Live on CrowdSupply! muArt - A UART converter finally done right

[pylo]

The campaign has ended and the product has shipped to backers.
Visit Crowd Supply to get yours too

UPDATE 5: Manufactured electronics arrived and work perfectly. Details in post #63. Also updated text below for current status.
UPDATE 4: Prototype cases arrived. They are great, details in post #58.
UPDATE 3: Added photorealistic render of planned plastic case. See post #42 for more details.
UPDATE 2: Added schematics as promised.
UPDATE 1: The forum replaced all my Greek "mu" with "?". I am now correcting it, remember wherever you see "muArt", "mu" is for the greek letter.

Okay, I know what you think - “Another one?!” - But in my experience, available UART converters, even the isolated better ones, still do not cater for all needs, at least surely not for mine, so I set out to make my own device. Let me introduce you to the muArt (“myuArt” - as a wordplay on both micro and UART).

My goals were to make the muArt safe, reliable, universal, and comfortable. Obviously these needs arise from experiences I've lived through. Let me elaborate on these points:

Safe
Accidents happen, especially during development, experimentation and debugging. Cables can be connected the wrong way, polarities reversed, two drivers interconnected, wrong voltages selected, or simply just a faulty component or a mistake in a design. These often mean the end of your computer’s USB port… or the converter’s… or your creation’s… or yours. I want my ideal UART converter to be safe for other devices and the user, even in the face of hardware faults or user mistakes like mixed up wires. Oh, and since it is sitting on my bench and sometimes (*cough*) with pieces of metal, solder and other boards laying around, I’d appreciate if it didn’t cause any shorts itself. To support this goal, the muArt has:

• Complete galvanic isolation, even for mains-level and higher voltages. Electric currents cannot flow between your computer and the other board. This spells safety for you and your computer, while also helping to keep noise levels down.
• Reverse polarity protection. The converter will survive even if you mix up the power pins.
• Over-current protection on all peripheral IO pins. With the muArt it is not a problem if you wire it up wrong by mistake. Both the converter’s and your other board’s pins will survive.
• A professional custom-designed case to prevent damage, shorts, and other troubles. Use, grip, handle, and transport the muArt with ease and without worries.

Reliable
A UART converter is one of my tools, and I should definitely not be debugging it instead of my creation. I want my tools to be reliable. I hate that minute while I’m trying to figure out why I get no data, just to find out that something (probably?) latched up and only needed a power cycle. It is frustrating when my terminal gets garbled up due to floating inputs upon device (dis)connection. And it is far from funny when my last converter turns out to be broken - either due to my own mistakes (see list above) or just plain who-knows-why. Did I mention USB protocol disconnections due to other devices on the bus doing their thing? To support this goal, the muArt has:

• Pullups on all inputs, so that disconnected pins don’t cause funny things like garbage random input, a ruined terminal app, or corrupt application state.
• Judicious filtering. Each IC in the converter is decoupled using multiple capacitors, both locally and in bulk. The USB signals and power are filtered, and sensitive pins are protected with RC networks.
• Complete ESD protection for all external pins. ESD damage is tricky because things can seem to work for a while after an ESD event, and you don’t need visible sparks to have them.

Universal
My ideal converter should give me everything I need for basic communication (obviously), also for handshaking, and importantly for flashing other boards. It should support all relevant voltages, notably 1.8V, which aside from working with very-low power devices, is a must-have for FPGA designers. The converter must also lend itself to low-noise circuits, enabling communication with sensitive analog systems without disrupting them. To support this goal, the muArt has:

• RXD, TXD, DTR, RTS, and CTS. Usable for flashing various ESP32 boards and multiple Arduinos such as the Nanos and Pro Minis. Hardware handshaking allows for reliable high-speed UART.
• A wide working voltage of 1.8 to 5.4V. Yes, also usable for FPGAs. Coupled with the galvanic isolation, you basically get “voltage auto-sensing”, meaning the converter will automatically use the correct voltage levels for UART communication. No more setting jumpers or “Which of these 5 converter boards was the one for 3V3?”. One converter to use everywhere.
• Suitability for communicating with low-noise circuits, supported by the filters, the isolation, and careful layout.

