Audio peak limiter for home-brew headphone amp

Hi Guys.
I have been using a homebuilt headphone amplifier for drumming gigs for about 5 years now.
I need to build/ design a fast limiter for those incidents when the sound guy does something stupid and blows my in-ear-monitors.
Anyone built something like this willing to share their work?

Fastest options, diode clipping, very efficient to limit peaks, RMS another story. Another option is to limit the rails of the amp and add a resistor at the output so certain max power could be delivered from the amp to the HP.

If you want a real limiter you may go for a FET or any other simple topology, Fet being the option 1, diode bridge and transconductance after I guess.

For FET:
1 Fet per channel, some caps and resistors for polarization and an opamp with a couple of diodes for control voltage.


For diode bridge:
4 diodes per channel, one resistor, same for CV.


A passive diode, nasty but working (tried out by myself, it could add attack release, ratio and even do asymmetrical compression)


For transconductance:
3 transistors per channel, some resistors, same CV.


A CA3080 or LM13700 also use this topology but already solved in an IC.

Times could be fixed for fast attack and release for taste.

/*images are for reference of the topology, not the final design that you should use, just to explain my self better, as amp you probably could use the one you already have 2nd and 4th scheme doesn't have the CV generator (rectifier and time constants)*/

JS

I lean towards a version of that first FET shunt limiter. Pretty simple and clean while not limiting. 

Looks like some spare parts in the schematic that may not be needed, and I might add a few different ones but a good starting point.

JR

joaquins said:

Fastest options, diode clipping, very efficient to limit peaks, RMS another story. Another option is to limit the rails of the amp and add a resistor at the output so certain max power could be delivered from the amp to the HP.

If you want a real limiter you may go for a FET or any other simple topology, Fet being the option 1, diode bridge and transconductance after I guess.

For FET:
1 Fet per channel, some caps and resistors for polarization and an opamp with a couple of diodes for control voltage.

For diode bridge:
4 diodes per channel, one resistor, same for CV.

A passive diode, nasty but working (tried out by myself, it could add attack release, ratio and even do asymmetrical compression)

For transconductance:
3 transistors per channel, some resistors, same CV.

A CA3080 or LM13700 also use this topology but already solved in an IC.

Times could be fixed for fast attack and release for taste.

/*images are for reference of the topology, not the final design that you should use, just to explain my self better, as amp you probably could use the one you already have 2nd and 4th scheme doesn't have the CV generator (rectifier and time constants)*/

JS

Click to expand...

Great way to tie a knot on these concepts. Thanks for the post!

Gustav

diydidi said:

Hi Guys.
I have been using a homebuilt headphone amplifier for drumming gigs for about 5 years now.
I need to build/ design a fast limiter for those incidents when the sound guy does something stupid and blows my in-ear-monitors.
Anyone built something like this willing to share their work?

Click to expand...

I would recommend a most basic implementation of THAT4301
http://www.thatcorp.com/4301_Analog_Engine_Dynamics_Processor.shtml
Check Fig.12 of the datasheet
http://www.thatcorp.com/datashts/THAT_4301_Datasheet.pdf
I would not recommend the FET approach, for two reasons:
First, it requires some tweaking and selection of FET. Conversely, a 4301-based circuit works first time, subject to proper implementation.
Second, the S/N ratio is not as good in a FET limiter (typically 70dB) as in a VCA limiter (typically >90dB), and that is a very important factor in a headphone application.

Hi.Thank you.
Looks promising. Have you built it?

diydidi said:

Hi.Thank you.
Looks promising. Have you built it?

Click to expand...

I built thousands of them, sold around the world, under the LA Audio brand.

abbey road d enfer said:

I would not recommend the FET approach, for two reasons:
First, it requires some tweaking and selection of FET. Conversely, a 4301-based circuit works first time, subject to proper implementation.

Click to expand...

I am not going to vigorously defend that exact FET shunt limiter circuit, but the topology in general has been widely used, probably more for guitar efx limiters than for vocal dynamics (because of distortion while limiting). The pinch off voltage for JFETs will vary so that very simple side-chain will show device sensitivity wrt threshold, so not repeatable enough for a commercial design. YMMV for DIY. 

Second, the S/N ratio is not as good in a FET limiter (typically 70dB) as in a VCA limiter (typically >90dB), and that is a very important factor in a headphone application.

Click to expand...

In general an above threshold shunt limiter can be very good for S/N and path linearity, when operating below threshold so not affecting the signal. I guess a shunt limiter with padded down signal and a bunch of make-up gain will suffer some noise floor degradation.  Alternately a VCA based comp/limiter will always have the VCA in the audio path so generally somewhat worse compared to a bare opamp running at modest noise gain.

FWIW modern VCAs are quite good and THAT application notes have inspired many commercial products. 

