Need ideas for a limiter circuit

[eschertron]

I can't just passively attenuate down to that level with the gain control and master volume cranked, or I will never be able to get a clean sound at full power.

It's not really volume that makes the class D power amp clip miserably, but max signal. What ever output headroom V3 has, it must be less than anything that clips the power amp. Even momentary peaks on an otherwise 'clean' signal. Regardless of how much quieter the volume appears.

With tube power amps, the momentary signal peaks don't immediately offend. In fact, there's some mojo to the compression/limiting that the power tubes do to the signal before it gets noticeably clipped. That luxury is not available with class D power amps. So the limiter circuit must provide a replacement effect to that expected soft compression.

[Gaz]

It's not really volume that makes the class D power amp clip miserably, but max signal. What ever output headroom V3 has, it must be less than anything that clips the power amp. Even momentary peaks on an otherwise 'clean' signal. Regardless of how much quieter the volume appears.

With tube power amps, the momentary signal peaks don't immediately offend. In fact, there's some mojo to the compression/limiting that the power tubes do to the signal before it gets noticeably clipped. That luxury is not available with class D power amps. So the limiter circuit must provide a replacement effect to that expected soft compression.

Yes, that's a really good way of putting it. I am already missing auto-limiting-tube-mojo. In the previous post I am referring to the master volume control, that when turned up, creates more signal voltage

[eschertron]

Yes, that's a really good way of putting it. I am already missing auto-limiting-tube-mojo. In the previous post I am referring to the master volume control, that when turned up, creates more signal voltage

Right. So you can turn the signal amplitude down for balancing heavily-distorted sounds versus clean sounds, but you can't turn it up to balance clean sounds versus heavily-distorted sounds. Make sense? The overall effective volume of the amp is less, if you take the clean sounds as the benchmark. This is one argument for why a SS amp needs to be 2x to 3x more powerful than an equally 'loud' tube amp.

[Gaz]

Right. So you can turn the signal amplitude down for balancing heavily-distorted sounds versus clean sounds, but you can't turn it up to balance clean sounds versus heavily-distorted sounds. Make sense? The overall effective volume of the amp is less, if you take the clean sounds as the benchmark. This is one argument for why a SS amp needs to be 2x to 3x more powerful than an equally 'loud' tube amp.

Yes, I certainly understand the dilemma. I am looking for possible solutions

[J M Fahey]

Ok, this is the very simple Ampeg/Crate Opto Limiter.



How it works:

* There is a string of resistors from +V to -V : R45 - R48 - Ground - R47 - R52
Values are calculated so R45 and R52 drop about 5V , so R48 and R47 must be scaled up/down depending on actual +/-V present, values shown match 50V rails but it can work with as low as +/-12V ones if necessary.

* so we have +V-5V at the junction R45/48 and its complement at the -V side.

* we have a diode net (D11-13-14-17) feeding the Opto LED, normally reverse biased, and it will pass current only if Speaker peak voltage "almost" reaches +V or -V rails.

* you use that lowered opto resistance as you wish.
Here they lower NFB resistance so gain drops; standard use is to put, say, 4k7 or 10k in series with amp input and LDR to ground.

* the beauty is that besides simplicity it´s *external* to power amp and does not care a bit about its internal circuit, I bet it will also work well with amps such as yours.

The speaker peaks never reach rails, so they never clip, and it´s auto adjustable: if rails drop or even mains voltage does, this peak detector self adjusts and tracks that.

[Gaz]

Thanks so much posting the schem and explaining that circuit to me! Super appreciate that. I am eager to try it out. Just have to get that opto...

One (perhaps dumb) question about the circuit: I plan on using the LDR in the lower leg of a divider like you mentioned. Does the LDR need to be ahead of the limiter input (SPK) like in the schematic? My guess is that it does, but my brain is going in loops trying to figure it out.

[J M Fahey]

Worst case, and if you can still buy a few CDS cells (already forbidden in Europe) you can cook your own Opto.

As of the limiter, it´s fully external to the amp, it simply compares amp peak output against rails and when too close, LDR attenuates, just like that.
Don´t worry about internal signal path, NFB loops, nothing, just a resistive attenuator at the input.

In fact, I made a small stamp sized PCB and add this to "other" amplifiers which need it, in general small PA, Keyboard or Bass ones, but specially Piano amps where you have very high peaks and the least distortion sounds horrible .... mine *all* include some kind of limiter of course, although this one is "too clean", I prefer Fets which add more "flavour".

[Helmholtz]

..although this one is "too clean", I prefer Fets which add more "flavour".

If a limiter reacts too fast, it produces (low frequency) distortion. Response time should be at least a few cycles of the lowest signal frequency for clean operation.

