Eric Edwards GW8LJJ has yet another handy shack accessory – a versatile audio processor suitable for most types of microphone.

 

This month’s project is an audio processor designed to accept a standard microphone − a crystal, moving coil, dynamic or electret type using the processor’s internal FET supply for the electret type. There is insufficient voltage to supply a capacitor (condenser) microphone because that requires a higher voltage and as much as 48V is common. The processor will, of course, accept a capacitor microphone with its own phantom power supply.

 

Power Supplies

The full circuit diagram is shown in Fig. 1. At the top of the diagram is the power supply circuit. The input is 12V (or 13.8V) and is connected to the input of an 8V regulator (VR3), which in turn supplies the voltage to a 5V regulator VR1. The 8V is also used to power the inverter (IC7660) U4 to provide a negative voltage for the negative 9V regulator VR2. The outputs from the power supply circuit are +5V and −5V and are the supplies for the Op-Amps U1 and U2, lowpass filter U3 and buffer/clock filter (part of U2). There is a link shown at the output of the 8V regulator so that a separate 8V supply can be used in place of the 12V. This was provided because the processor can be used with any other units that already have 8V available although probably it would be used with the shack 12V or 13.8V supply.

 

The Processor

This audio processor makes no claim to perfection but will provide a louder sounding signal than unprocessed audio and will produce a well-controlled modulation envelope.

The speech processor handles all the low-level audio processing, microphone amplifier, EQ (equalisation), compression and bandwidth limiting.

The speech compression part of the circuit is the design of James L Tonne W4ENE (ex-WB6BLD), and the author is grateful to James for his support and valuable insight into the processing of audio. His excellent compression circuit is reproduced unaltered. The pre-compression microphone amplifier and EQ stages are also inspired by James’ filter circuits although many component values have been changed. Readers may know of James from his popular ‘Elsie’ LC filter design program (see reference section).

 

Microphone Amplifier and Highpass Filter

The microphone input is at pin 3 of op-amp U2a and has a gain of about 22, set with the resistors R16 (22kΩ) and R15 (1kΩ). Capacitor C1 and resistor R13 form a passive highpass filter and R14 along with C2 a lowpass filter. Together they give the response shown in Fig. 2. It was modelled in LTSpice–IV. The op-amp U2a amplifies the microphone signal, driving U2b, a highpass filter with zero gain. U2c provides further amplification sufficient to drive the compression circuit. This stage provides a gain variation from +20dB to −10dB to cater for different microphone sensitivities and compression requirements. U2a and U2b circuitry is inspired by James Tonne’s processing paper, although the R/C values used differ.

The output from pin 1 looks like the blue trace in Fig. 2. The waveform is further modified with a sharper LF cut-off from 300Hz downwards in U2b, finally producing the red waveform.

At the input of op-amp U2a is a mute circuit provided by turning on transistor Q1 by applying +5V to its base. The purpose of this is to remove (mute) the processor during receive so as not to provide ‘splashes’ of modulation to the PA. The op-amp U2c is the compression drive and the aim is to find the right setting where talk power is substantially increased without it becoming obvious there’s a speech processor in use. Excessive compression merely increases background noise during breaks in speech and adds nothing to readability. There is only one potentiometer (RV1) to adjust on the processor board. RV1 controls the gain of the compressor driver amplifier. The gain of this stage can be adjusted over a 30dB range from +20dB to −10dB.

 

Audio compression

The audio compression part of the circuit is entirely the work of James Tonne. It’s an RMS-based AGC compression system with optimised attack, hold and decay time constants. James has chosen these well, and no pumping is apparent with sensible levels of compression. Audio from the driver stage (Pin 1 U2a) is fed to U1c, which has a voltage gain of 100 times or 40dB (defined by R4/R5). However, the level of audio at U1c pin10 can be controlled by the variable attenuator formed by R1 and D1/D2. The amount of attenuation is determined by how much forward bias is applied to the diodes. The diodes are in turn driven by the buffer U1a and inverting buffer U12b, which are connected to the AGC bus. The AGC bus voltage varies with speech amplitude, forming a control loop. The relationship between the incoming audio and the varying AGC bus voltage defines the compression characteristic.

The AGC bus voltage is derived as follows: The output of U1c feeds D3 and (via the inverter U1d) D4, forming a full-wave rectifier so that both peaks and troughs are captured. The output from D3/D4 will vary dynamically with audio level but to explain the circuitry, consider a steady DC voltage of say 4V. R8/R9 will halve this voltage as it is applied to D6. C7 will charge via R8 and D6 to a level close to 2V. C7 is across the AGC bus and the charging time-constant of C7 is the compression attack time. The 4V in our example also charges C8 via D5 and R10. Ignoring any diode drop in D5, C8 will charge to the full 4V, back-biasing D7. So, we now have 2V on the AGC bus and D7 reverse biased by 2V. Let’s now remove the 4V. C7 cannot discharge via D6. The only path is via D7 but this is reverse-biased. It is only reverse-biased, however, until C8 discharges, via R10 and R11, to a point where D7 can start to conduct. This is the recovery delay. C7 can then discharge via D7, R10 and R11. This is the recovery time.

