The audio processor and stereo encoder

Senin, 02 Agustus 2010

The audio processor and stereo encoder

The RC4200 analog multiplier used in this stereo encoder seems to have gone out of production. Before considering the construction of this circuit, make sure you can actually find this IC, or else be sure you know how to modify the circuit to use a different analog multiplier, such as the AD633 or other one you might find.

The textbook way of processing and encoding a stereo signal for FM transmission goes like this:

1) Take both channels and low-pass-filter them at 15kHz, with steep rolloff;
2) Apply pre-emphasis. Depending on the part of the world, it should have either a 75µs or a 50µs time constant;
3) Strictly limit the audio level to ensure that overdeviation cannot happen;
4) Create a stable, clean 38kHz sine wave;
5) Subtract the right channel from the left channel, and multiply the result with the 38kHz carrier;
6) Create a clean 19kHz sine wave, phase-locked to the 38kHz one;
7) Add the left channel, right channel, the (L-R)*38kHz signal, and the 19kHz signal, with specific amplitudes.

There are several ways to implement this algorithm. Modern factory made transmitters often do the whole thing digitally, in a DSP. But it's still less expensive and simpler to do in the analog domain. That can be done in various ways too, and far too many transmitters these days use ultra cheap, mediocre methods like hard-switched multipliers based on CMOS switches. They do work, but are very noisy! My design instead uses a true, high quality analog multiplier for that task. As a result, the signal from my transmitter is as good as the very best signals I can receive locally, and MUCH better than the bulk of them!

Here is the schematic diagram. You probably won't be able to read it at this resolution, so better click on it, save it in full resolution, print it, and refer to it for the following explanation. If you have trouble opening the large version, right-click on the diagram, so you can save it to disk, then open it using IrfanView or any other GOOD image viewer. This is valid for all drawings on this page. The full resolution drawings are large, and depending on the amount of memory in your computer, some web browsers cannot open them and will report a broken link.

The two single-ended line-level audio signals enter through feedthrough capacitors, and are welcomed by an LC low-pass filter to get rid of any RF that could be on them. In each channel there is a buffer stage, and then a combined pre-emphasis and soft limiter stage. The advantage of doing the limiting and the pre-emphasis in one step is that it avoids overdeviating from loud treble sounds, or having loud bass sounds flatten out the treble, without the need of a multiband limiter. The gain of the non-limited portion of the audio signals is adjustable by means of trimpots. Then comes a six-pole low pass filter that removes signals above 15kHz.

A 74HC4060 chip derives the 38kHz and 19kHz signals, as square waves, from a custom-made quartz crystal. Two resonant circuits using ferrite pot cores turn these square waves into very clean, low noise sine waves. Trimpots allow to set the levels, while the adjustable cores of the inductors allow precise tuning. Jumpers allow to disable each of these signals for testing and adjustment purposes.

A rather old fashioned, but low-noise and low-distortion analog multiplier chip modulates the L-R signal, produced by an op amp differential amplifier, onto the 38kHz subcarrier. This circuit has three adjustments for balance. Its output level is adjustable too. The signals that are necessary only for stereo can be disconnected for testing by means of a jumper.

The output adder combines the L signal, R signal, (L-R)*38kHz signal, and the pilot tone. The first two signals are fixed at this stage, while the (L-R)*38kHz can be adjusted by its own trimpot, and the pilot tone by the trimpot before its LC circuit. Then there is a final level adjustment, used to set the deviation of the transmitter, and then a buffer stage with low output impedance, that drives the output through a resistor to avoid instability from capacitive loads.

There is an additional circuit which consists basically of a dual superdiode detector with a time constant and driver with adjustable output. This circuit picks up the complete multiplex signal just before the final level control, and produces a DC signal to directly drive a small meter, for deviation indication. This is a most important tool for the transmitter operator to set the proper audio level during routine operation!

Here is the printed circuit board. Click on it to get it in high resolution.... It's seen "through the board", so you can print it directly and place the ink in contact with the copper to get a correct sided copper pattern.

The entire circuit is built on this single-sided PCB. Only a few jumper wires are necessary, so it's not worthwhile making a dual sided PCB for this.

