"Slick Trick" Rectifiers for Level Detection

[Kit]

Yes, its a clever simple circuit thats been around for a long time.

But most people arent aware of its existance.

I used a one transistor rectifier in the sidechain for an optocomp that i buildt.

I think I posted a schem. in the LA3A design thread.

Ah, there it is.

http://www.groupdiy.com/index.php?topic=13566&postdays=0&postorder=asc&highlight=la3a&start=45

Q2 is the rectifier.

 

[CJ]

check out opamp rectifiers also

if you want accuracy, no foward drop, for measurement etc.

 

[JohnRoberts]

Indeed this is an old trick and there are many more when you get into the nitty gritty of discrete design. I would point out that there is no free lunch and this circuit as drawn has a potential flaw or source of error.

The impedance reflected back at the base will be different when transistor is saturated or not, so instead of emitter resistor divided by Hfe it will drop dramatically, looking more like emitter and collector resistors in parallel. This will pump DC current into C1 that varies with signal amplitude.

There is merit perhaps as a peak detector and/or driven from a lower source impedance but these tricks are not really intended for precision applications.

Other stupid transistor tricks include using transistors backwards (swapping emitter with collector). Yes they work (sort of) but do some things better and some worse. One thing better is low on sat voltage. These were often used as cheap signal clamps in low end audio designs, before J-fets became more widely available. I don't recall off the top of my head, but there were even device numbers optimized for this application, like higher reverse b-e zener breakdown, since this now becomes a potential source of distortion in operation.

When I was on the steeper part of my learning curve I devoured old schematics to look for different circuit tricks. While old hand drawn schematics often contained mistakes, some circuits that looked like a mistake were actually clever utilization of some intrinsic device characteristic.

JR

 

[bcarso]

Toshiba has a bipolar part optimized for reverse-connection clamping. It has both high reverse breakdown and also fairly high reverse beta. I forget the number but it's something like 2SC28XX if memory serves. It's not that cheap, but has I believe lower capacitive loading in the off state than the equivalent on-resistance JFET.

I've used 3904's for the purpose but you have to drive them hard---the reverse beta is only 2-3 sometimes! And when off the base must ~float.

The rectifier unfortunately was part of a budget speaker I did for Harman, the Duet (a pretty nice value btw---one case where the industrial design served acoustics well; if only the transducer had a bit better linearity it would turn some heads). We had some versions ported quite low and boosted, and for simple textures it worked and initial demos had folks looking for the nonexistent subwoofer (shades of the latest Bose gobbling about their 400 dollar pair of computer monitors). However, IM just got too obnoxious for a lot of material so wiser counsel prevailed and bass extension was scaled back a bit.

So the circuit is proprietary, though of course not patented, and with some work could be reverse-engineered. Someone besides me could do this, given enough patience, but I can't legitimately provide it based on the covenants I agreed to. In fact I don't think I even have a schematic anymore---I left it behind when I left Harman.

However, the gist is: take the standard two-op-amp precision rectifier. Realize that when you know what the loading will be, the summation of 2:1 ratio inverted half-wave output from the first op amp with the input signal itself, which produces the full wave signal, can be done passively. Then substitute a common-emitter transistor stage for the opamp. Rude and crude and tweaky, but with known parts quite adequate for the job.

The actual speaker circuit provides a negative and positive full-wave-rectified voltage as well, by essentially putting resistors in series with both diodes in the classical half-wave rectifier opamp stage, and taking the two outputs from the diode-resistor junctions.

 

[bcarso]

That's quite different. Maybe that's the one that provoked my partial description of what I did for the Evil Empire (just kidding Sidney...).

 

[JohnRoberts]

Wayne: OK, that second one is a little more conflicted... looks like a sort of dual peak detector, but there will be interaction between the charge on the emitter cap and the integrity of the collector waveform. Oh well like I said, these are not precision, but cheap, and sometimes not too dirty to use.
---------------

Brad: re: the backwards transistor yup, I drove the base with a current source to switch it on hard but let it float when off. It doesn't matter if it zeners while attached to a high Z current source output. I think I've seen some value (cheap) designs where they drove it through a high value resistor from a high voltage, but as you mention the beta is generally not that great in this mode so it needs good drive. I inherited a design that had a 2N part number in it, but it didn't work so hot. I have no idea where my predecessor copied that. I ended up using a part designed for the app, they always seem to work better.

I think I've seen this trick done in some cheap tape decks to switch the head between record/playback.

JR

 

[JohnRoberts]

With a lot of these discrete tricks, you scale resistor values orders of magnitude apart to mitigate against loading effects.

I seem to recall an active pursuit of such tricks to optimize limited part count designs like those old Interdesign mono-chips where you had a finite number of devices to play with. If you run out of Xs and have Ys left over you see if you can do it with a Y.

