Log converter IC

[audiox]

I need a DC log converter for a level meter project. I found one that I designed a long time ago (made of transistor array and op-amps). It works fine but needs trimming. I thought that there must be available some (laser trimmed?) chip to do the same job but without manual trimming and component selection. It should be non-inverting (lin input 0...x volts/amps produces log output 0...y volts/amps) and provide 80dB range. Any suggestions?

 

[abbey road d enfer]

I need a DC log converter for a level meter project. I found one that I designed a long time ago (made of transistor array and op-amps). It works fine but needs trimming. I thought that there must be available some (laser trimmed?) chip to do the same job but without manual trimming and component selection.

That's basically what the AD536 does. I believe it's not anymore current but you may find some NOS. That's what's used in the SSL comp, I believe there's some info in the relevant threads.

It should be non-inverting (lin input 0...x volts/amps produces log output 0...y volts/amps) and provide 80dB range.

In essence, a log converter should produce an output voltage ranging from minus infinity (for no input signal) to whatever the volt-per-dB law suggest. For obvious reasons, there will be a lower limit and an upper limit to the actual output voltage.

Any suggestions?

The THAT2252 idles at ca. -500mV, with 6mV/dB conversion law, making it suitable for a range of ca. 100dB (usually scaled -80 to +20dBu). It normally requires no trimming for AC applications (DC is a tad more complex)

 

[JohnRoberts]

What kind of meter? Unless you are planning to drive a mechanical meter, you can just scale the comparator threshold resistor string for log scale.

The cheapest play to actually make an audio log conversion is the THAT chip, but IIRC it is missing the final stage of classic log conversions. This works fine for compressor or noise reduction applications interfacing with VCAs but you might want to tweak it some for precision voltage output log conversion.

I used to make them (log conversions) with a single transistor array and a few opamps, but that old soldier transistor array is long obsolete.

JR

PS: DBX used to make a nice log mechanical meter that displayed an extended range. I toyed around with making a cheap and dirty version using an IC companor IC (like NE570) to double the dynamic range of a typical VU meter. +3 becomes +6 and -20 becomes -40dB.

 

[jdbakker]

Any suggestions?

The modern way of doing it would be with an ADC+microcontroller (or even ADC/EEPROM/DAC, if that floats your boat). With the right converter it'll need less trimming and be cheaper than the analog equivalent.

That's basically what the AD536 does. I believe it's not anymore current but you may find some NOS.

A quick look at http://www.analog.com/ shows that the AD536A is still in-production.

JDB.

 

[Dan Kennedy]

There's also a fair number of rf chips that are log converters for signal strength info for cell phones and systems.

I thing Analog Devices makes a couple.

 

[audiox]

I am surprised. I thought there would dozen chips to choose from. I guess I have to use my old proven design...

Unless you are planning to drive a mechanical meter

Yes, I am driving a moving coil instrument.

I used to make them (log conversions) with a single transistor array and a few opamps, but that old soldier transistor array is long obsolete.

I have a bunch of CA3086s left, which I used in the original design. And NXP/Philips makes nice dual BC847BS which costs only a few cents each.

The modern way of doing it would be with an ADC+microcontroller (or even ADC/EEPROM/DAC, if that floats your boat) ...

I was 100% sure that you are going to suggest that...

A quick look at http://www.analog.com/ shows that the AD536A is still in-production.

Availabe from Farnell for 23 to 88 euros (!) depending on package. Surprisingly TO-100 metal can is the cheapest and plastic DIP most expensive.

 

[jdbakker]

A quick look at http://www.analog.com/ shows that the AD536A is still in-production.

Availabe from Farnell for 23 to 88 euros (!) depending on package. Surprisingly TO-100 metal can is the cheapest and plastic DIP most expensive.

Digi-Key has them for under 10EUR.

Most of the RMS-converter chips are AC-optimized, and drift (badly) with DC inputs. Hard to get 80dB range out of them. A chopper could help, but that gets very hairy very fast.

