Hi All...
I'm thinking about making multiband compressors / peak limiters with seperate bands for...
100-300Hz, 300-1kHz, 1-3kHz, and 3-10kHz.. Kind of a very much simplified old-type optimod. with only 4-bands.

So...
attenuator / amplifier -> BPF -> 4 o/p audio splitter -> 4xfilter/compressor/limiter sections -> summing amp -> O/P

Has anyone else made anything like this? There are plenty of simple limiters out there all are a big compromise and most make the sound seem nasty and obviously very compressed.

Any Ideas for a professional sound...? The idea is to make it a cheap disposable unit, one per TX

sterotool...and expensive sound card can process setup from stereo tool...

You don't need an expensive sound card. £30 usb one will do
Or raspberry pi

Sam. Your band splits are quite sensible, but in practice I've found that two bands can work well (with the split at ~300Hz).

The problem that you'll have is twofold - the attack time (necessarily a compromise - see later), and the overshoot problem. If you want absolute protection from over-modulation, the attack time of the limiter must be incredibly fast. The problem with this is that bass sounds (in particular) contain plenty of energy compared to the rest of the audio spectrum, and will tend to dominate the efforts of the level-sensing sidechain - hence the move to split-band limiting. The problem remains - if you have really fast attack limiting, your bass end can "self modulate" so that individual bass tones get wildly distorted. The problem is similar in the higher frequency ranges, because really fast attack leads to distortion. Orban tried to overcome this by using "soft" clippers and lowpass filters to get rid of the distortion artefacts.

IT DOESN'T WORK!

The Orban Optimod sounds truly horrible no matter how it's set up.

The only effective way to overcome the attack-time problem (and avoid overshoots that knock "holes" in your audio) is to use a delay-line. The incoming audio is level-sensed at the input end, and then passed through a delay line (I use ~1ms). The gain controller (VCA or whatever) is after the delay line. The limiter can "see the peaks coming" and pull the gain down in time. This effectively gives you a limiter with zero attack time.

There's a lovely BBC Designs Department paper on the design of a delay-line based limiter - I think I've got a digitised copy, and will post it here for the interested folks when I find it.

If done properly, a delay-line limiter can be a single-band job. I found that I got the best results with a two-way split (using a Linkwitz Filter configuration), and two differing recovery time constants for the bass and the rest. The delay line I use is an overclocked MN3007 charge-coupled delay line (in each channel of the stereo pair) clocked from a crystal oscillator. The attack time of each of the sidechains is also 1ms, so the net result is a zero attack time. The audio bandwidth of the delay line is certainly good enough for FM broadcasting, and the distortion and noise are low enough to make this a useful approach.

The use of the delay line completely obviates the need for any clipper!

Incidentally - Stereotool can be set up to sound OK, but it's not anywhere close to a good analogue limiter!

I was planning on using the NE570 or 571 (as I have some of them) as the comp/expander at the start and end of the chain (2:1 comp just after the input transformer, and 2:1 exp before the O/P amp/transformer. (with the filters/channel-co/compressors and peak-limiters between them). So more like this....
--
transformer
attenuator/amp/LED-indication
Band-pass filtering
2:1 compression
audio-splitter 1-in 4-out
4x filter-sections
4x channel-compression/peak-limiting/LED indication
4-input/1 output mix
2:1 expander
output amp and transformer
--
(which looks like total overkill)

The delay lines I have here are CCD types, I'll dig them out and look-up the specs.
The splits are an attempt at reducing the pumping effect of the limiter, and I was also toying with the idea of using a fast mild comp followed by a slow deep comp so two stages of comp per filtered-channel, But using one comp stage and a delay like may well be a better option.

The CCD delay lines I have were originaly bought for limiters on Mic's here and allowed the compressor VCA to reduce it's gain a fraction of a second before the audio gets to it... Is that the method you're suggesting Albert?

I know many people use soundcards and other software based limiters, But as my application needs the compressors at the transmitter rather than in the studio a hardware-solution is the better option in this case

Cheers....
S-T-D

The transmitters are AM and SSB so the requirements are a little differeent to those for an FM TX. For one the audio-bandwidth I'm looking at is very narrow, and heavy bass causes the mod-amplifiers to draw shedloads of current on AM !

Hi!

I made a 3 band audio compressor for AM bandwidths many moons ago using the MC3340P as the VCA (old but easy to use device). With a bit of tweaking of the low/mid/high levels and attack/decay times it sounded rather nice. I found that there was no point broadcasting any audio lower than 50 Hz on AM as most AM radios don't have speakers that work at those low frequencies. In reality, a high pass cut-off nearer 100 Hz makes hardly any difference to most AM listeners other than those listening through a hi-fi, even in cars it doesn't make that much difference.

In this design, the cross-over is done before and after each VCA. This means that any noise that the VCA's produce will only be heard in the band they are operating on. It also produces a much flatter frequency response.

A bit of HF boost before the clipper can be useful, as any of the harmonic distortion caused is gotten rid of by the final low pass filter. The LPF in this design has a cut-off frequency of 6500 Hz, and is 40 dB down by 9000 Hz which was what was permitted for LPAM in the UK at the time. There's also a phase compensator on the output of the LPF which gets rid of most of the overshoots caused by ringing in the final LPF and gives about another couple of dB of loudness before overmodulating the TX.

I know there is one of these devices in a garage somewhere to the west of London, connected to an AM transmitter, but I haven't seen it in years.

Having said all that, I'd still go for Stereotool (or similar) using a Raspberry Pi. If you're only using it for AM, then you won't even need a special sound card.

