Group Actions
Engagement
  • Author Author: michaelkellett
  • Date Created: Date Created
  • Views 2473 views
  • Likes 8 likes
  • Comments 16 comments
Related
Recommended

A Really Simple HF True RMS detector

michaelkellett

In a recent thread Shabaz was wondering how easy it might be to make a wideband true RMS detector  using a surface mount thermistor and resistor.

This is a first very crude and simple experiment to find out.

I soldered tiny bare wires to a 10K 0603 thermistor and a 100R 0805 resistor and glued the pair together. Its supported by its wires on a piece of strip board.

image

It needed a little bit of wind shielding since its a warm day and I had a fan on.

image

I drove the resistor with a range of pulses and settled on 0.66V at 0.01Hz 50% duty for plotting. This is 4.35mW. An 0805 resistor can take 1t least 100mW so we are operating at 13.6dB below full scale.

image

The response time is a bit slow but I think 0603 or 0402 parts would sort that !

The signal was clearly detectable but rather swamped by ambient thermal noise at 70mv (-43dB ref 100mW).

A differential sensor with an ambient sensing thermistor and better thermal design might well fix that.

Time for tea, more later.

MK

  • in reply to ggabe

    When HP were making the revered sensors such as the 478 to go with the 435 and similar meters the option for temperature control was less attractive. They also wanted to work up to 18GHz or more so were happy to use tiny matched thermistors with platinum/rubidium wires and all sorts of other exotica.

    They AC coupled the RF directly into the thermistor which caused a lot of problems, but is worth it because they wanted to work at such high frequencies.

    In a modern design, temperature control is really quite easy and if I can get away with a simple mechanical design with a single thermistor it has the huge advantage that it can be calibrated with DC. No temperature calibration (matching thermistors is a kind of calibration) is necessary.

    My design will NOT work at 18GHz (or even 1GHz) but it will work at DC.

    Some calcs based on the first one suggest that temperature control with 0.005C will do.

    MK

  • in reply to shabaz

    The 'rotating' roles are indeed interesting. Can become a tradeoff decision - I see the value of not needing to be in perfect balance, however there will be a thermal swing on every switching over from reference to DUT.

  • in reply to ggabe

    That's interesting.. I wonder, if the two were at the same level, then the two inputs (or outputs, but maybe better with the inputs) could be swapped around repeatedly (i.e. with MEMS RF switches or something), because if both detectors are now almost in equilibrium, then there's no need to wait in the order of thermal constant times, it can be a lot faster, and then it's an AC measurement that can be minimised until the values are equal.

  • I suggest making a pair of detectors - one to receive the signal, the other to balance to the same thermistor output level with a DC stimulus, just how HP did in their thermal RMS converters. It will bring another challenge - the symmetry of the two devices. Would be easy to characterize though, under DC conditions. 

  • in reply to michaelkellett

    Neat idea to have a testboard, the interleaving sounds interesting! Also, two resistors allows for more standard 100 ohm resistors to be used rather than 49.9 ohm. 

    Regarding the hole method, I just realized that diagram above had an issue, in that the right side had a lot of thermal conduction. 

    The lengths are so small at this scale, that some additional inductance will have no significant impact as your 100 MHz result shows:

    image