Passive soft limiters

[Guest]

Hey guys, I've found that I'm in need of some soft limiters for a/d overload protection, just to catch a few transients here and there. I found the discussion on the pris_m device, but for some reason I cant view the last page of discussion...
At any rate, my question is what formula do I need to use to choose the resistor value to set the knee of the limiter? Oh, and I am going to do a simple back to back zener pair. Sorry if this is a silly question, I feel as if the answer is dangling in front of me and I just can't grasp it.

Thanks.

Jason

 

[bcarso]

Need to know the impedance of the source going into the clipper and the impedance of the front end for the A/D to make an intelligent tradeoff.

Also, although I got some argument on the issue, IIRC, I would favor a zener in the middle of a ~1N4148 bridge over two zeners, unless you want to buy a batch of the latter and match them for breakdown voltage and forward voltage. The batch of 4148's I tried when I was challenged had a bimodal distribution of forward voltages but within each population they were really close. YMMV....

 

[tv]

some schottkys as a bridge?

 

[Guest]

Thanks for the info. Brad... I'll look into the 1N418. Since impedance varies somewhat over the frequency spectrum, would one measure impedance at different frequencies and average the results?

Also, what do you think about this?
http://www.diodes.com/datasheets/ds30108.pdf

I don't know anything about the manufactuing process, but I am guessing that the matching would be closer in something like this than in two seperate diodes...

Thanks for the help.

 

[magicchord]

I think LEDs work well for that kind of thing.

 

[AMZ-FX]

Look at the warp controls on my web site at:

http://www.muzique.com/lab/main.htm

regards, Jack

ps: The bandwidth limit will be reset soon...

 

[bcarso]

[quote author="JCMaudio"]Thanks for the info. Brad... I'll look into the 1N418. Since impedance varies somewhat over the frequency spectrum, would one measure impedance at different frequencies and average the results?

Also, what do you think about this?
http://www.diodes.com/datasheets/ds30108.pdf

I don't know anything about the manufactuing process, but I am guessing that the matching would be closer in something like this than in two seperate diodes...

Thanks for the help.[/quote]

If the zeners in the Diodes Inc. parts are fabricated monolithically they might have good matching, although nothing seems to be guaranteed. At least they will track in temperature.

With that in mind, DI and Philips both make two-chip dual transistors now that are plucked off of adjacent positions of the wafer. It's possible their emitter-base zenering is well-matched, although the nominal 6 volts or so is probably too high for this application without gain in front, which obviates the passive approach.

 

[Moby]

I found the discussion on the pris_m device, but for some reason I cant view the last page of discussion...

Same by me.. :? Is there some real conclusion about? I heard really nice comments about Pris*m clipper thing, so will be nice to know what's really inside :grin:

 

[NewYorkDave]

Is there some real conclusion about?

Here's my conclusion:

Don't run your levels so close to maximum and you won't have to stick a fuzzbox on your signal to prevent "overs." Set your average level at -18 or -20dB FS, not at -6 like many misguided people seem to do. You can set 'em even lower than that for material with a high peak-to-average ratio, such as percussion.

The thing many people just don't seem to get is that, unlike with tape, running the level very hot gains us nothing in digital recording. If you want your peaks to hit, say, -0.1dB FS, you can always bring them up after the fact. But you can't replace bits that are clipped off during recording.

I think many people, whether they grew up with tape or came along later, are trying to apply the old VU-meter "keep it close to zero" mentality to digital recording and it just doesn't work that way. 0dB= full scale makes sense from a purely technical perspective but intuitively it just doesn't feel right when you're used to monitoring on VU meters. I grew up with tape and VU meters and even though I work with audio equipment every day, I still have to make a mental readjustment whenever I'm setting levels on digital gear. -20dB FS = 0VU (in my facility, anyway).

It's funny... I remember when digital recording technology was first being sold to the public years ago; the biggest selling point was its dynamic range. And here we are, years later, and it seems most recordists are bound and determined to reduce dynamic range as much as they can!

 

[tv]

the main point was to get hotter signal, I guess, hence mentioning of converter driving. Clippers preserve punch more than limiters, at least in 4_4 kick genres ...

I guess the most effective way for a clipper element would be making a schottky bridge and putting a zener in it. At least I'd do it this way. Depending on how it's driven (transformer, transformerless), resistor(s) would need to be calculated accordingly.

 

[NewYorkDave]

the main point was to get hotter signal

Yes. But my point was, before adding clippers and so on to the signal path, one should ask himself: why do I need hotter levels? If you remove the "because that's what other people do" element from the equation, the answer might be surprising.

 

[Greg]

With lower quantization levels, it was necessary to get the signal as "hot" as possible to reduce quantization noise, but at 24-bit, quantization noise really isn't an issue as I see it. So I agree with Dave on this, that's it's not really necessary to get really hot signals into the digital recorder nowadays.

 

[soundguy]

[quote author="NewYorkDave"]Set your average level at -18 or -20dB FS, not at -6 like many misguided people seem to do. You can set 'em even lower than that for material with a high peak-to-average ratio, such as percussion.[/quote]

I think the simple fact to address is that if you are setting your system up hotter than -12db FS you are on thin ice, if your average level is at -6 FS youve really got to re evaulate your gain structure. How can you have that high of a level and not have your entire mix sitting in a limiter?

dave

 

[tv]

"because that's what other people do"

flattening sound for multimedia is even worse, it has to cut trough tiny computer speakers as well as pa-s for presentations etc.

therefore I mentioned schottky bridge...
I read some guitar stompbox diy stuff ... they often mention using mosfets wired as diodes for *softer* clipping - iirc it was mentioned somewhere that clipped waweforms aren't clipped flat as diode/led/zener clippers do, but retains waveform, although *scaled down*, so it's like clipped signal retains some *wiggles*. this would be a cool thing to test, except mosfets would need to be quite perfectly matched for stereo use which means - some work and a bag of moses. AND cheap irf510 is 50x more expensive than some diodes I can get ... math time.

I planned to breadbord something to hook up to my soundcard but I'm on a deadline so only blah-blah for now.

plus irfs could be mounted on sexy little winged heatsinks and with 2W or 5W resistors such clipper would look good. AND it would fit nicely between two aclass followers ... on heatsinks with 5W resistors . AND it would look even better if preceeded with pultec-ish bass controls with mustage (on smaller sinks) with 2W resistors. hmmm ....

 

[bcarso]

If you're going to go to that much trouble I would just purpose-design a programmable-slope/breakpoint kind of symmetrical memoryless transfer function. Use decently low impedances but don't burn up power in big buffers and power MOSFETs---for one thing, the signal-induced self-heating shifts will do strange things to the characteristic that no amount of heatsinking will eliminate.

The uC types could play games with memory lookup tables and multiplying DACs as one approach.

 

[tv]

I surely didn't mean to heat clipppers - I said it was for *the looks*

Nevertheless, I suppose clippers should be driven with low impendance and low enough resistor - there's bridge capacitance and irfs I mentioned are cca. 180pf Ciss.

But aclass drivers and followers - these should heat a little ...