First of all, I appreciated your original article, and have in fact used a variation of it. But I think your criticism here is unfounded. OK, I would personally agree the light bulb might be overkill. But I have found the use of a LED based LDR to control the gain of an OP amp stage (as a simple compressor / limiter) to be an extremely simple and effective solution, especially as part of a larger OP amp based circuit where there are likely to be a few spares. As a matter of fact, at least for my particular application, the behavior could not be more perfect!
Consistency? Yes... and like you said, "shopping" is a hazard. But the truth is, you can find bags of 50 LDRs for $10 when a sale comes out at one of those "goldmine" kind of outlets, and yes... you might have to screen them some (at least to find the ones with very close forward voltage on the LEDs). And, you likely will need at least one judiciously placed trim-pot in the circuit to maintain the behavioral consistency from circuit to circuit. But for your trouble there are some huge benefits:
- As you yourself mentioned, the inherent characteristic of LDRs is often perfect for several music applications, and if it does suit yours, you would have to add RC circuits with trim pots (or other controls) to make the J-FET method behave the same way.
- The LDR is linear. so you can pretty much work with any signal level. You don't need to drop a signal down to the JFET's "near" linear region followed by a ton of extra gain to recover the original signal level. And lets not forget about all the extra consideration and parts needed to assure good stability and low noise through those stages of signal transition.
- Finally, lets remember that when it comes to musical applications, one mans junk is another man's treasure. Of course all methods (LDRs, discrete solutions with a J-FET, or an integrated trans-conductance amplifier IC) are good to consider and understand. But even a hint of sarcasm because someone's chosen method is ancient history is simply not warranted. There is a reason why tubes are still popular, and while some of it is obviously hog-wash, a great deal is not!
re:Ancient technology
Heed oh Brethren of little faith, there be thousands of Fairchild linear MOS couplers in stock at Mouser.
They are pricey though at $3.00 in ones.
I still think the thermal bulb is difficult to beat. Shown here some modern studies of the light bulb compressor-limiter.
I only use discrete generic parts now. Have been burned so many times with 'final buy' and allication notices etc.
Heed oh Brethren of little faith, there be thousands of Fairchild linear MOS couplers in stock at Mouser.
They are pricey though at $3.00 in ones.
I still think the thermal bulb is difficult to beat. Shown here some modern studies of the light bulb compressor-limiter.
I only use discrete generic parts now. Have been burned so many times with 'final buy' and allication notices etc.
Keep your eye on this. They are back to their usual $1 each now, and you have to buy 50. But once in a while, like when they do their black Friday deals, they do a LOT better... less then 1/2 this price. Once in a while there is a dud, but for the most part they do work well. Ande note that they are small.. a nice feature if your board is tight.
http://www.goldmine-elec-products.com/prodinfo.asp?number=G15396B
Hey I'm defending you on the premise of using esoteric parts. I do have an incandescent based LDR here. For my app, these cheaper LED based ones work fine, and like you said... price! And for me, ensuring i can make a reasonable duplicate run is important too.
150k~400R is a nice range.
You won't have to set it off early for dull dynamics.
That one can do the opposite--a very lively natural sound compressor/limiter (which is more pleasant than clipping). The wide range lets us set the voltage switch-on threshold at nearly the same as the amplifier's clipping threshold, making it impossible to harm valid signal.
However, all optocouplers are different, and those broad range models are rarest. They also really confuse vendors, because the effectiveness differs without any significant difference to the printing on the casing.
Question: Did you try the dual-asymmetric ClipNipper DA for double quality at the cost of double the circuit? No bridge rectifiers added for that version. There's just two asymmetric circuits, one of which is polarity flipped. A sine wave test probably won't show what this is for. That's because music is an asymmetric signal.
You won't have to set it off early for dull dynamics.
That one can do the opposite--a very lively natural sound compressor/limiter (which is more pleasant than clipping). The wide range lets us set the voltage switch-on threshold at nearly the same as the amplifier's clipping threshold, making it impossible to harm valid signal.
However, all optocouplers are different, and those broad range models are rarest. They also really confuse vendors, because the effectiveness differs without any significant difference to the printing on the casing.
Question: Did you try the dual-asymmetric ClipNipper DA for double quality at the cost of double the circuit? No bridge rectifiers added for that version. There's just two asymmetric circuits, one of which is polarity flipped. A sine wave test probably won't show what this is for. That's because music is an asymmetric signal.
