Slew Rate Limiter

[barclaycon]

Another audible effect of slew limiting - ascertained from actually using the equipment in a recording situation, was in the original Focusrite EQ.
Rupert Neve slugged down all the 5534 op-amps with 47p caps between 5 & 8 instead of the normal 22p (and more usually 15p). He said he did this to match the slew of early Neve modules.
This resulted in a very 'dry' kind of sound - not harsh.
I read an article by Alan Sides who purchased a Focusrite desk for Ocean Way and he said exactly the same thing.
Oh and for those who want to try slew limiting, there's a circuit here:
http://www.all-electric.com/schematic/slewlim.htm
With some interesting comments.

 

[JohnRoberts]

The 4136 was not exactly a slow part... 1.7 v/uSec will not slew limit that easily... And the sonic consequence of slew limiting if it does happen is probably IM hash on loud cymbals and such, not a 'softer" vocal sound.

That's your 'sonic' observation from recording and mixing music is it John ?

No from my experience designing electronics...  The 4136 was a step up from slower, earlier opamps, and was popular briefly before being eclipsed by even faster 553x and TL07x families.  The speed of those faster parts was clearly faster than audible signals, but faster was better than marginal since marginal parts can have sonic consequences below hard slew limiting. You may want to check my math since I'm still working on my first cup of coffee, but 1.7V/uSec translates to a 30Vp-p sinewave at 18 kHz. If we target a factor of 2x for headroom we will still be hard pressed to test this edge rate with vocals. OTOH a close mic'd cymbal track is another story.

The HF content in vocals would mostly be sibilants (ss sound). While I wouldn't expect them to be loud enough to slew limit, heavy compression can raise low level sibilants to unnatural levels (thus the need for de-essers).  While I won't argue about what other people hear, in my experience, circuitry that is too slow for the program does not deliver a euphonious result.

JR

PS: While I have always had a speaker system on my design bench, and have designed studio products and recording consoles, I do not design by ear, with the possible exception of an early studio delay line flanger, and when working with dynamic processor time constants using non-linear side chain tricks. It was while working on an effect, in a studio, that I learned how unreliable my own hearing was, when my perception shifted dramatically before and after taking a short break. Maybe I have shit for ears but I think not... I do trust some tools far more than others. Ears are good for finding things you weren't looking for with bench tests.

 

[JohnRoberts]

Another audible effect of slew limiting - ascertained from actually using the equipment in a recording situation, was in the original Focusrite EQ.
Rupert Neve slugged down all the 5534 op-amps with 47p caps between 5 & 8 instead of the normal 22p (and more usually 15p). He said he did this to match the slew of early Neve modules.
This resulted in a very 'dry' kind of sound - not harsh.
I read an article by Alan Sides who purchased a Focusrite desk for Ocean Way and he said exactly the same thing.
Oh and for those who want to try slew limiting, there's a circuit here:
http://www.all-electric.com/schematic/slewlim.htm
With some interesting comments.

Increasing the compensation cap on a 5534 not only impacts slew rate, but also reduces the open loop gain at HF. In a high closed loop gain circuit this could actually impact the frequency response but more likely result is higher HF distortion (the input differential needs a larger differential voltage for the same output slew rate, or HF voltage gain). At very low closed loop gain this compensation might be necessary for stability, as under-compensated 5534 is not unity gain stable.

I am not inclined to argue about what other people hear, and I mean no disrespect for Rupert Neve.

I do have a sense for how the 5534s act as they were the go-to choice for bipolar opamps for a period back in the day, and are still respectable today when properly applied.

JR

 

[barclaycon]

If ever there was an industry where everyone has an opinion - it has to be this one!
There are no end of people who are willing to offer their view as being the definitive one, but I am always wary of those who describe how something will sound without actually listening to it, and I gave up arguing with people who don't have a body of work as a reference a long time ago.