[J M Fahey]

Yes, if you want clean operation

[g1]

So, it does not matter if you are sampling at the speaker output, or any other part of the circuit, you just use whatever rails for that part of the circuit?
In this case, you do not want to hear the harsh sound of the power amp clipping, so it would be best to sample in the power amp section, like in the example (post #40)?

[Gaz]

...I prefer Fets which add more "flavour".

I'd love to see an example... especially because I can't even find where to buy a vactrol. They seems to be obsolete!

[J M Fahey]

TO92 Fets are also going the way of the Dodo

You can still get CDS cells, non ROHS compliant of course, at none less than Adafruit: https://www.adafruit.com/product/161

Or ready made vactrols: https://modularaddict.com/parts/synt...s/max-vactrols

or: http://store.synthrotek.com/Vactrol

[Gaz]

TO92 Fets are also going the way of the Dodo

You can still get CDS cells, non ROHS compliant of course, at none less than Adafruit: https://www.adafruit.com/product/161

Or ready made vactrols: https://modularaddict.com/parts/synt...s/max-vactrols

or: http://store.synthrotek.com/Vactrol

Thanks! Those links are helpful. Are there anything in particular I should be choosing for the specs, or will all these sold from specialty synth sites be in the ballpark for this purpose? I realize I will have to adjust the top leg of the divider depending on the On resistance of the LDR in the optocoupler.

[J M Fahey]

You will only adjust resistor values to get 5V reverence below or above rails, nothing else.

Opto specs are important if you use them for something needing precision, such as tracking volume controls, Studio Compressors, tone controls, Auto Wah, Phasers or Jimi Hendrix type "chorus" , but here they are used individually and almost on an on-off fashion, so my experience is they are not critical at all.

Just for the record, pass, say, 5 or 10mA through the driving diode and measure resistance across LDR.

[Gaz]

Thanks, I got a couple ordered and will give this circuit a shot and report back!

[mhuss]

Sadly, Vactrols are also going the way of the Dodo (CdS!!! ). I started using MOSFET output optoisolators (e.g., NEC PS-7141 series) instead for switching. As Juan pointed out, if you just need a switch, linearity, etc., doesn't really matter.

[Gaz]

Sadly, Vactrols are also going the way of the Dodo (CdS!!! ). I started using MOSFET output optoisolators (e.g., NEC PS-7141 series) instead for switching. As Juan pointed out, if you just need a switch, linearity, etc., doesn't really matter.

Thanks for the advice, Mark. What application are you using the mosfet optos for?

[olddawg]

I used to use 4 diodes to limit the input to approximately line level. But they will hard clip anything over the propagation voltage. Another old trick is to take a signal off of the main amp input, isolate it with a cap, rectify it with a diode, filter it with a cap, and then use that DC voltage as a preamp output control voltage using a couple of transistors or an opamp. The problem is it isn’t instantaneous. When it turns on there may be something audible. You have to play around with cap and resistor values, fast diodes, etc. Time constants matter. There used to be an application specific IC used in a lot of high powered stereos that basically had a power supply pins, a ground, an in, an out, and a control voltage input. Sometimes it was used as a volume control too. The chip number escapes me.

[mhuss]

So far, just channel switching. The optos are available with 400v ratings. I attached a simple example.

[Gaz]

Cool. Thanks for sharing. Gonna put that in my bag of tricks for sure.

[Helmholtz]

I used to use 4 diodes to limit the input to approximately line level. But they will hard clip anything over the propagation voltage.

Any diode clipping can be made softer by wiring a resistor (e.g. 100 Ohm depending on source impedance) in series with each diode. (Also works with overdrive circuits.) But limiting will be less effectve as the resistors still allow for some level increase.

The problem is it isn’t instantaneous.

For clean limiting/compression, the circuit response must not be really instantaneous. This is why LDRs (photocells) work fine, as they react slowly enough having a HP corner frequency of 10Hz or lower.
Instantaneous limiting (like with diodes) inevitably produces distortion.

Time constants matter.

That's the main point, especially regarding attack. I think something in the 100ms range should do.
Too fast response (< 10ms) causes low frequency distortion, too slow response causes noticeable level overshoot. And instantaneous response produces distortion at all signal frequencies.

[Gaz]

Ok, this is the very simple Ampeg/Crate Opto Limiter.



How it works:

* There is a string of resistors from +V to -V : R45 - R48 - Ground - R47 - R52
Values are calculated so R45 and R52 drop about 5V , so R48 and R47 must be scaled up/down depending on actual +/-V present, values shown match 50V rails but it can work with as low as +/-12V ones if necessary.

* so we have +V-5V at the junction R45/48 and its complement at the -V side.

* we have a diode net (D11-13-14-17) feeding the Opto LED, normally reverse biased, and it will pass current only if Speaker peak voltage "almost" reaches +V or -V rails.

* you use that lowered opto resistance as you wish.
Here they lower NFB resistance so gain drops; standard use is to put, say, 4k7 or 10k in series with amp input and LDR to ground.