No... just single polarity. (see partial schematic below). Dual would have been better, but this seemed to work well. Also in the end, I didn't power the LDR directly from the power output, but as the last stage of my pre-amp going into the power amp. I used 2 spare OP-AMPS. Obviously after much experimentation, the OP-AMP passing the signal (Top) is an inverting amp with 100K for R-IN, and 220K R-F, with the LDR resistor across the 220K. The output, aside from feeding the power amp also is passed to the second (lower) OP AMP, this one configured as a non inverting stage. The feedback network consisted of 2.2K in series with 1uF to ground for RIN, and a 500K trimpot (parallel with 47pF for stability) for R-F. the output of that OP-AMP drives the LDR's LED through a pair of diodes (later I bypassed one), and a 10K resistor. Despite what the specs said on that LDR, it doesn't take much current to get it down into the 100s of kill-ohm range. Later I added the jumper to that OP AMPs output, breaking the connection to the 10K resistor. This allowed me a convenient point to tie the 10K to the wiper of a 100K trim pot, powered from a 3.3V LDR. So the 500K trim pot allows me to carefully adjust and select the Limit/Compress" point, so that the power amp virtually never clipped. Of course, since the first OP amp offered some gain, it created a compression effect too, so in the end the guitar signal sounded louder even though I was never clipping. Were I to generalize the circuit more, I'd likely make the gain going into the circuit adjustable too, but for this application the approx 2:1 gain was plenty. In the mean time, the jumper allows me to connect a switch across the jump point to turn off the limiter. With the limiter off, the 100K POT is needed to carfully bias the LDR to a safe output limit that also would seldom if ever clip, at the expense of a lower "apparent" volume.
There's no question a full wave rectifier might have done better, which is why my original circuit was tested with 2 diodes. When I found that at my available supply voltage I got ideal operation with one diode, I let the idea of a bridge go. Maybe I can find a shottky diode bridge for a version 2.
Last edited:
Another copy of the 10k, D, D, Opto string with the actives reverse polarity, could make it dual-asymmetric. Double the resolution at the cost of twice the number of optos. Good thing you got 50 optos.
Adding capacitance across the opto's inbuilt LED could make for a compressor; however, when the cap is missing, we'd have to call the circuit a soft-clipper. Light bulb based units don't need the delay cap, if the filament doesn't change brightness too quickly.
Also, bass guitar probably won't need the delay cap--a delay is for helping the treble quality.
If you wanted D1 and D2 sum 1v, some smaller size FR diodes might get close.
Adding capacitance across the opto's inbuilt LED could make for a compressor; however, when the cap is missing, we'd have to call the circuit a soft-clipper. Light bulb based units don't need the delay cap, if the filament doesn't change brightness too quickly.
Also, bass guitar probably won't need the delay cap--a delay is for helping the treble quality.
If you wanted D1 and D2 sum 1v, some smaller size FR diodes might get close.
Well the ground reference in all cases here is a true 1/2V point (made by another OP-AMP in a kind of "virtual ground" config, not shown). So I could use a diode bridge easily here to make it respond to both halves of the waveform. I think that would be a better solution as I'm out of "spare op amps" on this board ;-) It also eliminates the near impossible task of finding two of those OPTOs whose response curves match identically. But that means 2 diode drops, and having decided I liked the effect better with 1, I didn't go that route. But if low combined Diode drops are better, Schottky might be the path to improvement. Especially now seeing I could get them is a bridge part....
CDBHD140L-G Comchip Technology | Discrete Semiconductor Products | DigiKey
I guess this depends on what we mean by a compressor, which I'm sure you'd admit is pretty subjective. To me, adding gain at the INPUT makes it feel more like a compressor, but I guess my mental model of a compressor is simply a limiter that starts limiting at a lower signal level, but with the lost gain recovered. It seems to me that such a capacitor would only serve to slow the "ATTACK" time of the circuit, which makes it less useful as a limiter, but not necessarily a better compressor. It might also make any inevitable "pumping" effect worse.
Speaking of that, one improvement thats seems appropriate is to let some of the higher frequencies bypass the circuit entirely. That goes a long way to masking any pumping I think. Its easy to do, but again would require at least another OP-AMP as a 2 input summing amp, with one input being the compressor/limiter output, the other being the original input signal through a simple 1st order HP filter. But I can't think of any way to do that without an additional OP-AMP, can you?
I wouldn't mind the matching task when the reward is that good; and still fine even if you didn't match quite on (music is an asymmetric signal).
Oh right, if you're using it for soft-clipper/limiter function (delay cap missing), then a shottky bridge is fine. Try 4 of bat86 discrete bridge. That's a Signal schottky, so it holds steady at 0.3v. A Power schottky wanders between 0.12v~0.28v in this case. I actually like more flexible/current layover, for my own use. So, come to think of it, I don't know which part you'd want. Perhaps the helpful thing that I did was mention the difference.
