The linear Diode Detector

Hi James, I'm glad you posted the up0date on the amp. I have had thoughts about if I'm (trying) doing this I should make it as wide range as I can. The Boonton measures to 50 MHz. I haven't been above 1.7MHz.

Is the supply 20 Volts or 25 Volts?

What's my input capacitor value? Not critical or my 0.3pf?

The original meter is a 100ua FS with an internal resistance of 1,500 ohms. I would need to rescale the meter because of the nonlinear scale on the original. See nonlinear curve, >

I'm not even sure about the linearity of the meter movement,

at 1/2 scale, as best I could eyeball it, it had 47.5ua flowing, that is a 5% error. No way to tell if it gets worse with lower current.

Ideally, I'd like to find a meter with a scale marked 0 to 250. There is some pretty good software to make meter faces.

Thanks, Mikek

(+24V has the advantage of reducing the transistors' feedback capacitances a wee bit, but +12 should work fine too, when properly biased.)

Not critical. The buffer's Cin is 0.6pF. You could add 100pF in series for coupling and 0.6pF to ground to emulate a 1.2pF tube capacitance. Or you could put a capacitive-divider at the input, for handling larger voltage swings.

The meter looks decently linear to my eyeball over the plotted range, reasonably approximated as Reading = (20 + 2.3 * i (uA)).

Wouldn't you want to use the existing meter? I mean, if the aim is to revive old Boonton 260A's with a replacement for the no-longer-available tube?

If the goal were merely measuring 'Q' you could do that with a 'scope and an amplifier (using the 260A just as a signal source).

Cheers, James Arthur

Q1's emitter isn't peak-detecting like I'd expected. +/- 50mV input to the base only yields 21mV 'peak' detection at the emitter, less than half of the 50mV I'd expected. Weird. I'd figured it would just work. Can't afford to spend time on it right now though...

Cheers, James Arthur

Ya, I was asking about the low voltage supply, you had a 20 volt zener and then 25 Volts just above it.

Yes I'd like to, but I would think it should have a faceplate that is linear. Full scale (100ua) is marked 250, At 50ua the needle points to

First things first. A DDS with a keypad input is in the back of my mind, to replace the power oscillator, that changes frequency when the injection Voltage is adjusted, and vice versa. It's a pain to get the frequency and injection voltage adjusted, because one changes the other.

Mikek

_Much_ faster, if you want decent response at small signals. The amp has to slew at least a diode drop on each half-cycle, and small signals don't develop anything like its maximum slew rate. FET op amps need about a volt between inverting and noninverting inputs to reach maximum slew. Bipolars need more like 60 mV if the input stage isn't degenerated, but then their maximum slew rate is much less than a similar FET part.

Cheers

Phil Hobbs

As an example, here's a 1-MHz "perfect" rectifier using Schottkys to reduce the required swing, and a 1 GHz op amp (ADA4817). It steps the amplitude from 1 mV to 3.16V, 2 steps per decade, and normalizes the output so as to highlight the nonlinearity.

In a really perfect rectifier, the node NormRect would be a half-wave rectified sine wave of amplitude 1.0V. You can see that that's very far from the case, even though the op amp is hundreds of times faster than the input signal.

Cheers

Phil Hobbs

A GBW of at least 100MHz, with layout and parts to match. 100MHz may not be enough. Slew rate could severely limit accuracy -- 300V/us pops to mind as a bare minimum.

IIRC it was pretty good. Certainly better than something that's not compensated for diode drop. I didn't toss it up as "use this or die", I tossed this up as "use this idea to make an informed decision".

Prolly.

Oops -- Phil Hobbs says a GBW 1000 times the base freq, not 10. So, GBW of 10GHz, with components to match. That's crazy fast in my book. You got a part number you can quote?

huh, Thanks Phil, I couldn't open the PHParts.lib, but I think I get you point. You need to slew ~volts in a nanosecond.. or maybe faster. OK forgetting about a diode inside an opamp loop.

George H.

Sorry, forgot to delete the library. The ADA4817 model is actually there in the schematic, so you can just delete the .lib line. I fixed it on my web page, .

Cheers

Phil Hobbs

So, what's wanted is a full-blown six-transistor double-balanced mixer. That's a fairly good square-law device, for the DC output when you put the test signal in both LO and RF.

The kicker is, there's a bit of temperature effect on the gain. So, it'll take a thermostat, too. There's benefits to gain/saturation of the LO to make a square input, of course. If you game the squarewave LO amplitude that can do the temperature compensation, and the output turns from square-law to absolute value.

mp

If you know the timing of the zero crossings *a priori*, you can do a lot o f useful things like that.

AM detectors are really hard to get right if you want any sort of accuracy. I'm a fan of DLVA solutions for wide ranges at moderate resolution (1 dB u ncalibrated, maybe 2% with a good calibration signal).

You really can't count on more than 20 dB range per section in any vaguely accurate AM measurement.

Cheers

Phil Hobbs

What's a six transistor mixer? I built an audio (transformer) ring diode mixer. (diode connected transistors worked "better" than other bare diodes I tried.)

George H.

A normal Gilbert cell (minus the input diodes) such as an MC1496. A diff pair feeding two cross-connected diff pairs. (IIRC the 1496 also has a seventh transistor for the tail source.)

Cheers

Phil Hobbs

Or a resistor and thermocouple, which is what Boonton used originally. But it's a difficult mechanical arrangement to get right.

Be advised that temp comp "works" if temperature changes are VERY slow. Thermal coupling is very bad; expect 30min or more time delays and so lots of hysteresis.

Two nearby surface-mount parts that don't dissipate much power will be nearly isothermal on a PC board. A dual diode will be even better.

Are you getting much of use from this thread? I'm surprised there aren't more useful posts. I guess this is a harder problem to solve than I realized.

Asynchronous AM detectors are pretty hard. Over the OP's amplitude range, a fast comparator providing the LO signal to a diode bridge such as a Mini Circuits MPD-1, with the analogue signal going to the RF port via an attenuator will do a good job.

I usually use a few simple AM detectors with amps or attenuators to extend the range, digitize them all, and pick the one that's in its sweet spot. Assuming the ranges overlap some, you can use that to calibrate the relative gains.

Cheers

Phil Hobbs