A basic Limiter

[Shedbuilt]

I don't think it's likely to be because it's a 2N5458; rather than a 2N5459, but a few things spring to mind. Firstly, have you got the pin orientation correct. Secondly, is it definitely a genuine 2N5458 (and definitely a FET rather than a bipolar transistor). Beyond that, is the FET ok ?
It sounds likely to me, that the LED is being lit by DC, and as you mentioned that it's also not limiting, the issue could be around the FET; the LED driver op amp is DC coupled. DC could also be passed via one of the 1N4148s, from one half of OpAmp "B" (although as you said it's amplifying ok, presumably at least the upper / left hand half of "B" is ok, because that half is in the audio signal path.

[mike123]

Thank you very much Shedbuilt!
I will check these again and will put the results on the forum.

[rigmo]

hi guys.. question... is BA3308 so bad ALC?

[jvok]

Biggest issue I can see with the 3308 is it wants really low output voltage (40mV) to hit the 0.1% distortion spec. Which you can do easily enough but you pay the price of higher noise because you'll have to gain that level back up after.

All that said you end up running into similar issues with jfets or lm13700s too and they take a lot more tweaking to get right. So honestly for how simple the circuit is I'd probably put up with it. If you can actually get the chips anyway

[rigmo]

Biggest issue I can see with the 3308 is it wants really low output voltage (40mV) to hit the 0.1% distortion spec. Which you can do easily enough but you pay the price of higher noise because you'll have to gain that level back up after.

All that said you end up running into similar issues with jfets or lm13700s too and they take a lot more tweaking to get right. So honestly for how simple the circuit is I'd probably put up with it. If you can actually get the chips anyway

proposal, try with a faster rc-constant. Instead of 39o and 100uF, he would put 47o and 68uF. The difference would be about 15uS, which should correct irregularities faster and therefore less noise. No change in sensitivity. what you think pal?

[radium98]

https://www.youtube.com/shorts/sbOeuL_ifpk
that was a play with an old 1 year project. You can see the input with one led , and the output with the vu meter .

[rigmo]

https://www.youtube.com/shorts/sbOeuL_ifpk
that was a play with an old 1 year project. You can see the input with one led , and the output with the vu meter .

[rigmo]

Biggest issue I can see with the 3308 is it wants really low output voltage (40mV) to hit the 0.1% distortion spec. Which you can do easily enough but you pay the price of higher noise because you'll have to gain that level back up after.

All that said you end up running into similar issues with jfets or lm13700s too and they take a lot more tweaking to get right. So honestly for how simple the circuit is I'd probably put up with it. If you can actually get the chips anyway

What is story with LA3220 seams to be much better?

[Albert H]

To be honest, none of these "single chip" solutions is particularly good. You have to consider their original use - as recording level controllers for cassette recorders! Considering that the S/N ratio of a cassette was always abysmal, and that even the best quality recorders (Nakamichi and the like) couldn't do better than about 2.5% distortion, those compressor ICs were "good enough" for that kind of use. However, for broadcast purposes, you need to do a bit better:

If you use the simple single band limiter (based on the original Philips data sheet) using an NE571 and an LM339, you'll get around 0.4% distortion when limiting hard. In the usual run of things, you won't hear the distortion. I'll be happy to put up the circuit diagram for the variant on that circuit that I came up with - it includes pre-emphasis and an "overshoot" clipper, which completely eliminates overmodulation. Its operation only becomes audibly noticeable if the levels are really pushed, and it's crushing the dynamic range completely. Unfortunately, we all know DJs who have no idea about minding their levels.....

The next kind of limiter that I'd consider using involves the use of "transconductance" amplifiers - I use an LM13600 (a dual channel device) - but doesn't put the TCA in the audio path. The 13600 is used as a variable negative feedback arm around a high quality op-amp. If over-level is detected, the gain of the 13600 increases, giving extra negative feedback to the audio op-amp, and thus reducing its gain. Done right, this is a viable limiter, and I have a couple of designs that rely on this principle. The quality from this configuration is pretty good, and noise and distortion are both very low. Again I include pre-emphasis before the processor, and it really does sound very good. You will have heard any number of London stations using this kind of limiter over the last 35 years - I made and sold a hell of a lot of them!

The last limiter type that I'll describe today uses specialised audio attenuator ICs, and tends to be rather more expensive than the others, but sounds really good. The attenuator ICs that I like are made by the "THAT Corporation" and are the highest quality parts that are available these days. Their "THAT2018" is superb, and I've used them in commercial products for some years. They also do "Analog Engine" ICs that include VCA, RMS level detection and a couple of uncommitted op-amps, and these can make for relatively simple high quality compressors and limiters. They also publish copious design notes, making use of their ICs quite easy. If you visit https://www.thatcorp.com/design-resources/ you'll find plenty of design ideas and even a couple of quite simple circuits that you can build quite easily.