I used to make them (log conversions) with a single transistor array and a few opamps, but that old soldier transistor array is long obsolete.

I have a bunch of CA3086s left, which I used in the original design. And NXP/Philips makes nice dual BC847BS which costs only a few cents each.

Probably the best plan to get (close-to-)DC performance in an analog fashion.

The modern way of doing it would be with an ADC+microcontroller (or even ADC/EEPROM/DAC, if that floats your boat) ...

I was 100% sure that you are going to suggest that...

I can see why you'd say that. Let me go on record though that I prefer the elegance of a clean and simple analog solution over a DSP'ed one any day. Sometimes processing in the digital domain makes most sense, though, for reasons of power, money, flexibility and/or PCB area.

JDB.
[and I hated it when Elektor magazine started using microcontrollers in almost every project, even when logic and/or op-amps would have done just fine]

 

[JohnRoberts]

I have made a number of log conversions using the good old 3086 and you can make a passable RMS conversion (SQRT of intergal of X^2). With an opamp stage and a cap.  I think there's a few old national app notes that are pretty close. I'm sure we've discussed this before.

With a mechanical meter you could just about make th meter the integration stage in the RMS converter, if you want to be clever.

I used 3086 for the dB math in a Dugan algorithm automatic mixer back in the '90s.  I really liked that part back in the day. but I also like cheap micros.

JR

 

[abbey road d enfer]

Most of the RMS-converter chips are AC-optimized, and drift (badly) with DC inputs. Hard to get 80dB range out of them. A chopper could help, but that gets very hairy very fast.

I've had good results using THAT 2252, implementing their very clever and simple compensation and proper DC offset compensation.
Still I can't figure what application requires DC operation with 80dB dynamic range...?

 

[PRR]

It's been a while.... don't classic junction loggers want a tempco resistor? That was hard to score back then, and the company may have closed. Or is that not needed for meter-accuracy in a shirtsleeve studio?

There was a LM386 speaker-amp with extra transistors, and a plan to gimmick the power-amp to heat the chip to constant temperature, avoiding all tempco issues.

There's also the Neve diode string which gives near BBC PPM "log". With booster, rectifier, peak-catch, buffer, etc, the parts-count is unimaginably low by modern standards; almost as low as a PIC.

 

[JohnRoberts]

The biggest problem I recall from a dB meter I made back in the '80s that covered +20dB to -50dB (full bandwidth), was an error from the real resistance (Rbb) term in the base emitter junction.. I read more than a half dB high above +20 dBu. On the other end the errors were related to gain bandwidth of the rectifier.


JR

 

[abbey road d enfer]

It's been a while.... don't classic junction loggers want a tempco resistor? That was hard to score back then, and the company may have closed.

It would be easier today to use an LM335 to provide tempco.

 

[audiox]

Still I can't figure what application requires DC operation with 80dB dynamic range...?

I am replacing a faulty potted log converter in an old test instrument.

It's been a while.... don't classic junction loggers want a tempco resistor? That was hard to score back then, and the company may have closed.

The 1K +3500ppm/C in the National application note is difficult to get. But Farnell sells 2K +3000ppm/C and 2K +3900ppm/C. Connect them parallel if you want to use the old circuit.

 

[JohnRoberts]

For a piece of test equipment you may need to compensate for the Rbb error if you hit the log amp hard. Most test equipment uses input gain/pad switch so meter range is limited and this may not be a problem. In my applications my 0 dB was with a reference current fed junction subtracted from the input curent fed junction, so temp errors should be minimal around 0 dB.

Back then I scratched out a future design for a first order correction for the Rbb error, but never did a second generation of that product (TS-1) so probably didn't save my notes.

JR

 

[abbey road d enfer]

Still I can't figure what application requires DC operation with 80dB dynamic range...?

I am replacing a faulty potted log converter in an old test instrument.

Then I think the 536 is the way to go