There are some interesting arguments here about how things sound and how to improve them - some more relevent than others, but specifications are not the ultimate guide (e.g. 'high slew rate = better sound' is not absolutely proven).
I think the reason why old audio stuff tends to sound better is because, in general, it was designed by people who used it for a living.
It's also more pertinent to identify a great sound and then work backwards to find out how it was created.
Just one final example. I was working at a studio in NW London some years ago that aquired an original Neve 8078 desk. We decided to do some testing on it with a Lindos and a scope etc.
It 'looked' dreadful!
Bass rolled off early, Square waves had extreme ringing (probably due to all of the transformers), EQ curves looked strange etc. etc.
However, when we plugged in a microphone and did some recording - it sounded amazing!
It confirmed a lot of what I'd thought about Neve desks in that that they are not particularly nuetral, but distort sound in a very musical way.
Now you can opine about which bits of circuitry are responsible for which sound, but in the end it's whatever sounds good that matters. That's what we are chasing is it not ?

 

[Nishmaster]

That's what we are chasing is it not ?

I'm not sure how we got all the way over here, but I'm just chasing a guitar circuit.

Rupert later may have been trying to get back the mojo of his earlier work, but back in the day he was chasing performance just like everybody else. And, I'd be hard pressed to say he used it for a living. I don't think I've heard any records with a Rupert Neve mixing credit.

 

[JohnRoberts]

Sorry for my participation in this thread veer.

The objective vs subjective aspects of audio product design have been well debated for several decades and will not be resolved on an internet BB.

peace out...

JR

 

[Nishmaster]

Sorry for my participation in this thread veer.

No need to apologize, good sir. Your contributions are always welcome, veer or otherwise.

 

[barclaycon]

Re.

Rupert later may have been trying to get back the mojo of his earlier work, but back in the day he was chasing performance just like everybody else. And, I'd be hard pressed to say he used it for a living. I don't think I've heard any records with a Rupert Neve mixing credit.

Well you see, Rupert's body of work is that he designed and made recording consoles and equipment that sounded fantastic. That makes his opinions and arguments relevent (and I dare say he did make a good living out of it!)
Do you see the logic ?
Now if somebody said his consoles weren't good (because the transformers ring or the slew rate was bad) and if that person was someone without some kind of track record in this business - then why would we care!
I've encountered so many pundits in the past who express an opinion based solely on figures or specifications that I often think that some kind of perspective is needed and as a rule of thumb it's a wise move to look at what qualifies that person's opinion.

Your thread was titled Slew Rate Limiter right ?
I've been working on something similar - which is why I chimed in.
I'm not really bothered if somebody else here thinks I'm barking up the wrong tree - I'm actually getting some interesting results.
I recommend you look at the website I posted which has a useful circuit AND some interesting information.
I thought that might be helpful to you - sorry if it isn't.

 

[Nishmaster]

Your thread was titled Slew Rate Limiter right ?
I've been working on something similar - which is why I chimed in.
I'm not really bothered if somebody else here thinks I'm barking up the wrong tree - I'm actually getting some interesting results.
I recommend you look at the website I posted which has a useful circuit AND some interesting information.
I thought that might be helpful to you - sorry if it isn't.

Didn't mean any ill will by my comments.

I think we are chasing different dragons, is all. The circuit, in my case, isn't really meant to sound "good" in the traditional sense. I did take a look at that link (it's one I found earlier in my search also), but I didn't find it terribly informative. I'm not quite seeing how that circuit is working, but to me it looks like a current limiting type approach, maybe. He also mentions Iabc and C1 but doesn't mark them on the schematic. I'm not really an engineer so much, so I'm not sure what's going on there.

 

[JohnRoberts]

It's pretty simple.. a transconductance amp which is a current output amp, driving a cap to ground inside an overall feedback loop will current limit into the cap which is slew limiting, variable depending on the abc (amplifier bias current).

This is not unlike how an opamp works but the abc is the input long tail pair operating current and the cap is the compensation cap.

There really was a thread here several years back with much discussion and several circuit approaches..

JR

 

[PRR]

> I'm not quite seeing how that circuit is working, but to me it looks like a current limiting type approach, maybe.

Exactly right. A "flat" signal with limited current into a capacitor will slew-limit. It's a fundamental limit on any voltage-FB op-amp, or you can do it with an OTL which offers a convenient ABC pin.

