Slew Rate Limiter

[JohnRoberts]

Keep doing what you are doing... this needs to be your design.. What opamp you use is only an incidental detail.. first make the sounds you want.

To improve stability try some local HF feedback, say a small cap across r4 (5-10pF? ).

JR

 

[PRR]

> What could I run on 9v that would have similar or better noise performance?

Just use TL072. They are often in-stock at the nearest Radio Shack, for a rip-off $2 which is chump-change. Next time you commit to mail-order shipping, you can get ten TL072 for $2, which will build a lot of pedals.

In high-impedance applications, 5532 noise is not always lower than TL072. The BJT-in 5532 has huge current noise, the FET has none. And the TL072's noise is generally (in good design) comparable to the audible hiss from a Fender pickup, so even a "better" input will not necessarily be lower hiss. Yes, there ARE better choices for pedalboxes, but the TL072 is rarely a "wrong" choice.

 

[Nishmaster]

So, here's the next iteration of the design, using the OTA with the cap inside the feedback loop instead. The dual opamp approach proposed by Hickman just isn't a stable design, no matter what I tried. The HF stability cap across R4 was stable at some input frequencies, unstable at others, especially low clipped frequencies where you have basically DC at the top and bottom of each wave. Additionally, the pot there and R4 formed a peaking eq that was getting into the audio band at one of the extremes of the pot travel.

Did I make any glaring errors? I wasn't sure how bad offset voltage was going to be, so I've got .22uf caps between each opamp, as well as 1Ks. Do those values make sense?

EDIT: Also, is my output going to come with a DC component of 4.5V? If so, how can I solve that?

-Matt

 

[PRR]

The "+ input" of all opamps has to be biased. U1a has R1, fine. U1b is floating, bad. Replace R8 and C5 with shorts, you are good here. Likewise U1 probably wants something firmer than a couple diodes; again replace C9 with a short and it is probably fine (alternatively throw 50K from +in to ground).

That's just to get DC biasing close enough so it isn't jammed-up. Whether the circuit will "pass audio" or just distort massively, is a different can of worms.

 

[Nishmaster]

The "+ input" of all opamps has to be biased. U1a has R1, fine. U1b is floating, bad. Replace R8 and C5 with shorts, you are good here. Likewise U1 probably wants something firmer than a couple diodes; again replace C9 with a short and it is probably fine (alternatively throw 50K from +in to ground).

That's just to get DC biasing close enough so it isn't jammed-up. Whether the circuit will "pass audio" or just distort massively, is a different can of worms.

Heh. Don't I feel embarrassed. I was so focused on getting rid of offset that I forgot all about my bias. I'm trying to keep the wave as symmetrical as possible, so I kept C5, but moved R8 as 100K to ground. That way I'm not amplifying the first amp's offset and all the diodes are nice and happy because they get equal playing time. Of course, unsymmetrical clipping can sound good too, but we'll see how this sounds first. As an aside, I've never really understood the choosing of values for bias resistors. I used 100K, because that affected frequency performance the least, but what's the rule of thumb here? Obviously you don't want to load down the opamp too heavily, but that aside, does it really matter what large value R goes there?

What's scary is that it passed audio just fine in the sim as it was. I'm not sure if my virtual ground scheme is throwing SPICE, maybe. The slew limiter looks to be unity gain up until slew-limit, so a plus there.

I tweaked the values in the feedback loop of the first amp some, hopefully that can improve noise performance a bit.

 

[tv]

If I were you, I'd put a R-C combo between the second opamp and D3-D4.


-----------------------------------------------------
Note: I am totally clueless regarding OTA slew-rate limiters, but it occured to me that it should  not be so dissimilar to a simple two-opamp circuit, but with a R (or a pot?..) between the first opamp output and the cap -- into the second opamp input. (look at the schem I edited)

Or am I __totally__ wrong here?

 

[PRR]

> choosing of values for bias resistors.

Low enough to not make trouble.

High enough to not make trouble.

If too low, and you want bass, the cap value/size/cost may get annoying. 5532 will pull a 500 ohm load, but for 20Hz that means over 40uFd, which is a not-teeny electrolytic or a HUGE film cap.

5532 with a 1 Meg bias will have significant DC offset. Tolerable in some situations (just passing a signal, do you care if it sits at 4.5V or 4.0V?) and troublesome in other cases (as you say, the diode limiter gets obviously asymmetric with 60mV offset).

