Recreating a Marconi TF1375 AC µV meter

[6SN7WGTB]

Following my other post about modelling analogue meters, I have tried to design a circuit to replace the missing one in my spare TF1375.

It's just a fun project to bring the 'empty' but otherwise mint TF1375 back into life, and teach me about LTSPice in the process.

The basic requirement is to take a ca. 10-50µV AC input and drive the existing 50µA FSD analogue meter, plus provide an 'amplified output' which I have arbitrarily decided is 1V AC RMS for FSD. (Perhaps it should be 775mV...)

What I have come up with is attached. I'm not expecting anyone to fix it all for me, but I'd value general views on what I've done:

• 2-stage gain using LT1007s - to ensure that lowest practical input of ca. 10µV (due to meter scale) provides >5mV to...
• LTC1966 RMS to DC converter
• Buffer amp to drive meter
• Buffer amp to produce 'Amp Output' as noted above
• A smattering of Ge diodes which appear to give a 'curve' which almost matches the analogue meter µV scale vs. µA drive: without this mod the output is µA for µV perfect which is not what I need!

I have lots to do, and no doubt I will be guided to other things I've missed. Only my second use of LTSpice. I will add the -10dB attenuator by shunting one of the opamp feedback loops, and the LF cut is easily done I think.

I'm curious as to whether I could improve the bandwidth with a different RMS -DC converter, although right now the cascaded op amps probably limit me to maybe 500kHz. Putting aside practical design issues of course.

However, at the very least, might this work...?!!

[dobsonr741]

6SN7WGTB, did you make progress? I'm about to do the same on an HP3400A. I was planing to get XBUFF802, 2xLME49713 and LTC1968 in the signal chain, to recreate the specs of the original meter.

[6SN7WGTB]

Essentially no, as I have been focused on rebuilding one of my two TF1375 as original - which has worked very well.

I am still looking to do the 'spare' as a recreation.

One thing that has been bugging me is the potential noise floor. The original meter essentially attenuates every input range down to 50µV FSD to pass to the amplifier and meter.

It struck me that I might be more successful having everything attenuated to the 500µV level (where the noise appears less critical), and then have an additional op-amp for the more challenging 50µV input to bring it up to 500µV.

I've updated the LTSpice schematic. However, haven't really stress-tested it yet.

It might sound daft, but the exercise is showing me how competent the original seven Ge transistor design was...

I would welcome any thoughts as this may well be a good project to get back into over a mince pie or two...

[dobsonr741]

Noise is my biggest concern. What was the original Marconi bandwidth spec? My gut feel is the Linear parts you choose will be fine for a 40db SNR from 10Hz to 1MHz at 50uV full scale.

In my case the HP3400 bandwidth was 10Hz to 10MHz, with 10x overdrive-ability at 1mV full scale at 10M input resistance. True rms at 10MHz is my second concern, at the crest factor of 10. It's still an acrobatics achievement even today, if we stay on pure analog processing. For that end, I'm thinking ditching the LTC1968 and use AD8361 instead, with a little cheat of a small MCU based linearization and calibration.

[Kleinstein]

With a large BW the noise definitely becomes an issue. Ideally one could subtract the noise power of the amplifier, but this would have to be done as part of the RM-DC converter and not afterwards.

In the shown circuits with the LM1007 a major noise source is the 10 K resistor in the feedback network. For low noise and also to get a good frequency response the FB network should be lower resistance, more like 330 ohm and 3.3 K for a gain of 11.  Alternatively have a compensated divider with parallel capacitors.
The input of the op-amp also needs a DC path (e.g. some 10 M to ground).

Using an accurate RMS converter and than diodes to get back to the nonlinear scale is kind of strange.

For high frequencies and only an analog scale (and thus not that large a linear range) the AD8361 looks more suitable.

[6SN7WGTB]

Interesting replies - spurred me to get back into this.

Great point re the noise in the FB loop - will attend to that.

I do appreciate the point about the bodge to skew the scaling. But equally the point is that it then means the (inaccurate/non-linear) analogue meter will in fact reasonably indicate the 'accurate' result of the RMS converter!

The original Marconi BW was ±1dB 50 Hz - 200kHz as I recall. So pretty limited - it was a simple circuit with Ge transistors. Also it was averaging, calibrated for RMS, so non-sinusoidal required correction.

I did consider using an Arduino to do the 'calibration' curve for the meter. If I could create a decent BW up front I might consider that worthwhile doing...