"Slick Trick" Rectifiers for Level Detection

[Kit]

Moving into opamp territory.................

http://sound.westhost.com/appnotes/an001.htm

The usual suspects..............allthoug I must admit too having used the "simplified version" at the bottom of the page on several occasions.

But I usually insert the AC at the node between D1, D2 and leave D1 out of the NFB loop.

Keith, fascinating about the ABBA console. I didnt know they had a Harrison at Polar.

I always figured it to be a Hagstrom desk. :green: :razz:

 

[Guest]

[quote author="JohnRoberts"] but for your friend's combined meter I would make the peak release faster than the ave so the operator could visually discern when the meter indication is responding to peak or average, intuitively if not cognizantly. The faster release would also give it somewhat less weight visually for more accurate loudness estimation. [/quote]

I use one rectifier in my tube "magic eye" meters with 2 different capacitors: one small connected directly, and another bigger connected through a resistor. As the result, it shows slowly average level and peaks faster. However, I use one dual opamp with each tube (rectification and approximation) and one hi-volltage transistor to replace a first triode in the magic eye tube (base current variations instead of a big voltage swing on a grid).

 

[JohnRoberts]

[quote author="mediatechnology"]Thanks kit - I've used the last one too with good results.

Rod Elliot writes:

It is virtually impossible to make a full wave precision rectifier any simpler

With all due respect I think we should take up his challenge.

[/quote]

I consider it a given in design engineering to reduce circuitry to the simplest that will be functional with adequate performance. In precision rectifiers you often add parts to improve some performance characteristic. I have designed (modest) test equipment so spent a little time on precision rectification.

The popular basic FWR (from narional app notes) is two opamps with one diode clamped in one polarity, and contributing 2x the current in it's active polarity. This works and can be reasonably accurate, but will suffer from GBW issues at low level HF due to having to swing entire diode drops. An alternate approach is two polarities that are then rectified but you need to compensate for diode drop, making it faster but not very precision.

One of my favorite approaches was to use a single inverting op-amp with a transistor supplying feedback current in one polarity, and a pair of transistors connected as a mirror for the other polarity. The collector from one polarity is combined with the second collector from the mirror pair for precise FWR current output. This too suffers from the GBW limitations of the simpler 2 opamp FWR and you can mitigate with DC biasing the bases closer to conduction and I have even added voltage gain with a discrete transistor inside the opamp feedback loop.

I've used this in any number of front ends for log convertors and sundry precison applications. I could realize a 20kHz -3dB bandwidth @ -50 dBu with the bandwidth dropping by half for every 10dB drop lower. The current output was nice for wide dynamic range rectifiers as opamp DC offsets can become an issue (I was designing with circa '80s opamps).

For precision wide band measurements at even lower levels a 2:1 compander in front of the rectifier could scale the -3dB 20kHz bandwidth to -100 dBu perhaps limited by the performance of the companding but since you have a precision rectifier right there it would be straight forward to generate a gain control and send to a VCA.

FWIW, in my less than bench grade test equipment design (Loftech TS-1) we lived with the 20 kHz bandwidth down to -50dBu with no customer complaints. It would slightly under report noise floors but precise noise measurements was not it's primary mission. Frequency response from -40 to +20dB was excellent.

I must admit I gained a respect for big dog test equipment designers, but they had a few extra zeros in their BOM budget.

JR

 

[SSLtech]

[quote author="mediatechnology"]...Our UK dealer took me and "two of the guys from Abba"...[/quote]
Very possibly Leif and perhaps Michael Tretow, I shouldn't wonder.

Michael was -from all accounts- an incredibly creative guy, and worked musically with Benny & Bjorn with a phenomenal patience threshold also. Leif's technical ministrations were equally creative, and yet he never lost sight of the ability to distinguish between what was important and what was just technical frippery.

When Benny& Bjorn wanted more and more channels (and they loved the meter upgrade because it actually made it easier to spot "seven seconds' silence" on a track... (which would then -of course- get something punched in to fill the gap, and 'mult'-ed out in the mix... requiring even MORE channels to mix with!) so when the console got max-ed out too frequently, it was expanded to a 56 channel, by hacksawing two consoles together.

