High bandwidth, low distortion variable gain control?

[Alec]

I've grown bored with Things That Go Blink and little Arduino projects and I've decided to take on something new and a bit more challenging: high(er) frequency analog.

But my project has hit a road block. How does one do electronically variable gain (or attenuation) at 10MHz with minimal distortion? My first instinct was to use a digital pot- but the datasheets show that those sorts of things have nowhere near enough bandwidth. I would encounter the same problem if I put one in a opamp feedback loop. One prepackaged solution that I've found is the LMH6505. Unfortunately, it's a fairly expensive part and the Linear-in-dB thing would make it a little difficult to control accurately.  http://search.digikey.com/scripts/DkSearch/dksus.dll?vendor=0&keywords=lmh6505
A JFET operating in its low current range looked like it might work; but I can't figure out a way to compensate for the wide manufacturing tolerances without creating even greater problems.
(edit: I was trying to use an opamp to regulate the JFET's voltage drop and use a large cap to decouple the high frequency that it SHOULD NOT be trying to correct)
(edit2: and I just noticed that the decoupling cap should be on the opamp's sense line. Dumping the HF between the opamp and the JFET would effectively create a HF open loop and wild oscillations...)

Variable Gain sounds like it would be frequently encountered in circuit design. But my Google-Fu is apparently too weak to find the Simple Way of doing it. Is there one, or should I just go with the LMH6505?

For context, I'm building (yet another) AD9833-based function generator.

(Here is how I was trying to approach the problem. Never mind, the exact values and components. This is just the experimental concept stage.)

Full Size http://i42.tinypic.com/25i18hd.png

[Rerouter]

something like this ?
http://portal.tpu.ru:7777/SHARED/o/OS/Tab2/eumw06.pdf
 being a voltage divider drives the attenuator it looks like it would be very easy to control digitally or via opamp, then again it would appear that the signal is biased in some way from it so it may need a fully passive method...

just as a heads up i'm not much of an RF guy and this looked about right,

[Alec]

Hmm... p-i-n diodes. I hadn't thought if using those (in fact, I had never heard of them ).

But anyway, manipulating the effective capacitance of a diode, by playing with its biases, to effect its RF resistance sounds like it would work wonderfully for GHz range signals. But I need something that will work at much lower frequencies and across 5 or 6 orders of magnitude (and without requiring my MCU to do too much work to compensate for non-linear frequency responses).

But still, thanks. It's good food for thought. 

[vk6zgo]

PIN diodes are used extensively in RF work to do the job of attenuators.
Because they are extremely small,they can be incorporated into equipment with a minimum of fuss.

At this point,I would suggest having a look at some real equipment (not 'Net crap),to see what other people use in a similar application to yours.
The above suggestion may seem like cheating,but reinventing  the wheel isn't the most rewarding of activities.

[steve_w]

I assume you are doing a signal source.  A low risk design technique for this type of application can be achieved by buffering the output of the signal source and presenting a fixed load to the source.  Then use a resistive step attenuator to reduce the signal.  Think of it this way; the signal is at max amplitude and you choose how much you use rather than trying to develop a variable gain amplifier to get the signal level you want. 

advantages: you present a constant load to signal source this reduces pulling in the oscillator (a moot point if you use a DDS), frequency independent (to a point) and cheap (made of passive components and not as much complex design). 

Disadvantages: its big, its not continously variable frequency range drops off over high VHF to high UHF (depending on how good the design and construction are)

have a look here http://www.jfwindustries.com/catalog/Toggle_Attenuators-64-1.html for an example of a commercial one.

If you want to build the variable amplifier way let me know how you go, I'd be interested to see what you come up with.

regards

SW

[steve_w]

or you could just do this:

National Semiconductor LMH6502
www.national.com/ds/LM/LMH6502.pdf

It was the answer to a similar question to yours I take no credit.

regards

SW

[deephaven]

How about an MC1495 http://www.onsemi.com/PowerSolutions/product.do?id=MC1495
For best results use differential signals.

[blackdog]

Hi Alec,


Take a look at this data sheet, last page...
http://www.linear.com/product/LT1228

Kind regarts,
Bram

[Zad]

There are several methods.

You can use a resistor ladder, and something like this multiplex switch http://www.analog.com/en/switchesmultiplexers/multiplexers-muxes/adg706/products/product.html  In concept, it is similar to the PIN diode switch array.

Or you can use a multiplying DAC, http://www.analog.com/static/imported-files/overviews/AnalogMultiplyingDACs.pdf

Or a Variable Gain Amplifier. http://www.analog.com/static/imported-files/data_sheets/AD603.pdf

Or just do it old-school with a multi-way rotary switch and a bunch of resistors. 10MHz is practically DC!

[Alec]

Wow, thanks for the responses.

I've already found those. The point is that I'm trying to avoid using an $8 part. I'm also trying to get continuously variable control (or near it, 100+ steps) electronically.

The LMH6502, the linear version of the LMH6505, is not only expensive, but has worryingly low levels of stock on Digikey. This is a clue that I maybe shouldn't use it in a new OSHW design. LT's offering looks a little healthier, but is equally expensive.

The AD603 looked promising. But (correct me if I misunderstood the datasheet) it seems to only allow me to select one of the 7 or 8 taps. That's much too coarse for my purposes. However, if it allowed my to select any or all of them, for 8-bit control, then it might work. I've considered rolling my own with a IO expander, an array of FETs, and a bunch of resistors. The only question is what would happen when I try pushing a 10MHz square wave through it. If anyone knows of a 8-bit DAC with external reference that can handle having a high frequency signal driven into that reference pin then let me know.

I've also looked at optically coupling a steady control signal to a small variable conductance element in an opamp's feedback loop. Unfortunately, nearly all optocouplers use photo diodes or transistors and seem to be designed for on-off operation. I can't find anything with a gradual curve. Photoresistive optos only exist as obscure niche components. (and I've decided against rolling my own out of a LED, a cad-cell, and a bit of heat shrink. That would be a little hackish)

This thread has effectively answered my question. That no, I'm not just being thick. What I'm trying to to actually is that hard. There's no simple way to do it with a couple opamps and a few discrete transistors. So for now I'll just stick with the (slightly less expensive) LMH6505. I think I can manage the linear-in-dB thing.