Op Amp Questions, diode rectification and Bufferring...

[JohnRoberts]

e.g. ---> Electro-Mechanical Compression 

Basically ---> Imagine a typical VU meter, but rather than a needle, it has metal bar.  As the control voltage drives the bar up, gravity and general physics delays and restricts the way the bar moves... this restricted movement gets fed back to an amp as another control voltage.  So the "compression" occurs as the result of the mechanical process...  I have a feeling that it'd be a really nice compression curve......

Like I said - stupid ideas... but fun

Not so stupid. PRR suggested motorized faders for limiting.
There is a fundamental notion there, it's solid friction. The absence of solid friction in limiters results in LF distortion, which has been the subject of numerous attempts to get rid of. Hysteresis and dynamic time-constants have been used, but don't solve the problem entirely.
Electromechanical gain control solves the problem beautifully, but cannot achieve the promptness of reaction needed for true brickwall limiting.
Only recently digital processing of the side-chain has solved the issue, but once digital processing is involved, there is a small gap to go fully digital.

I am pretty sure the "slow" mechanical limiter is analogous to an electronic limiter with slow time constants. In electronic limiters we can be even more flexible.

I have wasted a lot of time on making transparent dynamics (mostly for use in Tape NR), and some for oscillator gain control loops (a similar source of distortion in oscillators).

I toyed around a little with the old trig identity  sin^2 u+ cos^2 u = 1...  We've discussed this here before.

In my judgement for analog control loops I favor a fast and slow time constant that is level dependent (large changes are fast, small changes are slow).  Human perception will mask a lot of the fast changes and steady sounds will be relatively clean. These days, digital side chain processing, or even better yet non-real time digital processing could be arbitrarily transparent.

JR

 

[jBam]

Hi all...

Thanks for the ongoing input on this thread - busy week at work already, but I've been spending my lunch times drafting up flow charts and descriptions.

I've hit another "my head hurts" moment trying to understand control voltages for constant ratio compression - hoping you guys can help!

(VCA Design)

Would a constant voltage offset be considered a constant ratio?  I've been angling for using the extracted peak info discussed above as my control voltage...  I don't think that's really even a problem, but I've been thinking that it would present a constant ratio / hard knee effect, but now I'm thinking it's not - it's a constantly varying ratio… But I've been working hard on "day job" stuff, and not sure if I'm just confusing myself...

Please refer to the attachment for a sketch discussing this question...

If my approach ends up being "constantly varying", then that's fine - "nice and smooth" ... butI'll hit the drawing board again to get a hard knee (steady cv generated once threshold is met) in there somehow (already have some ideas)...

I've been really confusing myself today haha…

[EDIT (after driving home thinking about this stuff)…  But wait - it HAS to be reactive to be a true ratio… so it can't just be a constant voltage once triggered…  otherwise there is no definition of "how loud" the signal is over the threshold, and therefore no way to apply a true ratio - e.g. a constant voltage would be a constant  (e.g.) -6dB constantly, irrelevant of dynamics above threshold…  therefore not a true ratio…  I think (hopefully) my intended solution with varying voltage control is the way to go! …  But please - assist and comment away!!…]

 

[abbey road d enfer]

Please refer to the attachment for a sketch discussing this question...

I see your main embarassment comes from the fact that you don't fully understand how the side-chain reacts to signal variations and above all, how the gain cell reacts to the control voltage.
Old style gain cells (transconductance, FET, optocoupler, diode bridge) they all have a somewhat hyperbolic response, i.e. gain varies inversely with CV, so above threshold, the gain equation is roughly G=k/(Vout-Vth), so when Vout>>Vth, the gain expression is purely hyperbolic, yielding constant output since gain varies inversely with signal. When Vout is close to Vth, the gain is unity (I'm deliberately isolating the static gain (make-up gain) for clarity. So the compression ratio varies from 1:1 to infinity:1.
Ther are other factors that reduce the actual upper value, but it's basically what happens.
VCA compressors have a different behaviour; the gain control law is in decibel-per-volt; that means that a variation of let's say one volt of the input signal will have different action whether it's from 1V to 2V (6dB) or 10V to 11V (only ca. 1dB).
Then you have to consider the topology, feedback or feed-forward.
Feedback reduces the slope of the compression curve since any compression results in reducing the control voltage.
Understanding compressors is a vast subject; I'm sorry I can't resume it in a single post.

 

[jBam]

Understanding compressors is a vast subject; I'm sorry I can't resume it in a single post.

