Podcasting since 2005! Listen to Latest SolderSmoke

Friday, November 5, 2021

The Importance of Keeping the Noise FLAT


A few days ago I put up a blog post about using a noise generator (in my case my cheap FeelTech sig generator) and my TinySA spectrum analyzer to look at the passband of a crystal filter.  I was using the 9 MHz filter used by Dean KK4DAS and the Vienna Wireless Makers Group.  The idea is simple: insert broadband noise into the input. The filter should pass more of the noise that falls within its passband.  The TinySA should let you see this.  At first, I was pleased that I could clearly see the passband.   I thought I had succeeded.  See above. 

But I was bothered by something.  Look at that bump in the passband.  It should be close to flat across the top.  

I decided to take a look at the same filter with my NanoVNA.  Here I was not using a noise generator.  The NanoVNA sweeps the filter using and looks at output in the Log-Mag mode.  Here is what it looked like (below): 

That was much better.  But why the difference?   Tony Fishpool G4WIF suggested that my noise source might not be putting out noise at the same level on all frequencies.  I took at look at the noise output of the FeelTech sig gen in the range of the filter passband (with some above and below frequencies for reference) and I found that the flatness of this noise depended a lot on what frequency I had the sig gen set to.  I tuned it around a bit until I found a setting that produced a flat noise output in the desired frequency range.  Then I went back and swept the filter with the noise and the TinySA again.  Here is what it looked like with the "flat" noise: 

Better,  I think.  Closer to the passband displayed by the NanoVNA.   

Tony points out that these Chinese sig gens don't really put out random noise -- they give us predictable noise.  Dean said "Predictable Noise" would be a good name for a rock group.   I said they could open for my favorite: "The Ceramic Spurs." 


  1. Thanks Bill

    Digital white noise generators are pseudo random whether the random number generation gets done via code, or by logic gates plus and a clock. The best digital [expensive] versions run long sequences with well defined amplitude.

    For decades, builders have relied on analog noise sources since they prove easy,cheap + provide true random white noise. Often this means reverse biasing a BJT with very low current so it functions as a zener diode. A transistor may produce more noise than an actual zener diodes. 1 or 2 op-amp stages are needed to boost the noise amplitude into something usable depending on your application.


  2. Now that's very interesting. I don't have a white noise generate here so I've not been able to run this type of test.
    I've always used my HP network analyzer to sweep it.
    I will have to look into making a noise source.
    Any suggestions?
    G0MIH /HS0ZLQ.

  3. Some of the commercial RTL-SDR up-converters have a built-in zener-based noise source (the Ham-It-Up converter is one), presumably because the dongles can be used as (really-cheap) spectrum analyzers. I just bought one of the Ham-It-Up units without the noise-source option because I want to embed it in a (also really-cheap) "homebrew" SDR receiver that will live on my operating bench and not my lab bench. I may or may not put together a dedicated noise generator using the Ham-It-Up design.

    There's not much to it. The zener has a breakdown voltage above 6V to get the avalanche noise, and from there it's just a few stages of broadband amplification. I need to look into Vasily's idea that a reversed-biased transistor might be better than a zener. Either way, having such a source will help as a "sanity" check on the NanoVNA, just as you used the VNA to check the results you got with the TinySA (though in this case not the SA's fault). I love those little "good-enough" analyzers, and I like the idea that one can be used to check the other.

    Of course I could just use my 40-meter antenna as a source. With a 37dBm noise floor (about S9!!!), it's barely usable as an antenna, and because I'm getting the same every-50MHz spike many others are now getting, I can also use the antenna as a marker generator! Keep on the sunny side of life, I always say. --Todd K7TFC

    1. K7TFC,

      What is this "every-50MHz spike" you mention? Scott NM8R

    2. I think he meant "every-50kHz spike"? That would be harmonics of typical switching power supplies that run at several tens of kilohertz.

    3. Oh, yes, not 50MHz. 50*K*Hz! I guess it's coming from a switching supply, but not in my house--at least I haven't been able to find it. It's not conducted noise coming from my own power supply because it's there on battery power, too, and it doesn't matter if the rig is open or buttoned-up in its shielded enclosure. It's coming in through the antenna. In truth, it's no more obnoxious than some over-driven blowhard parked on the same frequency . . . actually, it's less obnoxious.

    4. Thanks for the clarification. Could it be possibly entering on the shield of your antenna's coax? I've encountered that before and wound common mode choke and put it in the coax at my operating point. It worked nicely. For that I used a mix 31 core that is almost 4 in in diameter overall. I also keep on hand a couple Nick's 31 split fare right beads. You can open them up and wind five or six turns then simply close it and see what impact that has. They are pricey, but great for troubleshooting and RFI mitigation. Good luck. I know the feeling... NM8R

    5. I apologize. Voice to Text struck again and I published before proof reading. 'Mix 31' and 'split ferrite beads' should be substituted appropriately in the my reply above.

  4. Hi Bill

    Lot of choices on Ebay


    Jim AB9CN


Designer: Douglas Bowman | Dimodifikasi oleh Abdul Munir Original Posting Rounders 3 Column