will return you to Dick Campbell's Home Page
The MM3B design is in response to many queries about a microphone preamp module that can operate into a balanced input with or without phantom power. Users wanted to have a 9-volt battery powered version but also wanted to have the preamp work from phantom power if it were available.
An example where this is useful is for touring musicians who use electret-microphone acoustic pickups on their instruments. Ninety percent of the time they plug into a board with phantom power and can forget about the battery. But, for that other 10% -- --
A common request is for a size that will fit into a 19mm (3/4 inch) diameter tube, such as the extended barrel of an XLR connector.
The microphone input is at one end of the pc board and can accomodate either a 2-wire or 3-wire electret capsule, or a dynamic microphone by eliminating the electret powering components. Of course, it can also be used as a general-purpose amplifier, for example for raising -10dBu (single-ended) to +4dBu (balanced). There is a provision to add a precision 4.5 volt regulator for applications where capsule excitation voltage must be held constant during battery decay.
The battery terminals and balanced output terminals are at the other end of the pc board.
The lowest gain is two, being unity to each leg of the differential output. Many users have requested a provision for a clickless switch-selectable low-frequency rolloff at 100Hz.
Now the answer to the first question --
The MM3B uses the LM833M dual opamp, although many others could be used for specific applications. This ancient chip has all the right stuff -- low noise [ 4.5nV/Hz^-2 ], modest battery drain [ 5mA@9V ], and sufficient slew rate [ 7V/µS ] to avoid transient intermodulation under heavy load.
The dc current drain is important for long battery life, but is even more important when the power source is 48 volts behind a pair of 6.8K resistors, which is the standard phantom power arrangement called varient P48. This is a severely limiting power source and juggling amplifier performance against such meager resources is a real engineering challange.
The phantom power current limiting resistors in the MM3B are 2.7K each feeding a bypassed 9.1V zener diode. These resistors must be precision components as they appear across the audio output circuit. The steady-state current drawn from the phantom supply is about 10mA.
The LM833M draws about 5mA at idle. If loaded with 500 ohms (differential) the audio output before clipping reaches 2.4V RMS, or 11.5mW, or over +10dBu, with a dc current drain of about 8mA. Even at this exceptionally high output drive the zener diode has not run out of gas so the chip dc supply voltage will be constant through this dynamic range.
One difficult design issue involves selecting the output coupling capacitors. The MM3B uses a newly-available large ceramic monolythic with type X5R dielectric and a whopping 10uF capacitance at 16V in a modest 1210 SMT size. Using the previously mentioned operation at 2.4V RMS across 500 ohms, the output coupling capacitor is passing nearly 10mA of peak current. Even at this current level, the 2nd order distortion at its low-frequency peak is below 0.3%. It ain't perfect, but it's a huge space saving and avoids the problems with back-to-back electrolytics. Remember that the MM3B can run on either battery or phantom power causing both dc polarities to appear across the output capacitor. The output capacitors are bled to ground with a pair of 100K resistors so we don't play "Pop Goes the Woofer" when plugging in. There is a series 100 ohm ballast resistor in each output leg as well making the differential output impedance 200 ohms (hey, it's supposed to look like a mic!).
The single-ended to balanced circuit is not of a cascade type, but uses a precision 2:1 voltage divider to develop the drive for the two amplifier stages, one inverting, one non-inverting. Therefore the circuit constants are identical around each stage. Precision components are used where dictated by the desired accuracy of the balanced output in minimizing common-mode signals. The primary low-frequency rolloff control is via the coupling capacitor to the voltage divider. There are actually two capacitors, one that can be switched in parallel with the other, providing the selectable rolloff function. The "open" capacitor is bled with a 100K resistor making the selection clickless.
A ferrite bead followed by a 200pF capacitor to ground gives some EMI protection to the input circuit.
In the event that a regulated voltage is desired for the electret microphone, a MAX6145 can be installed. This 4.5V precision LDR is followed by a 1K series resistor and a 10uF || 0.1uF to ground. Under battery operation, this assures constant microphone excitation as the battery runs down. For normal performance audio use this feature is almost never needed.
Performance
At this writing, I am still messing with prototypes, i.e. pre-production boards with final components but not the final pretty production boards. Some of the measurements are difficult to make and sometimes I scratch my head wondering if what I am seeing is the amplifier or the measuring instrument. But that's life in an audio lab when dealing with low-noise precision items.
Frequency Response. The lowest input impedance mixer I have seen is one of my own - a very old, indestructable British baby from Audio Devices. It has a 1000 ohm differential input inpedance on balanced mic circuits. That is an unusually low value by today's standards. Under this kind of load, the MM3B has a -3dB point at 22Hz, governed by the output capacitors. With a more normal 10K input impedance the -3dB point is about 12Hz, governed by the internal coupling capacitor.
The high end beyond 25kHz is not well documented yet. The MM3B is unconditionally stable under any capacitive loading. Simply from the output stage topology, one can calculate -3dB at 20kHz with 0.04uF across the terminals. That's about 800 feet of the usual shielded twisted-pair microphone cable. I am not concerned. However, I have not yet tested the MM3B with heavy capacitance loading between the output legs and ground. More on that at a later time.
Noise. I have measured the noise floor in several ways and I have the following data to offer. These data are spectrum level, i.e. one-cycle bandwidth output voltage expressed in dBV with the input shorted. The values are for a differential output connection. The MM3B is running on a bench supply.
| Frequency | dBV |
| 50 | -116.0 |
| 100 | -120.0 |
| 200 | -123.5 |
| 500 | -128.5 |
| 1000 | -131.5 |
| 2000 | -133.0 |
| 5000 | -134.0 |
| 10000 | -135.0 |
| 15000 | -136.0 |
Fit a smooth curve through these data points. From these data one can calculate any other noise value based upon bandwidth or filtering. For example, if I apply the 'A' weighting curve, and calculate the effective bandwidth as 10kHz, I come up with an overall 'A' weighted noise output of about -106dBV or about 5µV.
Two things to remember: these are conservative numbers, and my instrumentation noise floor is not all that far below these values.
Distortion. The main source of harmonic distortion seems to be the output coupling capacitors. To my knowledge, no one has definitively explained the distortion mechanism in ceramic capacitors. Luckily, it's mostly 2nd order, and it's far below human perception levels.
With one volt into a 10K differential load, the total harmonic distortion remains below 0.05% above 500 Hz. It is about 0.1% at 200Hz, and rises at about 14dB/decade to 10Hz.
I will do a much better job on distortion statistics when I get production quantities of the MM3B.
Errors. The board layout uses 0804 SMT size in the locations for precision resistors that will accomodate 0.1% tolerance parts. However, using 1% parts delivers impressively low errors. I am talking here about output common-mode errors that can be explored by examining carefully each leg of the differential output. In a half-dozen prototypes assembled with 1% resistors, I have not seen an amplitude error of more than 0.02dB above 25Hz. Typical amplitude error at 10Hz is 0.3dB. The worst case phase error I have seen is 7.5 degrees at 10Hz. The low-frequency errors are attributable mainly to the output coupling capacitors.
Clicking this red button
will return you to Dick Campbell's Home Page
Copyright (c) 1996-2001 Dick Campbell
This page was created using WebEditPro
-- Last revision: 11/30/2001