PIRA Simple Limiter

[Albert H]

Sinus does this amplify the weak signal and attenuate really the high one , i built one 10 years ago , but did not work , if you can make a little video later and pass on youtube channel .

Yes! It gives quite a large audible boost to the quiet sections, This can also be seen in my measurements!

At an input of 250mV the gain is nearly double! (500mV)

That's NOT a limiter - it's a compressor!

A Limiter has no effect on the audio passing through it until a level threshold is reached, whereupon the "sidechain" (the part of the circuit that derives a DC voltage proportional to the level of the audio) drives the attenuator into gain reduction to try to keep the level no louder than the threshold level. The reasons for using compressor and limiters differ:

A compressor is used to reduce the dynamic range of the audio passing through it - quiet parts are made louder, and loud parts are made softer - it's a sort of automatic volume control, and is really useful when the audio you're trying to broadcast is of wildly differing level. It's quite common to have a slow-acting "AGC" (Automatic Gain Control) at the beginning of a broadcast airchain to try to reduce the effects of inept audio production!

A Limiter is used to prevent over-modulation. A transmitter will have a maximum amount of audio that can be applied, and so a limiter is used to prevent this critical level ever being passed. This prevents interference to adjacent frequency stations (on FM) by keeping the maximum deviation to 75kHz. Any more deviation causes interference, and will also give gross distortion in the receiver.....

Unfortunately Limiters aren't perfect, and very fast transients can sneak through, causing over-deviation. This is handled (in most cases) by applying clipping to the audio. Clipping is (generally) horrible, because it's adding distortion! The better way of preventing transient problems is to introduce a delay-line into the audio path. Less than a millisecond of delay can be enough - just enough time for the level sensing sidechain to react! The level sensing is done at the beginning of the delay line, and the attenuator is at the output of the delay line....

[Albert H]

Here's a cheap and simple limiter and clipper:

limiter++.png

You can easily add pre-emphasis by putting two 1n capacitors in series across the 100k input resistors!

[jvok]

I have to say that 5mS sounds short?? But is still recognisable by the ear! The Limiter kicking in can be physically heard!

I wonder if some of what you're hearing is the DC offset shift when the limiter kicks in. It looks pretty bad in that waveform you posted. This is a problem you often get with fet limiters, and rod went to a lot of effort to avoid it in his design.

This is handled (in most cases) by applying clipping to the audio. Clipping is (generally) horrible, because it's adding distortion!

I've spent a lot of time experimenting with this and found that the clipping is barely noticeable as long as you set the clipper a couple of db above the limiter steady state. You lose some loudness but a couple of db isn't worth worrying about imo. Of course a delay line is better but a lot of people don't want the extra cost, especially now the digital delay ics are all obsolete. Or you could do the whole thing in software for free I guess

[jvok]

The limiter on the ESP page is pretty good - fast attack, reasonably low distortion, and an adaptable decay (recovery) time. I've used versions of this type of circuit for years. The only real problem you're likely to encounter is matching FETs for a stereo pair, and commoning the sidechains so that the stereo image remains consistent when the gain reduction kicks in.

The other issue with a simple limiter like this is that a big bass thump will dip out the midrange and treble. This will make that audio "pump" - really not too good with most modern forms of music!

The "pumping" can be mitigated by splitting the audio into bass and the rest, and separately processing them - you're going to need four matched FETs! The outputs of the bass and mid / top limiters (for each channel) can be combined after processing, and the pumping problems will be largely eliminated.

One thing I found works well is giving the preemphasis its own limiter after the main limiter. I used a voltage controlled filter driven by its own side chain which rolls off a variable amount of the preemphasis. At full limiting the preemphasis is fully cancelled out but it never cuts any further than that.

The signal path is main limiter -> preemphasis -> variable deemphasis. I got the idea from an old bbc research paper from the 70s and it works really well. Having the preemphasis after the main limiter stops it from being over sensitive to treble. Pumps a lot less although still not quite as good as a multi band limiter.

[Albert H]

That's a trick I've tried - the results are rather underwhelming when compared to (even) two-band limiting! I partially agree with you about clipping - with the threshold set sensibly, and with some distortion cancellation trickery, it can be (subjectively) OK. My clipper uses 2N7000 FETs in the negative feedback path around an NE5532 op-amp - varying the bias to the FET gates can make the threshold "track" a couple of dB, reducing the severity of the clipping distortion.

One processor I designed way back in the 80s "loudness wars" stripped off everything below 160Hz, clipped that bass end severely, and then low-pass filtered the harmonic crud that the clipper generated. It was pretty good for "pop" music, and we ended up with the "loudest" thing on the band! The only real downside was that it was really fatiguing to listen to for any length of time!

