This article details how I modified my Paia 6710 Vocoder for much better noise and sound. Here's a picture of the finished product:

Front panel L-R: Instrument input, instrument level, hole for peak indicator LED, mic input, mic level, hole for peak indicator, HPF switch, Main output, mic threshold, bypass switch, output jack, power indicator.

Here is the original schematic, which is referenced below:

Re-housing

I built up its' own power supply and mounted the whole assembly in a rack case. The circuit board is slightly too wide to fit, so I filed off a few mm on either side; now it fits snugly into the case. The original front panel has controls that are also outside the dimensions of a normal rack case, so I decided to make a new front panel.

While I was doing so I realised that a lot of the cool extra functions the vocoder has, I don't use; so I removed the effects loop, 'Loop level' control, 'Mic direct' control, second output for stereo, and the Bypass switching.

Bandpass filters

One of the first things you will notice when using other equipment with a stock Paia Vocoder is that it's really hard to set the levels up. You're supposed to do it by turning things up until you can hear distortion, and then turning them down a bit! I suppose this makes sense, but it leaves you with no headroom. When I looked at the types of levels that were happening inside the unit I almost had kittens :-D

The initial mic preamp has a gain of about ten, which is fine, and the initial instrument pre has a gain of two. So far so good. Measuring the signals coming out of the bandpass filters, I was amazed to see ten volts peak to peak with normal signal levels! I did some maths and figured out what was happening for the bandpass filter section, using these online calculators:

http://www.captain.at/electronics/active-filter/
http://sim.okawa-denshi.jp/en/OPtazyuBakeisan.htm

The 571 compander chips that are being used as VCAs and level detectors are specified to have a max input voltage of 1.4VRMS. So with gains that high in the bandpass filters, to keep things within spec, you would have to limit the input signal to 6mV RMS!!! This also explains why the unit has so much noise. We are cutting a 1.4V signal down to 6mV, and then amplifying it back up to 1.4V eight times.

Fortunately modifying this was easy, and required no changes to any capacitors. If you look at the calculators I linked to above, and compare the filter below to the one used in the original schematic above, you can see that the calculators have an extra resistor to ground at the input of the filter:

This 'R2' resistor is actually quite important, and has a major affect on the gain.

After modifying the filters to have 1M input resistors and 4k7 resistors to ground as in the circuit above, the noise dropped considerably, and I can now use actual line level input signals! By itself it is a great mod. You'll notice that the 6kHz filter was an anomaly in the original circuit, by using a 2M2 input resistor here I was able to bring it into line with the other filters.

Buffer amps

The levels coming out of the preamps are still quite weak, however. The Input 1 and Input 2 level controls have resistors before them, so even when you run line level into them you are only getting a portion of this out.

The way the 571 part of the circuit works, it is hooked up as an 'expander' – 1dB of level change on the microphone will cause 2dB of level change on the output. I hooked up a sine wave signal to both inputs, found the centre frequency of one of the filters by watching it on a scope, and by changing levels determined that you get exactly 2V P-P output when you have exactly 2V P-P coming in from both the mic and instrument preamps. So ideally you want your input signals at 0.707V RMS, to give a 0.707V RMS output.

I changed the mic input stage and removed the 'boost' switch, since I'm only ever going to send it line level signals from a mic preamp. I hardwired a 33k resistor from pad 'K' to ground, giving a gain of two which is the same as the instrument input stage, and replaced R30 with a jumper to retain all the signal after the mic pre.

For the instrument input, I jumpered R1, to retain all the signal level from this input stage.

I then added a voltage follower immediately after the Mic and Instrument volume control wipers. This was to provide a low impedance output to drive the eight bandpass filters. I'm not sure if it made any difference to the sound, but it made it a lot easier to predict the signal levels throughout the unit. You can see it on a little brown daughter board in the photo at the top of the page.

Mixer levels

The output resistors mixing the various bands back together after the VCA stages are a mixture of 3.3k and 10k. I swapped the 3.3k ones all out for 10k resistors, figuring that any tailoring of the response I could do on the microphone signal rather than forcing a 3x gain boost on the 1.6kHz, 2.3kHz and 2.8kHz bands. The resistors in question are R3, R4 and R5.

High Pass Filter

After doing some experimenting with a graphic EQ in line with the Mic signal, I found that you get the most natural sounding vocoder response by using a 6dB/oct high pass filter at about 1kHz. I added a switch across C80, to switch between the stock 5uF value and a 0.039uF to give a 3dB point of 1kHz. You can use this calculator to find the cap value you need:

http://www.muzique.com/schem/filter.htm

With everything stock, R10 plus the Mic level pot, in parallel with the Mic direct pot, give 4000 ohms. I missed the Mic direct pot first time and ended up with a value way out for this filter.

Next steps

I do actually intend to do more to this vocoder in future; I'm currently using it with a mic preamp that has a built in compressor, which I want to build into the same casing, and also it needs 'peak' indicators on the inputs so you can tell when you have the correct signal levels going in.

I hope this helps someone with their Paia Vocoder. It's a cool piece of gear and sounds great, especially once you do the bandpass filter mod!