Mic. pre-amp circuit. Discrete or IC?

[Zero999]

I've been working on a mic. pre-amp for an intercom system. The circuit will be fitted as close to the mics as possible, to boost the signal, before being send down a cable which goes past several noise sources.

I've designed a discrete solution, which gives reasonable performance and has the advantage of only using two wires (C1, R1 and V1 can be at the other end of the cable), but I'm tempted to use an IC, such as the NJM2783 or NJM2783 because it should give lower distortion, although it will need another wire for power. At the moment the 8V is coming from a regulator. I might increase R1 and use 24V, to give more gain.

Has anyone had experience with this?

[fourtytwo42]

Nice and simple but doesnt your agc rectifier need a higher input impedance or a buffer to avoid distorting the output ?

[Zero999]

Nice and simple but doesnt your agc rectifier need a higher input impedance or a buffer to avoid distorting the output ?

The rectifier has a very high steady state load impedance: 10M. There will be some distortion, for a short period of time, after the circuit is  turned on, or after a sudden increase in signal amplitude, but once the C5 has finished charging, the impedance looking into C6 is very high.

I have tested the circuit, both with a signal generator and mic and the performance is OK. I've also tested the NJM2783, which gave lower distortion, but it needs three wires and will probably use up more space, after all the passive components have been added.

[Damianos]

To reduce the distortion in the discrete circuit add some negative feedback. Try to remove the C2 or place an emitter resistor.

EDIT: I now see that from C2 is also the supply for the microphone; so, by removing the capacitor, there is a need to manage for that.

[SiliconWizard]

I'd suggest something like the MAX9814 ( https://www.maximintegrated.com/en/products/analog/audio/MAX9814.html )
You'll most likely get better performance with a much reduced BOM and cost.

[Zero999]

To reduce the distortion in the discrete circuit add some negative feedback. Try to remove the C2 or place an emitter resistor.

Yes, both will reduce the distortion, at the expense of gain.

I tried removing C2 and it didn't provide enough gain. I could increase the gain at the other end, but with the gain so low, the limiter will never kick-in.

I'd prefer not to have an emitter resistor, because the DC bias would become sensitive to the collector current, which I might want to vary later, without changing the circuit. As is, the collector voltage varies from 3.2V when R1 = 220R to 2.2V, when R1 = 1k2.  There will be an anti-click/squelch circuit on the other end, which will use the change in bias voltage to operate, so I'd prefer not to adjust that either. It's a shame it's not practical to use an emitter resistor and bypass it with an inductor, due to size constraints.

Another possibility is connecting a resistor between the output (C1's negative side) and Q1's base.

I'd suggest something like the MAX9814 ( https://www.maximintegrated.com/en/products/analog/audio/MAX9814.html )
You'll most likely get better performance with a much reduced BOM and cost.

That's a nice IC. The only downside is the package: I need something which can be hand soldered.

[SiliconWizard]

A TDFN/EP package is definitely hand-solderable if you have an hot air iron (even a cheap one such as Atten ones).

If you're just prototyping or intend to make only very few units, you may even use a breakout board (you'll find one at Adafruit for instance).

[Audioguru]

I see the Schottky diodes overloading the transistor the moment they conduct which causes clipping until the compressor begins to reduce the input  level.
The attack time of the compressor is very slow which lengthens the clipping duration.

Use an audio opamp for the preamp and another audio opamp (and maybe a transistor to charge the capacitor quickly) as an active rectifier.

[floobydust]

You need a low noise, low distortion gain stage that behaves well when overloaded, a compressor or limiter stage, then a line driver.
Doing all of that with one transistor is a great idea but perhaps use a couple more, or op-amps or a dedicated specialty IC.

Tried the SSM2166 and found it terrible. High noise and distortion, and the IC shares some internal biasing so the mic preamp made other sections of the IC misbehave. Tossed that design in the garbage, and instead went with premium THAT4320 which has a compressor and limiter with low noise and distortion. Top performance but complex and expensive.

If you used a two-transistor gain stage, and then a push-pull output stage, that would be best  to get a few volts of signal.
Car/SUV NAV headliner microphones Toyota uses a dual op-amp design. One for gain/filtering and another for voltage follower, but no limiter.


How long is the output cable run?
Beyond about 3m/10ft I would suggest using a balanced output.