The lazy/nice thing about TL072 is that values over 1Meg give "no" offset, so you can use-up those 0.01u caps and meg-resistors littering your bench.

> it passed audio just fine in the sim

Depends on the sim. Was the opamp model realistic or Ideal? Did the sim pre-run with all caps open to find DC op-points, or just plow in? For some circuits, the simmed initial op-point "works" but the DC currents drift over a period of minutes... it can take a very long run to discover this.

_I_ say it must work on a cocktail napkin before you boot the PC. And that if you can't compute it on a slide-rule (preferably on a match-book), it isn't really going to work. But I'm going out of fashion.

 

[Nishmaster]

Low enough to not make trouble.

High enough to not make trouble.

I suppose a lot of that comes down to experience with the parts too. I did switch over to TL07x for this project, as of the last schematic, so hopefully that will solve a few issues also.

Depends on the sim. Was the opamp model realistic or Ideal? Did the sim pre-run with all caps open to find DC op-points, or just plow in? For some circuits, the simmed initial op-point "works" but the DC currents drift over a period of minutes... it can take a very long run to discover this.

_I_ say it must work on a cocktail napkin before you boot the PC. And that if you can't compute it on a slide-rule (preferably on a match-book), it isn't really going to work. But I'm going out of fashion.

The model was ideal, yes. I was using LTSpice, which doesn't have an 553x or TL07x model specifically. I'm assuming that it didn't just plow in, but I didn't run the sim long enough to check for drift.

You're not going out of fashion. I'm just not good enough yet to do in on a cocktail napkin, much less a matchbook. I'm slowly getting there, though. I definitely appreciate the help.

I'm using a TL074, so I've got an extra amp on chip that won't be doing anything. What is the right way to terminate it, or can I just leave it float?

 

[JohnRoberts]

In general float the output. To prevent inadvertent output toggling, terminate the inputs to any convenient voltage.

JR

 

[Nishmaster]

Does my virtual ground count as a convenient voltage?

 

[JohnRoberts]

is it nearby?

Is it within the + and - PS voltage extremes?

is it truly convenient?

JR

 

[Nishmaster]

Heh, thanks John!

It would appear to be all three of those things. Parts are in the mail, I'll let you all know when (if?) it passes audio. Thanks again to everybody who chimed in.

-Matt

 

[Nishmaster]

Well, finally pieced this thing together last night, and it passed audio, sort of. It seemed like I was having a biasing problem (splatty, gated sound). All the voltages at all the pins appeared ok, up until the output of the OTA. I was getting nearly 8 volts there, which didn't make a whole lot of sense to me, since the + pin of the OTA was getting the proper 4.5ish volts bias (it was exactly at virtual ground). I've double checked the layout, checked various continuities to make sure I hadn't somehow connected V+ to the output of the OTA, but the layout seems good.

So, I figured maybe I burned the chip up a little, as I soldered it directly to the pcb instead of using a socket. I desoldered the chip and soldered on a socket, but now the chip gets extremely hot (too hot to touch). I did use a fresh chip (a couple, actually). My gut feeling is that I probably didn't burn the chip up soldering it, but perhaps whatever is happening now was happening before, as I didn't really check to see how hot the chips were getting right when I first powered the thing up.

I'm kinda out of ideas. Any suggestions on what to do or where to look? I've got no scope at the moment, just a DMM.

 

[JohnRoberts]

The OTA should never get very hot.. it's ouptut is naturally current limited.

If the schematic is similar to above it looks like 100% negative feedback so output of OTA shoud not be more than input always less if lagging.

Chips getting too hot to touch is usually a symptom of mis-wiring, or even parts in backwards..  check for something blatant.

JR

 

[Nishmaster]

You were right John. I had been staring at the problem for too much time. As soon as I took a break I saw it.

I had copied the pinout wrong from the datasheet to the schematic and I had V+ and V- reversed. Works just fine now!

Now how does it sound? Kinda bizarre. Not as smooth as I was expecting; the action of the slew limiter has a somewhat envelope filter-y quality to it at high gains. There are definitely sounds to be had, though, so I'm in the value tweaking stage and we'll see where I end up. This PCB is a mess of green wires!

 

[xmvlk]

And what about using of dominant pole:
http://radio.feld.cvut.cz/~vlk/schemo/fltdompol.gif