interpolation VS wbuart32

My interpolation tutorial has a section of generating interpolation filters using the Remez Exchange algorithm. It starts at about page 39, and has some nice pictures to help illustrate the concept. It's a bit of a different approach from the standard Parks-McClellan approach to Remez, since the tutorial shows the design of an M-band filter rather than a generic lowpass. This puts additional constraints on the filter design, to the point where the Parks-McClellan approximations don't necessarily make sense any more. In other words--it's strictly Remez.

I have built a lot of open-source C++ tooling for design verification. You can find a lot of my C++ models posted on my Github. Example C++ models include: UART, SPI/DSPI/QSPI Flash, SD-Card (SPI-based interface), VGA Video, Ethernet MDIO, PS/2 mouse, OLED display, SDRAM and more. (I've even simulated PLLs using C++ models ...) I have also written extensively about doing so at ZipCPU.com.

Try this one.

For example, some time after I built my own first serial port transmitter and receiver, someone tried using them in composition: A host (i.e. PC) would transmit a bunch of data, get received by the FPGA, processed by the receiver, and then the data would be sent back to the host via the transmitter. This is a hard test to get right, and my own design failed at the task. (It only returned every other byte!) What I learned from this is that the transmitter must take exactly (10*BAUD_CLOCKS) to transmit a byte. That also means that READY must be high on the last clock cycle of the byte to avoid falling behind. Let's just say that my own serial port wasn't (initially) up to the task. Among other things, the serial port receivers output "VALID" was only one cycle long, and didn't get latched anywhere if the transmitter wasn't ready for it. As I recall, we spent many days scratching our heads at the problem. Eventually, the person using my FPGA switched his host to sending 2-stop bits and things started working. Later, and only a long time later, did I ever find the off-by-one bug in the STOP bit state. My point? Serial port composition is an exacting test, and therefore a good one to work with. You ought to try it.

The core component to the single simulated to TCP port can be found here, in uartsim.cpp. You can find a companion software program that will forward the same TCP port to a hardware serial port here, under the name netuart.cpp.