Podcast: Play in new window | Download
Subscribe: Apple Podcasts | RSS
- Chris just got back from Teardown 2019. Shows next week will be with guests that he interviewed during the conference, while Dave is away.
- While there, Chris did two workshops, both with FPGAs and the open toolchains.
- The first was a workshop with our sometimes co-host Piotr Esden-Tempski. The workshop involved the 1bitsquared iCEBreaker.
- This used the Yosys toolchain by past guest Clifford Wolf (holy moly that interview was two and a half hours!)
- It also uses NextPNR, which was discussed by Piotr, Clifford and Dave at Chaos Communication Camp
- Altera is now Intel FPGA, if you didn’t remember.
- Git paper
- VHDL vs Verilog
- “Cross platform code” works between two separate chipsets meant for USB C
- USB C PD
- Dave was actually discussing cross VENDOR code, because of the standard register format between the chips.
- Chris also took a workshop with Sean “Xobs” Cross that involved 3 levels:
- Micropython
- C written for a RISC V processor (soft IP core)
- Writing hard logic for the FPGA
- Xobs is co-creator of the Novena laptop and the creator of the Fomu project.
- The Fomu was inspired by the Tomu, a project of past guest Tim Ansell
- Raspberry Pi 4 announced with new specs. Dave not impressed.
- The new Sparkfun Artemis Module puts TensorFlow lite onto a Cortex M4 processor with Bluetooth.
- This is similar to the Audeme project (in a much smaller form factor) by Gerald and Bertrand.
- Dave has made a video about how voice recognition was done in the 80s.
- Udemy is moving to a model where creators must be exclusive. The video is by past guest of the show Robert Ferenec.
- Chris is looking at buying a simple VNA.
- The “Buy/Sell section” of the EEVblog forum links to auctions. Dave recommends finding labs that are shutting down.
- Past guest Derek Kozel suggested to look at the HP 8753, based on this wonderful reference site.
- Beware the phrase “Pulled from (a) working environment”
- One last thing from Teardown, Kate Tempkin did an amazing talk about USB analyzers and software around decoding USB traffic.
- Building a STM32 open source multimeter.
- Simone Giertz (yep, also a past guest!) built a “Truckla” and showed how she did it!
Thanks to Billie Ward for the image of the “fantastically overpriced junk”
Aside from the PocketVNA and the miniVNA Tiny/PRO.. has anyone tried building Henrik’s open source VNA?
http://hforsten.com/improved-homemade-vna.html
tomtektest did a 4 part review miniVNA Tiny:
https://www.youtube.com/playlist?list=PL2u6aYAXZ0Z_7lxh-ByKuh1GH08HPIy5u
I think Dave is right: if it is free, most users don’t care if the FPGA toolchain is open source. But some aspects were not mentioned in your podcast. If it is open source, you could inspect all code and no hidden information could be injected into the bitstream, for example a backdoor on some reserved pins to read out the internal SRAM. But this might be only interesting for a small number of people. I think the real benefit of an open source toolchain is for the vendors.
As Dave mentioned in the interview, there are probably thousands of programmers writing the big official 20 GB IDEs. If there were an open source toolchain for everything, from synthesis to placement and routing, the vendors wouldn’t need that much programmers anymore, and can spend more money on the things that make money for them: developing and selling their FPGAs.
Ideally they would all publish their bitstream specification and all contribute back to the open source projects. But they could still add their secret sauce to the open source project without releasing it, if it is a BSD like license, and only releasing it in binary format. But all the work like synthesizing VHDL/Verliog would be the same for all vendors, and sharing it as open source would benefit every vendor.
Some vendors don’t already use their own synthesis code, but just integrate Synplify, like Gowin, for their nice FPGAs with 8 MByte integrated SDRAM, with their (free) YunYuan IDE. The step toward using an open source synthesizing program instead is not big, and vendors who do this will have an advantage because of less programmer cost, and less third party program license cost.
It is like with GCC. Companies like Microchip are using GCC inside their IDEs, and other vendors as well. No need to re-invent the wheel again and again.