Saturday, April 17, 2010

MEP_Rx1.3 Added Bob's Comments about adders

Here's this morning's additions to the doodles.

Thursday, April 15, 2010

FPGA 1.2 drawing and questions

Here's our notes from this evening's conversations about the MEP receiver, using an FPGA.
-Michelle W5NYV

----- Original Message ----
From: Paul Williamson <>
To: Bob McGwier <>
Cc: "" <>
Sent: Thu, April 15, 2010 9:15:07 PM
Subject: Re: [Mep-dev] FPGA 1.1 drawing and questions

This sounds really useful! Can you provide more details, or a reference to a paper, or the source code for an implementation, or something we can refer to?

73. -Paul KB5MU

On Apr 2, 2010, at 7:16 AM, Bob McGwier <> wrote:

> In the FPGA prior to the costa's loop you will need a nonlinear
> equalizer and DC elimination. The QSD, done in this case with some
> fairly high speed switches and baseband filtering most likely, will
> have both amplitude and phase imbalance in the I/Q legs and further, it
> is frequency (rate of switching the sampling "capacitors") dependent.
> I have devised an algorithm which really makes the difference in the
> performance of these devices (out of necessity). Especially for digital
> signals, this is imperative.
> Suppose Z is the incoming signal AFTER the QSD. Z* (complex conjugate)
> is taken, and a filter is applied to Z*, and the applied to a delayed
> copy of Z. This sufficiently eliminates the image rejection imbalance
> to make it an unimportant part of the communications system.
> An ASSUMPTION of the algorithm is that there is NO DC component, so it
> must be eliminated before F(Z*)(n) is applied to Z(-n) and the
> reestimation of F is done.
Mep-dev mailing list

Saturday, April 10, 2010

Add NEON support to FFTW - draft

We started working through the existing code in the simd directory of fftw, and here is what we learned about simd-sse.h. Our comments are in black serif text. We got about halfway through before other chores interrupted.

The plan here is to understand what the current set of simd code is doing, and then write code for neon to present for review. Several simd processors are already supported by FFTW, so learning how they work will help in creating code for NEON.

If you would like to participate, grab the latest version of FFTW source code (we're working with 3.2.2), go to the simd directory, and walk through the code. If you are already a black belt in FFTW simd code, then by all means speak up - we'd love to accelerate the process! We'll keep working on simd-sse.h until we understand it, then will move on to another header file. When all the header files are understood, we'll write down what we think a neon header file must accomplish.

-Michelle W5NYV

Thursday, April 8, 2010

Minor revision to Neon Test Tutorial - font change

If you didn't like the Papyrus, then I hope you'll like Helvetica.

I'm a fan of the movie, myself.
-Michelle W5NYV

Potestatem obscuri lateris nescis.

Evaluation board ordered - Direct Quadrature Demodulator from Skyworks

Here is the data sheet for the evaluation board ordered for MEP and other local microwave groups that may want to experiment on 3GHz with direct digital downconversion. The frequency band selected was 3200-3900MHz.

Datasheet here:

I need a recommendation for an A/D converter to feed our I and Q signals to. The FPGA would pick things up from there.

-Michelle W5NYV

Potestatem obscuri lateris nescis.

Wednesday, April 7, 2010

Neon Test Tutorial - published version

Here is the published version of the NEON test code tutorial. Next step, add neon support to FFTW.