Measuring IP3 with cheapskate equipment

Discussion in 'Homebrew and Kit Projects' started by N5HXR, Jul 19, 2021.

ad: L-HROutlet
ad: l-rl
ad: Left-2
ad: L-MFJ
ad: L-Geochron
ad: abrind-2
ad: MessiPaoloni-1
ad: HRDLLC-2
ad: Left-3
  1. K7JEM

    K7JEM Ham Member QRZ Page

    Two different kinds of IP3 are being compared. Usually, we see an output IP3, but there is also input IP3. Don't compare one with the other. The BGA2866 lists IP3 as being close to 20dBm (this is higher than the mixer, at +15):

    N5HXR likes this.
  2. N5HXR

    N5HXR Ham Member QRZ Page

    OK, here's that result. Set levels on the siggen to get S9 from the meter on the FT891, and compared 6.999 MHz with 6.899 MHz, which should put the tones at 1Khz (I think?). I did tune around just to make sure I couldn't hear something if it was just a little bit off somehow. Also checked 7.1 and 7.2 MHz, and they looked identical the the 7.0 and 6.9, respectively.

    No audible tone at all, event at max RF/AF gain, just in case it might just not show up on the meter.

  3. SM0AOM

    SM0AOM Ham Member QRZ Page

    Then you have an IMD suppression of at least 60 dB, which is more than what the oscilloscope plots show. It appears that the IMD is generated in the oscilloscope.
    N5HXR likes this.
  4. SM0AOM

    SM0AOM Ham Member QRZ Page

    My systems designer's (as usual) take on this is that it depends (as usual). To start with, you are never benefited by worrying.
    I reserve worrying for really important things.

    A JFET before the mixer will both reduce oscillator radiation and lower the noise figure, which may be significant if you have a small antenna. On the other hand, a small antenna delivers lower signal levels, which lessen the requirements on the signal handling properties.

    People using reasonably sized antennas usually are content with a noise figure of 15 dB and an IP3 referred to the antenna connector of 0 dBm. This translates into a three-signal selectivity of about 75 dB, which is about what is usable in today's RF environment.

    HF receivers of "yesteryear" had IP3:s in the -10 or -20 dBm ranges, and noise figures of 5 dB. In quiet locations or using very small antennas these low noise figures sometimes were justified. The users did not complain anyway.

    Today, I see very little point in making a receiver with lower NF than 15 dB and a three-signal selectivity much better than 80 dB. This is mostly caused by the fact that HF spectrum occupancy has become so much lower, and the remaining interfering signals so much "dirtier", especially those within the amateur bands.

    N5HXR likes this.
  5. WB3BEL

    WB3BEL Ham Member QRZ Page

    No. If you did not change the amplitude of the signal generator, but merely inserted a 10 dB attenuator, and then observed a different intermodulation level on your oscilloscope/spectrum analyzer, then at least some of that IM change is only in the oscilloscope/spectrum analyzer.

    Everything else was the same. You only changed the level into the analyzer and you see different IM means some of that IM is generated in the observing instrument.

    Perhaps I am misunderstanding your experiment.
  6. N5HXR

    N5HXR Ham Member QRZ Page

    This is helpful info, thanks! I pulled down AppCad and am playing with different numbers... some of the interactions between stages were surprising, so I guess I'll have to hunt down the math underneath and figure out why it's sometimes counterintuitive :).

    Now... throwing noise figure and three signal selectivity at me is too much ;-). At some point, I have to actually build a thing! I looked at what it takes to measure noise figure, and I don't think that's on the radar anytime soon.

    I'll go read up on noise figure now...
  7. N5HXR

    N5HXR Ham Member QRZ Page

    Nope, I don't think you're misunderstanding it -- and I think we've worked out that it's the oscilloscope that is the source of the IMD.

    I dug out my TinySA, and tried its OIP3 measurement mode. I think it's not nearly as good of a spectrum analyzer as the NanoVNA is a VNA, so I had to push the two-tone signals much farther apart.

    But once I got it kind of settled in, with no obvious weirdness, it does not have the IMD peaks that I saw on the oscilloscope FFT. It also shows an OIP3 for this BGA2866 amplifier of 15.2dBm, and I suppose this is the best number I'm going to get from my equipment in the home lab.

    So I guess the short conclusion is that the FFT mode on the oscilloscope is touchy, and probably can't get me reliable IP3 measurements. I think the TinySA might do better for me, but it doesn't seem to have a tight enough resolution to test input signals that are very close. So I guess as long as I'm testing a wide bandwidth device so that I can put the tones far enough apart to get a good reading on the TinySA, that's my ticket going forward...

    In the meantime, I also picked up some new topics I have to research, so it's been a good thread.
  8. SM0AOM

    SM0AOM Ham Member QRZ Page

    Three-signal selectivity is a measure that can be directly derived from noise figure, IF bandwidth and IP3.

    It is often used by systems designers as it gives a direct view of how a receiver performs in an environment of strong in-band signals.

    In the 1980s, when I learned the trade, a very good ($10000 or $15000) HF receiver had a three-signal selectivity of about 85 or 90 dB and this was considered necessary for enduring the RF environment of Cold-War Europe when large directive antennas were used. 500 kW broadcasters using 16 dB curtain arrays were not at all uncommon.

    The "only $5000" ITT-Standard Radio receivers that my first employer used for building systems had about 75 dB selectivity and did sometimes overload in this environment, but using the switchable attenuator usually solved the problem.

    Compared to the US, the situation was at least 20 dB worse.

    Dr. Ulrich Rohde, DJ2LR wrote a lot about this in Ham Radio Magazine during the 1970s.

  9. SM0AOM

    SM0AOM Ham Member QRZ Page

    For your guidance, this is how it is measured (from the SRT receiver specification):


    If the interferers are not clean enough, these figures becomes meaningless. To measure this on a above-average receiver, it
    takes two crystal oscillators or signal generators at least in the HP8640B, R&S SMLR or HP8642 class.

    N5HXR likes this.

Share This Page