Comfortable
Readable labels on pins never hurt (sarcasm). LEDs should let me know when there is activity. Since UART communication is probably the first thing you try on board bringup, some extra feedback or GPIOs can be extremely practical. Mechanically, I want to be able to actually grab the thing without making shorts all over the place and without making my jumper wires come loose. To support this goal, the muArt has:

• A general-purpose in and a general-purpose out pin that you can control directly over USB, parallel to the UART lines.
• Four LEDs, each of different color to make them immediately distinguishable. One lights up if power is connected, two indicate RX/TX activity, while the last one is connected to the general-purpose in, giving you visual feedback about its state without special software on the host computer.
• Four-wall shrouded header. Gives protection from shorts by tipped board or dangling wires. Also gives a nice firm surface to grip the converter by, again without the risk of shorts or ESD damage from your fingers.
• Readable descriptions printed on the PCB to help you remember the most important things about usage.

Wow, this turned out lengthy, sorry about that. So, where am I with the project? The design and layout is done and manufactured prototypes have arrived, they have been tested, and every feature works as it should. Now I would like to crowdfund a batch if there is enough interest. I’m trying to get about 200 pre-orders, because that amount would suffice to make the price just barely reasonable (estimated crowdfunding price is around 30$). Preparations for the campaign are ongoing so the campaign is not online yet, but I will let everybody know once it is.

Do you like the muArt? Would you buy one / would you support the crowdfunding campaign ? I'd be happy to see your feedback

[ataradov]

I'm not sure what you are doing with UART that is so hardcore, but 3.3V FTDI cable covers all my requirements.  And it is very compact and reliable enough.

I would like to see some design files before I commit to something with so many promises.

[os40la]

Sounds Interesting. I have been looking around for something like this for a reasonable price. What speed would we be looking at?. From my research galvanic isolation seems to limit the speed.

[Mr.B]

I would like to see some design files before I commit to something with so many promises.

Me too.

[pylo]

I'm not sure what you are doing with UART that is so hardcore, but 3.3V FTDI cable covers all my requirements.

The muArt is not about industrial or any "hardcore" usage. It is just one board that is universal enough to be used everywhere while giving you some comfort and handy features. I think the FTDI cable you mention is a very good example to support my campaign, actually. It might be robust enough, but each cable only supports a single voltage, and each cost about 22$. That is near the price for my board, but from FTDI you might need 3 of those cables (different voltages for different projects), so you've already spent over 65$, and you still don't get all the handy stuff from the muArt, like the DTR pin (needed for flashing ESP32 boards), the GPIOs, LED activity feedback etc. If your hardware projects are different enough, you will have 3-5 different converters in the end and have paid 70$, whereas you could just have bought a muArt. That is what this project is about, not about being hardcore. EDIT: The safety and robustness features are mostly just protection against common user mistakes that most other boards do not provide.

Alternatively, you can go to Amazon or eBay and buy those 2-5$ chinese import boards, but those are not nearly as robust as the FTDI cables, and otherwise lack the same features.

I will publish schematics, at latest sometime during prototype manufacturing, possibly earlier. There is nothing magic about the board that, just simple stuff that anybody could do. I think it is just that most converter boards try to be as cheap as possible, while I prefer to be higher quality even if costs a bit.

[pylo]

Sounds Interesting. I have been looking around for something like this for a reasonable price. What speed would we be looking at?. From my research galvanic isolation seems to limit the speed.

Correct, optocouplers tend to be the limiting factor when it comes to speed. That is why the muArt uses magnetic coupling, and theoretically it should achieve 3Mbaud/s (supported by both the USB bridge and the isolation). While I calculated the filters for 3Mbaud, those could also limit speed if I made a mistake. So my intention is 3Mbaud, but let me test my prototypes first before making any promises about this.

[pylo]

By the way, the key components are the CP2102N and the MAX14932. The latter will be replaced ofc should the 3Mbaud speed turn out to be unachievable.

[spanner888]

The auto voltage sensing has got me especially interested.

