Log Amp

TI/Burr-Brown LOG1xx (several types, all a bit pricey)

Best regards, Spehro Pefhany

the classic an-31 ?

-Lasse

If it has to be high BW and precise I had good results with the AD640. Ok, has voltage in/output but I guess you know how to make a current from/to that :-)

Before looking at the price make sure you are firmly seated:

We used to use AD-640's many years ago but now use the AD8307 (no compressed IF) and AD8309 (with compressed IF) prices are under $10 plus only need one for 70+ db. 640 I think you need two for 70 dB plus more expensive.

I needed over 90dB at really good linearity with two in a row. That design was a real white-knuckle ride. But yes, depending on the range Tim needs there may be better deals to be had. It's just that I know the AD640 works really well. But the price, ouch, ouch.

We redesigned some of our boards to specifically get the AD640's converted to the 8309's. We were using 4 per board and went to two

You must really pay attention to input filtering and not allow stray signals onto the input. Basically the filter must be placed as close to the input pin as possible. We also used shields around the circuits. If he need 60-70 dB or less and depending on IF freq, shields may not be needed. Our IF was 60 MHz.

It's baseband, and not a terribly high bandwidth at that. Cost is an object -- the customer will be happier with an op amp or two and a couple of transistors.

Thanks -- I'm keeping this here pdf.

OK; start with the "standard" non-comp logamp using a grounded base transistor for the feedback (emitter resistor please for BW purposes). How about a second one with a reference voltage for input, and then a summing amp to get the compensation?

This may be over complicating things, but an A/D convertor - lookuptable - D/A convertor might be an approach as well.

If low BW then Brent's solution with a LUT may be best. Opamps plus diodes will be a rather drifty concoction.

Another very low cost solution can be RF receiver chips. Some have an analog RSSI output and you could let the whole rest of the chip lie fallow.

Some thirty years ago, when fast ADCs were expensive and hard to come by, we used an 8 bit unit in the front end of a blood chemistry analyzer. Was measuring optical density as picked up from a photomultiplier tube. We needed better than 3 decades of range and measurement was log base10 anyway. Went right from the photomultiplier anode to the current input of a (ADI) log amp and then to the ADC. Was able to get more useable range this way.

My all time favorite log amp design was a (now obsoleted) National part that had an opamp and three (?) uncommitted transistors. The opamp was set in a temperature control loop driving one transistor as a heater, another transistor was the die temperature feedback and the third transistor was used with an external opamp as a logger. Idea was to get the die temperature stabilized significantly higher than the ambient. Darned if I can find the apnote now.

Found this one- Doesn't say it was authored by Jim Williams but I recognize the characteristic scope photo

Ah, here it is... The LM389. Couldn't find the later apnote but did find this one

Hope that helps. While the LM389 is no longer available, transistor arrays are and can be used with external opamps.

And that's the problem, availability. I remember some panic at my first employer when a log amp was obsoleted, pretty much out of the blue. And later a similar thing happened at a client that had thought life was all peaches because of a nice chip. Don't rely on boutique chips :-)

Maxim has the MAX4206 and MAX4207.

Good luck in finding any production quantities ...

It looks to me like these are log amps WITHOUT temperature compensation for the logarithm gain. On page 10 of the MAX4206 data sheet, (kT/q_e) becomes (K) and temperature is ignored thereafter.

You'll need to use the builtin op amp and make an inverting amplifier with gain set by one resistor with zero tempco, and the second resistor with PTC character (resistance proportional to temperature). And that PTC resistor has to be isothermal with the chip.

The data sheet has temperature specs, and the block diagram shows temperature compensation. Page 11 talks about temp compensation.

The bigger problem is that it's Maxim, from 2004.

John

I was waiting for that comment. We have some Maxim parts in our products but I haven't heard of any problems lately. Then our builds are in small quantities. I remember the problem being in their small order desk. They didn't have normal distributors and you needed to find a company credit card to purchase from them. Today Mouser and Digikey carry them.