[jvok]

I'd like to try multi band limiting combined with variable deemphasis at some point. Seems to me that the preemphasis should still be processed separately to the limiter even if you have multiband.

My biggest issue with multiband is the parts count (and cost!) quickly gets out of hand when you start adding multiple limiters, crossovers etc, especially in stereo. Fair enough if you're a big budget commercial station but as a pirate the cost is everything. And something seems wrong when your diy project costs almost as much as something by innovonics. I guess stereotool & co are the answer but doesn't leave much for the likes of us to play around with.

My clipper uses 2N7000 FETs in the negative feedback path around an NE5532 op-amp - varying the bias to the FET gates can make the threshold "track" a couple of dB, reducing the severity of the clipping distortion.

I have a design (which I'll post some day...) where the clipping points set by a pot. If you turn it all the way up til there's obvious distortion then back it off a bit it cleans up very quickly. I was surprised by just how much you can get away with but maybe I just have lead ears. The clipper itself is nothing special, but I used diode connected bjts because they clip much sharper. I found normal diodes were too soft which just means the distortion creeps in way before you hit the final limit. A hard clip gives you back some headroom.

[radium98]

can you please post a clearer image close for the corrected version , i would like to see if mine is correct , if no i will correct by vision from the one you modify , thanks Sinus as always

[sinus trouble]

No worries Radium!

My version is identical to the PIRA but i used some electrolytic capacitors instead of non polarised!

I will post more pictures soon!

[radium98]

thank you as always , also me i put a pcb soon and will see what is happened , and what missed , i did it a very long time .but i like to see the videos you post Sinus

[radium98]

Sinus please when can you post the more clearer pictures please ?

[radium98]

that is mine . What is wrong for not working . Note that the amplifier stage lm386 is working .

[radium98]

Any chance to help , anyone have a look ?

[radium98]

I have found the error. Now I do have audio
But the level is too slow I mean the attack time .so what is the best value to make it fast for 10mohms and 10uf capacitor ?

[tvtv]

The limiter on the ESP page is pretty good - fast attack, reasonably low distortion, and an adaptable decay (recovery) time. I've used versions of this type of circuit for years. The only real problem you're likely to encounter is matching FETs for a stereo pair, and commoning the sidechains so that the stereo image remains consistent when the gain reduction kicks in.

The other issue with a simple limiter like this is that a big bass thump will dip out the midrange and treble. This will make that audio "pump" - really not too good with most modern forms of music!

The "pumping" can be mitigated by splitting the audio into bass and the rest, and separately processing them - you're going to need four matched FETs! The outputs of the bass and mid / top limiters (for each channel) can be combined after processing, and the pumping problems will be largely eliminated.

Here's one way of doing it.....
Split-Band Limiter.png

I'm thinking a doing a PCB for this circuit, with a few additional options provided on the board: -

• Pre-emphasis at 50µs or 75µs on the way in,
• An option to introduce delay lines to the audio paths to allow "zero attack time" processing,
• Audio filtering to get rid of content above 15 kHz,
• A post-process clipping option.

The 15 kHz filters and the delay lines would be on additional plug-in boards, and I've designed a very high quality stereo coder and an optional RDS board to work nicely with this unit. I also planned to add LED bargraphs to show "gain reduction" (the uninitiated love blinking lights), using more LM339s, and even a silence alarm.....

The whole package would be functionally equivalent to something like the Inovonics "David" processor / coders, and would sound a whole lot better than most UK commercial radio ststions!

do you have a clearer image to shar4, its not clear what some of the ic's are, even those which do have numbers on them , is this intetional? also, the ines that are blank which are 74 and which are 72 or 71v?

[yellowbeard]

The chips are op-amps TL072, TL074, LM339 and compander NE571. Opening the image in a new tab and reading the text solves that problem you have...
download/file.php?id=2594

[tvtv]

The chips are op-amps TL072, TL074, LM339 and compander NE571. Opening the image in a new tab and reading the text solves that problem you have...
download/file.php?id=2594

Thanks, I’ve tried viewing the file but if you had not said LM339 there would be know way I could make that out, how do I know which of the blank opamps are which from the list? Sorry to be a pain, but I trying to create this schematic so I can test it

[yellowbeard]

There are 13 TL type op amps, if I were building it I would use TL074's to keep the board size and chip count down. If you are just running a simulation then use whatever you like, they are the same op amp for most purposes - the different last number just refers to how many op amps are onboard each chip. Have fun now!

[tvtv]

Thanks