Hmmm... a "signal" diode. Yes I do see the lower VF on the specs. Thanks. I'll have to see if I can find something like that in a bridge version, and also as an SMD part. As to the wandering of the power version, it may not be a problem. Note how little actual current should be expected to drive the LDR's led, through that 10K resistor. I haven't worked with schottky diodes a lot (hell I just learned to spell it right this past week!). But in my limited experience, their VF is pretty low when the current draw is minimal. And if you think about it, a little upward wandering of VF as the current rises will off er a softer knee on the limit/compression curve.
As to the other circuit you describe... duplicating the drive circuit for the two halves of the music waveform, I guess I don't see the advantage. Even with a delay cap on each LDR, how is it accomplishing anything that rewiring to use a bridge diode wouldn't solve? You'd probably have to recreate the WHOLE circuit (both op amps and LDRs), each with opposite facing diodes and LDR polarity, and then mix the two outputs in a summing amp. Then you would truly be compressing/limiting each waveform polarity separately. That is, if that is the what you're after.
Yes... I'm totally aware of how asymentrical music is (did you notice my waveform description of two notes an octave apart when combined?). But personally, it seems to me that applying completely independent compression to the (+) and (-) parts of the waveform and then recombining is a form of distortion. I don't know what you'd call this kind of distortion, but I can't imagine any advantage. Better IMHO to re-wire the original circuit to use the bridge diode, and then just use capacitive coupling to ensure the overall signal finds balance, when the the waveform is asymmetrical.
Although they're both limiters, the compressor and soft-clipper aren't really interchangable. Mainly, the soft clipper is MUCH prettier for bass (albeit seems less powerful due to suppressing initial impact too quickly) and the compressor is MUCH prettier for treble (albeit the treble is less loud due to smoothed/reduced noise/switching content).
Apparently, our circuits are related by parts, but not by function. Since mine doesn't have a soft-clipper function inherently, I added the LTP Soft Clip accessory circuit to the amplifier (either that or baker clamps will do it).
My suggestion is even less convenient.
I never get very far in such considerations before that ideal of Bi-Amp pops up.
The only problem with bi-amp is the years of work I put into making crossovers, which would have to be modified.
Each mono track is like two different tracks--the asymmetrical upswing and the asymmetrical downswing. Now let's say that only the upswing exceeded, resulting in the need of a modification. In that case, it would be wrong to also modify the downswing portion of the signal.
This is of some concern with the compressor (cap delayed led, or just a bulb); However, I think it is not an important consideration for the much faster soft-clipper.
Edit:
If the point was for louder, forced symmetry will also maximize both of the amplifier output devices, instead of only one of them. So, that really is more power. That's not why I did that. Actually, the reason I used a forced symmetry scheme in ClipNipper-DA is for lower distortion (in not modifying a part of the signal needlessly).
I honestly don't know if those concerns apply to a guitar amp. I've been using mine for music replay.
Last edited:
It's one of those thing where the bottom line would have to be what it actually sounds like, which definitely is a subjective "ear of the beholder" thing. You're explanation of not affecting the part of the waveform may make logical sense, but I think you're missing an important point.
I only mentioned the guitar because its much easier to visualize individual tones combining when there is a single instrument involved. But when you allude that listening to music is different from listening a guitar, you have to admit something is missing in that argument.
Just consider and envision that two musical notes an octave apart are going to approximate a single frequency waveform with a second harmonic added. That waveform will look like a sine wave with one half segment diminished by the second (X2) waveform, and and the other half segment augmented by that 2nd waveform. That's because the X2 wave will be out of phase for 1/2 the X1 cycl3e, and in phase for the other.
So my argument is simply that whatever that LOOKS like on a scope is also representative of what it is SUPPOSED to sound like. And if you just compress the half that is in phase, you must be changing the natural harmonic content.
But again, one man's distortion is another's nice sound.
I've just been accused of being too logical, for the first time ever. I sincerely do appreciate this and am most grateful.
It is important to compare varieties so that we can say we've done our homework. Comparison is actually what makes this easier. There's a vast array of sophisticated measuring equipment that is incomparable. However, comparison by a vast array of means and is valid in most of them. All you've got to do is try your available options. That part is inconveniently necessary.
As for my part, and historically speaking, the Volumax 1.0, which got a lot of DJ's fired because of asymmetric voice, was supplanted by Volumax II which had twice the parts so as to force symmetry ONLY (!!!) during excesses, and resulted in the re-hire of some of the previously fired DJ's. My synopsis may have a shortfall. Look it up for yourself if voiceband audio during hard use or maximized conditions was of some relevance to your application.
More importantly, avoid fell swoops if you want hi-fi. I didn't design a very effective compressor, mainly because minimum change did a lot of good without a lot of damage. That is entirely linear. When more needs to be done, stacking is in order. I've only ever proposed the version 1 and 2 compressors, assisted by a soft clipper when needed.
