Limiter / Compressor

[thewisepranker]

I'm an analogue sort of guy - for me digits and audio don't mix too well (most CDs sound really horrible these days because they're all deliberately clipped).

(snip)

Most crucially of all - I do not need to use a clipper! There is one at the output of the board to prevent really gross overloads getting through, but the distortion will be so nasty that even the most deaf DJ will back off the levels!

Any chance of some schematics please?

[Albert H]

Any chance of some schematics please?

Certainly - when I'm back in the UK. I'm having a holiday somewhere nice and warm (helps with the aches and pains when you get to my extreme age!)

[thewisepranker]

Are you back in the UK yet?

[Albert H]

Today!

I'll dust down the main computer and locate that which you seek....

Word of warning - the circuit I used includes a pair of MN3007 delay line chips. These are about as common as a Croydon virgin, so don't moan if you can't get the bits. They are available on line, and there's a "second source" manufacturer of them - I'll look out the details.

[thewisepranker]

Wouldn't dream of it.

[wolfmanjack]

Im also interrested in your schematics Albert! It would be the last component in our little summer radio project (many thanks to 3m amateur for all help) my station also is a bit flat and not as loud compared with the commercial stations,im not interrested in contribute in loudness war,just a bit more "punch" in the sound. I am a toe tapping Linn/Naim audiophile and like that kind of sound ideal, i would like to have nice balance between phase rythm attack and timing, you know that sound that give you the feeling as if the music just rolls out of the loudspeakers.

I have a 15w chinese transmitter that gives me just all of the above, but i don't dare to use it since we looked at the signal on the spectrum analyser..

[thewisepranker]

Bump!

[yellowbeard]

Oh and by the way...

Code: Select all

http://www.donberg.ie/descript/m/mn_3007.htm

[Maximus]

I wouldn't want to be as loud as some of the big stations.
I'm appalled with sound of radio 1. They've got things cranked up way too high, to try and compete with the other stations which employ half deaf nutcases to tweak their processors.


Sent using Tapatalk

[bristolpirates]

Let's not put all the blame on radio processing, some of the blame needs to go on record labels who insist that mastering uses as much compression and limiting as possible.

When the tracks you are playing are already as compressed and limited as possible, you should make sure they aren't really hitting the radio compressors at all for them to not sound worse than they already do.

[Albert H]

Most modern CDs are clipped to hell. It's interesting to compare oscillograms of the original 1985 "Brothers In Arms" by Dire Straits (the biggest selling CD of the time, and credited with starting the CD revolution) with the 2012 re-release. The new one has almost no dynamic range at all - it's into clipping for most of the album. The thing that these mastering "engineers" really don't consider is that clipping IS distortion!

The distortion that's added by clipping is truly horrible and also adds extra upper frequencies to the sound, making it "brighter". Aphex is a process that does this sort of thing to a narrow band of frequencies (usually around 1 kHz is chosen) and is added to a mix to make vocals more prominent. I usually refuse to add deliberate distortion like this!

Many broadcast processors use crude diode clipping to prevent overshoots causing over-modulation. The conventional compressor / limiter has a finite attack time - sometimes into the milliseconds - during which time the over-level will be allowed through. Often, a basic compressor / limiter will over-compensate, pulling the level down too far, then recovering the gain so that the sound "pumps". This is really unpleasant to listen to!

Another problem with basic compressor / limiters is that the speed of reaction of the "side chain" (the level-sensing circuitry) can be fooled by low frequency musical content that you do actually want to transmit. The processor can (effectively) amplitude modulate the bass - adding another form of distortion.

The ONLY way to prevent these problems is to build a compressor / limiter that "sees the peaks coming". This involves the use of a delay line in the audio path. The level sensing is done before the delay, and the control is after the delay. This allows the effective reduction of the "attack time" to zero, allows the use of (relatively) slow-acting, but very accurate side chains, and can also prevent the low frequency distortion problem and overshoots.

The delay needs to be a few milliseconds - similar to the time taken for sound to travel a few metres from a loudspeaker to your ears, so is (in most cases) imperceptible.

There are three ways to achieve this delay:

The oldest method was to use a bank of "all pass" filters tuned to a range of centre frequencies across the audio band, and scaled to give the same delay at all frequencies. This is easily achieved in the bandwidth needed for medium wave, but more difficult for FM audio. It also demands the use of high precision components and is highly complex. If you want good quality results, you have to use expensive, high quality op-amps (and lots of them!). The earliest versions of this approach used L-C filters - using wound inductors!

The second approach is to use an analogue delay line - a "Bucket-Brigade Device". This uses a chain of FETs and capacitors (all built inside a chip) to pass little "samples" of the audio. The clock frequency and the number of "buckets" determines the length of the delay. Nyquist's Theorem tells us that the clock frequency has to be a minimum of twice the maximum modulating frequency. For FM, that means our lowest possible clock frequency (giving the longest delay) is 30kHz. I usually use 64 kHz or 72 kHz since either is easily derived from standard crystals. The IC that I use is the MN3007 (1024 buckets) which gives a delay of 8 ms at a 64 kHz clock.

