FTDI-be-gone

Well, it turns out that Windows still needs a driver. I've put the driver installer package in the files section for those of you who don't run a better operating system.

I actually made this for the isolated variant, because I didn't want to complete the matrix by making an isolated RS-232 variant. What it is is simply an RS-232 level converter and a DB-9 on a board that plugs into the 10 pin connector for the TTL level variant.

It doesn't matter whether the power is jumpered for 5 or 3.3 volts. The converter chip works either way.

It's designed to have a 5x2 .1" DIP header mounted on the bottom side of the board, and plugged into the UART board in such a way that the DB9 is facing the opposite direction of the USB-C connector.

I've updated the design to switch from micro-B connectors to C connectors. In principle, this doesn't really change the design very much at all, except to add the two CCx pull-down resistors. The board has to by physically slightly larger but it's not that big a deal.

The TTL variant has also changed to supply all 8 signals rather than just 4. There is now a 10 pin DIP header that has the same pinout as a DTE DB-9 RS-232 connector would, with pin 10 being the power supply pin. Although this is not the industry standard pinout (the so-called FTDI 6 pin connector), it's far more versatile, and you can use SIP jumpers to make any pinout you might want anyway.

The latest boards for the RS-232 variant have a little problem.

I switched from using !SLEEP to using !POWER for the !FORCEOFF pin. Well, !SLEEP and !POWER sort of have opposite meanings. I had intended to use !POWER because it is an indication of the chip either being in USB suspend OR not yet being successfully enumerated. The trouble is, it's low when that happens. !SLEEP was low when in USB suspend. !POWER is high during USB suspend.

It turns out, though, that I can go back to using !SLEEP for the RS-232 variant, because prior to enumeration, you're allowed to pull up to 100 mA while in suspend you're only allowed 2.5 mA. Using !SLEEP to power off the level converter, therefore, is necessary, but turning off that chip before enumerate isn't.

I also can salvage the boards I have now by remapping the GPIO pins with Cypress' configuration utility.

This won't affect the TTL variant boards that are on the way, because the MOSFET that gates power for the output header really does need an active-low power signal. So there's that.

I've been thinking some more about the TTL (as opposed to the RS-232) variant. The TTL variant is designed to have the traditional 6 pin SIP header output.

Many traditional FTDI style interfaces just connect the Vcc output pin directly up to the output, which in principle violates the USB specification, as you're not allowed to draw significant current before the host has properly enumerated the device.

The fix for this is to switch the Vcc pin on and off with a MOSFET driven by the !POWER pin of the UART. In addition to this, an AP2331 current limiting switch can be used to protect the UART and the host from inrush or other problems. Since the CY7C65213 doesn't run internally at 3.3 volts, if we want to provide 3.3 volt functionality, we'd need to add an LDO. That's actually just as well, since we want to support up to 200 mA of supply current from the Vcc pin. Because of that, we'll specify in the bus enumeration parameters that we require up to 250 mA of bus current. You can do this using Cypress Semi's configuration utility (it's part of their SDK for Windows).

Anyway, take a look at the schematic in the PDF and let me know what you think.

The current version works fine electrically, but some of the silkscreen wound up on the wrong layer, so I've fixed that. Another minor issue is that the GPIO I chose to use for the !FORCEOFF line is defined as !SLEEP, but there's another called !POWER that's a slightly better choice. Lastly, the 10 uF cap is just a little too close to the USB connector for comfort, so it has to move just a touch North.

Other than that, the design is perfect, and the next step is designing the 3D printed enclosure for it. I'm not sure how that's going to happen, as I have zero skills there.

The first boards from OSHPark are on their way. There is on error on the boards, but it's not very important - the "Geppetto Electronics" is on the top silk layer instead of the bottom, so it's interrupted and covered by all the components. The next version will fix that.

The OSHPark shared project board is here, but because of the QFN chips, you should use a stencil for the pasting. I got one at OSHStencils. You can make one from the paste gerber using the EAGLE files in this project's file section.

The TTL variant is being processed by OSHPark right now.