"Slick Trick" Rectifiers for Level Detection

[mikep]

[quote author="JohnRoberts"]

For the record if I was doing this today it would be one <$5 micro...

Tempus Fugit.

JR[/quote]

hi john, which processor, or family of processors are you using? Ive been spending a fair amount of time working with 18F PICs myself.

 

[JohnRoberts]

Yup.. I'm using the 18F (microchip) in my current drum tuner and have a development board for the dsp 16 bit family which I haven't fired up yet.

I'm a little weary of the limits of dealing with only 8b word length.. life comes in more than 256 flavors.

JR

 

[PRR]

> Sorry about the size and two images

370KB PDF file.

 

[Crusty2]

[quote author="JohnRoberts"]
...For the record if I was doing this today it would be one <$5 micro...

JR[/quote]

You should do it. I think there's a huge market for a "poor man's AP". I mean, what's out there right now?

 

[JohnRoberts]

I have a (short) list of products I could do but I should point out I have a full time gig plowing a virgin field at the moment (electronic drum tuning). I find researching stuff that isn't in the books (or any books I could find), quite engaging.

Low end audio test gear is a potential niche product but 20+ years later I doubt it would support the price it did ($300 in 1980 dollars), and to do it right it needs to be in nice tooled up case with perhaps a custom LCD display. With modern low cost DSP solutions it's not out of the question to make a hand held 8903 killer.

Another serious consideration wrt this kind of gear is, "can you do it with software in a PC" ? To answer, mostly yes. Hard to sell a $300 box when you can cover it with freeware or low cost software and a sound card.

I found designing the TS-1 an interesting experience. We take test equipment for granted but designing that piece to even relaxed performance standards on a budget was a stretch.

Over the years I have seen some nice pieces come and go. dBx made a nice AC voltmeter in the form factor of a VU meter, but reading out on a log scale so it could display over a 100 dB range.

Years later Buff made a similar meter but with more bells and whistles (noise weighting curves, etc).

Heathkit had a huge low end test equipment business.

To some extent these were either too specialized, and/or mooted by low cost test gear coming from the east (Japan before China).

So I find it interesting but not commercially viable. Not that I'm looking for work.

JR

 

[EZ81]

Is the value of the 5k resistor significant at all (assuming ideal opamps)?
Either D3 is conducting, then it just adds some extra loading for the left opamp's output. Or D3 is not conducting, then the +in of the right opamp is grounded (to the virtual ground) through R11 anyway. So it looks useless. Or am i missing something (maybe a real world effect...) here?

I would like to see if people can figure out why the resistor values are in that ratio...

I can't explain how the professor came up with this :shock: , but it seems to work on paper. Thanks :thumb:

 

[CJ]

Gain + 1 in the denominator is all I can think of.

 

[bcarso]

The 5k is the right value for having the same magnitude of conduction current when a given diode conducts, and hence for matched diodes a symmetrical output swing at the left-hand opamp output [EDIT wrong---the swing is asymmetrical]. It's not ultra-critical, but it is intentional. But when I went back and dredged this up I thought it was incorrect---the fun (if you don't get out much) is figuring why it is correct.

The circuit grew out of trying to understand some of the errors in the standard two-opamp precision rectifier.

 

[bcarso]

That circuit might be better called the symmetrical diode-current rectifier. And it doesn't really address all the errors due to diode charge, both due to diode capacitance and (for regular PN) reverse-recovery charge.

To do a bit better use a true differential amp.

 

[JohnRoberts]

Good thing I was given a pass, I didn't even consider the diode C.

Funny, my approach in such circuits was always to bias up the diodes to be closer to conduction, having to swing almost a volt at every zero crossing is always a gain bandwidth bummer.

It looks like opamp input bias currents will be in imbalance so at idle, one diode may be slightly on. Is there any reason to consider varactor like reverse voltage effect imbalancing the diode capacitance?

I typically ended up running the result of rectification into a virtual earth for logging or integration so the NI topolgy in second stage is not as accommodating for my applications.

JR

 

[bcarso]

Sure. Again, the objective was to have the diode forward currents in conduction the same magnitude. This is only one aspect of the error, and maybe not the most important one.