In the same way that SSL discovered that balanced bussing was suddenly a necessity, Leif found that the mix buses had to be balanced, so that was another thing which had to be fairly promptly accomplished. (...-You know the words to that tune, I'm sure Wayne!)

Did you know also that Leif originated the "monitor level feeds routing matrix" mods, for more FX sends at mixdown time...? -SSL took that and added it into the modules, I think when the 4000B begat the 4000E.

Most of Leif's stuff was apparently was done in (short) gaps between sessions, or -more commonly- overnight, since Polar was a fairly busy place at the time.

Keith

 

[JohnRoberts]

This weekend I'll look for an old TS-1 shem.. If I can find that as I recall I did the FWR and log conversion with one cheap transistor array and a few opamp sections.

Real tricky part was converting the log output voltage into a digital readout in dB.. It also had a frequency counter so my dB convertor was a V to log to F

If I recall last time I looked at I had trouble following it, and I designed it.

Funny cheap cmos counter trix and dividing the mains freq for my time base, etc.

I wish I had a cheap micro I could use back then like today..

JR

 

[bcarso]

[quote author="JohnRoberts"]with one cheap transistor array
JR[/quote]

Oh do I miss the CA3086/3046, particularly the STMicro version where they used a vastly improved process---negligible popcorn noise, high beta. Amusingly their datasheet was a photocopy practically of the original RCA parts, but testing revealed something waaay better.

But, not a good use of foundry time/space.

 

[bcarso]

[quote author="mediatechnology"]
I could use a matched diode set right now. [/quote]

Dual JFETs will work if all you need is a dual.

 

[PRR]

> It is virtually impossible to make a full wave precision rectifier any simpler...

But who wants a rectifier? I usually want a Peak Catcher. Most "precision" rectfier schemes won't peak-catch without added monkey-motion.

The lone PNP in this plan plus the next NPN implements a peak+peak catcher with over 40dB range with precision ample for most practical audio. Operation is not obvious, and masked by other functions like log-bending. But roughly: current out equals absolute value of current in. Read this on a resistor and buffer to a capacitor.

Of course even with input booster and output buffer there's fewer transistors than in a TL071.

 

[JohnRoberts]

I find rectification useful for lots of stuff, while for meters I used both peak and average.

In my current (day job) I find averaging over multiple cycles much more reliable than capturing peak information for identifying resonances and making level measurements. Although I'm mostly pushing digital poop around. No diodes were electrified performing this rectification.

JR

PS: Yes I miss the old arrays, but that's old school discrete design... :guinness:

 

[JohnRoberts]

That is reminiscent of "synchronous rectification" a technique used to improve efficiency of low voltage high current power supplies. In a 5V supply you're throwing away 10% in the diodes. A lot of modern stuff is running from 3V or so. Of course the switching element must tolerate the negative swing but in low voltage supplies that is also modest.

JR

 

[bcarso]

Seems to me there are some integrated parts that accomplish this now, that I saw several months ago.

 

[Guest]

Thanks for the good idea, I'll use now FETs to rectify 12.6V for filaments, instead of Shottky diodes I used before! :thumb:


...or FETs in parallel with Shottkies, gates controlled from 600V p-p...

Hmmm... Some of them have Zeners already... How good they are on reverse polarity, when drain and source swapped?

 

[JohnRoberts]

That looks like a glorified version of the diode clamped two opamp FWR. Unless there's some extra circuitry not shown to hold the output of A1 close to conduction during the part of the waveform when the transistors are cut off that circuit will suffer from lousy low level HF performance with A1 pegged to the - rail and having to slew - supply voltage before it can start conducting again each cycle. A typical speed up bias circuit could be as simple as two signal diodes joined at their cathodes with a resistor to -supply. one anode ties to A1 - input. The other anode to A1's output. This should hold the two transistors just cut off but ready and close when it's their time to conduct.

It still would have to slew a part of a diode drop each zero crossing so will still be somewhat limited for precision at low level HF.

JR

 

[JohnRoberts]

Sorry about the size and two images but my scanner couldn't cut it as one.