* the beauty is that besides simplicity it´s *external* to power amp and does not care a bit about its internal circuit, I bet it will also work well with amps such as yours.

The speaker peaks never reach rails, so they never clip, and it´s auto adjustable: if rails drop or even mains voltage does, this peak detector self adjusts and tracks that.

Ok, so I got the parts in, and got this functioning... kind of. I am using a NSL32SR2 opto, and subbed 1N4007's for the 1N914's (I don't believe it would matter in this circuit).

Here's a schem of how I have it setup. I can get the opto to turn on with signal, but need help setting where that happens. I am using the LDR in the bottom half of a divider (as you suggested) feeding the limiter. My goal is have this kick in when signal reaches 3Vp. Does that mean I drop the the +/- rails to be 3V? Still a little confused, but think I've almost got it

[J M Fahey]

Why 3Vp?

I thought we were talking about limiting a big nasty Class D module, hundreds of watts RMS, so speaker out NEVER reaches power rails but stops 1V shy of them, never clipping.

This power amp limiter does not know or care about what the preamp does.

[Gaz]

Why 3Vp?

I thought we were talking about limiting a big nasty Class D module, hundreds of watts RMS, so speaker out NEVER reaches power rails but stops 1V shy of them, never clipping.

This power amp limiter does not know or care about what the preamp does.

Ok, I see, so I need to be taking the input from the speaker output? Someone else asked about that earlier as well, so I was confused. It reaches about 60Vrms (~84 peak), so I take get the +/- rails from the preamp B+, and drop them down to 80vdc or so? Then I use LDR the same way as the bottom of a voltage divider at the preamp output?

[J M Fahey]

Ok, I see, so I need to be taking the input from the speaker output? Someone else asked about that earlier as well, so I was confused. It reaches about 60Vrms (~84 peak), so I take get the +/- rails from the preamp B+, and drop them down to 80vdc or so? Then I use LDR the same way as the bottom of a voltage divider at the preamp output?

1) Forget the preamp, period.

2) re read and apply the first 3 asterisked answers in post #40 , it´s all there.

If you do not understand some bit of it, ask.

Forget RMS volts, all we care is rail volts and setting voltage references 5V away from them.

Original design applies to +/-51V ... what are *yours*?

[Gaz]

1) Forget the preamp, period.

2) re read and apply the first 3 asterisked answers in post #40 , it´s all there.

If you do not understand some bit of it, ask.

Forget RMS volts, all we care is rail volts and setting voltage references 5V away from them.

Original design applies to +/-51V ... what are *yours*?

Yes, I reread that post many times, and I do get the concept now about 5 volts away from the rails. I will have to measure the rail voltage now that I understand that.

My main confusion now is what is the purpose of the LDR? I thought it needed to go in the preamp to lower the signal when the limiter kicks in.

[J M Fahey]

The LDR is floating and can go anywhere.

Typical is to put a resistor in series with amp input and the LDR from amp input hot to ground, so "between preamp OUT and power amp IN" but it can attenuate at an earlier stage.

In a nutshell: the *detector* part is Power amp business, and uses +/- rails and speaker output.

The attenuator part may be between preamp and power amp or at an earlier Preamp stage.

*IN THEORY" , the LDR might even be in parallel with Guitar/Bass passive volume control but for other reasons (not bringing the fox inside the hen house) I prefer not to bring powerful *speaker* signal towards early sensitive stages.

But ... but ... I´m bringing the LDR, not speaker signal.
Whatever, LDR wires live in a loud signal environment (LED current IS part of speaker signal) and parasitic capacitance or inductance may some of bring it backwards, so ... I play it safe.

After all, many amps have radio interference problems ... and we are talking about a signal carried through insulating air and measured in micro Volts ... or less.

[Gaz]

Thanks for taking the time reiterate - this is all clarifying. My only issue now is trying to measure the positive and negative rails on the class D module since I have no schematic. Maximum output voltage is listed on the datasheet as 76Vp. Does that mean that is the rail voltage as well? Here is also a shot of the block diagram if that's helpful to find a measuring place.



Once I find that out, I'm wondering if it's advisable to tap off the module for the detector, or create the detector voltage from elsewhere so I don't inadvertently damage the module...

[eschertron]

Thanks for taking the time reiterate - this is all clarifying. My only issue now is trying to measure the positive and negative rails on the class D module since I have no schematic. Maximum output voltage is listed on the datasheet as 76Vp. Does that mean that is the rail voltage as well? Here is also a shot of the block diagram if that's helpful to find a measuring place.



Once I find that out, I'm wondering if it's advisable to tap off the module for the detector, or create the detector voltage from elsewhere so I don't inadvertently damage the module...