I've had installers screw up and put in the wrong kind of cable. Either no shield at all, or one overall shield, instead of a dedicated shield for the microphone signal. The cable can also get pulled alongside mains cables, or alongside the speaker cable, adding to the headache.

[Zero999]

I see the Schottky diodes overloading the transistor the moment they conduct which causes clipping until the compressor begins to reduce the input  level.
The attack time of the compressor is very slow which lengthens the clipping duration.

The attack time is fairly fast. There is distortion when it starts limiting, but it doesn't last for long (a few cycles at 1kHz, according to the simulation) and is still much better than continuous clipping, which would happen with no AGC. I think some distortion, when the limiting kicks in is unavoidable, irrespective of the design. In any case it wasn't noticeable, especially as in real life, it's the lower frequencies which typically cause the limiting to kick-in.

Use an audio opamp for the preamp and another audio opamp (and maybe a transistor to charge the capacitor quickly) as an active rectifier.

That's a sensible suggestion, the problem I have is space. Ideally I want to put this on a headset and I don't want it to be too bulky.

You need a low noise, low distortion gain stage that behaves well when overloaded, a compressor or limiter stage, then a line driver.
Doing all of that with one transistor is a great idea but perhaps use a couple more, or op-amps or a dedicated specialty IC.

Tried the SSM2166 and found it terrible. High noise and distortion, and the IC shares some internal biasing so the mic preamp made other sections of the IC misbehave. Tossed that design in the garbage, and instead went with premium THAT4320 which has a compressor and limiter with low noise and distortion. Top performance but complex and expensive.

If you used a two-transistor gain stage, and then a push-pull output stage, that would be best  to get a few volts of signal.
Car/SUV NAV headliner microphones Toyota uses a dual op-amp design. One for gain/filtering and another for voltage follower, but no limiter.


How long is the output cable run?
Beyond about 3m/10ft I would suggest using a balanced output.

I've had installers screw up and put in the wrong kind of cable. Either no shield at all, or one overall shield, instead of a dedicated shield for the microphone signal. The cable can also get pulled alongside mains cables, or alongside the speaker cable, adding to the headache.

The cable is much longer than 3m. More like 10m. I've done some tests with the single transistor amplifier, without the limiter and had reasonable results, even using unscreened, non-twisted, cheap and nasty bell cable. The problem is different mics have different output levels, people speak at different volumes and I've had issues with clipping with one mic. I may ditch the AGC and trim the gain accordingly, but want to investigate the idea first.

A TDFN/EP package is definitely hand-solderable if you have an hot air iron (even a cheap one such as Atten ones).

If you're just prototyping or intend to make only very few units, you may even use a breakout board (you'll find one at Adafruit for instance).

Hmm, I don't even have a stock of solder paste. I wonder if I could dead bug it?

[Yansi]

What about adding an emitter follower after that?  Will have much lower output impedance if biased accordingly.

[BrianHG]

One trick I used to use to get rid of mic POPs, or excessive volume without alc was to mimic a vacuum tube pre-amp's soft rounded peaks.  It was nothing more than 2 red LEDs in parallel, one forward, one backward, with the line signal fed through a 1k resistor (I believe it was 1K, might need to tune this value), then that point was buffered out through a j-fet follower to drive the audio amp's line in.  The current curve knee on the red LEDs was a smooth curve unlike a diode limiting circuit which has that hard knee at 0.65v.  The effect was a low, but tamed audio since the mic-amp stage before the 2 red led's had more than a +/-12v output range, though the led's limited that down to just below line level.

Yes, this was not a true automatic gain circuit, but, it never suffered the volume pumping in and out effect of such a circuit, yet produced a really comfortable effect for speech.

[Zero999]

Thank you for all of the replies. Sorry for not getting back earlier.

The discrete solution was chosen because it only requires two wires: one for 0V and the other to carry both the signal and power. The biasing will be done in a separate circuit, which will integrate the mixer and anti-click functions. The AGC circuit will not be used at the moment, but will quite likely implemented later. In the end dynamic mics have been chosen, rather than electrect, because that's what's already used elsewhere in the system. Here's the final schematic, showing the optional AGC. If used, the values of C3 & R4 may change, to alter the release time.

I've got the pop suppression sorted. The pre-amp and mic will can be hot swapped and this circuit is very effective at suppressing the clicks. I tried simulating it, but I doubt it's that accurate, because I don't have all the models and it relies on the leakage current through D1 to turn off the MOSFETs and I doubt that's modelled very well.