TBL means, apart that I have seen this designation with a lot of chinese RF junk, but have had a hands on experience with some extremely hot running chinese noise source board with I think a couple of ERA-5 amplifiers, that produced "something" (not anywhere close to flat by any means) up to some hundreds MHz, then steeply and wildly rolling off.

I should have image somewhere...

What current do you run through the EB junction of the BFR93? (or, respectively, what voltage and resistance). Guessing higher resistance with more volts is better.
It seems promissing, so I will try to repeat your result. A broadband noise source even within 2 dB is good enough for a homegamer, if it can do at least those 3GHz, even better those 6!

//EDIT: Found it! The upper trace is directly the output of the chinsy RF noise source. Well, lol. Not flat anywhere it seems. (Ignore the lower trace)

BG7TBL is a Chinese call sign, however, there is no information about the holder of the call or even if the call was ever issued.

The lower trace looks rather like my BG7TBL "nose source" :-(

Over 10 MHz it's flat enough to be usable with an SDR, but obviously not suitable for anything more demanding.  And even with lots of heatsink it still gets quite hot.

[EVS]

What current do you run through the EB junction of the BFR93? (or, respectively, what voltage and resistance).

Various BJTs have different breakdown voltage. For BFR93, BFR520, BFT25, 2SC3356 and similar it's about 4.5-5.5V (depends on the current).
Therefore, for a supply voltage of 9V a resistor of 1-2 kOhms will be a good starting point.

Guessing higher resistance with more volts is better.

Well, that's not exactly true. The ohms and volts of more than necessary does not make sense.

[Yansi]

Guessing higher resistance with more volts is better.

Well, that's not exactly true. The ohms and volts of more than necessary does not make sense.

I'd think it is, because too low resistance would I guess load the noise source too much and eat some of the noise power generated, wouldn't it? Of course going to silly high values will not help.

[rhb]

If you've not already done so it's important to read this article by Franco Rota:

https://www.rf-microwave.com/app/resources/uploads/diodes/VHFComm_NW303.pdf

Here is more information and nice color pictures of the prototype and the final design.

www.eme2008.org/papers/Lab%20equipment/i5sxn_Noise%20generator.pdf

Franco presents some important information about the self resonance effects of SMD device orientation and the issue of enclosure resonances.

I'd like to have a precision noise source without having to spend $$$.

A question for EVS:  Would a precision current source flatten out the spectrum shown on the SA?  It seems to me that current variations during the sweep might account for some of the variation in apparent output.  Of course, averaging would eliminate those.  Could you show an expanded vertical scale with and without averaging?

[Yansi]

With a $$ hefty  noise diode such as the NS303, everyone can build a nice noise source.

I thought we were talking about cheap deigns?

[Wolfgang]

Its not only the noise diode, but also the Rogers PCB and a milled case ...
If you can do all that, you could probably also pay for a commercial noise source on eBay.

If it does not have to be a Keysight or Noise.Com I can recommend a RFD2305 from G8FEK.
Thy are calibrated and not too expensive.

[EVS]

Would a precision current source flatten out the spectrum shown on the SA?

No. Source DC mode almost doesn't affect the ripples inside the operating frequency range.
And... Low ENR (10-20dB) noise souce and generator with more than about 60-70dB are completely different devices.

[rhb]

If someone will create the board files, I'll make enclosures.  I have a mill and a 20 ton press.  What I have in mind is making dies and forming the RF cavity from aluminum sheet which is then thermally bonded to any available heatsink such as from scrap PC PSUs.  The Rota design should lend itself to being formed from  aluminum.   So I could make several dozen enclosures in about an hour of working time at very low cost. 

Materials would run $2 for 0.062" x 2" x 5" to  $4 for  0.125" x 2" x 5".  That's a top and bottom piece that clamps on either side of the the PCB with screws which would also clamp  heat sinks to the assembly.  I'm certain I can form the 0.062 with a 20 ton press, but the 0.125" might require a heavier press.  For this operation it would cost under $200 to build a 50 ton unit.

The dies would be the male equivalent of the female cutout in the Rota design. A metal box filled with soft rubber is then used to form the metal to the shape of the die using the 20 ton press.  It's an excellent process for low relief shapes.

I'm happy to do such enclosures for any OSHW noise source someone designs.  For testing one can use a rectangular enclosure and RF damping material  and once the circuit is developed then design the formed enclosure and eliminate the damping material.

Except for my power meter,  all my gear stops at 3 GHz.  I'm really not interested in going much higher, and certainly not any time soon.  So the Rota design is somewhat overkill for my needs.