Other features also sound great.

I have heaps of cheap ones already, but will definitely pay that sort of $A.

[janekm]

Personally I avoid FTDI like the plague after their driver debacle (and their obnoxious distribution policy), so I'm glad you went for CP2101 

Lately I've started to just place the Holtek HT42B534 USB-UART chip on my boards since it is literally the cost of 5 pringles chips and works well, this way I can get users in the field to read out diagnostics without having to ensure they have a separate cable / module.

It's fairly rare for me to need an isolated USB-serial adapter (though there's been one case) so I might hesitate to pitch in... I suspect this project may turn out to be a little niche (but maybe 250 orders is realistic). Good luck anyway, I'll be glad to see this exist in case I need it in the future 

[pylo]

The HT42B534 is new to me. Looks interesting, the IC has good features and is priced very aggressively. But I cannot find a reliable distribution channel (actually the only place I've found it was alibaba), and also no information about a working driver for Linux. If the latter does not exist, that'd be a big and serious no-go for me.

[pylo]

I added a render to top post. Physical board size is less than 53 x 24mm. Real-life images follow once I get the prototypes.

[Cerebus]

... and also no information about a working driver for Linux. If the latter does not exist, that'd be a big and serious no-go for me.

It shouldn't need one. Quoting from the landing page for the device from Holtek's website: "Communication Device Class (CDC) for communications and configuration". CDC class drivers are completely generic (by design) and I can't think of a major operating systems that doesn't natively support the CDC profile for a basic serial port straight out of the box. I'd want to test it before relying on that, because you can't be sure that they have implemented the standard properly, but if they have it should "just work^TM".

[janekm]

... and also no information about a working driver for Linux. If the latter does not exist, that'd be a big and serious no-go for me.

It shouldn't need one. Quoting from the landing page for the device from Holtek's website: "Communication Device Class (CDC) for communications and configuration". CDC class drivers are completely generic (by design) and I can't think of a major operating systems that doesn't natively support the CDC profile for a basic serial port straight out of the box. I'd want to test it before relying on that, because you can't be sure that they have implemented the standard properly, but if they have it should "just work^TM".

Indeed, that's why I like it. At least on relatively recent versions of MacOS and Windows 10 it just works without any driver install needed (older versions of Windows need a .INF file to tell it that yes, it can go ahead and use the default CDC driver). I would not expect any issues at all under Linux either, though I haven't had a reason to test that yet.

A quirk of using the CDC class is that it will be considered a USB modem by the OS, but I don't think there are any practical downsides to that. It's possible to write to the chip to overwrite the default name it comes with.

In terms of distribution, it's true that Holtek don't have much "western" distribution (their chips mostly go into low-cost consumer goods, for example Dave is using a Holtek LCD driver chip in his latest project), but you may be able to get the chips directly from Holtek, or they should be able to tell you who their local distributor is. But they're a long established Taiwanese company so not a fly-by-night outfit likely to disappear any time soon (and in fact they have very little volatility in their offering, they don't sunset products very often from what I've seen).

[BNElecEng]

Hi pylo. I'd be interested in a couple of your boards. Please let us know how your prototypes turn out and when the crowd sourcing campaign starts.

[pylo]

@BNElecEng: Thank you, sure, I will let everybody know.

@Everybody: Please do let me know if you are interested. You need not promise anything, I'd just like to know if you think such a board could be interesting. Critique is also welcome, that is the whole point of asking for feedback.

[pylo]

@ataradov: As promised, I've attached schematics to the first post.

[ogden]

@Everybody: Please do let me know if you are interested.

Thank you, I am fine with non-isolated converters I have. Also I do not agree to "done right". Mine generic non-isolated converters have better insulation than yours. Why? - They are inside heat shrink tubes. Due to offset USB connector and UART connector which is not edge-mount, your board is not compatible with heatshrink tubing.

[T3sl4co1l]

Le sigh, USB still being used badly. 

Just tie shield and GND directly to GND.  Only bad things happen by introducing impedance.  Shield in particular should be tied to circuit ground plane as soon as possible, and in as many places as possible.