One of the biggest errors is just the capacitance of the diodes. There are ways to compensate for those to some extent.

 

[EZ81]

@bcarso re. R12=5K: thanks.

What determines the gain when D3 is conducting?

I'll try, but the result does not comply with [quote author="bcarso"] [...] and hence for matched diodes a symmetrical output swing at the left-hand opamp output.[/quote], as the voltage swing at the left-hand opamp seems far from symmetrical. Maybe someone can help with that?

gain from input to R12:
- (R11 || (R8 + R9)) / R10 = -2/3 (inverting)

gain from R12 to output:
1 + R7 / (R8 + R9) = 3/2 (non inverting)

overall gain = -1 :?:

 

[bcarso]

The voltage swing at the left hand output is not symmetrical. Only the diode current magnitudes are---I went back and corrected the original misstatement.

To do both a true diferential amp on the right is needed (and then get rid of the 5k).

 

[JohnRoberts]

I vaguely recall a trick rectifier that was published in one of the design rags, way back when. It was using a CMOS (?) or some new something MOS opamp where the output stage would pull to the -rail and the input remained operational down to -rail and below.

I don't recall all the details but they powered the opamp from only the plus supply but fed it from a bipolar waveform. The output clipped at ground, performing one half wave rectification, a second opamp section inverting performed the other half wave...

It wasn't very practical, but was clever. An example of non-linear thinking, good for rectifiers.:grin:

JR

 

[JohnRoberts]

Yup, that's it... Imagine trying to figure that out from just the schematic

JR

 

[bcarso]

Yeah, basically the same thing except with no effort to equalize diode currents (which is probably a small effect most of the time anyway).

I like one example in one of his books in that series (maybe it is that one) where he has two diodes, two resistors, and one opamp, but posits that the input signal is a current rather than a voltage.

If currents were our typical "currency" that would have appeal.

EDIT: Note that if we also had a lot of "true differential" input-output amps it would work awfully well, as the current source would have the same small voltage swing magnitude independent of polarity.

I've come to the conclusion that the best way to do rectification is with a basic multiplier quad cell and a fast comparator with complemented outputs. But that's not something simple to do as it requires matched quads etc.

 

[bcarso]

A better transistor array (like THATs with good PNPs and dielectrically isolated) would make that circuit work a lot better too.

I was amused to see the datasheet for the 3096 that comes up from Intersil saying the part is obsolete, but maybe you can use the HFA3096. Maybe, if you can manage to avoid GHz oscillations...

 

[JohnRoberts]

From a quick glance it looks like one of his PNPs in the FWR is drawn in backwards (the lower one Q2?). This FWR like so many others will still suffer GBW issues at low level HF as it must swing a full +/- diode drop at zero crossings.

I have used the same general approach to make RMS convertors for high end VU meters, but never for large scale production. The relationship for RMS is square root of the integral of X squared.

That circuit FWR the signal with Q1 and 2, then squares the signal by doubling log voltage of Q3 then converting back to current with Q4. That squared current is integrated by 10uf across A4 and then SQRT of result performed by Q5.

I typically did it all with an all NPN array.

JR

 

[Guest]

[quote author="mediatechnology"]

I have no idea what any of it actually does. I can quit at any time if you don't find it useful.[/quote]

It may give an interesting guitar effect. I would add an opamp with couple diodes in feedback before this thing for some sustain distortions and compensation of crossover distortions in input transistors (Dioded transistors as drawn are not enough to convert small signals).

 

[Guest]

[quote author="mediatechnology"]Thanks wavebourn. I'm kinda interested in lashing up the FW current rectifier part. I find the current mirrors interesting. JohnR posted his Loft TS-1 schematic here that used mirrors in a different rectifier configuration as well. You might want to have a look at that one: There's a lot of clever stuff in there. Let me rewind and I'll find the link....

In case you missed it: http://www.groupdiy.com/index.php?topic=24631&postdays=0&postorder=asc&start=54[/quote]
I saw that already, thanks!
I did not comment it because it is worthless now for my practical purposes, what I have satisfies me completely (I use log conversion before rectification). However, it is not a true measurement systems, but good enough to see presence of a signal and approaching it to a clipping point.