Rev

I made also 4 band processor with NE571
First a input stage to control the input with a VCA then a NE571 as a Leveler then a other VCA to contol the levels to the 4 band crossover and the 4 times NE571 after that a som aplifier with a adjustable multiband clipper and that go to a 15KHz lowpass with adjustable loudness clipper
And as option a DSP stereo/rds encoder
A picture of the prototype and will post a picture when it's made in the 19"rack unit with all the ledbars

I have more MC3340 devices than NE570s..
As folks here have used both devices in your projects before...
In your opinion which are the better to use in this application..

I'm tempted to use 3340s as I have more of them, BUT NE570s as I've used them before....

Cheers// S-T-D

Have a look at the automatic level control chips used in tape recorders. The BA3312N limits nicely and it can be made to sound good when tweaking the time constants. I'd use a simple passive RC crossover at the input and a summing op-amp at the output with a couple of diodes for ultimate clipping of the signal.

spent an hour or so yesterday going through the stock of components here, have loads of Op-Amps etc..
While searching for CCD-Delay lines I found a batch of 12 Mitel DTMF decoders, so I will find a suitable project for them in the next few months....

Looks like this first-version is going to be based on op-amps and a handful of MC3340s. so active-filtering and 4 frequency-channels. Later versions may use delay-lines (when I find them amongst my spares!).

Below is the process. It's a bit chaotic. Graph-paper and ideas, What can I find in my spare-parts stock, and then design a circuit-board...

RF-Head wrote: ↑Sun Sep 05, 2021 12:00 pm I made also 4 band processor with NE571
First a input stage to control the input with a VCA then a NE571 as a Leveler then a other VCA to contol the levels to the 4 band crossover and the 4 times NE571 after that a som aplifier with a adjustable multiband clipper and that go to a 15KHz lowpass with adjustable loudness clipper
And as option a DSP stereo/rds encoder
A picture of the prototype and will post a picture when it's made in the 19"rack unit with all the ledbars

Really nice work

SamTheDog wrote: ↑Sun Sep 05, 2021 5:04 pm I have more MC3340 devices than NE570s..
As folks here have used both devices in your projects before...
In your opinion which are the better to use in this application..

I'm tempted to use 3340s as I have more of them, BUT NE570s as I've used them before....

Cheers// S-T-D

For AM, the 3340 is fine. The difference between the ICs is that the 3340 is just the VCA part, whereas the NE570/1 has two gain cells and two rectifier sections. The 570/1 also has distortion compensation, but the internal op-amps in the chip are horrible (about as good as a 741). You have to remember the uses that these ICs were designed for - the 570 & 571 were designed for telephone trunk companding, and the 3340 was designed as a cheap and simple solution for a television remote volume control! The MC3340 is just an 8-pin packaged version of the ancient MFC6040 which came in a bizarre 6-pin package. You'll notice that two of the pins on the '3340 are not connected!

The 570 was a great IC for guitar effects!

The multi-band 3340 circuit earlier in the thread is somewhat based on a design of mine from the early '70s. The circuit originally used MFC6040s and I used to hack them together on bits of Veroboard. A number of '70s AM stations used them. There's also a version with a bass clipper (and a lowpass filter to get rid of the harmonics). I also built them with "all-pass filter" delay lines - a whole stack of series connected filters configured for the range of frequencies you wanted, with a delay time of ~500us.

The delay-line approach is to delay the audio going into the gain control cell, whilst doing the level detection before the delay - you effectively get a zero-attack time limiter. The delay is imperceptible - it's just like being an extra couple of feet from the monitor loudspeaker!

I decided to dig out the old circuit and give it a modern twist using SMD (another lockdown project). Most stuff I've sent off for SMD has come back with various rookie errors but this one worked first time, though I then spotted two resistors were in the wrong place but swapping them fixed it.

It is a 3 band compressor (LF:30 - 200 Hz, MF:200 - 3200 Hz, HF:3200 - 15000 Hz, with 18dB/octave cross-overs) with clippers for the LF and HF channels. You can set the thresholds of the three channels and there's some innovative tweaks on there to reduce the decay time of the HF and LF channels if their attenuation exceeds that of the MF channel. It also does the pre-emphasis so that the HF compressor/clipper is actually a proper HF limiter, and provides 15 kHz LPF. You can link 2 boards together to make a stereo compressor/limiter.

All it needs is a final clipper and final LPF. It sounds as good as the old design (which was the basis for it).

The only down-side is that it needs a very stable and very clean +/- 8V supply. I'm working on an updated version which gets around these problems.

Rev

I have a two-band NE571-based circuit here that's working pretty well. I left out the delay lines for this first-stage prototype, and the attack time is <1µs, and the decay is fairly long at 250 - 850 ms. With judicious tweaking, it can sound pretty damn loud (subjectively) without letting the mod get too high. I have used a couple of green bi-directional LEDs as final overshoot clippers (just in case) and their action is virtually inaudible, whilst allowing another couple of dBs of "loudness". They're inside the feedback loop of the final op-amps.

The filters aren't as aggressive as the Rev's as I went for 12dB/octave, and I used "constant power" Linkwitz topology (like I have in some commercial products).

Some years ago I was involved in the design of a processor / coder that was designed to take on the "Goliath" of Orban, with a better-sounding product at about one eighth of the price of the competitor. That used PWM for the gain control - possibly overkill for pirates - but interesting because it eliminates channel matching issues in the gain controlled cells!