Noise and Stuff in Consoles:
More About Specifications
by John Roberts
Last time I suggested how to interpret power-amplifier specifications. This issue, I'd like to delve into mixers. Since mixers are usually marketed by features rather than specifications, there is a little less hype-but still enough to shed some light on. Before I get into the actual discussion of mixer specifications, I'd like to review what a mixer does, and what, from a design point of view, the hard parts are.
The most common requirement of a mixer is combining multiple input signals into one or one pair of usable output signals. An actual mixer (desk, console, etc.) will do many more functions, but combining is number one. Functional requirement number two is conditioning signals so they can be combined. This usually requires amplifying low-level signals from microphones or musical-instrument sources up to some nominal level. The second most common conditioning performed on signals is equalization. Depending upon the specific application, there may be several different reasons for frequency-response shaping, but in general, it's to correct for errors and/or make the signal sound better (that's what it's all about). Finally, the mixer provides signal routing. In the case of recording, you may want to route multiple signals to and from a tape recorder. In a live-performance environment, you will probably want to send additional mixes to monitors. In both applications, you will probably want to send signals to and from effects devices to add reverb, etc.
The goal of any properly designed mixer is to provide all these functions with a minimum of effort by the operator and with no loss of signal fidelity.
To better understand console specs, let's figure out what the hard parts are and concentrate there. In general, the most apparent sonic artifact is noise. Even the best designed mixer has noise. To better understand this, let's look at what noise is and where it comes from. There are probably three significant sources of noise in consoles: mike preamps, summing buses, and cumulative channel noise.
Without getting too theoretical, you have thermal noise (a.k.a. "Johnson" noise) and electronic (semiconductor) noise. Thermal noise is caused by (you guessed it) heat. Heat causes electrons to bounce around. Electrons bouncing around in a resistor generate small, random voltages. This thermally induced current is characterized as a power that varies as a function of temperature. The larger the resistance, the more voltage, even though the power is constant. The voltages are very small, typically measured in nanovolts, or a decimal point with 9 zeros. Thermal noise can generally be ignored when dealing with line-level signals. However, when dealing with low-impedance microphones, output signals can fall in the microvolt to millivolt range (3 to 6 zeros after the decimal point), so thermal noise is important. But before you decide to run out and put your microphones and console in a refrigerator, you need to understand that thermal activity is relative to absolute zero Kelvin (-273° C), and room temperature is in the 300s. Therefore, dropping even tens of degrees will not make a significant difference. The important thing to deduce from all this is that a perfect, noiseless microphone preamp, if such a thing were possible, would amplify the noise of the microphone's resistance (normally 150200 ohms). A lower-impedance microphone, while having less noise voltage, would also have less signal and require more gain ... I think you see where this is going.
The second type of noise is electronic noise. There are several causes, but these are usually lumped together for analysis. Electronic noise is usually referenced to the input of a circuit or device and stated as a noise voltage in series with the input of a perfect noiseless amplifier and a noise current in parallel with the input dumping into the input and feedback network's impedance. For analysis, this is converted to a voltage and multiplied by the circuit's gain.
THE MICROPHONE PREAMP
Now, using these new analytical tools, let's first look at microphone preamps. Because we are dealing with relatively low-source impedances (150200 ohms), feedback resistors are usually kept low. Note: a design tradeoff may occur when you make the feedback network very small to keep noise down, because it requires a very large-series capacitor to deliver full low-frequency response. In fact, one of our "friendly" competitors cut corners there and has microphone preamps that are -3 dB down (half power) at higher than 50 Hz. This is only measurable on a mike preamp set for full gain, so they're getting away with it, so far.
While many manufacturers (Peavey included) specify their preamps in EIN (Equivalent Input Noise), which is computed as output noise divided by gain, I prefer a specification call NF (Noise Figure). Noise Figure states how much more noise you have than a theoretically perfect preamp. My preference for NF is that it's easier to interpret. Over the years I've seen many specifications of EIN that were lower than theoretically possible! Just for the record, the theoretical minimum is something like -132 dBm. Some of these may be honest mistakes, as theoretical noise levels are defined in power, and conversion to voltage leaves some room for interpretation as to what the actual impedance is at the input of a preamp, etc. The important point is that the state of the "Peavey" art right now in mike preamps is running around 12 dB NF. That's important, because if you could find a quieter preamp, even for a hundred times the price, you probably couldn't hear the difference. We beat our closest competitors by about a dB; however, you'll be hard pressed to hear that. Look for an EIN in the -130s but not more than -132, or better yet, a noise figure less than 3 dB. More likely, you will find such sonic shortcomings in our competitors' products as premature roll-off at high gain/low frequency and lack of headroom (e.g., the gain can't be turned down low enough to accept high output microphones direct, when miking a mortar for the 1812 Overture).
THE SUMMING BUS
The other noise problem in large console structures involves the summing amps. Here the problem is due to the insertion loss and the need for compensatory gain, not because you're applying a lot of gain to a small signal. Note: in virtual-earth summing structures, the signal may appear to sum at unity gain with no loss, but in fact the summing amplifiers' EIN is referenced in series with its input and amplified by a "noise gain" term. Without getting into the math or specific circuit designs, the rule of thumb is that most summing structures have a noise gain of N+1, where N is the number of channels being combined. This is generally not a big problem in small mixers, but as you get into tens of channels, the summing-amp noise can become significant. For this reason, we actually use discrete low-noise transistors in our Unity(TM) 4000 series summing amps, and we do measure a few dB quieter than the other guy. The thing to be careful about when looking at a summing-bus noise floor is that a realistic number of channels are connected. There are two (maybe more) design philosophies regarding summing topologies: one "backgrounds" signals not assigned to the bus; the other just connects them as called for. In the "backgrounded" system, the noise gain and thus the noise floor is constant; backgrounding facilitates a balanced bus for minimum crosstalk and hum pickup. The non-backgrounded (open) bus appears quieter in listening tests or measurements with only one or two channels punched up. Don't be tricked. Measure your board the way you're going to use it, with N channels assigned.
The last and final noise source is a tricky one, because at first glance it doesn't appear to be significant. Suppose your well-designed (Peavey) console has an individual channel noise of -96 dBu (obviously, the mike preamp is down, fader at unity, no EQ, etc.). This sounds pretty good, until you add up 32 of these. This kind of noise addition is called incoherent (that doesn't mean nobody understands). Incoherent noise sources sum as the square root of the sum of the squares - easier to do than say. To follow through with the example, 32 channels at -96 dBu through an ideal noiseless summing amp combine to -81 dBu. On the other hand, if the channels' noise floor were coherent (such as an identical power-supply hum in each channel), they would sum linearly, resulting in a final noise floor of -65.9 dBu. Unless you have a poorly designed console or an incredibly large structure (the biggest Peavey AMR(TM) board has something like 120+ inputs to the 2 bus), this will be swamped by even one channel of a theoretically perfect microphone preamp at 60 dB of gain ( 132 dBm + 60 dB = -70 dBm).
When working with consoles, noise is a fact of life. Accept what can't be eliminated (-132 dBm x mike gain), but don't accept what can be designed out (noisy buses, poor bus structures, hum, etc.). I hope I shed a little light on a confusing subject.