Conversely, a differentiated signal into a voltage-limiter then integrated will slew-limit.

The sweetness of the OTL plan is that you can vary the current-limit, symmetrically, very easily, without any change in the non-slew-limited action.

The sweetness of diff-clip-int is that it is all standard parts. And real serendipity when existing parts can be re-bodged to serve the function (making my high-less horn a bonus instead of a drag). The drawback is that to change the slew limit without changing non-slew action you must trim two parts or clever-up some form of complementary coupling.

Over the main frequency band, it is all the same. (There will be differences at extreme freqs. Differentiators can be very fussy.)

 

[Nishmaster]

So this is what I ended up with:

First stage is the clipping stage, second two opamps form the slew limiter. R5 is a 250K log pot (drive control), R11 and R8 are a 25K lin pot (slew-limit control), and R15 and R16 actually ended up being a 100K log pot (all I had left, volume out control). I just threw this together kinda quickly on some stripboard, but I'm thinking I should have considered the layout more heavily. Also, I ended up picking up 5532s for the opamps, which aren't happy on +-4.5v, so instead of the battery divider pictured, I'm using two batteries. I left out the filtering caps, though. Some problems:

* Lots and lots of hum. Not sure why, I'd assume batteries would be fairly noiseless. My guess is that I'm rectifying something on the way in, since it's not in a box yet and I didn't use shielded cable anywhere. What would be the standard locations for using shielding and how should I connect it (shield to both ends, just one end, etc.)? I've got the oft quoted Rane grounding guide but it's not really helping me for some reason. Should any of the pots have shielded wiring?

* One half of a 5532 is left unused, and I'm not sure how to properly terminate it. Right now I have it floating.

* The slew-limiter section oscillates when the pot there gets close to fully 25K to ground. I originally didn't have R6 from the wiper to the next opamp, but SPICE indicated that it would oscillate pretty badly if I left that out. I'm wondering if I should up the value there, but it decreases the range of my limit control, so I'm trying to avoid that. The 5532 half that's floating is not one of these two here.

Any ideas?

As an aside, how do people do prototype work on stripboard? This little thing took me a good three hours to put together, and I'm certainly not looking forward to having to bodge this thing together again just to get it to a noiseless state, and certainly less happy if it doesn't sound good yet and I have to tweak the circuit any. It seems like breadboard isn't really going to help me in the oscillation or noise game, but I'm sure those more experienced than I have some tips for this kind of thing.

EDIT: Something else I also thought of...I know 5532s can be finicky, should I be using some decoupling caps here? Or perhaps a completely different opamp model is more appropriate.

 

[tv]

What's the role of the R10?


for "alpha" prototyping, I usually make a ratnest - either "floating in air" or on a piece of a pertinax that I carved some copper "islands" into.

For a "beta" prototype, I usually use perfboard, and try to make a circuit in a visually appealing way, and solid (i.e. not easy to break apart).


I'm sure you have seen this:
http://www.diystompboxes.com/smfforum/index.php?topic=78319.0

 

[Nishmaster]

Had not seen that tv, thanks. R10 is to keep a minimum gain on the opamp and to avoid it going to open loop. Not sure if it's strictly necessary, but it seemed like a good idea.

 

[PRR]

You want C1, not R5, to dominate gain.

R5 should be as close to infinity as the opamp bias current allows.

Drive may be adjusted with R7.

I'm not understanding U2 U3 but maybe coffee will help.

9V total is awful teeny for 5532. It's near-dead at 7V, and really wants a lot more.

Shielding: of course it hums with no box out on the bench between the fluorescent lamp and the TV set. You may line a metal lunch-box with a Daily News, drop the breadboard in, clip the lunchbox to circuit common: that sometimes gets 98% of the hum down. Closing the lid helps, often you can leave it open a crack for the wires to come out. Keep pot-leads SHORT, preferably inside the box (that's why they put the knob outside and the actual pot inside).

 

[PRR]

This requires a dual 500K or 1Meg Audio taper pot (stereo volume pot).

