Automatic Gain Control Schemes

[KI4YAN]

So, furthering the "Common Gate Jfet" thread, I'm on to the next part of the signal chain-the variable gain amplifiers. Currently I'm planning to use LMH6505 variable gain amps-these will go 0 to 80dB of gain, and are controlled by a 0-2v signal, and respond in a linear-in-dB fashion. Laying out the circuit and PCB layout exactly as per the datasheet, the design is even targeted at 9Mhz, which is my IF frequency.

BUT, how to generate the control voltage? Here's what I am thinking:

The IF strip must handle SSB and AM signals, voice and data modes, and will be used on receive and transmit. Since my VGA's are linear in dB, is a log detector (say, the AD8307 commonly used as a wattmeter) with it's linear-in-dB output going to be a good way to go? Since I'm not requiring the AGC detector to provide any kind of audio or recover any kind of signal, just tell me what the power of the signal is, I don't see a downside.

I'm going to be using 2 of the VGA's, with a noise filter between them. (a wide, low-Q 9Mhz bandpass filter) I want the first variable gain amp to do most of the heavy lifting for gain, and the second amp to operate as kind of a "limiter" to provide the diode mixer with a constant RF level. This way, I would hope to get a constant mixer output, since both inputs are held at a constant power and the mixer is in the flat part of it's loss curve. From there, the mixer output will either be sent to the audio chain, or sent to the transmit chain. Does this mean I need two detectors, or will a single detector, sampling the IF as it goes into the mixer, be enough to feed two different AGC amplifiers to drive the two IF amps?

[WU2D]

That AD8307 should give a great deal of gain control. See http://www.ad5gh.com/if-amplifier-board-hybrid-cascode/

And you can let it sense the IF and form a loop and set your slope and time constants. You may want to consider also sampling early in the chain, before selectivity where you may see strong off channel signals that will spoil the front end just as much as on channel signals.
Except you do not see them in the IF. 

[KI4YAN]

That's kind of what I was thinking.

Since both the gain curve and the AGC curve will be linear-in-dB, it would only be a matter of scaling and level shifting to get the appropriate voltages for the amplifiers. A potentiometer to set the gain of the level shifting op-amp could be used to change the gain curve a bit, putting a "knee" in the gain curve to the second IF amplifier.

That way, one amplifier will have the linear-in-dB response, and the second will have a knee in the curve to reduce it's response, to a certain point. Not 100% sure how that will work out, but I'm fleshing this system out as I go.

[WU2D]

Yes you will have a real modern AGC system that you can taylor. I have been away from that stuff for a while now but in the late 90's I was the Applications guy supporting the AD608 and AD607 chips. As you might remember, the AD607 was a pretty complete receiver with a mixer and linear IF strip and I and Q demodulator on a chip for AM and SSB and you could do synchronous AM detection with it. The 608 was the one for FM with the mixer and the Log IF back end. AH but the days of analog are gone....

[KI4YAN]

So here's what I have now.



On the far left, is a wide bandpass filter between the first variable gain amp (out of frame to the left, but exactly the same as VGA2 in the middle) and the second VGA.

The variable gain amplifier, LMH6505, is configured for single-supply operation and thus has some biasing resistors in place. Maximum stage voltage gain is currently set to be 17.6dB. (I think. Max gain is 0.940 * (R28 / R27), which comes out to be 17.6dB by my math, however the datasheet calls out this specific combination at 26dB...) The maximum gain can be adjusted from 0 to 100, so I can (and probably should) increase R27 to 100 ohms, and R28 to 2.7K, to bring gain up to 25dB or more. The datasheet seems unclear on what is the "max gain", but mentions that the gain control input should be low-pass filtered.

C1 may not actually be needed.

R29 is fitted as per the LMH6505 datasheet to present an output impedance of 50R.

The RMS-2+ mixer's RF port on pin 4 receives the signal from the variable gain amplifier. Since this mixer will be asked to output down to audio, the IF port on pin 5 will be the audio output in receive mode, and the RF output in transmit mode. It's specified for DC-1000Mhz operation so should be no issue.

The BFO/Transmit Carrier Oscillator is applied to the mixer's LO port through a series resistor and a pi-network pad. 1K is used as a placeholder value, R8 may well be a 0 ohm link and the value of R5, R6, and R7 will be adjusted to match the applied LO signal level. Even though this is a Level 7 mixer, very good performance can be had at a LO level of +4dBm, while reducing LO feedthrough.

Now we get to the sticky bits. Port 5 of the mixer, the IF port, needs that broadband 50R termination for best performance. The AD8307 Log Detector's input impedance is 1.1k, and needs a 52.3 ohm resistor to match to 50R. I happen to have a strip of that value so I've used it here.

