Hi all,
I made some further progress. I discarded my previous prototype and made a new one using better JFETs, this time the 2N4856A. According to its datasheet it is especially designed for chopper applications (well, I don't quite understand in which parameters a chopper JFET should differ from an ordinary JFET).
And with the new JFETs, indeed, the waveforms at the chopper output look much cleaner and nicer. However it is difficult to measure because my scope (Agilent 54832B) is very noisy with those small voltages.
I was able to get more stable and reliable readings, and the zeroing of the sensor also works more reliable. However, the linearity is still quite bad; at low input power, it is okay but towards increasing output power, linearity degrades quickly and at the worst point (0 dBm input into the detector) it is almost 3dB off. My detector can measure -30dBm to 0dBm with usable accuracy. Above it becomes a bit nonlinear - could be due to the diode no longer operating in the square law region. But I will further investigate in the subject. Probably I also need to add the shaping network mentioned in the 8484D service manual.
@ Dave NR1DX
I am testing the diode detector with the HP 436A calibration output. And I also test at various frequencies down to 1MHz and up to 2.6GHz with the 8663A signal generator. However, I think later it would be desirable that my detector works to, say, 6 to 10 GHz. I think otherwise it is not very useful. On the other hand it would also be good if it works to below 1MHz.
@ Rainer DF6NA
Thanks for the links. I will probably order some of those connectors when I am ready for a 'real' prototype! looks not toooo expensive.
@ Mike N2MS
I think for the diode sensors (like 8484A, 8481D) it is not possible to replace the diode without a highly sophisticated workshop and tools. I have never seen one of these sensors from the inside, but I expect wire bonded black magic inside; the service manual even says that the position and shape of certain wires is important, otherwise the calibration goes nuts.
@ Dave NR1DX
The thermocouple has the advantage that it really measures power dissipation, i.e. RMS. Since it does the measurement via temperature, it is basically independent of the waveform and even independent of the frequency - as long as the mechanics is reliable and precise enough, you can measure each frequency with this, even DC. The HP sensors have an internal DC blocking capacitor, so this is why they don't work down to DC.
@ bownes
yes, like AD8307 or so would be nice as well. However, I am not sure whether it is possible to connect these somehow to a HP meter like the 436A. For instance, the AD8307 (or some other LogAmps) need a power supply, which is not available on the existing interface to the sensors, except if you abuse some signals, e.g. the chopper signals and then rectify them etc.
Besides that, you need to mimic the behaviour of the thermocouple/diode detector signals.
This is why I stuck to the 8484A/8481A design.
@Dave McGuire, AK4HZ
You said you have already done what bownes suggested? was it working? could you tell a bit more about this? :-)
