Audio Processor

[stax356]

I'm trying to figure out the correct way to build an audio processor that processes the signal before entering the stereo encoder.
As far as I understand it should include the following function blocks in the following order.

1. Front end
This is where the input signal from the audio source comes in. We should start with a simple RF input filter. Then an input converter from balanced to unbalanced signal and possibly some input level adjustment.

2. The audio signal should then go to a subsonic high pass filter to remove anything below 20Hz.

3. The next block is the automatic signal leveler. It is a slow-acting ALC system that should work both ways, amplifying too weak signals and reducing too strong ones. As an option, it can also include a noise gate.

4. Audio Compressor. Is it possible that block 3 and 4 are combined into one so that the same circuit performs the functions of both ALC and compressor?

5. Preemphasis 50/75 uS.

6. Low pass filter 15 KHz.

7. Fast limiter plus clipper.
There are 2 options here.
- Simple broadband limiter and clipper.
or
- Split band. In other words, the frequency spectrum is divided into 2, 3 or more parts with a crossover. Then for each lane we have a separate limiter plus clipper. Finally, we sum the frequency bands.

And with the thus processed audio signal we attack the stereo encoder.

My questions are:
Are these all the required function blocks or am I missing something? Is there a block that needs to be removed?
Are they lined up correctly?
For example, where is the pre-emphasis best placed?

[mpx]

It would benefit from some input conditioning filters like DC bias removal and azimuth correction

[Albert H]

Stax - your building blocks are essentially correct, but I find that the compressor section really isn't necessary if the ALC is working correctly. The ALC I use is slow-acting, and exists (primarily) to overcome the sometimes arbitrary range of audio input levels from differing sources.

The limiter is the most critical part of the whole deal - you need very fast attack times, but without distortion creeping in. The problem with very fast attack in the bass end is that you can end up with "self-modulation" distortion, caused by the time constant of the attack being similar to the very low bass frequencies you're trying to pass....

Multiband really is the way to go, but you really can get good results with just two bands - <~200Hz and >~200Hz. This is enough to prevent the highest energy low frequency "thumps" (like bass drum) causing the midrange and treble to "dip" in sympathy. My latest cheapo processor works on this basis, and - if set up sensibly - it sounds really good, and there's NO chance at all of over-deviation. I used the THAT2180 attenuator IC, and it sounds great!

[stax356]

Albert,
thank you for your time and explanations. Over the past few days I have read most of the threads here on the forum and I have paid particular attention to your comments and observations.
I want to build a pirate radio with a really nice analog sound, no distortion or over pumped sound and I have ideas for some of the blocks but I'll need help with the others. I will read more about THAT2180
Could you help me with advice and guidance?

Let's clarify the topology first, and we will comment on the specific blocks at a later stage.
As far as I understand, your suggestion is:

1. Front end
2. ALC
3. Preemphasis
4. 15 KHz Low Pass
5. Dual band fast limiter with THAT2180

Is this the correct sequence - first the pre-emphasis and then the 15 KHz low pass filter?
Or first 15 KHz filter then pre-emphasis?

[Albert H]

That order will work, but you need to add a couple of final stages:

The stereo split-band limiter will do most of the level control, but you'll quickly discover that it can miss the fastest transients, so you need a final clipper to prevent the breakthrough of these, and then a further low pass filter to shave off the harmonics that the clipping inherently generates.

The '2180 is relatively expensive, it's "nice to have" (it's very high quality), but you can certainly get really good results with an operational transconductance amplifier (like the LM13600 or 13700) in the negative feedback loop of a higher quality op-amp (but cheap like an NE5532 or LM833). This configuration is easily capable of around 24dB of attenuation and adds very little noise or distortion.

I've also successfully used other methods of attenuation:

• PWM-controlled choppers (using FETs or CMOS switches), where the pulse-width of the control signal controls the amount of attenuation - my favourite method in many ways
• Opto-FETs (like the H11F3) though the level that can be applied is very low, leading to the need to attenuate on the way in, and then amplify, which adds noise
• CMOS ICs used as enhancement-mode FETs (with great accuracy between channels because all the FETs are on the same die and so are identical), but this suffers the same noise penalty
• Simple FET analogue "shunt" control, where the FET is used as a voltage-controlled resistor (again with the level limitations imposed by the devices)
• Even the classical "long-tailed pair" of transistors where varying the current through the pass transistors changes the gain of the stage.....

[stax356]

Albert,
I just looked at THAT2180 - it can be bought for 14EUR per piece. It's true that it's an expensive chip, but it's not so much that I can't afford it. However, this is not a commercial project so the budget is not a concern for me. If 2 units are used for stereo ALC + another 4 units for two-band stereo limiter, if the result is good it will be worth the investment.
A cheaper option with LM13600/LM13700 can be considered. I think I have a few of these as well as quite a few of the NE570/NE571.
I was thinking about your idea of a zero response time limiter using the bucket brigade device MN3007. It is better to spend more money and effort to get a technically superior limiter, than to struggle with the consequences of the clipper and its distortions.
I'll draw the block diagram, then the individual blocks I have in mind for discussion, and then I think I'll move on to experiments.

[stax356]

Here is the block diagram:

[stax356]

Here are the first two blocks:
I would greatly appreciate any criticism or suggestions.

[Krakatoa]

The first op-amp needs 10k from non inverting input to ground to be a true differential input.
The filter stage around the second op-amp I don't know, but its gain is set to greater than 1 and may lead to oscillation - better try it in a real world circuit.

[stax356]

Thanks Krakatoa,
I missed the resistor while drawing the schematic. I fixed the mistake.
About the filter - I took it from the input part of the RANE-23S crossover, where it was made this way.
I will of course try it out if it works without any problems.