Yes absolutely sure it is …  Thanks for the detailed reply.  Just to clarify - I'm not really trying to understand a particular type of compressor… mores hypothesizing about ways to affect and control dynamics.

i.e. The idea I have is not based on anything other than a possibly flawed interpretation of possibilities within electronics, combined with a reasonably good understanding of "what SHOULD be possible"!

Hopefully this thread is fun for others too!  I'm bound to be talking about a few non-conventional things, and like I've maintained in previous comments - it's quite possibly just a silly idea 

You certainly seem to have a pretty solid understanding of these types of things - to assist / clarify, and hopefully squeeze a little more good advice out of you… I'm currently working toward a feed forward style arrangement, and it'll be entirely solid state.  I think the main part that's confusing me is the logarithmic dB world (which is strange - because I'm pretty switched on in understanding things like this… it's just when I add in CV control to to it all I'm spinning around in circles!!)…

So, with what you suggest, CV control within a solid state VCA arrangement is in dB / V?  So a constant dB drop would represent a ratio drop in level…?  Hmm… that's interesting, but slightly annoying haha…  i.e. - and I know it's a flawed crappy example - the "1" option in the attachment is actually presenting more of a "ratio" control (e.g. -3dB).  The "2" option, would have a widely varying response, greatly increasing compression the more the signal goes over the threshold?

I can work with that, but - back to sketching up crap and scratching my head re: getting a constant hard knee ratio control in there!

Any other comments are welcomed!



[OR is there some kind of antilog VCA that I can use my extracted peak for a more consistent control?… might search around]

 

[CJ]

have you read up on Blackmer and his stuff?

http://www.google.com/patents/US3681618

http://www.thatcorp.com/History_of_VCAs.shtml

looks like Jakob has some stuff

www.gyraf.dk/schematics/VCAs_Ben_Duncan.pdf

Patent Links

http://www.google.com/patents/US4225794

 

[jBam]

have you read up on Blackmer and his stuff?

http://www.google.com/patents/US3681618

http://www.thatcorp.com/History_of_VCAs.shtml

looks like Jakob has some stuff

www.gyraf.dk/schematics/VCAs_Ben_Duncan.pdf

Patent Links

http://www.google.com/patents/US4225794

Cheers CJ… I'll get reading… also cheers for the patent links…

Side topic (or not really - still related to dynamic control options)… does anyone have any good links to RMS levels for dynamic audio… I'm getting sick of searching and just finding the old "straight line at the average through a sine wave" picture (literally high school stuff)…  I'd love to see an RMS reading across a broad dynamic piece of music, and even just things like a Kick Drum…

Cheers all!

 

[abbey road d enfer]

[OR is there some kind of antilog VCA that I can use my extracted peak for a more consistent control?… might search around]

Sure is, just check THAT RMS detector, and of course read all CJ suggests.

 

[jBam]

Yes - been reading through most of the THAT page over the past few days actually!… I think I need to reread though - I think I'm expected RMS of dynamic audio to be a little more "responsive" than what I seem to be realizing (particularly after reading through concepts like how it relates to heat dissipation through a load).

I've always expected RMS to be a very forgiving signal for compression, and pictured it as a self correcting arrangement in feedback form in my device…  But still sort of figured is would loosely follow e.g. the wave of a kick drum like a really lagged / delayed wave follower.  The "wave follower" curve in my attachment isn't intended to be RMS though of course  ;D…

My little gadget is currently (supposed to be) focussing on peak control.  The overall unit is intended to be an experimental multifunction dynamics modification box…  A Levelling Amp and multi-level-band (multi threshold) dynamics processor (both expansion or compression through separate threshold bands); and any other thing that I feel like throwing in there…  (Including an "oops" button - which I'll talk about another time … )

I'm still convinced it will work nicely IF I can build it (which will require help here I suspect!) - in fact, maybe I just need to start building a basic version to hear and experiment before I decide what "work" really means haha…  Will hopefully get a flow chart together this week and post it up.

btw… those patents are SERIOUSLY good reading CJ thanks!! (or, at least, the theory is… not the "said input into said whatever" wording …)

 

[jBam]

Hi all - hope the week has been well… Crazy busy at work for me, but still had enough time to pop together 99% of my flow chart - forgot to print to pdf though, so only have a hard copy which is kind of useless here haha…

Outside of work hours ---> I've been getting into modeling stuff in SPICE, and I'm learning a million miles an hour.  Spent most nights just messing around with basic op amp circuits, going "hmm.. ahh.. ooh ok…" adding thing here and there, and modifying parameters.  Then last night I stayed up a little too late, and plugged together around 30% of my design…

And it bloody well works!… or at least in part.  I've been able to "very" accurately extract the peak out of an audio signal… Tonight --> work through the rest of the extraction design I have in mind, which is a little more complicated, and then I'll focus on the front end with a wave follower…  Hmm this still probably makes little sense without the flow diagram!  But the point is - this is cool stuff, and it seems that I'm on track.

