1. Do not specify and tolerance the PassBand center frequency and bandwidth, instead define without tolerance the center Fo and width B of the PassBand in which other parameters are specified.
2. Specify the minimum insertion loss in the defined PassBand.
3. Do not specify passband amplitude or phase ripple. If amplitude is weighted then reference the ideal weighting function (e.g. Taylor 40dB 100MHz). Specify instead the key output compressed pulse characteristics: pulse width, sidelobes and s/n mismatch loss. If max noise band width is important then specify it.
4. If an expander, specify the impulse response pp amplitude flatness inside a defined interval of width T centered at T0, and the subsequent compressed pulse response of an ideal compressor.
If a compressor, specify the compressed pulse response for an ideal rectangular FM input of width T centered at t=0, and chirp rate T/B us/MHz or an analytic expression for phase(t).
5. Specify the compressed pulse at zero doppler:
maximum -3dB pulse width
maximum sidelobe for |t|<T
theoretical gating sidelobes = -20*log10(TB)-3 dB
maximum spurious for |t|>T
theoretical feedthru = CWFT-10*log10(TB)+6 dB
theoretical triple travel = CWTT -10*log10(TB)+3 dB
maximum s/n mismatch loss
if doppler is significant, then specify these parameters additionally at maximum doppler
6. Specify compressed pulse maximum widths at multiple levels only if truly necessary, and then only at ~10dB above the specified sidelobe level. The main pulse shape is not easily modified. Such specs can drive yields down and costs up.
7. Specify out of band rejection only if truly necessary. Define without tolerance the StopBands in which rejection levels are then specified. Do not extend the StopBands below Fo/2 or above 2Fo.
8. Test is performed in a 50 ohm system with return loss usually unspecified. If necessary, specify the minimum return loss either at F0 or in B. Return loss will always degrade near the PassBand edges. A 1dB increase in input and output return loss will cause a 1dB increase in insertion loss. Do not specify return loss outside the PassBand, assume it’s zero.
9. A polynomial is least mean squared fit to the un-wound measured phase(f), which is then displayed as the deviation from that polynomial. Midband delay (T0) is derived from the first order derivative of that polynomial at Fo, and chirp rate (S0 = T/B) from the second order derivative at Fo.
10. Dispersive filters are usually ovenized for temperature stability. If TCV is the temperature coefficient of SAW velocity, and dTemp is the temperature range/2, then dTemp must not exceed:
for LFM dTemp < .2/(T*B*TCV), and for NLFM dTemp < .1/(T*Fo*TCV)
11. Specify the operating temperature range with care, an excess can drive yields down and costs up.
12. Non-operating temperature range can be -55C to 125C.