Post Go back to editing

Division Error above 18 GHz

Category: Hardware
Product Number: HMC862A

Good Morning,

I am writing because I have had a problem with the HMC862A divider used as a prescaler for a PLO, where I am trying to obtain division in a bandwidth of 10-20 GHz.

At first glance, looking at the divider without any other components or path, just turn it on. A small oscillation is visible when no RF is applied, which disappears when RF is applied.

I have ensured that the supply voltage precedes the control signals, the input is single ended with 100 nF capacitors attached to the component, and I would not consider the RF path to be long. The output was previously configured as differential, but when testing the divider to study its behaviour, it was set to single-ended output, where the 100 nF capacitors are also attached and the negated-OUT signal ends with 50 Ohm resistance. It is also currently configured with an N=8.

While conducting tests, I noticed that the splitter was not performing well above 16 GHz, even though I had completely isolated it from the board and used a generator to sweep the 10-16 GHz frequency range to see its effect and ensure that it was receiving the maximum power it could handle (+10 dBm). But as soon as it reached 16 GHz, it stopped performing the division and only noise was visible on the analyser. I tried replacing it with a new one in case it had broken. With the new one isolated from all the components on the PCB (I bypassed the input path to the splitter and connected an external cable to feed the generator signal directly into it, as I did with the one that reached 16 GHz, and did the same with the output, but feeding it into the analyser), I increased the splitting range to 18 GHz, but it still did not reach the maximum that, according to the technical data sheet, it can reach.

I tried reducing the power, and the effect was that the frequency at which it was able to perform the division was getting lower and lower. I also tried changing the input capacitors to 1pF to see if it would perform better at high frequencies, but I didn't notice any improvement. I changed the division to N=4, but the same effect continues.

I am attaching a video in which you can see how, with a power of 5 dBm in the generator (with cable losses it will be less), before 18 GHz the division is lost, and then by increasing the power it is possible to reach 18 GHz, but there comes a point where even at maximum power the divider cannot perform the division above 18 GHz. Furthermore, when it is at frequencies such as 17.9 or 17.8 GHz for a while, it must be so close to the limit that the division ends up disappearing, and you have to stop feeding the component and wait a while for it to reach 18 GHz for a few moments.

Any idea why it can't divide beyond 18 GHz? Any possible solutions to the problem? Or Is there any possible replacement for the HMC862A?

Thank you very much.

  • I'm attaching a schematic I drew based on what you wrote, in a separate post. Please take a look at it and let me know if it looks correct. Is it correct that you are using 5V for VCC on both VCC pins? And what caps do you have on those pins. I'm guessing that INbar is ac-coupled to ground using a 100nf cap. Is that correct. 

    This testing is happening on your pcb, not on the ADI evaluation board, right? Are you able to see how much current the part is pulling in general as you adjust the divide ratio?

    From what you describe, it appears that the part is struggling to get input power as you increase the frequency, that is, you increase the frequency, lose division, increase the power, get it back, increase the frequency and lose division again. 

    Can you let me know the additional details of the circuit? On my side, I'll dig around to see if there is a history of this issue coming up before. 

    I'm guessing that you do not have an ADI evaluation board to do stand-alone testing? - Eamon

  • Sketch of your schematic as described in your post. Let me know if this matches. 

  • I found out that these divider parts can be prone to self-oscillation when a low input signal has been applied.

    See Application Note: 

    AN-1463 (Rev. A) (analog.com)

    To eliminate this effect, an offset resistor can be applied on one of the inputs and ground. Try a 40k Ohm initially to see how this effects the output spectrum.

  • If it is correct that I am using VCC=5V for both pins, I have 2.2uF and 1nF capacitors very close together for each pin (as recommended in the datasheet).
    Exactly, INbar is ac-coupled to ground with a capacitor.


    It is not the ADI evaluation board, it is the chip directly on a prototyping PCB in the prescaler path.

    I have been able to isolate the component and see its consumption, is the same that the datasheet in all format of division. Therefore, the component is consuming properly.

    The circuit is a PLL into which MHz is fed and goes to a phase detector, whose signal goes to a loop with which to vary the frequency of a VCO capable of operating from 10-20 GHz. Subsequently, to achieve locking, a prescaler is used in which a division is made with the HMC862A and then another divider, the HMC705, is used to reach the MHz frequency, which then enters the phase detector for comparison.

    And, no, I do not have an HMC862A evaluation board to test it.

    The schematic is the same as the one you attached in the image, with the difference that I have a 100nF capacitor at the INbar input and the power supply capacitors I mentioned earlier, which I use for each power pin.

    I have tested the application note you sent, and when I insert the 40KOhm resistor, the oscillation that occurred when the component was turned on is removed, but in exchange, the division is reduced by a few MHz. Even so, the oscillation seems to be resolved! Now all that remains is to be able to divide up to 20 GHz

    If you have any other ideas, please do not hesitate to share them, because I do not know where else to turn.

    Thank you very much.

    Marcos.

  • Unfortunately, I cannot put my hands on a board to test this on the bench. I've reached out to some colleagues who know the part better than me for assistance.