Most modern design seems to come down to which IC do I use?

JR

 

[bcarso]

[quote author="JohnRoberts"]With a lot of these discrete tricks, you scale resistor values orders of magnitude apart to mitigate against loading effects.
....
Most modern design seems to come down to which IC do I use?
JR[/quote]

It's a lot more work, but taking loading effects into account in detail is something I do a lot, to enhance performance and in some cases reduce cost. It depends on the expected volume of units---if it's 100k, then a few more hours to save a few cents and get better performance can be well worth it. Provided of course the almighty schedule doesn't slip.


As far as design amounting to IC selection, I probably told this story before in here, but one day a guy came in for a job interview and let us know almost immediately that it was a lark---he was happy with what he was doing and didn't like this company that was interviewing much anyway. But he went on to talk about how he'd considered hardware design early on (he was maybe in his early 30's), but decided that it consisted of opening up databooks and looking through them for parts (implying integrated circuits in the context at hand), and then stringing these together. Essentially zero creativity he said---he went to software instead, where he found ostensible fulfillment.

I told him I could understand that point of view, and I took some comfort in knowing that it was fairly prevalent---and that as a result, I would likely always find work.

 

[JohnRoberts]

I believe "value" engineering or engineering for lowest possible cost while delivering good performance is under appreciated. In many ways engineering the old VW bug was a far greater accomplishment IMO than 200 mph ferraris.

These days the value engineering calculus has pushed me over the line to use a dreaded digital chip to replace a bunch of analog circuits with incredibly useful extra flexibility.

In a strange turn of events my hardware design these days is dominated by "which IC do I use" but I can still cobble together the analog glue as needed to do real things in a real world.

JR

PS: I don't find software design any more or less creative than circuit design. In some ways it is very similar. You can be clever or crude in either medium. In software design you don't pay as high of a penalty for crudeness unless you run short of clock ticks, or memory.

 

[Kit]

In many ways engineering the old VW bug was a far greater accomplishment IMO than 200 mph ferraris.

Exactly!!

This is for me, the very essence of electronic circuit design.

Simple clever little circuits that do a lot for not much money.

Probably the main reason I still bother to build stuff.

I cant find the schem. now, but we should check out the Studer 189 limiter sidechain for some (more) clever discrete circuit design..............

 

[bcarso]

I tell my clients that I try to use the appropriate technology for the job, with a judicious mix of analog and digital. I collaborate with a very good firmware guy when there's a lot of coding involved. And such people are indeed (potentially) equally creative as anyone else.

Depending on product volume, a digital-heavy approach can make a lot of sense, particularly if the specification is a moving target.

OTOH, since I left Harman they have more-or-less abandoned analog-domain EQ in one division, simply because there's no one there that knows how to do it quickly and well! They are leaving money on the table like crazy, but since no one is rubbing their nose in it (given no comparison available), and since deadlines are more-or-less met and the Time-To-Market-God is being appeased, who cares? Audio quality is at the mercy of codec quality and in some cases "digital" amplifier quality, but again who's listening? :razz:

I was pleased when a vendor of a switchamp chip used in the JBL On Tour battery-powered portable said that a Far East outfit had attempted to reverse-engineer that last product for which I designed the electronics. They gave up---it was too difficult to figure out. Mind you, this was something with no custom ICs, no scuffed-off part numbers, no epoxied areas, no custom code---although it was a four-layer board, there were no buried vias. It has an abundance of discrete transistors, and only a few ICs, because it was cheaper to do it that way than use expensive audio-quality low voltage opamps, or power-hungry DSP and codecs (although of course great strides are being made in such parts, and the proper design approach might differ if the design were started today, about three years later).

But it did take some serious design work, a lot of simulation on top of knowing what one was doing to begin with, and it was justified by the product volumes >100k. Had the volume been a few thousand pieces a year, things would have evolved differently.

 

[Guest]

One cap and couple of diodes will rectify and double a voltage...

 

[Kit]

Thats also an old slick trick for audio rectification in limiter circuits.

It looks like the Boss crew might have had a sneak peak at the Dynacomp/Ross pedal schematics.

Hey, Wayne maybe this thread should be called "Old slick tricks for level detection." :thumb: :thumb:

 

[Kit]

"Slick tricks."

I like it. Kinda sounds like an 80´s rockband that never made it.

Dyna/Ross-.comp schem:

http://www.geofex.com/PCB_layouts/Layouts/d&rpub.pdf

 

[JohnRoberts]

This seems to have morphed from how few parts to how many can I use to make a rectifier? :wink:

JR

 

[JohnRoberts]

Lets not count how many transistors in a 339/4 :wink: What you have there is a precision diode.

