Rms converter to match HP3400A?

Not quite sure what you're asking here.  Is the 'converter' an add-on to the 3400A that you want to build from scratch and put into your meter?  Or do you just need an analog r.m.s. converter with those specs for something else? Or are you looking for a complete equivalent to the 3400A that does all measurements in the analog domain?  Old National Semiconductor app-note books had designs for r.m.s. measurement, although I'm not sure they went to 10MHz.

I have a 3400 with burned out thermocouple converter. I want to rebuild the guts between the input BNC and the analog meter. Building AC amplifiers with 10MHz <1% tolerance is a no issue. The rms converter with 5% error at 10MHz is. Also the crest factor. And again, the parts should be new, and not from an eBay thermal converter surplus.

Right, I see your problem.  The meter having ranges in (I assume) 10dB steps probably means that the converter has to be really accurate over only a 20dB range or so, it's probably sitting somewhere in the meter circuitry where it's working on voltages in the tens to hundreds of millivolts or more. But you're right, the 10MHz requirement is a tough one.  Most of the r.m.s.-ing circuits I'm familiar with use the nonlinear characteristics of transistors in some feedback circuit, which if built with op-amps that can slew up a storm might be good to 10MHz.  Never built one though.  Best of luck!

www.analog.com/.../1968f.pdf

LTC1968’s bandwidth is not more than 1MHz. 

Frequency response is up to 15MHz.

-3dB is 29% error. The 1968 samples at 2MHz, there will be a response for sure, just accuracy and frequency response will be off of a measurement grade.

My current thinking is an AGC looped AD8361, like on fig38 of https://www.analog.com/media/en/technical-documentation/data-sheets/AD8367.pdf, with the exception I’m about to use a linear VCA, not a decibel scaled, so that the control voltage is the rms output.

I'm also thinking creating two signal paths, for low frequencies up to 1KHz with traditional log-antilog converter or LTC1968. Not sure though how to manage an accurate transition between the two.

Those are interesting parts. I have three  HP435B with no sensors and no prospect of getting any.

It would be nice to get them doing something useful.

The AD eval board for the 8367 is a bit pricey (>£100) but cheapo boards for both designs on the data sheet are available from Aliexpress. The pair cost me £32 which is a bit more reasonable.

Won't arrive to July. May work, may not but worth a punt. At the worst I can replace the chips with AD ones

MK

I've always wondered if it could be possible to buy a 50 ohm (or 600 ohm) 0402 resistor, and super-glue a 0402 thermistor onto it, rotated 90 degrees, and it could be extremely sensitive even without much amplification, since the thermistor would be able to sense even micro-watt level changes and be fast-responding maybe, since it's so tiny. However I know nothing about how such a thing could be insulated, and if it would even be accurate, and how to calibrate! I've never through to experiment since I didn't have a real use-case, just curiosity.

I've got 0402 resistors and thermistors (2k and 10k) - no time today but I'll give it a go.

Also got 0603 Pt sensors but I think they do 0402 as well.

MK