Wednesday, July 16, 2014

Modifying BG7TBL noise source

I own an old IFR-A7550 1GHz spectrum analyzer. It works fine, but lacks tracking generator, so its usage is somewhat limited.

The tracking generator for my unit is difficult to find, and very expensive, so aimed by Iban, EB3FRN, I explored other solutions, for example a noise source.

I bought this noise source in eBay. It works at 24 volts, and produces a lot of noise. Around -30dBm @ 3MHz BW on HF but around -50dBm around 1 GHz. But after playing a bit with it, a problem become evident. It gets hot. Very hot. Really hot!. It gets so hot it starts to works erratically after only 10 or 15 minutes.

I assume the 24 volts operation comes from some industry standard, but for me this is a problem. It would be much nice to work from 12 volts. So in this way, I could even install it inside the spectrum analyzer instead the tracking source.

The noise source design is classic. The PCB is so clean you can deduct the schematics at glance. A zener diode produce noise and three cascaded MMIC to amplify it. A small attenuator is placed at the output, to prevent weird impedances to the last MMIC I assume. The MMIC are ERA-5, and according to the datasheet, they work at 5 volts, so a 12 volts conversion could be made if I change the polarization resistors to 110 ohms.

The zener diode is a 22 volts one, so I replaced it with a 5 volts one. I also replaced the MMIC resistors with 100 ohm 1W ones (close enough). After the modification, I tested the unit at 12 volts. The unit did not produce a lot of noise, and the power level fall off very quickly, being barely usable around 1 GHz. I tested some currents along the zener diode and noise peaked at 15mA, but level at 1 GHz was noticeably inferior to the one at HF frequencies, something around 40dB weaker. It seemed the zener diode had a lot of capacitance.

I tested some zener diodes, with different voltages at different currents. Each diode peaked at a different current, but all of them exhibits a more than significant noise power reduction at 1 GHz compared to HF frequencies, from 20 to 50dB, according to the diode used. I tested plastic, glass and even SMD zeners, but none of them worked as expected.

Suddenly, I remember what is commonly used to generate RF noise. It was not zener diodes, but transistors! More precisely, the emitter-base union reverse polarized (yes, into the zener area). I took a BFR34A UHF transistor and placed in the PCB replacing the zener diode. Now noise was (almost) flat from almost DC up to 1 GHz, but somewhat weak. Playing a bit with the current, I got the maximum noise using a 2k resistor, about -35dBm @ 3MHz bandwidth. With this 2k resistor, the voltage across the emitter-base junction is around 5.5 volts, and current is only a bit over 3mA.

To provide adequate shielding, I put the noise source inside a box donated by Luis, EA4BGH, a nice metallic box from a faulty RF amplifier. It was almost the right size.

The noise source inside a box.

So, in few words, to modify the BG7TBL noise source to 12 volts operation, and improve it sightly, simply do:

  • Remove R8, R9, R10 and R11 resistors (all of them are 360 ohms)
  • Remove D2.
  • Put 100 ohm 1W SMD resistors on R9, R10 and R11 (right size is 2512)
  • Put a 2k resistor on R8 (don't need to be a 1W one)
  • Place a BFR34A transistor in D2. Base to ground, emitter to R8, collector unconnected. You can experiment with other transistors, even with common ones like 2N2222, but remember to tweak R8 for maximum noise. Small size and very short legs are preferred for the transistor (SMD?)

That's all. Now don't apply more than 12 volts to your noise source. The final noise spectrum produced by my modified unit is this:

Noise spectrum produced at 3MHz bandwidth, 0 - 1 GHz.

It has some ripple, produced by the metallic box cover, and maybe the SMA-BNC adapter, but even with that, is much flatter that the noise produced with the unmodified noise source. This ripple can be wash out with the built in compare function in the spectrum analyzer, so it will not be a problem.

The next step is to install the noise source inside the spectrum analyzer, and use the original tracking generator BNC to output the noise. Stay tuned!


After playing a bit with the noise source, I noticed it could easily drop my WiFi connection. The reason? It was self-oscillating. The metallic box acts as a tuned cavity, and its resonant frequency is 2.4GHz +/- load. Put a high gain amplifier inside a cavity and you will get a nice oscillator. I have seen it oscillating between 2.3GHz and 2.5GHz with the aid of my WiFi Spectrum Analyzer.

To damp the oscillations, I glued a flat ferrite bar from an AM radio inside box, about at the center of the box. It worked just fine, with the benefit of an even flatter noise spectrum.

Miguel A. Vallejo, EA4EOZ


  1. The original untouched board has another con, which is the noise is not fully gaussian. while this not may be a problem for a poor-man's tracking source, it is a problem if you require the gaussian distribution. Observing the noise with a high speed scope I saw that the noise is saturated at one end (and consequently there is some dc level at the output). I guess this over gain related issue can be tackled by your solution - I admit I didn't yet tested your changes.

    1. Probably it has too much gain. If you really need a gaussian noise source try to remove one of the MMIC amplifiers, the one close to the output. It should work.

  2. Hi Miguel.
    Great post. Thanks.
    I just measured a 2014 12Volt version of the BG7TBL. See Aliexpress. Check for 2014... on the photograph. Price is 2..3 € higher.
    It has a step-up converter to generate 39 Volts on the zener diode.
    Flatness is quite good. I can send it to you if you whish.
    Un saludo.

    1. Hi Eduardo,

      Please, send the plot to me. I consider buying one of those. How does it look in the range 2GHz to 4.4GHz?

      Thank you!

      Kind Regards,

  3. Since this is probably the most visited page about this noise source, I'll just leave a note here.
    For those exceptionally lazy persons like me, you can just replace the zener diode with 12V one. Cut its pins as short as possible and solder it.
    Power the device with your usual 12V wall-wart (mine gives 12.3V actually) and voila, no more that "hot PCB smell".
    Sure, MMICs will run underbiased thus less gain, but RFExplorer still shows approx. -50db at 1GHz. Maybe smaller diode with less capacitance would help (the one I used is larger than original).
    But as a quick fix for those without 1W SMD resistors at hand I think it's OK.
    / de RA9OEK

  4. Another thought for oscillation suppression. Marie Callender Chicken Pot Pies have a sheet of microwave absorber in the box in order to heat up and brown the pie in the microwave. It's a thin metallized sheet of plastic and I think it would also do a good job of being a lossy item in the cavity.
    Charles R. Patton