Remember - FM radio is supposed to be a quality medium, and done right can sound amazingly good. Don't compromise your audio quality by using cheap and nasty audio gear!

[Albert H]

A bit more info....

These days, most stations get programmes delivered to them by various means, from DJs working in their own home studios. This tends to mean that programme average levels can vary by quite a lot. It's really quite tricky to get everyone to record with the same levels and the same frequency response!

To cope with this, I tend to use a fairly complex airchain:

The first stage is a slow-acting AGC. This is used to "average" the programme, and basically brings all the sources into line level-wise. I also include a low-level expander circuit (the opposite of a compressor) with a low threshold - if the incoming audio drops below a pre-set level, the expander assumes that it's silence, and mutes the audio path, eliminating any background noise, and preventing the compressor stage from bringing up the low-level background noise in the mistaken belief that it's programme! You may have heard this effect on some commercial radio stations - on loss of a source, the background hiss swells up until it's quite loud!

The next stage is a three-band compressor. I use split-band processing so that loud bass sounds don't cause the midrange and top to "dip". You'll hear this effect on many badly produced records (though some numpty producers claim that it's "an effect") - bass drum beats can cause the other instruments and the vocal tracks to "dip" rhythmically. This sounds truly horrible and is just an example of studio ineptitude!
I find that the compression ratio usually needs to be around 2.5 - 4 : 1, and as long as it isn't hammered, it's reasonably transparent in use. This will shave off the worst excesses of the DJs before we get to the final stages:

I then use a split-band limiter. Again I want to separate the bass end from everything else, and the limiter doesn't do anything until you reach around PPM 5.4. After that point, the limiter kicks in with a ratio of over 25 : 1, and nothing "over the top" gets past it. Some jurisdictions insist on a final clipper to nail any overshoots, but in most instances - unless you have a real idiot at the controls (I'm looking at you, Peter), I don't ever get any overdeviation even without the clipper.

I don't agree with the Orban approach, by the way, introducing clippers all over their circuits. Granted they never allow over levels to leave their airchains, but they sound truly horrible!

My commercial designs use PWM control of audio levels. This basically consists of an electronic switch shorting the audio path to ground. The switch is operated at a high ultrasonic frequency, with a variable mark / space ratio. The switch is followed by an audio lowpass filter to prevent the switching getting through, but it's a nifty way of controlling levels, and channel-to-channel matching doesn't rely on specially selected components, making construction cheaper and quicker (it also eliminates a lot of the calibration requirement).

So - as you can see - getting a good-sounding FM signal isn't trivial!

[jvok]

Huh I had no idea 571s were still in production. SOIC only but still.

I have a limiter design that uses two lm13700s. One is a wideband limiter and the other does a separate stage of limiting on just the preemphasis, so you don't get overshoots OR pumping on loud treble, which is the flaw with single limiter designs. I'll post the schematic and gerbers when I get round to fixing the layout mistakes.

I found the trick with post-clippers is to set the limiter threshold 3db below the clip point. That way you don't get any audible clipping even when really overcooking the levels. You loose a little bit of loudness but its barely noticeable and worth it imo to avoid the clipping.

[Albert H]

Jvok, I'd be interested in seeing your approach. You're entirely right about the limiting and clipping thresholds. To be frank, I've never wanted to squeeze the last dB out of my airchains - over-processed audio (like "Heart" and some of the other commercial offerings in the UK) is really fatiguing to listen to, and tends to make listeners tune away quite quickly (well - that and their crap content!).

Back when I worked in the USA in the early 80s, there was a "loudness war" going on between broadcasters, and audio processors with controls like "Density" became the norm. Many of these stations sounded really horrible, and when our little mid-market station changed the way it sounded so that there was an emphasis on quality rather than "loudness", we won a huge number of listeners from other, "louder" stations!

[sinus trouble]

These days, most stations get programmes delivered to them by various means, from DJs working in their own home studios. This tends to mean that programme average levels can vary by quite a lot. It's really quite tricky to get everyone to record with the same levels and the same frequency response!

Albert is correct! However, I still believe there is NO perfect "limiter"

As a DJ myself for many years, Talent is often dismissed by most people! (Anyone can play records! Blah Blah!)

From my experience! DJing live for a crowd is no different to radio, Yet i have no "Fancy" Audio processing going on!

An experienced DJ selects the right tunes, beat matches and most of all! Keeps the levels in check simultaneously!

An experienced DJ also instinctively knows how to put on a show!

Maybe some people should look at thier presentation skills instead of a quick automated solution!

[Albert H]

Sinus - you're right that there's no substitute for talent, but no amount of mixing ability (and "golden ears") is going to keep modulation right all the time.