JR

PS: I never used a JFET limiter in a commercial design but back in the early '80s I wrapped an OTA around an opamp to make a combination limiter/noise gate and that was very quiet. 

The Rod Elliott one doesn't work.
I don't think even think he has built it. Its a design by some other guy in AUS. Anyone got it working before??????

diydidi said:

The Rod Elliott one doesn't work.
I don't think even think he has built it. Its a design by some other guy in AUS. Anyone got it working before??????

Click to expand...

Which one is the Rod Eliot one? (OK I just googled and that first one shows up on his website too. )

While it is not exactly like I would do it, it looks like it "should" work. One obscure aspect of that design is that the source of the JFET is biased up to about +5V DC (10k and 4.7k divider).  On Rod Elliot's version there is an * on the 4.7k R6 suggesting you can tweak that value to change the limiting threshold. According to the JFET data sheet the cut-off voltage for 5457 is between -0.5V and -6V. Since that schematic starts with a nominal -5V on the gate, a 6V part would never cut off. Of course the threshold could vary that same several volt range from device to device. 

You could eliminate the distortion canceling network in the side chain to make it a little simpler (R11,C3,C4) but that distortion cancellation technique can make an audible improvement. I've used similar tricks in JFET shunt noise gates and this cancellation helps. (Note: in a noise gate the JFET shunt is only active only during low level signals so that much level padding in front of the JFET is not as useful. In a limiter the JFET is conducting during loud parts so higher distortion than a noise gate where the JFET is off during loud parts. 

Indeed running a TL07x at 30+dB of gain may not be the quietest configuration ever, but that is probably a trade-off to keep distortion down.

When you say this one doesn't work, what it doing wrong? Making R6 a little larger will at least facilitate higher than 5V pinch off parts.

Do you actually want to make this one work... ? I'd debug by looking at DC voltages around the JFET and CV circuitry.

JR

diydidi said:

The Rod Elliott one doesn't work.
I don't think even think he has built it. Its a design by some other guy in AUS. Anyone got it working before??????

Click to expand...

Are you sure? I mean, did you build it and it didn´t work? Sorry for asking, but I personally have a long history of getting circuits not to work. ;D And the fault was not always *cough* with the circuit.
I guess, it could be made to work
I have another document with an interesting fet limiter(haven´t built it). With lots of background info on what is happening. Check it out.
http://users.ece.gatech.edu/mleach/papers/limiter.pdf

I expect a limiter published by Marshal Leach (RIP) to work also.

They both use the same (old) trick to reduce distortion in JFET audio shunts, namely add 1/2 the AC voltage on the drain to the gate control voltage. A simple trick that helps, reduce distortion... for the guys who think they would like distortion, think of all the parts you can save.  ;D

JR 

L´Andratté said:

diydidi said:

The Rod Elliott one doesn't work.
I don't think even think he has built it. Its a design by some other guy in AUS. Anyone got it working before??????

Click to expand...

Are you sure? I mean, did you build it and it didn´t work? Sorry for asking, but I personally have a long history of getting circuits not to work. ;D And the fault was not always *cough* with the circuit.
I guess, it could be made to work
I have another document with an interesting fet limiter(haven´t built it). With lots of background info on what is happening. Check it out.
http://users.ece.gatech.edu/mleach/papers/limiter.pdf

Click to expand...

Yes I built it. Tried different Fets. still nothing....

What about this circuit??

diydidi said:

What about this circuit??

Click to expand...

A great classic. Excellent noise performance. Does not offer real control of attack and release times (dominated by characteristics of the opto-coupler), but workd surprizingly well.

diydidi said:

What about this circuit??

Click to expand...

Yup an old classic and the same topology I used for my early '80s noise gate limiter but I used an OTA in place of the LDR.

Note: I would probably upgrade that design to use a modern opamp. Almost anything you can plug in the same socket will beat a 1458 for noise and audio performance..

JR

diydidi said:

Yes I built it. Tried different Fets. still nothing....

Click to expand...

As I suggested earlier troubleshoot the DC voltages around the JFET.

At idle with no audio input the JFET should be pinched-iff so source voltage at +5V dc or so, gate at 0V dc. With loud signal applied the gate voltage should climb up toward the 5V to turn on the JFET and increase attenuation.  In general you do not want to operate the JFET with the gate diode forward biased or a higher DC voltage than the source.

Are you getting good gate control voltage and still no gain reduction, even trying different JFETs??

JR

diydidi said:

What about this circuit??

Click to expand...

The only thing that could be argued against this circuit is that optocouplers have slower response than fets. It is just a very tiny click, the one situation I found it really annoying was with...headphones : But NSL32-sr3 is really quick otoh...
For simplicity I would decide for this one (if the attack is acceptable).