I say "very" above, because I'd kind of like it to be even MORE precision.  I used the following as a starting point for rectifier:

http://sound.westhost.com/appnotes/an001-f4.gif

and there's low level "wobbles" in the rectification, and a little spill over into -V.  This is expected I guess from what you all have been suggesting, and from what PRR has said regarding signal level (very low is affected; high is not --> which is reflected in my analysis output).  But the "low level" signal is still important in my design (to a degree), and I'd like to / prefer to absolutely minimize "wobble" (by wobble, I mean it doesn't perfectly trace the original to zero, and then perfectly reflect back up +… I'm aware perfection in reality is a myth, but the deviation is noticeable, although small).  My current thinking is that:

(a) well it probably doesn't matter… but…
(b) this is the first rectifier I've even looked into, and the first time I've used a SPICE simulator AND I never even knew what SPICE was until this week… SO…
(c) it's most likely the diodes, rather than the op amps I'm modeling… but
(d) well.. I still don't really know what I'm doing… I'm just trying to approach this with logic

What I'm getting from the model is:

1kHz Sine signal with 1V PtP amplitude ---> this is the input signal...

After Full wave rectifying, I get:


- A little peak past 0V of -30mV
- An "erroneous" irregular wave up until around +60mV
- I also notice one peak near 0V looks a little different to the next, but then it repeats (like a cycle of two different near 0V occurrences…  I'm a little stumped on why that is!)

Then the rest is spot on (or enough for my extracted peak to be very nicely extracted… but with little wobbles at the 0V mark [my peak extractor leaves only peaks poking up positive away from 0V]).

Sorry guys… no pictures at the moment.

Because my design has at least 3 rectifying stages, I'm kind of keen on getting a super precise outcome here...

--------> 

So my question is: is a little low level wobble likely to be unavoidable?

Is the wobble definitely diodes (e.g. voltage drop), or could it be a whole range of things including OpAmp selection (sorry - I know that's a broad question :-/)

If it is likely diodes, is it like the friction analogy I used earlier - like it takes a little energy to push through, but then it flows with ease?  If so I have a few crazy idea up my sleeve hehe…  maybe push a constant voltage through, and then ditch it the other side?? (probably more stupid ideas from me  ;D)

Referring to older posts:

This is what I meant by precision rectifier.

JR

John - you seem to have some experience here --> I've read through a few of your older posts on similar topics from several years ago.

Or anyone else really? CJ?…

"anyone got a super duper precision rectifier for me"?  ??? :

------> In the meantime, I'll keep cracking on with the current design ---> be well!

 

[jBam]

Ok ---> All issues solved.  A big combination of generic mistakes

Most particularly grounding issues and a poor OpAmp Choice...

Output of my current "peak extractor" model / Schematic looks great!

Big Thanks to PRR for the suggestion of boosting signal through rectifier and then attenuating... dramatically reduces the rectifying inconsistencies at low voltages.

Refer to attached for a SPICE model output (this output is not attenuated after rectification, but it helps emphasize the results ....)

----------------

- Original Sine at bottom...

- Pink is Upper Peak from a given threshold;

- Green is a peak from a lower threshold (Green minus Pink gives us a "middle"part of the signal - Light Blue)

- Original rectified signal (basically the Red wave) minus Green gives us the remaining "low level"portion of the signal - Dark Blue.

- The low portion + Middle portion + peak portion = Original rectified signal (Red)... with very very minor inconsistencies at the "joint"where they are summed.

---------------

So - what the hell does this mean?

It means that, pre summing all portions back together, I now how full control of the amplitude and +/- of a signal's "peak", "middle", and "Low" levels... 

Although this experiment is currently just on a sine wave signal, the intent is to use this technique on a wave following circuit [EDIT: actually - every time I've said "wave follower", I mean something more like an envelope detector] that is tracing dynamic audio as an approximation of its level.  With this I can now extract level portions of the signal and then modify each section and then stitch it back together as a control voltage for use in dynamics control...

Compression / expansion of any section should be possible.  I still need to figure out linear / exponential VCA configurations when using the extracted / modified CV to then control the output level... but that's for a nother time.

At the moment I'm pretty pleased