The third method is to use a digital delay line. These are often complex, add distortion of their own and are probably overkill for the kind of thing we're trying to achieve.

Unfortunately, BBD ICs are getting rarer, but there's still a good supply of MN3007s.

I'll put circuit details up here when I've transcribed them from my notebooks and tidied up the diagrams that I have stored electronically.

[bristolpirates]

The thing that these mastering "engineers" really don't consider is that clipping IS distortion!

I wouldn't blame the engineers either. If they want work from the labels, they are pretty much forced to do this. It's the label bosses insistence that drives it in the music industry, and likewise the station bosses insistence in the radio industry. I'd agree that any engineer that does it by default without being told has come up in the age of the loudness wars and hasn't got a clue.

[Mike Volume]

I usually use 64 kHz or 72 kHz since either is easily derived from standard crystals. The IC that I use is the MN3007 (1024 buckets) which gives a delay of 8 ms at a 64 kHz clock.

I wouldn't have thought the frequency of operation would be particularly important here, and an MN3101 would be adequate? Can I ask why 64 or 72 kHz?

[Albert H]

Those frequencies are easily generated from standard crystals, give useful delay times, and are not related to 19, 38 or 57kHz!

[Analyser]

I was having a look at that BBD MN3007 BBD IC and to me the quoted distortion of 0.5% seemed quite high, so I've had quick look online to see if there was something more modern with better performance. I've found the PT2399 which may be of interest.

http://www.experimentalistsanonymous.co ... PT2399.pdf

[Albert H]

The PT2399 is very poor quality, and isn't capable of the short delays we need. The other problem is that the delay is determined by an analogue CCO (the current out of pin 6 determines the delay). It's not really possible to synchronise two of them for stereo. The 2399 was invented for cheap Karaoke machines, to add echo to the vocal channel. Guitar effects people like them (if they're not too discerning). The specifications for noise, frequency response and distortion in the datasheet are pure fiction!

The MN3007 quotes 0.5% distortion - that's the "untrimmed" figure. I get quite a lot better than that by precise adjustment of the bias. It's best adjusted on a 'scope, feeding in a nice 1 kHz sinewave. You turn up the level of the signal generator until the sinewave is clipped slightly. You then tweak the bias for symmetrical clipping. The delay line then has the maximum headroom it can have and minimum distortion - typically under 0.1%.

This is the delay line that was used in the legendary (and insanely expensive) Technics scratch remover. This worked by detecting the huge common-mode spikes from scratches, and turning off the audio (at the back end of the delay line) for a few microseconds. With careful adjustment, this would eliminate the worst of scratches!

You'll notice that the MN3007 has a good low noise floor, and the delay through it is very accurately determined by the clock speed. The original Panasonic specification claimed a maximum clock rate of 100kHz - if you want really short delays, you can go very much faster than that!

[Sietedj]

On the limiter Pira I have to say I have two installed and I do not like the sound they have. Change much, loses sharp and the sound dirty for my taste. I connected directly without the limiter and much better sound note. For my part I would not recommend.

[Albert H]

The basic Pira limiter is fairly crude, and has too slow an attack to prevent the clipper coming into action on every rapid peak. This will give distortion!.

Good news - I've almost finished transcribing the delay-line limiter from my notebooks. I also found a pretty nifty modification which allows the use of unmatched FETs in the gain control stages, using the BBD clock through a PWM circuit to act as the voltage-controlled attenuators. You have to cascade two of the circuits for best results, but the end product is remarkably good. It's probably much too complicated for "pirate" use (and I'd like to keep a commercial advantage for my products!) so I'll post the more basic processor here. The results from this are better than most commercial processors, and have very low noise and distortion. It also doesn't "pump" or "breathe".

[LeeCavanagh]

The basic Pira limiter is fairly crude, and has too slow an attack to prevent the clipper coming into action on every rapid peak. This will give distortion!.

Good news - I've almost finished transcribing the delay-line limiter from my notebooks. I also found a pretty nifty modification which allows the use of unmatched FETs in the gain control stages, using the BBD clock through a PWM circuit to act as the voltage-controlled attenuators. You have to cascade two of the circuits for best results, but the end product is remarkably good. It's probably much too complicated for "pirate" use (and I'd like to keep a commercial advantage for my products!) so I'll post the more basic processor here. The results from this are better than most commercial processors, and have very low noise and distortion. It also doesn't "pump" or "breathe".

I couldn’t find your basic compressor ‘ limiter when I tried the search, but as someone posted about it today perhaps you could point us in the right direction, or post again (which isn’t ideal I know), why can’t I find it in the search ....

[teckniqs]

viewtopic.php?f=11&t=1501&hilit=basic+limiter