The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Simulators >> RF Simulators >> QPSS Oversampling with non-50% duty cycle
https://designers-guide.org/forum/YaBB.pl?num=1223563305

Message started by fz2101 on Oct 9^th, 2008, 7:41am

Title: QPSS Oversampling with non-50% duty cycle
Post by fz2101 on Oct 9^th, 2008, 7:41am

I'm simulating a quadrature mixer using QPSS Flexible Balance. The LO signals are rail-to-rail digital signals with non-50% duty-cycle. In fact they have a duty cycle of 25%, with clk1 high (0 to T/4), clk2 high (T/4 to 2T/4), clk3 high (2T/4 to 3T/4) and clk4 high (3T/4 to T).

The settings are as follows:
freq = 100.15MHz, Mxham = 3, Ovsap=1, Tstab=yes.

Strangely, the spectrum (as well as the ifft) of one of these LO signals looks very different from the others (clk2). It has a different DC value and lower magnitude in the fundamental.

I increased the Mxham, and the simulation ran out of memory.

I increased the Ovsap value from 1 to 2, and this seems to solve the problem. What is the reason for this? Does this mean if I change the duty-cycle to 10% I have to increase the oversampling ratio to 5?

Title: Re: QPSS Oversampling with non-50% duty cycle
Post by pancho_hideboo on Nov 22^nd, 2008, 6:41am

fz2101 wrote on Oct 9^th, 2008, 7:41am:

The settings are as follows:
freq = 100.15MHz, Mxham = 3, Ovsap=1, Tstab=yes.
Strangely, the spectrum (as well as the ifft) of one of these LO signals looks very different from the others (clk2).
It has a different DC value and lower magnitude in the fundamental.
I increased the Mxham, and the simulation ran out of memory.
I increased the Ovsap value from 1 to 2, and this seems to solve the problem.
What is the reason for this?
Does this mean if I change the duty-cycle to 10% I have to increase the oversampling ratio to 5?

I think HB_Order=3 is too few for expressing waveform of non-50% duty cycle.
So you should increase HB_Order, e. g. HB_Order=7.
Since HB-QPSS of Cadence Spectre is very inferior than Agilent ADS or Agilent GoldenGate,
I don't use HB-QPSS of Cadence Spectre.
http://www.designers-guide.org/Forum/YaBB.pl?num=1205223090
HB-QPSS of Cadence Spectre consume huge memory compared to Agilent ADS or Agilent GoldenGate.

But if you don't have other simulator than Cadence Spectre, you have no choice.

I don't know details and relations about HB_Order, HB_Oversample and HB_Sample_Number in HB analysis of Cadence Spectre.
The followings are for HB analysis of Agilent ADS.
HB_Order=3, 7, 15, 31, 63, ....
HB_Oversample=1, 2, 4, 8, 16, ....

HB_Sample_Number=2**Int[log2{(2*HB_Order+1)*HB_Oversample}+0.5]

In your case, HB_Order=3.
HB_Oversample=1 --> HB_Sample_Number=8
HB_Oversample=2 --> HB_Sample_Number=16
HB_Oversample=4 --> HB_Sample_Number=32

HB_Sample_Number=8 is too few I think.

Before doing total simulation, check validity of HB_Order and HB_Oversample by simulating only local signal and observing waveform using IFFT.