It is unity-gain, transparent, for all signals below the slew limit.

A 1KHz 10V sine (shown) can pass un-altered, or be squashed to sub-Volt triangles. At max slew, simple non-cymbal speech/music (even "music") should pass hardly-touched; yet be zig-zagged at minimum slew. This lets you experiment to find what cap and pot values "work" for your signals, signal-level, and ear-taste.

The 10Meg DC resistor is too much for 5532. Use TL072.

You could scale all impedances down to use a BJT-in chip, but the input impedance would strain most sources and the diode clipper, forcing two buffers.

 

[Nishmaster]

You want C1, not R5, to dominate gain.

R5 should be as close to infinity as the opamp bias current allows.

Drive may be adjusted with R7.

I'm not sure I understand why. The only reason I can see is to ensure that the frequency rolloff of C1 is not dependent on the "drive" control, but that doesn't bug me so much.

I'm not understanding U2 U3 but maybe coffee will help.

Basically (I think), you've got the clipper/integrator idea there (I think). The R11/R8 pot fully towards U2's input means the gain of U2 is set by R1 and R4 in parallel. When the pot goes the other way, the gain of U2 is 25K (the pot) in series with R1, parallel with R4, driving the amp to clip. The integrator following enforces the max-slew condition (I think). It's the same circuit that was in my first post, just drawn slight differently. Actually, there is one error, and that is that R6 should be connected straight to the - input of U3 instead of being in series with R1. This is just my guess on the behavior as explained by Hickman in his book, page 70.

It's pretty darn close to unity gain for everything under slew-limiting, certainly seems like it follows the input just fine.

http://books.google.com/books?id=toedBeJPSNcC&lpg=PA68&ots=h6Uengs1En&dq=slew%20limiter%20hickman&pg=PA70#v=onepage&q=&f=false

9V total is awful teeny for 5532. It's near-dead at 7V, and really wants a lot more.

Yes, but as I mentioned, I've got 2 batteries, one for each half of the supply, so 18V.

Also, thanks for the lunchbox idea! Hadn't thought about something so simple, and so available, and as the biggest bonus of all, so roomy.

 

[tv]

>>> 9V total is awful teeny for 5532.

For guit. circuits, at 9V the 5532 sounds quite fatter than other opamps, and sort-of "relaxed", but with some authority (words...). Not really battery-friendly, but with AC/DC bricks, it works fine....

 

[JohnRoberts]

While I resisted the temptation to say just do it this other way instead... I admit to not being comfortable with your design also..

Looking at your link I see why..  one part is misconnected (on your schematic). R1 should connect to the other end of r6. Does you circuit agree with your schematic?

5532 is not an obvious choice for battery. In fact an OTA based approach could be pretty thrifty on supply current.

JR

 

[Nishmaster]

Thanks JR, I had caught that above. My circuit does currently agree with the schematic.

Actually, there is one error, and that is that R6 should be connected straight to the - input of U3 instead of being in series with R1.

I just made that change in SPICE and now I've got myself an oscillator instead when R8/R11 pot is fully towards U2, and I can't really figure out what purpose it is supposed to be serving. This whole configuration seems rather unstable in general, perhaps ditching isn't a terrible idea.

While I resisted the temptation to say just do it this other way instead... I admit to not being comfortable with your design also..

Feel free to let me know if I'm barking up the wrong tree or doing something stupid!

5532 is not an obvious choice for battery. In fact an OTA based approach could be pretty thrifty on supply current.

Yeah, but when I bought them, I didn't know what an obvious choice was, so these are what I have at the moment. What could I run on 9v that would have similar or better noise performance? In addition to having something unique to add to my guitar palette, I was also hoping to not have to actually spend nearly as much as I would on a boutique pedal, so I'm somewhat hesitant to solve the problem by adding costs.

I must also admit to not really understanding how to implement the OTA approach properly, especially in regards to setting abc, so that's why I've avoiding pursuing that path thus far.

I know this is probably training wheels type stuff for some of you, but I really appreciate the help, gents. Thanks so much!