AEi Systems
5933 W. Century Blvd. Ste 1100
Los Angeles, CA 90045
+1-702-625-0568
aeisystems.com
This month’s articles cover two hot topics: SSO Noise For Space Qualified (RTAX) FPGAs and SPICE Modeling
SSO Noise Effects in RTAX FPGAs
For some time, SSO performance has been a source of concern for designers due to
poor packaging performance and limited performance data for the space grade devices.
Good design practices and guidelines are needed, but without detailed data and models,
designs are having real issues.
AEi Systems has developed a methodology for measuring the effects of SSO on the
PDN performance of FPGAs and carried out laboratory testing using one of the most
commonly used parts, the Microsemi RTAX2000.
The point of it all - AEi Systems is using the data to create accurate high-fidelity
SPICE models correlated to SSO Noise performance that can be used to perform nominal,
best, and worst case SSO analysis on our customer’s particular designs. The models
can be customized to each FPGA configuration (strength, quantity and proximity of
I/Os, board and decoupling capacitance).
AEi Systems has completed testing of the Simultaneous Switching Noise (SSO/SSN)
performance of the Microsemi SX32 device, a prelude to testing the RTAX2000. A report
on the SX results can be found here, http://www.aeng.com/power_integrity.htm under
“SSO Noise Effects in RTAX FPGAs - Part 1.”
Lessons Learned from SPICE
(Downloadable Presentation)
This presentation highlights many of the pitfalls and issues surrounding IC vendor
and EDA software vendor supplied SPICE models, as well as SPICE modeling in general.
Several examples of poor IC manufacturer models are discussed including those for
Power Mosfets, Voltage References, and Linear Regulators.
The presentation, originally given at the Aerospace Corp.’s Space Power Workshop
Presentation on May 8, 2014 has been enhanced with additional notes and explanations.
There are many SPICE models available online. Unfortunately, availability doesn’t
necessarily equal usability. The majority of these models have no documentation
or detailed description of the model’s capabilities. More importantly, the limitations
are not discussed, at least not in detail. Rarely is there data correlating the
model performance to test data. This all leaves one to wonder what the model is
and isn’t good for and over what operating conditions is the model valid.
There is so much at stake and yet, despite the lack of model correlation, these
models are often presumed to be correct and used without validation because they
are provided ‘by the vendor.’ Some simple simulations, as shown in this presentation,
reveal this to be a dangerous assumption.
Results To Be Published
The RTAX2000-CGA624 device testing will be complete in just a few short weeks. RTAX2000
test data will include:
Background and Motivations
The results of the FPGA testing show the severity of SSN effects and demonstrate
the accuracy of AEi System’s SSO modeling techniques. Our lab measurements show
SX32 SSN in excess of the amounts advertised in the manufacturer’s application notes.
With thirty-two simultaneously switching I/O the noise produced is all but guaranteed
to disrupt static output logic levels, even under best case temperature and voltage
conditions. Under worst case conditions ground bounce and VDD droop exceed 2 volts,
wreaking havoc on digital interfaces.
The collection of test data is necessary to refine models which are used to predict
the performance of various design configurations. Improved modeling allows designers
to avoid potential issues by identifying them early, when they can be mitigated.
The results of our RTAX2000 measurements will be used to improve the fidelity of
our SSO modeling done in SPICE. This modeling will allow us to analyze the SSO performance
of any customer-specific design.
In order to test the RTAX2000 we’ve designed a sophisticated test platform allowing
accurate measurement of SSN and its effects, under many operating scenarios, all
controlled by our custom test software. Our flight-like, 16 layer test board features
fifty RF connector test points allowing us to measure not just SSN, but also part
package parasitics and power plane impedance. We are able to perform similar testing
for any programmable device of interest to you and your projects.
Cheaper than in-house resources
How can we help you? Contact us for more specific information or a no-obligation quotation.