"tdfir" is the opcode to enable the "Time Domain Finite Impulse Response" processor in Dimension Pro. Basically, it performs a time-domain convolution between the incoming audio and the given impulse.
Time-domain convolution and Frequency-domain convolution (as seen in SIR, Perfect Space, Gigapulse, etc.) are processes aimed to get the same result, but through different approaches. In time-domain convolution, each input sample is multiplied by each sample in the impulse, and the results are added to get one output sample. In this procedure, the required computational power grows exponentially with the impulse size. Therefore, it is very CPU intensive. The main benefit is that as there isn't any Time-Domain to Frequency-Domain conversion, it is extremely precise so it reproduces the impulse characteristics perfectly. Also, it can easily be made zero-latency.
In Frequency-domain convolution, the input signal is first converted to Frequency Domain (using a 'transform', often referred as FFT (fast Fourier transform). Impulse is also converted to Frequency domain, and then both are convoluted. As there's a small number of 'bins' for the input signal, this operation can be made substantially less computationally intensive. Huge impulses can be handled in this way by modern computers, such as hall reverbs, delays and other naturally-decaying places. This type of convolution does often impose some latency in the input signal.
Dimension Pro features a time-domain convolution processor, enabled by the 'tdfir' opcode. It has a handful of parameters, all very simple:
<effect>
type=tdfir
tdfir_impulse=Impulses\1x12, 421-1m.wav
tdfir_gain=-3
tdfir_dry=0
tdfir_wet=100
tdfir_impulse=
This defines the impulse file to be used.
tdfir_gain=
The convolution process could substantially alter the RMS volume of the input signal. This opcode allows for compensation, in dB.
tdfir_dry=0
tdfir_wat=100
Dry/Wet mix, the values above are the ones we use for all 'pass-thru' type effects.
While DP doesn't feature a Frequency Domain Convolution processor, there's a lot of info on impulses, how to create those and how to use in the 'bible' of convolution processors in the web:
http://noisevault.com/nv/ Hope this helps.