[ibis-macro] Let me clarify the benifit of getting rid of Use_Init_Output along with Digital Input to Tx GetWave
- From: "Walter Katz" <wkatz@xxxxxxxxxx>
- To: "Ambrish Varma" <ambrishv@xxxxxxxxxxx>, "IBIS-ATM" <ibis-macro@xxxxxxxxxxxxx>
- Date: Tue, 6 Oct 2009 16:47:43 -0400
Ambrish,
We have proven that the ?Digital Input? to Tx GetWave is required.
The whole concept of Use_Init_Output was introduced because Kumar and
Cadence rejected requiring the model to implement the convolution of the
impulse response of the Rx linear filter in the Rx GetWave. Kumar repeated
his objection to doing this convolution in a recent IBIS-ATM meeting.
If Use_Init_Output is True then the flow of using an externally generated
waveform (e.g. Spice simulation) as input to the Rx GetWave is broken. It
breaks the flow because the EDA tool is required to convolve the output of
the Rx Init with the input to Rx GetWave. The input to Rx GetWave already
contains the loss of the channel, so does the output of Rx Init.
The argument that model writers are not capable to do this simple
convolution is condescending to say the least.
I would expect anyone who is capable of writing a model is capable of
implement the code described in
http://en.wikipedia.org/wiki/Overlap-save_method
(Overlap-save algorithm for linear convolution)
H = FFT(h,N)
i = 1
while i <= Nx
il = min(i+N-1,Nx)
yt = IFFT( FFT(x(i:il),N) * H, N)
y(i : i+N-M) = yt(M : N)
i = i+N-M+1
end
So if we get rid of Use_Init_Output, we do not need Init_Returns_Filter
(although SiSoft will be generating models with this as a Model Specific
Parameter, even if it does not become a Reserved Parameter).
I have already spent many man months of effort on Analog Input to Tx
GetWave, and Use_Init_Output, and frankly I do not want to spend any more
time on it.
Walter
Walter Katz
303.449-2308
Mobile 720.333-1107
wkatz@xxxxxxxxxx
www.sisoft.com
Other related posts:
