Regulator as a compressor?

[preciousmolina666]

does this work?

[vigilante397]

Does that look like compression? Looks like clipping to me. There is a very big difference between voltage regulation and compression. So a pair of LM317s could be used like this as clippers, but they would kill your signal, as most silicon clipping diodes have a forward voltage of 0.7V, and the datasheet says there must be at least a 3V difference between input and output, so it would be like clipping with a 3V forward voltage diode

[iainpunk]

This would add clipping, which is a form of compression, but it would also introduce crossover distortion to your signal. Crossover distortion occurs because you need a certain voltage before the ic starts to conduct. Crossover distortion sounds really nasty. I wouldn't recommend you use that.

[antonis]

Crossover distortion sounds really nasty.

Many "ancient" push-pull amps just claimed you their official enemy.. 

[Rob Strand]

Many "ancient" push-pull amps just claimed you their official enemy..

I'm a passive-ist I can only wage war on capacitors and resistors.  I have a beef with those inferior little SMD capacitors that don't have the full capacitance at the rated voltage.

[EBK]

Crossover distortion sounds really nasty.

Many "ancient" push-pull amps just claimed you their official enemy.. 

I was thinking about that as well.  I vaguely remember a lab exercise in college where the objective was to use feedback to eliminate the dead spot in a power amp.  Sadly, I cannot recall more details without looking it up, bit I suppose knowing that there is an answer is almost as good as remembering what it is. 

[PRR]

> use feedback to eliminate the dead spot in a power amp.

If the amp's "dead spot" is truly dead, then no NFB will happen in the dead zone, it will still glitch until you get out of the dead zone and have more open-loop gain than the closed-loop gain-set.

NFB works if gain available is "very large" compared to gain needed. When gain goes away, so does NFB.

Yes, as a lab experiment you can "show" that crossover is "reduced" with large nominal NFB, by bringing the shoulders of the notch closer together. The ear is still annoyed.

This also applies to power amps which use bootstrapping in the Vas stage to get "high gain". If the final stage has a dead-spot, the bootstrap goes dead, gain falls way off right where you would like lots of it. (There are complex chips which use bootstrap for larger swing but get gain other ways; these can do OK.)

[Rob Strand]

If the amp's "dead spot" is truly dead, then no NFB will happen in the dead zone, it will still glitch until you get out of the dead zone and have more open-loop gain than the closed-loop gain-set.

The other way to look at it (which is equivalent to what you said) is the dead spot is like decreasing the gain.    If the overall gain is then low the distortion produced by the "dead-spot" is not reduced enough by feedback, so it is visible on the output.  You can see this on those simple class B amplifiers where the bases of the output stage are tied together.

http://1.bp.blogspot.com/-C5QubPhV7aM/VOOhmNS8KCI/AAAAAAAAAng/3mqkOc8MCHQ/s1600/Figure%2B1.JPG

[EBK]

Ok, I had to look it up because you guys were pooping on my memory a bit.   

There should be an asterisk after "no" in that figure, but I'm stubbornly ignoring that for now.   

[Rob Strand]

There should be an asterisk after "no" in that figure, but I'm stubbornly ignoring that for now.

If you look at the waveforms you often see slew limiting in the crossover region.

The precision rectifier has got the same problem.   The output is hardly precise at high frequencies especially for large signal swings.  Sometimes the signal shape is barely recognizable as a rectified sinewave.
http://www.next.gr/uploads/135-10162.png

[PRR]

> Ok, I had to look it up

Yes, just like that. From 20 feet away, you don't see the glitch.

Look inside at the opamp output. It is slewing 1.2V "infinitely fast" to leap-past the dead zone.

But the infinite-fast opamps are never in stock. Even if they were, a big kick still won't poke a fat power device into instant action. You just have smaller glitches, and during them all fine detail is lost as the opamp catches-up the output's gross error.

If I am sensitive, it is because this idea "feedback fixes crossover" was too popular in my youth. I bought into it, but with doubts and growing understanding.

[EBK]

I did actually feel a little bad about finding a picture that said "no crossover distortion!"

I knew that marketing folks couldn't even get away with that claim, but I didn't mean to hit a sore spot.  Sorry about that. 

[anotherjim]

...slew limiting in the crossover region....The precision rectifier has got the same problem.