I'm NOT suggesting that compression and limiting are "an aid for the technically incompetent" - they are an essential to prevent over modulation. No professional or commercial station would ever be licenced without rigourous, unburstable limiting - on either FM or AM.

You could (I suppose) just clip the peaks of the audio with a couple of diodes to ground, and this would prevent over-modulation. It would also sound truly horrible!

[sinus trouble]

Sinus - you're right that there's no substitute for talent, but no amount of mixing ability (and "golden ears") is going to keep modulation right all the time.

I'm NOT suggesting that compression and limiting are "an aid for the technically incompetent" - they are an essential to prevent over modulation. No professional or commercial station would ever be licenced without rigourous, unburstable limiting - on either FM or AM.

You could (I suppose) just clip the peaks of the audio with a couple of diodes to ground, and this would prevent over-modulation. It would also sound truly horrible!

I completely agree!

It will always be a compromise between DJ competence and technology working together!

Lets look at the Limiter as a "Safety Net" Not a solution to poor broadcasting!

Detailed preparation of your audio files and vinyl etc... Will produce a better listening experience!

[Albert H]

Completely agreed!

[mike123]

Hi,

I build this limiter a while ago but have some questions.
If I'm informed correctly the 3k3 resistor (at the right site of the schematic) is for the release time?
I experimented with this resistor and increased the value to about 4k1.
I would like to change attack and release time.

I also disabled the two 1N4148 diodes because it's clipping way to much.
It might be they kick in too early, before the circuit starts limiting?

Also is there a way is disable the limiter function without bridging the circuit?
I would like to switch it on/off to check how much is actually limiting.

The circuit is also very picky which FET is used. I tried a 2N5458 because the 2N5459 was not available, but it did not work very well.

[Albert H]

The 3k3 resistor that you've ringed is part of the LED indicator circuit. It sets the threshold at which the LED lights. Attack time is determined by the 4k7 connected to the junction of the the two diodes. Release time is set by the 3M9 resistor. With hindsight, I'd make the 3M9 resistor a bit lower - 2M2 or 2M7. The 4k7 resistor in the source of the FET can be increased to as high as 10 k, and reduces to as low as 1k - it adjusts the threshold somewhat. You could make that resistor a 10k preset, and adjust until it "sounds right".

If you want to bypass the limiter part (you'll still need the pre-emphasis and the filter, switch your output to the output of the second op-amp. You'll need another level control (because the level there will be different to the output of the limiter section....

Like this:

AH_Mono_Processor - with limiter bypass.png

Remember - all audio processing is a compromise.

The very best broadcast limiters (better than Orban etc) all use delay lines in the audio path. The sidechain (the level sensing circuit) monitors the input of the limiter, but the gain cell is after the delay line. The delay is usually 300 - 500 µs, and is enough to eliminate the "attack time" of the limiter. Done right, this leads to an amazingly transparent limiter that doesn't produce any of the nasty audible artifacts of a simpler limiter. My current commercial limiter design uses PWM gain control, three bands, and a 320µs delay in the audio path, which is inaudible - it's just like being a foot or so further from the speakers! I have managed to completely eliminate the overshoot clipper , which reduces distortion still further. There's even LED bars indicating input levels, gain reduction in each band, and output levels!

[mike123]

Thanks for the extra info!
I will try to tweak for the setting I feel are the best for me.

I'm very interested in audio processing. Specially the analog versions and I don't mind single band with that pumping sound

I builld the pira compressor years ago (did not like it very much): https://pira.cz/hlimste.htm

Tried the much used 80's compressor which was in the Elektuur (see page 12.24):
https://www.piratenpraten.nl/schemas/El ... YQBp6Fow58

Did you already spot this 6-band compressor (price does seem low for a multi-band)? :
https://www.ebay.nl/itm/134194051392?ha ... R-y1tdKSYQ

Do you sell your processing?

[Albert H]

Hi Mike

I design processors for an American broadcast equipment manufacturer. I sometimes build limiters and other processors either for my own stations or for friend's recording studios. I don't build gear for "private" broadcasting any more!

That three-band stereo limiter for <€170 looks good, but the attack time is described as "2ms" - this is far too slow for broadcast use (my gear goes into limiting in around 200µs - 10 times faster than the one advertised. I'd be interested in the approach that the designer has used.

I experimented with various filters and eventually came up with a "phase linear" circuit that had adjustable turnover points. This then drives three stereo limiters (using transconductance amps in the negative feedback loops of high quality op-amps), and the resultant signals are combined in a simple mixer stage. I added lots of blinking LEDs - input levels, output levels and gain reduction - because the users like blinking LEDs!

Audio processing is often just a matter of taste - when you've overcome the over-modulation problem, you've got to develop your station's "sound"!