I'm following this thread with intent... I'm going to build something like this for a class D power module as well. My take is that the power rails are whatever is being fed to the module. In my case, it's a 24vdc single-ended supply so it'll end up looking like +/-12vdc to some arbitrary signal ground. The takeaway is that the rail voltage reference is the very rails themselves, not something derived from the rails. I think

[J M Fahey]

Not too good news.

Pity we got to know only on post#63 which Class D amplifier are we talking about.

Not only it should have been mentioned from the beginning, but also a direct link to its datasheet should have been shown:

https://icepower.dk/download/1652/

no mention of schematic or internal circuitry but from specs, connectors and interconnection with slave units (which is most unusual), it can be reasonably guessed that:

1) no +/-V rails whatsoever so Ampeg/Crate limiter is unusable here.

2) 700W RMS my *ss
It can meet that spec only for 5 minutes, only starting from cold (i.e. : "5 minutes once a day")
Of course, since FTC rules no longer apply, under pressure (and probably $$$$) from consumer products vendor associations, you can publish any rating as long as they hold for a few seconds, so ...

3) actual RMS power ... which it can sustain without shutting down ... is 270W RMS ,into either 4 or 8 ohm

4) you do have a direct DC out from power supply, meant to power auxiliary units (as in a triamplified PA cabinet) ; it´s +43V DC, at an unspecified current consumption.

5) almost forgot: even if we had +/-V split supplies, we do not have a single speaker out to limit but two, since output is bridged , two hot, no ground, so it´s another nail in the Opto limiter coffin.
Oh well.

6) so I guess you´ll have to apply Plan B: build a conventional limiter, as fancy or simple as you like, use it between preamp and power amp, drive power amp to clipping (check on a scope screen) and then adjust limiter threshold or level out until clipped signal goes down to normal, and further drive increase does no make output clip; readjust as needed and leave it there.

Oh well.

EDIT:

I'm following this thread with intent... I'm going to build something like this for a class D power module as well. My take is that the power rails are whatever is being fed to the module. In my case, it's a 24vdc single-ended supply so it'll end up looking like +/-12vdc to some arbitrary signal ground. The takeaway is that the rail voltage reference is the very rails themselves, not something derived from the rails. I think

Sorry, same thing as with Gaz´s amplifier, this very clever limiter does not apply in single source bridged output amplifiers because it´s connected rail to rail and detects where a single speaker out "almost" reaches one or the other.

[eschertron]

Sorry, same thing as with Gaz´s amplifier, this very clever limiter does not apply in single source bridged output amplifiers because it´s connected rail to rail and detects where a single speaker out "almost" reaches one or the other.

gnats!

I appreciate how a wider voltage spread helps minimize the effect of the diode voltage drops. If I have a signal that's 3vpp, not much room to control.

I'm likely to play with a J111 as a control element at the lower signal voltages I expect to see.

[J M Fahey]

Excellent idea, had forgotten J111

Just ordered 1000 of them at Mouser in TO92 case, which has gone out of manufacturing, I guess they are practically "auctioning" the last ones.

As "obsolete" as they are, they are less "obsolete" than CDS cells , at least they are not forbidden.

Not good enough as amplifiers, but definitely usable as audio switches, compressors, and phasers.

[Gaz]

Ah, Juan, sorry I am such a solid state neophyte, wasting your time with all the particulars of this circuit. Hopefully it will help someone building a proper "lead sled"

You're right that the Class D 700W is bullshit. I actually got my hands on my bandmate's Mesa D800 Subway that uses the same Icepower module, and it does only 600Wrms into 4 Ohms when the limiter light comes on. All this amp has (from some minimal circuit tracing) is a passive zener limiter. It actually clips before the limiter light comes on, which is under 600W. This thing does not like to be run with a continuous sine wave as you mentioned, and while I haven't gotten into thermal shut down I can feel it heating up, even with a fan, extra heat-sinking, and the chassis open. Mesa says to only run with the clip light going approx 25% of the time, and their published spec of 800W is with 10% THD. I don't really understand THD measurements, but on the scope coming from tube amps that's what I would personally call noteworthy clipping.

I've had pretty good luck trying a back to back 2.7V zener diodes at the output of the preamp, but I'd like a little more control, or precision as it goes into clipping a little earlier than I'd like (albeit it is a smooth transition).

[Gaz]

gnats!

I appreciate how a wider voltage spread helps minimize the effect of the diode voltage drops. If I have a signal that's 3vpp, not much room to control.

I'm likely to play with a J111 as a control element at the lower signal voltages I expect to see.

What circuits have you been looking at? It goes without saying, but I'm looking for more ideas now too

[J M Fahey]

Oh, don´t worry, discussions are aways worth it because lots of new ideas appear .

Pity this simple clever limiter which covers 99% of amps, from TDA2030 to 1000W monsters met the 1% which does NOT use dual rails.