Envelope Follower Advances

[R.G.]

There have been a few generations of sensing and control ICs since I last designed an envelope follower. In designing some instrumentation I ran onto some RMS to DC converter chips which differed from the ones all the way back into the 1970s by being affordable and easy to apply. Linear Techology and Analog Devices now offer RMS to DC converters that are under $10, some under $5 in ones. RMS to DC is arguably as good or better than peak detect for an envelope detector.

There ought to be (wink, wink, nudge, nudge) some good designs to do with these chips.

[tubegeek]

1. Check your forum messages ^above.

2. Check whichever email the reply form for your Vox Thomas stuff goes to

3. Check keen@eden.com if that still exists

4. I also left a message for you using the reply form at visualsound

I'm coming to your neck of the woods mid-October and would very much like to buy you a plate of BBQ or whatever. I hope this message finds you in finest fettle.

tubegeek@earthlink.net or jeremyepstein@earthlink.net or jer.eps@gmail.com or 917 589 5054. Any of those will work!

[EBK]

THAT Corp. sells some VCA chips with RMS detectors as well (they call them Analog Engines).

[smallbearelec]

There have been a few generations of sensing and control ICs since I last designed an envelope follower. In designing some instrumentation I ran onto some RMS to DC converter chips which differed from the ones all the way back into the 1970s by being affordable and easy to apply. Linear Techology and Analog Devices now offer RMS to DC converters that are under $10, some under $5 in ones. RMS to DC is arguably as good or better than peak detect for an envelope detector.

There ought to be (wink, wink, nudge, nudge) some good designs to do with these chips.

Hi R. G.--

THAT Corp. now suggests their 4305 over the 4301. Neither available in DIP, unfortunately, but not expensive, as you mention.

http://thatcorp.com/datashts/THAT_4305_Datasheet.pdf

[Rob Strand]

The rms chips have been used in many new-ish places over the past 25 years or so.   One place is stabilization of sine oscillators.

One point worth mentioning is the attack and release time constants are linked.   You can only set one time constant on the chip.  That means adding another filter after the rms chip, it will need to have diodes.   If you poke around other THAT chip app notes and datasheets you might stumble on some circuits with time constant stuff and how to set thresholds on compressor/limiters.

[EBK]

In the app notes somewhere, THAT describes a non-linear capacitor circuit for the time constant, which is quite nice. I've seen similar in the old compander chip notes.

I'll dig it up later if no one beats me to it.

[pruttelherrie]

In the app notes somewhere, THAT describes a non-linear capacitor circuit for the time constant, which is quite nice. I've seen similar in the old compander chip notes.

I'll dig it up later if no one beats me to it.

I think you mean this one: http://www.thatcorp.com/datashts/dn114.pdf

[Sooner Boomer]

I just found an Analog Devices AD536 from a long-ago research project.  I think I'll throw a breadboarded circus together and see how it works.

[EBK]

In the app notes somewhere, THAT describes a non-linear capacitor circuit for the time constant, which is quite nice. I've seen similar in the old compander chip notes.

I'll dig it up later if no one beats me to it.

I think you mean this one: http://www.thatcorp.com/datashts/dn114.pdf

Yep, that is it (no pun intended--I hate that company name, again, no pun intended  ).


Costello: I'm asking you who's the company that makes the Analog Engines.

Abbott: THAT's the company's name.

Costello: That's whose name?

Abbott: Yes.

Costello: Well go ahead and tell me.

Abbott: THAT's it.

Costello: That's who?

Abbott: Yes.



Another description of the NLC is in here:
http://www.thatcorp.com/pedals/4316%20Battery-Powered%20One%20Knob%20Squeezer.pdf

[cjlectronics]

Looking at the Analog Devices LTC1968, https://www.analog.com/en/products/ltc1968.html#product-overview  The minimum RMS input value is 5mV.  How does that translate into minimum frequency.  Obviously, as long as there is a input frequency that is 5mV RMS the IC can convert to a DC voltage value.  But is there a minimum frequency where this IC would limit out on.  I didn't see that in the datasheet.

[R.G.]

@CJ:
The LT part is one of the ones I was thinking of. I'll take a look. the low frequency limit may be zero - DC has an RMS value too.

To do real RMS, you have to compute the DC +AC value of the waveform. Maybe they did it right and have response down to DC.

[Rob Strand]

Obviously, as long as there is a input frequency that is 5mV RMS the IC can convert to a DC voltage value.  But is there a minimum frequency where this IC would limit out on.  I didn't see that in the datasheet.

It does DC, see the list on the first page "True RMS AC + DC Measurements".   At low frequencies the chip will produce an output.  The lower frequency limit is set by the *error in the RMS measurement*, which is more to do with being able to interpret the output as RMS.    RMS is defined as an averaging integral.   The averaging time is set by the averaging filter cap.     

There is a point where you have to change your interpretation of the input signal being:
- a slowly changing "DC" input where the chip tracks the DC level, and
- an AC signal where you can compute the rms, which just happens to have a DC offset.

The averaging time sets the time scale where you interpret it one way or the other.   There is a fuzzy zone where it is unclear which case it is and this is where the idea of measurement error at low frequencies creeps in.

[PRR]

As said, LTC1968 will work down to zero frequency. (This has NOthing to do with minimum voltage; why would it?)

In many applications we do not want to know about the (boring) DC, only the alternating part of the input. We put caps in. Pages 10+ show the error versus frequency for several sizes of caps. There's certainly no trouble at any musical frequency.

LTC1968 (and RMS converters in general) apparently has a "minimum input voltage" because internal offset errors degrade accuracy to less than you want. Page 6 "DC Transfer Function Near Zero" shows that few-mV signals may come out either high or low of true zero. Without knowing what is really inside this too-complicated chip, I am picturing several devices with ~~1mV offset error each, making a few-mV zone of "poor accuracy" where the output can't be taken as Truth.

[Ben N]

I recently picked up a THAT4301 in DIP configuration from https://www.profusionplc.com/parts/that4301p20-u, but it was pricey (~$13), and their supply is obviously limited (172 when I logged in).
I also came across this--take your chances if you must, I don't mind throwing down a dollar or two on questionable LM301s, but I'd be wary of this.