Not sure if that's the cause. I assume Rob you're thinking of the sudden, "square" change in phase angle (practically vertical) at the zero crossing to the flat line "dead" period.

Precision rectifiers, at high signal frequencies (dont need to be super high, ultrasonic is bad enough), suffer from poor phase alignment resulting in skewed waveform when the 2 half cycles get stitched back together. The phase shift caused by the stray and device capacitance needs to be identical for the positive and negative half cycle rectifiers. Some PFWR circuits are better balanced than others, but I can't think of a perfect one.

[antonis]

dont need to be super high, ultrasonic is bad enough

More close to destructive, for some appilications..

[Rob Strand]

Not sure if that's the cause. I assume Rob you're thinking of the sudden, "square" change in phase angle (practically vertical) at the zero crossing to the flat line "dead" period.

Yes that!

Precision rectifiers, at high signal frequencies (dont need to be super high, ultrasonic is bad enough), suffer from poor phase alignment resulting in skewed waveform when the 2 half cycles get stitched back together. The phase shift caused by the stray and device capacitance needs to be identical for the positive and negative half cycle rectifiers. Some PFWR circuits are better balanced than others, but I can't think of a perfect one.

Yes not much around.   There was a better one in EDN (maybe 1995) where you bias the diodes so the opamp doesn't have to swing so far.  That one is kind of better wheel but is basically the same circuit.  Beyond the straight rectifier you have rectifier+averaging filter and rectifier+peak detector types where you have more options to get around the slowness.
A really clever peak detector was by HP on page 29 of,
http://www.hpl.hp.com/hpjournal/pdfs/IssuePDFs/1980-09.pdf
The idea undoes the effect of the diodes.  It was used in a tweaked form in the HP 5335A counter (see schematic).  They used it in counters and CRO's for the auto-trigger stuff.  IIRC there was a patent.

[Edit:  here, it has more detail
http://www.google.com/patents/US4362996]

[PRR]

> A really clever peak detector was by HP

Clever, fast; not what I'd call precision. At least the simplified form seems to be deaf to low levels. (Since this is a trigger, and you don't trigger on zero, this is OK for that application.) And it is half-wave; again, for trigger. (There is surely a +/- inverter ahead so you can select negative trigger levels.)

Yes, "precision" is a loose concept, depending what you trying to sell. 20dB range may serve many load indicators (if loafing below 10%, we don't care). I find 50+dB useful for some audio. 96dB would be logical in 16-bit digital files (though taking the bits direct avoids analog slop). I rarely care over 20KHz but as that HP rig shows, 200Mhz is done in other shops. You can make it as good as your cleverness allows, often good for a product. You can't remotely approach a "perfect" or even "all purpose" rectifier.

[Rob Strand]

Clever, fast; not what I'd call precision.

Technically I wouldn't call it precision either but it does ride the border more than some other designs.
(I nearly edited the post to put in a caveat.)

At least the simplified form seems to be deaf to low levels.
(Since this is a trigger, and you don't trigger on zero, this is OK for that application.) And it is half-wave; again, for trigger.

From what I rememeber if you tune the bleed currents it doesn't do too badly for 10's of  mV.

(There is surely a +/- inverter ahead so you can select negative trigger levels.)

Normally two are used. One positive and one negative.  The trigger is adjustable between the two limits.  There's a couple of common trigger schemes.  If you look at the HP 5335A counter schematic it shows the real circuit which has all the tweaks.

The Tektronix scopes used a similar idea with the old emitter follower peak detector.  Not as fast.

Yes, "precision" is a loose concept, depending what you trying to sell. 20dB range may serve many load indicators (if loafing below 10%, we don't care).

Agreed.

96dB would be logical in 16-bit digital files (though taking the bits direct avoids analog slop). I rarely care over 20KHz but as that HP rig shows, 200Mhz is done in other shops.

The digital domain is a whole new ball game.   There's a lot of patents for "fast" peak detection and triggers in the digital domain.   Then there's the issue of predicting the true peak not the peak of samples.   HP and the like really know their stuff.

[mac]

I experimented a bit with a 317, nothing really special, just for fun,



There must be a post too.

edit:
the ic provides DC, is not constant when you play.
the signal voltage from the ic should not go lower than the 317 limit.
the upper side of the wave is more "limited" or whatever 

mac