I have accumulated my own collection of favorite precision rectifiers over the years, but these days for me it's about how many line of code does it take.

JR

 

[SSLtech]

Now, that's a "Slick-Trick" indeed! :wink:

I spent an hour or so catching up on lost time with Leif Mases (philosopher, producer, engineer and analog designer, of Abba/Maselec fame) at the New York AES, and he did mention a rather slick "greater of two' trick which he came up with in the 1970's for Abba's Harrison console at Polar Studios:

The bargraphs were VU, but they worked with both analog AND digital devices, and people wanted something more generally usable from the metering. -What was devised was a DUAL detection system; one of which was average/RMS and the other of which was peak detecting. The peak detection output was then dropped by some (presumably empirically-arrived-at) amount and then fed to a "greater-of-two" combiner to the bargraph.

The result was a meter which was able to work happily with either the 32-track 3M digital machine or the analog 24-track (Studer, I seem to recall) and ALWAYS gave a usable indication of level... Apparently the users ended up trusting the console meters over and above the machine meters, since they just felt "right" in the end, with no overly complicated vu-plus-peak/bar-plus-dot display... just a bar which you tried to keep out of the red...

Keith

 

[JohnRoberts]

[quote author="SSLtech"]Now, that's a "Slick-Trick" indeed! :wink:

I spent an hour or so catching up on lost time with Leif Mases (philosopher, producer, engineer and analog designer, of Abba/Maselec fame) at the New York AES, and he did mention a rather slick "greater of two' trick which he came up with in the 1970's for Abba's Harrison console at Polar Studios:

The bargraphs were VU, but they worked with both analog AND digital devices, and people wanted something more generally usable from the metering. -What was devised was a DUAL detection system; one of which was average/RMS and the other of which was peak detecting. The peak detection output was then dropped by some (presumably empirically-arrived-at) amount and then fed to a "greater-of-two" combiner to the bargraph.

The result was a meter which was able to work happily with either the 32-track 3M digital machine or the analog 24-track (Studer, I seem to recall) and ALWAYS gave a usable indication of level... Apparently the users ended up trusting the console meters over and above the machine meters, since they just felt "right" in the end, with no overly complicated vu-plus-peak/bar-plus-dot display... just a bar which you tried to keep out of the red...

Keith[/quote]

Yup, digital was a problem for many in the early days.

Peak will always be higher than average (except for when it isn't) so this is basically ignoring peak until it becomes some arbitrary X dB above the slower average. This will allow you to compress the scale to better accommodate the extra dB needed to indicate digital FS. In my meters I made the peak and ave release time the same so crest factor could be seen better, 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. If all the meters were used for was keeping out of the red you could rescale peak meters and be done.

I appreciate there is utility in KISS and making smaller evolutionary steps are better received than the radical jump to peak only meters required for digital media. I will resist a knee jerk defense of my two patented simultaneous pk/vu meters, one of which was one of those dreaded dot and bar designs. My criticisms aren't important wrt a 35 YO design so I'll keep them to myself.

JR

 

[SSLtech]

Oh no 'criticism' was intended John, -But Leif wanted the result to be 'invisible' and not have the user even think about the difference. i doubt that he used an overly-molasses-dipped decay though... I do take your point, and it's an excellent explanation that you give.

Actually, I remember the first time I saw a bar-plus-dot RMS/Pk meter... I called my 'Yoda' Peter over, and said:

Now...THAT's a cool meter!!!

We both advocated peak metering -even for analog machines- and one of our SSLs had black sifam peak meters specified when ordered new. -For lots of applications, you DON'T want tape compression, and the PPMs meant that you got back what you put to analog tape, MUCH more than VUs did. (although you got more hiss to go with it! :wink

Keith

 

[JohnRoberts]

No worries. I've actually heard that criticism before and product design must look at things from the POV of the customer. When I first came out with my meter, digital was a fairly rare curiosity so it was arguably TMI for the typical customer who preferred mechanical VUs. You didn't have to look at them unless they were clicking against their stops.

Since them I see two trends. Digital has become pervasive making peak information more important, and we have culturally learned to adapt to more clutter.

Think back to what your computer monitor looked like back in the '70s compared to today. Watch a few minutes of business channel TV and there's two or more tickers rolling under the picture, maybe a news crawl, and bugs showing the current DJIA/NASDAQ levels. Even entertainment programing has teases for other shows and all kinds of spare moving parts going on along the bottom of the picture.

So these days we are better equipped to manage complexity, and more in need of it, and my patent (which I didn't even own) has expired of old age, so enjoy.

JR

PS: An early prototype I made using vacuum fluorescent had 100 segments, 1 dB per segment, and if anything it was less distracting than lower resolution displays.