But now how to pick off both an audio signal, AND the Transmit RF source, which will range from 3Mhz to 30Mhz? My first thought is a 1Mhz diplexer, sending the below 1Mhz to the audio system, and above 1Mhz to the transmit chain. This also eliminates the need to relay-switch the IF chain output.

The log amp is configured as a power detector with a 25mV/dB output, with the 0 intercept at -84dB. I've used this exact circuit before as a power meter driver and it works very well, and is fairly accurate (good enough for the girls I go out with, anyway!) Power cleanliness is important though, as the circuit is just as sensitive at 60Hz as it is at 600Mhz, and still has some decent response at 1Ghz...and any of these signals will cause the AGC as drawn to respond and reduce the gain of the VGA's. I figure this will give Audio AGC on receive and RF AGC on transmit.

Now, coming out of the log detector is a simple rail-to-rail opamp, which can operate from a single supply voltage no problem. I haven't gotten too far into this yet, as you can see. The log detector output is 25mV/dB, and the VGA's gain control input is...25mV/dB. Perfect. Just needs to do a level shift, and a low-pass filter to have this thing working at a basic level.

If the level shift is configured at unity gain, inverting, then the default state of the VGA should be maximum amplification. (0V output from the log detector should mean +2V output from the opamp, which should give maximum gain from the VGA) I need to level shift it up to a +6V to +8V range, as the 0-2V variable gain input has to be biased up when running from a single supply (or I need to redesign the thing for a +5/-5V supply, which honestly wouldn't hurt things.)

Questions:

What do do about the output match for the RMS-2 mixer?

What to do about trying to make sure that the VGA closest to the mixer will lower it's gain first, then we start lowering gain at the VGA before the noise filter?

How much IF gain do I really need? As drawn, and if I believe the datasheet instead of my math, I'll have 56dB of gain before the mixer, maximum. This seems a bit low. Like half as much as I should need low.

[KI4YAN]

Revisions.

How much gain do I really need? All of it. ALL THE GAIN. How do I get it? Partitioning. I've got a simple 20dB gain opamp in front of the 26dB VGA. This gives maximum 46dB gain, and minimum -60dB of gain, per each variable gain stage.

A modification of the common gate Jfet amplifier kept impedances the same, but increased gain to 12dB.

The total receive chain:
Antenna->Bandpass Filter->Mini-Circuits RMS-2+ Diode Mixer->JFET LNA->SSB bandwith filters->JFET LNA->fixed 20dB gain->-80dB to +26dB VGA->Noise Filter->fixed 20dB gain->-80dB to +26dB VGA->Mini-Circuits RMS-2+ Diode Mixer->total max gain 97dB.

This doesn't seem like enough, and seems like too much at the same time. I need to figure out what the front end overload level will be, and determine the dynamic range of the system I guess.

[W4AMV]

Hello,

Consider 1 uV the threshold point for AGC to react and
300 mV the level required for efficient detection. Then the
required voltage gain is 110 dB. If 60 dB of AGC range is
available then you would be able to track a signal input up
to -47 dBm with no distortion. AGC action should occur on
the back end stages first and lastly the front end to maintain
noise figure and sensitivity. This is a very simplified outline!

You need to be able to track how the gain, noise and distortion
tracks with the distribution of AGC. Find a cascade receiver analysis
program or write one into a spread sheet. There are several excellent
work sheets available. They will permit you to tune interactive with the
gain and watch the noise, sensitivity and distortion re distribute itself
through the receiver. Highly recommend to get a handle on what you have
by this method before cutting metal.

Alan

[W4AMV]

I suggest you visit this site:

http://www.hp.woodshot.com/

And install AppCAD on your PC.

I have written scripts in Mathcad for doing the type of analysis you desire. However, the noise analysis routine in AppCAD is excellent and easy to learn. There is an example of a dual conversion receiver in one of their example files. Good reference to start. See the file attached.

Alan

[KI4YAN]

Right now, there is no amplification in the RF portion of the receiver. Only selectivity. the JFET LNA's are mostly present to overcome filter loss and conversion loss from the mixers.

In total, 80dB of AGC action should be attainable, both the detector and the VGA's are able to manage this. I think limiting it some may need to happen, but not sure.

I was hoping to avoid digital gain control, but it is looking more and more attractive. It's very easy to change the method and distribution of gain when it's digital. Just run the detector output into a ADC, and feed the digitally managed AGC out through independant DAC's for each amplifier. 8 bit DACs over a 0-2 volt range would give a gain step of 0.32dB per DAC increment.

I was hoping to avoid digital gain control, but the more I look at it, the better it would integrate in with different demodulation/modulation modes.

I've downloaded AppCad, I'll look through it tonight.