On a sidenote:
This is a circuit that I found in Teemu Kyttala´s ´Solid State Guitar Amplifiers´
page 257, which is a ´detectorless limiter´, basically a clipper, but with very soft response.

´Practically there is a thin line defining, which circuits exhibit clipping and which ones soft limiting, as both terms may essentially mean the same thing. The Figure 6.65 displays some examples of circuits commonly considered as “limiters”. The circuit a) has been used as a “compressor” in some Fender and Music Man amplifier models. Some Crate amplifiers utilize it as a distortion circuit. The circuit varies the voltage potential across collector-emitter junction according to drive signal at transistor’s base.  Basically, the circuit is nothing else but our well known “rubber diode” (VBE-257 multiplier) configured to control two anti-parallel “transistor diodes”. Conventional diodes are also fitted in series with the collectors to prevent reverse conduction. The circuit is usually seen in conjunction with a unity-gain inverting amplifier – however, it naturally works in the passive shunting form as well. When passive, the circuit is non-inverting – just like an ordinary diode clipper. Properly tweaked, the circuit can accept a huge range of input amplitudes without going into hard clipping.´

Click to expand...

L´Andratté said:

On a sidenote:
This is a circuit that I found in Teemu Kyttala´s ´Solid State Guitar Amplifiers´
page 257, which is a ´detectorless limiter´, basically a clipper, but with very soft response.

´Practically there is a thin line defining, which circuits exhibit clipping and which ones soft limiting, as both terms may essentially mean the same thing. The Figure 6.65 displays some examples of circuits commonly considered as “limiters”. The circuit a) has been used as a “compressor” in some Fender and Music Man amplifier models. Some Crate amplifiers utilize it as a distortion circuit. The circuit varies the voltage potential across collector-emitter junction according to drive signal at transistor’s base.  Basically, the circuit is nothing else but our well known “rubber diode” (VBE-257 multiplier) configured to control two anti-parallel “transistor diodes”. Conventional diodes are also fitted in series with the collectors to prevent reverse conduction. The circuit is usually seen in conjunction with a unity-gain inverting amplifier – however, it naturally works in the passive shunting form as well. When passive, the circuit is non-inverting – just like an ordinary diode clipper. Properly tweaked, the circuit can accept a huge range of input amplitudes without going into hard clipping.´

Click to expand...

Click to expand...

The two back to back transistors in the feedback path are configured as "Vbe multipliers". A Vbe mutliplier behaves like a zener diode in the forward direction.

How softly this clips may depend on the Hfe of the devices and resistor divider impedance (larger resistors=softer knee). IMO a clipper or "soft" clipper is not a limiter but perhaps in the world of guitar amp efx.

JR

> limiter for those incidents when the sound guy does something stupid and blows my in-ear-monitors.

So it should do "nothing" normally?

But should limit when HUGE signal hits the head/ear-phones?

And you don't care what the limiting-action sounds like? You just want to avoid damage?

And you don't really need "fast", relative to typical sound-squashers. 0.01 seconds of transmitter over-modulation is ugly all over town. Headphones can take huge power for a second, it takes several seconds for "loud" signals to burn them out.

I'm thinking what you might really want is a cut-out. 3 Volts for 1 Second, a relay drops-out, headphones disconnected. Probably a timer so that 2 seconds of not-huge signal lets the relay pick-up again. A fairly complicated thing to design, build, and calibrate.

> built something like this

Yes. For years I protected some costly (to buy and also to get off the ceiling) speakers with a few passive parts wrapped around the power amplifier. Below threshold it does "nothing". Above threshold it applies a quasi-RMS limiting with uncertain (depends on the LDR) time-constant. Sound gets a bit squishy mucky, and you don't want to work the system above limiting, which in this case was the idea. (I also had a "nice" limiter before this to take peaks off; this kludge was the back-up for idiot operators or gross accidents.)

Here's the simple form. Assumes headphone amp can stand an LED across its output (it really should, if it can drive low-Z phones).

LED across driver shines on LDR which works against resistor at amp input.

Any generic LDR will be fast enough to protect drivers. Release time is uncertain, but likely fast working against 5K-10K resistor (too fast for "nice" limiting, but you don't need/want "nice").

LDR resistance should come below a few K Ohms when you hold the LDR against a bright LED.

Limiting is 1.6V peak. This may be ample for the more sensitive in-ear monitoring plugs. If it won't go loud enough, try two LEDs in series for 3.2V peak. The extra LED may be mounted on the panel so you know when it's working.

Deluxe variations include back-to-back LEDs so both sides of the waveform are covered, or a FWB before the LED(s) for the same reason (but adds 1.2V of threshold), or a 100 Ohm pot so the limit level may be dialed-up well above 1.6V (I used that when working 75W speakers off 300W amps).