At the risk of bragging there are several IMO clever bits in that design.

The FWR I was talking about is in middle of first image. That rectifier was -3dB @ 20kHz, at -50dBu (-3dB @ 10kHz at -60dB etc).

The current output of the rectifier feeds into a log convertor to provide a dB output. The voltage from the opamp used to generate the FWR also feeds a frequency counter.

Some perhaps interesting areas:
* frequency counter actually counts zero crossings so LF can be counted in 1/2 second.

*Log (dB) output compared to a zero dB reference and converted to frequency for + and - dB readout.

* simple voltage controlled osc runs over <20Hz to approx 30kHz range with <.2% THD (not trivial).

*note trick PS in top left corner. High current 5V rail unregulated is fed from switched rectifier to eliminate dissipation of dropping full unregulated supply. Pulled a lot of heat out of the unit and allowed operation from smaller transformer. Today this would probably be done with a switcher.

* sine wave output uses 51 ohm buildout resistor and is normalized to dB meter input so swept impedance measurements could be calculated from drop across the resistor (look up table was provided in accompanying book)

* frequency counter had div by 10 auto range. Nothing special about counter. Grabbed time base from mains frequency which was problematic for customers who did battery powered conversions.

While I apparently checked and signed off on that schematic I make no representations about the absolute accuracy, but it is a working design with thousands sold and in the field. I have one still working on my bench.

Trust me when I say there are many hours of design in that one schematic. Some of the things that may look like unneeded complexity are for first order temperature stabilization. Sine wave oscillator was not perfect over wide temp range but decent.

Biggest customer criticism of original design was desire for finer frequency adjustment. IMO that was more an artifact of having a frequency counter hanging off the output than really needed, but the customer is always right and a later version had a trim.

Enjoy

JR

 

[bcarso]

That's cool

It looks like the substrate diode of U3 would be getting a bit forward-biased, but I guess not enough to create a problem?

I often wished they had put that chip in a 16 lead package and given us uncommitted emitters everywhere. But whadaya gonna do, as Ben Gebhardt says.

 

[StephenGiles]

Many thanks for that, I have joined the two halves:
http://www.4shared.com/dir/4185394/f049a7cf/TS1.html

I hope you don't mind me linking this to Aron's Stompbox forum, because I think there may be circuitry of interest to guitar synth builders - tracking VCOs and the like.

 

[JohnRoberts]

Brad the substrate diode(s) are wrt the collectors. The substrate (13) has one collector connected to it, and all the rest clamped at higher voltages. Perhaps it isn't obvious but that rectifier input and therefore the substrate is sitting at around -4V DC. Most of the collectors are up around ground.

I've run into a number of quirks with that array (in other designs) even when you follow all the rules, but the substrate was AFAIK under control.

One error term that I considered cleaning up in a later generation design was an over reporting at the upper limits of dB range due to deviation from true log law cause by a resistance term in small array transistor emitter. Error was fraction of a dB at +20dBu so close enough for government work, but a first order correction was possible (I don't recall if it involved another opamp, but a small R in series between 7 and 1,14 could quantify that voltage error term for subtraction. Note companding 2:1 in front of that circuitry would have also reduced that error to <1 dB @ + 40dB
=============
Wayne: This is a schematic that we printed up and openly distributed to our dealers and service shops. I don't know if this product is still being built. I know I haven't received a penny related to design, the book, or support for over 20 years so I couldn't care less. This doesn't contain any patented technology and AFAIK wasn't copied by anyone. I don't know the details of how or what rights Gold Line acquired from my old partner(s) but they made and sold these for quite a few years.

I can't guarantee how long I will host these up on my site so copy if you'd like. I see little problem with non commercial use. The design is clearly published (by selling units) so design has been public domain for a couple decades.

JR

 

[JohnRoberts]

Stephen.. be my guest..

As Brad's comment suggests some of the parts can be a little tricky so if copying be careful about changing things too much without understanding what's going on.

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

Tempus Fugit.

JR

 

[StephenGiles]

Many thanks

 

[NewYorkDave]

The TS-1 was a great li'l box. At the first MI repair shop that employed me, we had 'em on every bench