Constant RF power output circuit?

[setq]

So I've been building myself some new test gear. Frequency counter is done. Now I have a multi-range colpitts oscillator that spans the HF bands and IF frequencies I need on the bench and it mostly works. Unfortunately at the moment it's just a buffered FET output and the amplitude is somewhat variable. This results in different RF power being delivered to the load attached. I'd like to get this to kick out 10dBm power out across the board so I can plug a step attenuator into it and get a known power out.

I assume this is AGC territory? But what do I need to look at reading-wise to solve this and what approach should I take? I don't have any op amps with a high enough slew rate to handle this (bit out of op amp territory anyway) so would rather solve it with a CA3096 array or something as I have that on hand. Any pointers appreciated!

[setq]

Excellent - that looks just the ticket. Thanks for the suggestion. I want to avoid any specialised ICs for this as it's a one off piece of relatively primitive kit.

[David Hess]

Use two diode temperature compensated bridge to detect the output and then implement a servo loop to control the bias current or supply voltage to the oscillator to control the output.

[C]

setq

If you think on the parts your building, you could be creating modules that can be used many places. By building in modules, at a later date you could say "this is poor crap, I can do better" an when you do build new module you can use it in many places making all a little better.
 The AGC you need for your oscillator could be used in many places for example. Here an output to a step attenuator, others places is OSC to mixer, RF to mixer, IF level and the uncommon overload preventer for RF to many outputs.

Might note also that for AGC, size of error is what is controlling gain. Feedback adjusts gain to sample point. Past sample point, A little feed forward can try to remove the size of error change past sample point.   

You might also want to keep your eyes open for places that you can get an inverted copy of wanted signal. You can gain in some areas when this is used.

[Alex Eisenhut]

Here's an old generator you might get inspired by:

http://w140.com/tekwiki/wiki/191

It's basically what David says. The Tek manual is full of good information in the Circuit Description section.

[David Hess]

The Tektronix SG503 is a more modern albeit still old example of what I had in mind but the Type 191 schematic may be easier to follow if you can get past the vacuum tubes.  Maybe think of them as n-channel depletion mode MOSFETs. 

Jim Williams used the same idea several time with a two diode bridge operating as a temperature compensated detector but he used an OTA (operational transconductance amplifier) as the gain control element.  See the schematics on pages 24 and 26 of Linear Technology application note 61 for an example of this.  If you have control of the oscillator, then the separate gain control element is not required.  Or a simple multipliers could replace the OTA.

[KJDS]

If you just want a constant output at a fixed level, then run another common emitter output stage in compression. Setting a suitable bias current will fix the output level and if you use a PTAT biasing scheme then it will be temperature stable. I'd then add an emitter follower to give a low output impedance.

This will save having a detector and control circuit.

[JohnPen]

The Scotties Spectrum analyser project includes some useful calibration test gear that may help.  The device called a Leveler will provide  an almost flat response up to typically well over 500 Mhz.  It was designed by Sam Wetterlin using an AD8367.  This is the link  http://www.wetterlin.org/sam/SA/MSACal/CalAccessories.htm .  You will probably need to download the 'Express' s/w to see the schematic.  It will provide ~-8 dbm to which you can then add a gain stage to bring it up to the level you require although the design of the latter may limit your total frequency range.

[K9BF]

It's called an ALC or automatic level control. Any newe Ham radio has one. Just generate a DC signal from RF output circuit, compare it to a reference voltage and feed the result to a double gated FET. It been done this way for decades.

[JohnPen]

Agreed ALC has been around for many decades.  However if you want a precisely calibrated output level over a wide frequency range typically up to 500 Mhz the 'leveler' provides that.  The input level can be anywhere between -30 dbm to 0dbm for up to 500 Mhz or -15 dbm to 0dbm at 1Ghz.  Sam's quoted typical output levels are as follows.

1 Mhz        -7.70 dbm
40 Mhz      -7.80 dbm
175 Mhz    -7.60 dbm
400 Mhz    -7.80 dbm
900 Mhz    -8.60 dbm
1000 Mhz  -8.85 dbm
1050 Mhz  -8.85 dbm

Of course this may be an overkill for normal  RF usage but it is pretty essential when trying to accurately calibrate a piece of home built test equipment such as a spectrum analyser.  As always it is what do you really need.

John

[Wolfgang]

If you insist to homebrew, you could use a Bruene bridge, a diode detector or an AD8307 detector, and a PIN diode attenuator to make this.

Some samples:

https://electronicprojectsforfun.wordpress.com/rf-module-gallery/swr-bridges/a-bruene-bridge-directional-coupler-with-direct-rf-output/