In Baluns

A while ago we wrote a product feature for Microwave Journal for our isolation baluns. Basically we showed why they were better than 4 port VNAs for high speed differential testing.

Check out the article here.

These baluns are now available up to 67 GHz:





Recommended Posts
Showing 4 comments
  • Pedram Kermani

    Hi Doug,

    I need a setup to measure the output of a differential amplifier which has unbalanced DC offset on it. I want to measure the width band noise of DC to 100Mhz on the output and only considering the power level. Can I use the BAL-0003 in combination with DC blocker in the input (if needed), and then connect the output of the Balun to the spectrum analyzer? FYI: the output of the amplifier is 130mV only.

    Kind regards,
    Pedram Kermani

    • Doug Jorgesen

      Hi Pedram,

      Whether or not you can use the BAL-0003 comes down to your amplifier.

      First the DC. The BAL-0003 is DC short to ground on all ports. It does not require a DC block, but your amplifier probably does (to keep the bias) so I would recommend a DC block. The positive and negative sides need to be phase matched, so you should put DC blocks on both sides.

      Next the RF. Because the isolation of the BAL-0003 is only around 8 dB, this means that the positive and negative sides of the amplifier will see 8 dB of crosstalk between them. Will this affect your noise? I don’t know. The way that I would conduct the experiment is to measure the noise using the BAL-0003 like you suggest (into the spectrum analyzer) and then measure it again with phase matched 3 dB pads on both the positive and negative ports. If the noise level drops by 3 dB then you know that you have a good measurement. If it drops by more (or less) then you know that the feedthrough is causing some kind of noise contribution.

      At 3 GHz I wouldn’t worry too much about phase matching the components as any two identically constructed components should be close to phase matched.

      Finally you can measure the single ended noise from the positive and negative side and compare them with the balanced noise as a sanity check. The difference will be the common mode noise that is rejected by the balun, so the balanced noise should be lower.



      • Ellis Harrison

        Hi Doug,

        In a very similar experiment, I am measuring wide-band noise of an amplifier with a 100R differential CML driver output up to 2GHz. I am connecting the differential outputs of the amplifier into a BAL-0003, and then the Common port of the BALUN into a spectrum analyser.

        The issue I am having, is that when combining noise powers of each amplifier output using the BALUN, I see ripples in the measured noise profile on the spectrum analyser at frequencies >500MHz. If I measure the noise single-endedly from the amplifier (other output terminated 50R), directly into the SA, the noise profile is smooth with no ripples.

        This looks like some sort of impedance mismatch issue to me. The differential output of the amplifier is not a perfect 100R source impedance out to high frequency; it actually suffers from impedance ripple at frequencies >100MHz.

        Why is it that the single-ended measurement looks clean without the BALUN, but differentially with the BALUN you get ripple in the noise measurement? I have also measured using an active differential probe, and the measured output its clean. The issue is the probe doesn’t go down low-frequency enough for my application, hence the use of a BALUN.

        Many thanks.

        • Doug Jorgesen

          Thanks for the comment. The issue sounds to me like a balun isolation problem. I would take a look at this post on how balun isolation can eliminate ripple in differential to single ended measurements.
          It is an odd thing, requires me to do some mental gymnastics, but the Sdd22 is not actually a ‘true’ return loss. It is actually a mathematical construction of the reflected waves on two different ports. This is why a lack of isolation between those ports can show up as a poor differential return loss.

Leave a Comment