A thorough treatment of the subject can be found in the following (among others) references:
Today, most audio systems operate on the voltage transmission model. Sources have low impedances, and loads are many times this impedance so that they bridge the source. You can model the connection of source and load as a voltage divider. Looking at it this way, you can see that the majority of the source voltage appears across the load, and only a small fraction is lost in the resistance representing the source impedance.
The usual application for a pad is to attenuate the output of a microphone that is too high for the dynamic range of the following microphone preamp. Using a matchedimpedance pad here is not optimum (this is not to say that it won't work) because even microphones expect to have their output bridged by the input impedance of the microphone preamp.
Another issue here is one of coloration. The microphone and preamp operate together as a system. The input impedance of the preamp (which is not resistive) varies with frequency and this interracts in a complex manner with the output impedance of the microphone (also not resistive). If there are transformers involved at either end, that's just an additional factor in the equation. This complex interraction causes coloration, which may be good or bad, beneficial or harmful. It's one of the things that make different preamps and microphones sound different. The point here is that you can minimize the change in coloration caused by inserting a pad by paying attention to this detail and designing the pad to mimic the conditions present before its insertion. The easiest parameter to mimic, and the one that is the biggest contributor is the impedance that the pad presents to the microphone, and the source impedance that it presents to the microphone preamp.
All of these configurations may be made into balanced configurations by simply mirroring them. For example, a balanced Tpad (known as an H pad) has resistors in series with both sides of the input, and both sides of the output, but with a common resistor (R2) bridging the midpoints of the series legs. The series resistors are half the value used for the unbalanced case. In the following diagrams, the notation R1/2 means "the value of R1 (in the unbalanced case) divided by 2."
Unbalanced  Balanced  Remarks 
Match input or output but not both.
Asymetrical. Input and output can not be exchanged. 

Symetrical. Input and output can be exchanged.  
Symetrical. Input and output can be exchanged.
R1 = R2 = line impedance 
These configurations may be made adjustable by making the various resistances variable. The nontrivial part is that the resistances have to change according to a particular relationship which is not linear. To get an adjustable Tpad, you must buy it as such. An adjustable Lpad has 2 variable elements, whereas an adjustable Tpad has three. The BridgedT pad is popular because two of the resistors (R1, R2) are fixed and never change value (they are always equal to the pad impedance), so making this configuration adjustable is just a matter of making R3 and R4 variable. It is possible to buy a readymade pot that can be used as a bridgedtee attenuator but they are expensive and subject to minimum order quantities, making them nearly impossible to purchase. It is easier to make a bridged tee attenuator using a rotary switch but this also makes the attenuation steps discrete.
The formulae for finding the resistor values needed for the T and BridgedT configurations can be found in any of the references mentioned at the beginning of this document. Since these configurations are most often used in impedancematched systems, they don't concern us beyond basic familiarization. For the purposes of this page, we'll concentrate on the Lpad, Upad and Opad.
As mentioned before, the Lpad is the classic voltage divider circuit using two resistors to reduce the input voltage by some amount. Both the U and O configurations are derived from this (if you're a little circuit savvy, you've probably already figured this out). For example, to convert an L to a U pad, you divide the series arm (R1) by two, and these become the new series elements for the Upad. Taking this further, you get to the O configuration only when the pad input impedance is significantly higher than the impedance bridged across the output (i.e. the following mike preamp). The extra resistor simply shunts down the input impedance to better approximate what the microphone sees with the pad connected to what it sees without the pad. There are still subtle differences, such as any reactive elements present at the preamp input, and these will not be mirrored by the pad.
I'll show you how to find the resistor values for any arbitrary value of attenuation for an Lpad, Upad, and Opad. Then, I'll put a few of the usual suspects into a table. The math involved is not complex; if you have a calculator I invite you to follow along.
Excel spreadsheet creates KFactor table.
Practically speaking, the absolute loss is determined by the parallel combination of the shunt resistor and the input impedance of the mike preamp. Since the preamp bridges the shunt resistor, the actual loss is slightly greater than calculated (less than 1dB). For loss values less than 20 dB, you may have to raise the value of the shunt resistor to make the series resistors large enough to not excessively load the source. In that case, the contribution of the preamp's input impedance to the attenuation error will increase.
Another matter is the impedance seen by the microphone. This is the U vs O question. Since the mike is connected across all three resistors, it sees an impedance of (680 + 680 + 150) 1510 ohms. Close enough. For higher loss ratios, you can add a shunt resistor (R4) across the input to lower the input impedance to something closer to the mike preamp's input impedance. Calculate the shunt resistor by finding the resistance needed to parallel the pad's input impedance to come close to the mike preamp's impedance.
Use a variation of the parallel resistor formula: Rtotal = 1/((1/R1)+(1/R2)). Solving for R1 we get 1/((1/Rtotal)(1/R2)). The calculator says 2238 ohms. 2200ohms is the closest 5% value.
As mentioned previously, resistors are only available in particular values (unless you happen to own a resistor factory). The easiest way to find the standard values is to consult a catalog, such as the Digikey catalog. A computer program (PC only, not mac) is also available (freeware) to help make this selection.
Of course, for some things, you need to be right on. For this, you may need to use multiple resistors in series or parallel to build the value needed.
It's helpful to have two different colors so that you can keep pin 2 and pin 3 separate (to maintain polarity through your pad). Solder a third wire to pin 1 of the male connector, and trim all three so that they would stick out about an inch beyond the end of the shell of the tube when the male connector is inserted into the tube. Before final assembly, strip and tin the ends of the wires. Now insert the male connector into the tube, leaving the wires sticking out of the female end. Tighten the setscrew on the male insert. Now solder the three wires sticking out to pins 1, 2, and 3. Ensure that the wire connected to R1 ends up at pin 2 at the female end, and that R1 connects to pin 2 at the male end. If you're using R4, connect it across pins 2 and 3 of the female connector. BTW, the female end is the input.
In use, it is technically better to connect the pad between the output of the phantom power network and the preamp input, so the resistors from the pad aren't between the phantom source and the microphone. If your condenser microphones don't require much current (like the Neumann Fet80series), or if they have a wide operating voltage range (like the AKG C451) then this consideration is strictly academic. My advice is to try it before worrying about it.
n.b. These formulas will get you close. Usually, getting the attenuation within 0.1dB of some arbitrary value isn't as important as just getting some value of attenuation that is close to what you need. Getting the attenuation value within a gnat's ass of some arbitrary value requires taking into consideration the actual source and load impedances.
n.b. Again, this will get you fairly close to the actual attenuation value. Getting more exact numbers requires taking into consideration the source and load impedances.
Microphone Pad Values (U or O)  

dB loss  k  R1  R2  R3  R4  Config  Remarks 
10  3.16  510  470  510  U  Impedance values are a compromise.  
11.79  3.88  10k  3.6k  L 
+4dBu > 10dBV pad
Values for line level operation. Use Lpad schematic. (Unbalanced) 

15  5.62  510  220  510  U  Impedance values are a compromise.  
20  10  680  150  680  U  
25  17.78  1300  150  1300  3300  O  
30  31.62  2200  150  2200  2200  O


40  100  7500  150  7500  U  Use to convert line level to mike level; i.e. use a mike input as a line input.  
Note: K is the voltage loss ratio for the pad. The resistors for the L and U configs follow the ratio K1. 
If you order the pad board kits from me, calculating the resistor values is included in the price. (within reason, my interpretation). Of course, you're welcome to do it yourself.
Last modified 02/19/2014.