And this is a crude parts overlay, just to see where a part goes. Exactly which part goes where, is something you will have to work out with the schematic! Don't be lazy!

About the components: All of the critical resistors are metal film, 1% tolerance, both for stability and for low noise. The operational amplifiers are a low distortion, low noise type, except for the opamp of the metering circuit, which is a simple BiFET type. All trimpots are high quality multiturn units. The capacitors are mostly polyester, but in the low pass filter I used 5% silver mica ones, simply because I had a big lot of them and could match the values very well! Matching the capacitors is a good idea, because their 5% tolerance is a bit wide for obtaining the optimally flat filter response. In uncritical places you will find ceramic and electrolytic capacitors. The chokes are dipped ones removed from a junked VCR, but similar ones can be bought new. The ferrite pot cores came from the stereo decoder of an old (wooden boxed!) radio, which I got in a condition too incomplete to restore. I don't have information on them, so you will have to select your own cores and calculate the number of turns to obtain the inductance stated on the schematic. Just be advised that the pot cores MUST have a significant air gap, in order to be stable enough. The crystal can be ordered from JAN Crystals, specifying a frequency of 2.432 MHz, fundamental mode, parallel resonant, 30pF load capacitance, HC-49 holder, with standard temperature, stability and tolerance ratings.

You have to understand this circuit to be able to calibrate it properly. And you need an oscilloscope, of course! The process starts by presetting all adjustments to their mid points, applying a +/-15V power supply, and an audio sine wave of 1kHz to both channels, at a level of 1V peak-to-peak. Set R5 and R23 for exactly 4.5V p-p at the outputs of the low pass filters, as noted in the diagram. Then you adjust L4 and R44 repetitively while looking at the output of U9A, tuning the coil for maximum signal and the trimpot for exactly 4.4V p-p. Then you apply the 1kHz signal to only one input of the board, and you short the other input to ground. With the oscilloscope at the output of U11A, you should see a classic two-tone signal. Now you adjust R60, R61 and R62 repetitively for best ground centering, symmetry and linearity. This is easiest to do by using a dual channel scope and putting the other channel on the input signal to the analog multiplier (output of U6A), superimposing the two traces. After adjusting the gain of the scope channels, the modulated two-tone signal should precisely fill the 1kHz sine wave.

Now install a jumper on JP2 and put the scope on U6B's output. There you will see the sum of the 1kHz signal and the dual-tone signal coming from the multiplier. Adjust the level of the (L-R)*38kHz signal with R55, so that it is exactly equal to the 1kHz signal's level. That's very easy, because when the setting is right, the 38kHz signal always moves between zero volt and the instant level of the 1kHz sine wave. So, you only have to adjust the trimpot to get this zero volt line nice and straight! If you have never built a circuit like this, you may not understand now what I mean, but it will become clear immediately when you are playing with the adjustment! Be sure to make this adjustment with best precision, because the good stereo separation of this encoder depends on it!

Now remove the jumper on JP2 and install it on JP1. Apply the 1kHz 1V signal to both channels. Tune L5 for maximum 19kHz signal, and set R45 so that the pilot signal on the scope is about 10% the amplitude of the 1kHz signal. Now place the two scope probes on the outputs of U9A and U9B, remove the jumper from JP1, and retouch L5 to align the phases of the two sine waves, so that the zero crossing happens at exactly the same time. Increasing the scope gain on the 19kHz signal helps in getting the waveforms more parallel to obtain better precision.

R68 will be adjusted once the exciter is complete. For now, just set it to about mid range, which will give about 1V at the output. If you already have your meter for the deviation metering (any panel meter from 10uA to 1mA full scale should work), you can draw a scale for it and adjust R73 so that it reads 100% deviation (or 75kHz, whatever you prefer). Do this with a signal of more than 1V applied to the inputs, so that the signal is being limited. By the way, the reading should be the same regardless of whether you apply the audio signal to only one inputs, or to both. When there is no audio input, the meter should read about 10% of the full deviation value. This is the pilot tone, and you might want to mark its level on the meter.