Simpler attack/decay circuits produce LF distortion because the AGC line is modulated by low-frequency ripple. The recovery delay prevents this from happening. With the component values used, the attack time is 3ms, recovery delay 5ms and recovery time 10ms. JP2 disables the AGC bus. This allows a higher audio level to be passed through the circuitry for modulator line up. A constant tone has a high RMS value so the AGC bus settles to a level that gives a lower output than normal speech, making it impossible to fully modulate a transmitter on a whistle or single tone despite a high modulation level being produced on normal speech. Disabling the AGC bus also allows the gain stage to be driven into clipping should adventurous users wish to experiment with a more brutal approach to loudness enhancement (luckily U3 prevents too much spectrum carnage!). The AGC bus voltage can be monitored from JP2; useful for compression level monitoring. The voltage could be applied to, say, an LED bargraph driver chip to indicate compression level. Note that any circuitry added at this point must have a high-impedance input so as not to load the AGC bus.

 

Lowpass Filter

Audio emerging from the compressor has undergone amplitude compression and frequency response tailoring but so far there is no ‘brick-wall’ filtering to limit higher frequencies. Voice frequencies will still be present that would result in an unacceptably wide modulation envelope. U3 remedies this. U3 is an 8-pole switched-capacitor filter that can be adjusted to define the overall bandwith of the audio. With eight poles the roll-off is very steep. A single capacitor, C9, alters the internal clock frequency that defines the filter corner frequency. There is a 100:1 clock-to-corner-frequency relationship. Reducing C9 raises the clock frequency thus raising the corner frequency. A value of 68pF gives a corner of about 4500Hz. For a bandwidth within 3kHz it is recommended to fit a 47pF capacitor across the 68pF to bring the HF cut-off to 2.9kHz. U2d buffers the audio, with C10 acting as a clock filter to remove any chance of U3 clock leakage.

 

High Frequency Response

I have prepared a table of capacitor values that can replace C9 on pin 1 of the MAX294 chip. The following changes affect the HF response as follows:

Cap: 68pF, 6dBV reduction: 4.3kHz, zero output: 4.7kHz.

Cap: 47pF across existing 68pF for a cut-off at 2.9kHz.

Replacing C9 with the following values will provide the frequency cut-off as shown.

Cap: 100pF, 6dBV reduction: 3kHz, zero output: 3.3kHz.

Cap: 122pF (100p + 22p in parallel), 6dBV reduction: 2.58kHz, zero output: 2.8kHz.

Cap: 150pF, 6dBV reduction: 2.2kHz, zero output: 2.4kHz.

 

Reducing the LF Response for SSB

The following change can be made to affect the low frequency response:

Replace C5 and C6 both 47nF (HPF) with 22nF capacitors for slightly more LF reduction at about 300Hz.

 

Other Uses

This processor can be used as a preamplifier and filter in conjunction with a direct-conversion receiver, thus removing the sometimes annoying hum and also tailoring the output for speech. As a microphone processor the output is normally taken to the line input on the transmitter but where there is only a microphone input, it can be connected there. However, it is suggested that the output from the processor is attenuated. This is easily done by connecting a linear potentiometer, say 50kΩ or thereabouts, across the output terminals of the processor with the slider and earth connections connected to the microphone input socket on the transmitter. If you use a very high gain microphone as found on some headsets, the input gain can be reduced simply by fitting a 1kΩ ¼W resistor across the original 22kΩ (R16) across pins 1 and 2 on U2a. This reduces the gain of that stage to unity (R16/R15 = 1kΩ/1kΩ = 1). This would not normally be needed but I had to do that when using a headset for my FreeDV transmitter soundcard.

 

Is There a Kit?

I can offer a full kit or just the PCB for this project so send me an e-mail for details.

 

Reference

I very much appreciate Dave GW4GTE letting me produce this article and using some of his descriptions as taken from a comprehensive manual for this processor. This manual, a free PDF, can be found on Dave’s website:

www.s9plus.com

LTSPICE: Visit the website below for an excellent paper on audio processing. Go to the homepage for much more information and free software.

http://tonnesoftware.com/appnotes/speech/speechamp.html

 

This article was featured in the July 2018 issue of Practical Wireless