[Albert H]

You might find that the second stage filter will "ring", particularly around the turnover point, as you're going for a slightly higher order filter than I used.

My filter uses two 150n caps in series feeding the non-inverting input, and the gain just needs to be 4dB (1k feedback, and 1k8 from inverting input to ground is a good approximation). The resistors are both 6k8 - junction of the capacitors to the op-amp output and 6k8 non-inverting to ground. The turnover is almost exactly 20Hz, at ~-6dB/oct, with little chance of ringing. You can cascade two Sallen-Key filters like this for -12dB/oct, but I found that one stage was generally sufficient. I also dimensioned the input capacitors to provide a free passive highpass characteristic into the de-balancing buffer.

Another thing - it's well worth adding a couple of back-to-back 3V9 zeners from each input to deck. These will go a long way to protecting your op-amps if static (or microphone phantom supply!) gets plugged into your input stage. I know it sounds like overkill, but you really can't account for idiots!

Your 10µ capacitors are bigger than I'd use, (I like the additional high pass filtering) , and you can use un-polarised capacitors, which are almost always better for audio.

Your input RF filter is a good idea, by the way.

[stax356]

Thanks Albert.
The note about the input capacitors is very relevant. I reduced them to 2.2uF non-polar.
So I will be able to use polyester or polypropylene capacitors.
Regarding the input protection, as far as I know the NE5532 has a pair of opposing diodes on the input and is well protected. But just in case, I also added the zener diodes to a table.

[stax356]

Albert.
If I understood you correctly, your proposal for the subsonic filter is the following circuit.
But I think there is some error because with these values the cutoff frequency is 156 Hz.
Maybe the resistors should be 68K each and that would give a cutoff frequency of 15.6 Hz?

[stax356]

http://sim.okawa-denshi.jp/en/OPseikiHikeisan.htm

[stax356]

I came across the following schematic for a subsonic filter.
It is from the input part of a HITACHI HMA-8300 amplifier.
It is similar to the one proposed by Albert but is a little more complicated.
Are there any advantages with this connection?
There is also an ON/OFF option.
If I'm not mistaken, the cutoff frequency is even lower - around 10Hz.
Isn't it better to be higher - 20 - 25Hz or even 30Hz?
I don't believe there are many music records with such a range.
I still doubt that there are radios and home systems that can reproduce them.

[Albert H]

I've always found that 20 Hz is about right - the Sallen / Key filter starts to roll off around 30 Hz, and is quite gentle at first. You're right - there's virtually nothing recorded below about 25Hz, so the 20 Hz filter is a good compromise. You're right about the values, by the way - should be 51k each!

I'm aware of the "protection diodes" inside the NE5532 (I used to work for Philips), but I wouldn't want to rely on them. I'd put the zeners on the input side of the 10k resistors. The zener protection has saved me from embarrassment on many occasions! Way back, we used "fusible" resistors in series with the inputs, with the zeners immediately after them - they'd vapourise quickly and protect the rest of the circuit.

[stax356]

Thanks Albert,
I totally agree with your arguments about the filter. I will change the values of the resistors for a cutoff frequency of 20 hertz (51K) and move the position of the zener diodes.
I will redraw the schematic when I have time.
I'll be traveling for the next few days, I'll fix it when I get back.

And what will you recommend for ALC?
I looked around on the internet and the following options caught my attention:

1. ALC with NE570/NE571
They are still easy to find. I think I have a few of these. The built in op amp is bad but I can put in an external NE5532. I'll draw a schematic to discuss it.

2. LM13600/LM13700 in the feedback of a NE5532. I have a few of these as well and they are also still easy to find and cheap. I don't have a specific schematic.

3. THAT4301 - It looks pretty high quality and there is a specific design for ALC with this chip on the site. Besides, it's all in the chip. The downside is that it has been discontinued. It is still available but it is expensive - about 40-50 dollars per unit. I will need 2 for the stereo.

4. THAT2180 is also expensive and hard to find. About 14 euros a piece, for stereo I will need 2 pieces.

5. I also found 2 schematics in Elektor magazine that I found interesting. I'll show them too.
The special thing about them is that they use a combination of analog and digital techniques. The regulation is done by switching in small steps resistors that change the gain of an operational amplifier.

[Krakatoa]

In my opinion, your best bet (cost/complexity wise) is to try the schematic of the Broadcast Warehouse compressor-limiter, that used the LM13700. Increase the time constants and don't use the clipper. This way you will get it working very similar to a typical AGC.
If you want lower figures of noise and distortion by using a proper VCA (since OTAs are quite noisy and require careful trimming) you don't need to pay for premium THAT Corp. VCAs. There are clones of those from "cool audio" for half the price.
Even there is a quad VCA package inspired on the obsolete SSM2164 (altough the core is different tech. than the THAT2180 series) easy to use, with ok performance and for the same price as one unit of the 2180s.

[Albert H]

Cool Audio make a cheaper clone of the SSM2164, called the V2164. I bought a tube of them a couple of years ago - they're not the greatest, but they're acceptable. I used them in a quick split-band limiter I built for a station recently - I made it clear that it was a prototype! The performance was much better than I'd expected.

[stax356]

Thanks Krakatoa,
I was also thinking about the same Broadcast Warehouse scheme but as a limiter. I will try it with a larger time constant.
I'll play around with a few options. I ordered THAT4301 and while I wait for it to arrive I will experiment with NE570 and LM13700 because I have those chips.

Thanks Albert H,
An SSM2164 or V2164 with this quad VCA seems more suitable for a multiband limiter. I will probably try this option and choose the most suitable one.