Although there is a lot more that could be done beyond that point, it tends to sound like the digital solutions utilized by One Direction (when overdone). An out, seems to be to bi-amp, since suitability varies by frequency. That applies to either small or large signal.
So, your question about parts count just got answered. Buy more op-amps.
Just included that part, but I did read your entire.
Listen, I wasn't trying to downplay your logic or question your experience here. I'm only raising one small point, in a friendly way, and that is the merits of altering top or bottom portion of a musical waveform separately. I do grant you that the argument for or against likely does depend on the source material. You are dealing more often with full range musical source, and I am more often dealing with the processing of a single instrument, the guitar. My visual description of the resulting waveform of two guitar notes an octave apart was just to illustrate a point. It is similar to what you'd see on a scope, even with just a single note, being rich in harmonic content. It is generally a an often complex, but periodic and repeating waveform, which is a different case. And in THAT case, I can see a strong possibility that applying compression just to one half of the waveform or the other, like you say on an 'as needed" basis, might have a less then desirable effect.
I'm not going to defend my history as a designer or my choice of what pieces of various measurement equipment are needed for verification of correct operation. Any designer knows that apart from all measurements, many is the time a calculated and logical approach to audio processing does not yield the expected result when you listen. The converse is also occasionally true, though Murphy's law often favors the first case.
I was focusing heavily on the various applications of LDRs in limiter and compressor circuits when I found this thread, not realizing your examples were representative of specific products I assume you have designed. Having proven the approach you're advocating under real use cases makes it an approach I'd definitely want to try at some point, to do my own A/B comparison.
And by the way, I've seen some VERY impressive digital approaches, sometimes offered for free. You've no doubt seen "Stereo Tool" plugin for Winamp? The below is a very old version, but I've used it live to control the relative levels of my backing tracks at shows, and it is very impressive. I wonder if the author's latest versions use the symmetry approach you speak of. It might eeven be built into the old version, I'll have to look more closely.
Stereo Tool 2.10 - free Winamp plugin
In any case , for my particular application though, I think just moving to the schottky bridge will be the last needed improvement.
One thing I really will have to do (which you'll find amusing)... I need to build an LDR test circuit, to sort the huge variation in these "el-cheapo" LDRs, and categorize them. They are all usable, but applying approx 8 microamps to the LED side (5V though 100K actually), output resistance on the Cadmium side ranged from 1/2 megohm in some units, down to 2K in others. Fortunately I'm only finding abut 3 categories here, and they are worth the effort for the price (especially when on sale). But were I to expect huge sales, at some point a circuit with enough trim pots to work in ALL cases would get old fast. At some point I'd have to move to the more expensive (but much more consistent) "Vactrol" equivalent. Either that or start making my own with known LEds and cadmium cells.
Last edited:
This is not na easy tasc!!
I´ve been spending time testing some limiting circuits and, in general, the result is poor .
The good ones are expensives, using nsl32 or vactrol similar.
At the beginning, i was thinking about using fet, but all disappointed me, too much distortion.
Now i´ll try ne13600.
I´ve been spending time testing some limiting circuits and, in general, the result is poor .
The good ones are expensives, using nsl32 or vactrol similar.
At the beginning, i was thinking about using fet, but all disappointed me, too much distortion.
Now i´ll try ne13600.
The task isn't difficult; but, if the need is great, then the task might be multi-stage.
The task isn't expensive; so, if you spend too much, then inferior performance is what you will get.
Please post what has gone wrong, including the circuits, the schematics, and far more detail.
Rather beautiful compressors do exist and although I have made two of them, that fact does not make me an expert. Even so, more details on the subject are welcomed and requested.
I'll agree that any solution using OPTOs is likly to have less distortion over a wider dynamic range then a FET solution. I couldn't find anything about the NE13600 you mentioned, but those nsl32 optos look very much like the ones I described from the surplus places. I have found them to be a good and inexpensive building block, BUT... I have found them to have VERY widely varying characteristics. Since I needed repeatability, I found it important to do some pre-testing and separate them into categories. Originally I was just looking at small differences in the LED diode forward voltage drop, but later found a much better method. Use a regulated 5V supply to drive the LED side through a 100K resistor, (about 32uA) and measure the resistance on the C-cell side. In doing so, I grouped mine into devices showing 2K-5K, 6K-12K, 30K-50K, and another group up over 250K, all with that same input! Anyway, sticking with one group (I stayed with the 2K-5K group) made it possible to get very consistent results from each circuit instance. My trim pot was still necessary for sure, but at least I was able to reliably duplicate the desired behavior.
OH OK... a transconductance amp! I messed with some TC based compressor circuits many many years ago. It may be that I was choosing bad parts or maybe just designing poorly, but I found them to be unacceptably noisy. I guess LDRs could be also depending on how they are used. More experimenting needed there for me.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.