There are a few use cases where enclosure ground should be used to handle ESD, with a little isolation between that and internal ground, but such is the domain of experts.  It's definitely not the way to handle a dongle, which is its own ground period.

(And yes, I wish I could address this towards whatever appnotes you read that promoted these exactly wrong ideas.  Pass it on, if you will.)

Tim

[pylo]

Mine generic non-insolated converters have better isolation than yours. Why? - They are inside heat shrink tubes. Due to offset USB connector and UART connector which is not edge-mount, your board is not compatible with heatshrink tubing.

Mind you, we are talking about two different cases of isolation. The heatshrink tubing you refer to does not actually solve any of the problems the galvanic isolation in the muArt does. Galvanic isolation prevents dangerous voltages or high currents from reaching your host computer and isolates noise, a heatshrink does not do these. What the heatshrink does, it prevents damage (by ESD or shorts) to the uart-converter itself. Galvanic isolation protects not the converter, but you and everyhing else.

That said, I was already thinking about adding heatshrink tubing too (not instead), and I'm not sure the offset USB receptacle would prevent it. However, it is surely no problem to modify the board layout to make it centered. Maybe I'll do a quick layout revision just to be sure. Thank you for pointing this out.

[Cerebus]

@Everybody: Please do let me know if you are interested.

Thank you, I am fine with non-isolated converters I have. Also I do not agree to "done right". Mine generic non-insolated converters have better isolation than yours. Why? - They are inside heat shrink tubes. Due to offset USB connector and UART connector which is not edge-mount, your board is not compatible with heatshrink tubing.

I think you're confusing insulation with [galvanic] isolation - the two are not the same thing.

Oh, and the heatshrink thing - what?. I hardly think that someone not explicitly designing their product to suit bodger's methods is a problem.

[ogden]

I think you're confusing insulation with [galvanic] isolation - the two are not the same thing.

Oh, really? - Thank you for clarifying 

I meant insulation indeed. Funny that my native language (which more or less obviously is not Englisch) does not have separate words for isolation/insulation.

Oh, and the heatshrink thing - what?. I hardly think that someone not explicitly designing their product to suit bodger's methods is a problem.

You may work using bare wires and bare PCB's of your tools all over your desk - I don't mind. Personally I want my tools to be insulated. If you think your product will be used only by beginners which will never work with/on expensive equipment, then indeed make that adapter as inconvenient as possible.

[Monkeh]

So put a 3D printed case on the thing. Wrapping stuff in heatshrink fairly sucks.

[ogden]

So put a 3D printed case on the thing.

So I have to buy 3d printer to get that thing into case? - No, thanx
Actually 3D-printed enclosure would benefit from edge-mount connector as well.

Wrapping stuff in heatshrink fairly sucks.

Visually maybe, thou translucent does not look that bad:

https://www.tindie.com/products/Saimon/debug-board-usb-to-uart-with-voltage-translator/

[pylo]

Le sigh, USB still being used badly. 

Just tie shield and GND directly to GND.  Only bad things happen by introducing impedance.  Shield in particular should be tied to circuit ground plane as soon as possible, and in as many places as possible.

There are a few use cases where enclosure ground should be used to handle ESD, with a little isolation between that and internal ground, but such is the domain of experts.  It's definitely not the way to handle a dongle, which is its own ground period.

(And yes, I wish I could address this towards whatever appnotes you read that promoted these exactly wrong ideas.  Pass it on, if you will.)

Tim

Well, it's not just about an appnote. There are plenty examples (forum posts, schematics, appnotes) advocating the way I did it too. Certainly the most common two schools are 1) on peripheral side connect the shield directly to ground (like you said), and 2) through some network (like I did). Frankly I don't have the resources to do actual EMI testing - on that note, if anybody can help me with that even if non-officially I'd be glad to send a few boards - but based on my research this is not so bad. Maybe one can argue if it could be better, but by leaving the schematics like this, I can always switch to your solution by not placing the cap and replacing the resistor with a 0-ohm. So this is certainly the most versatile. FYI, I made sure to connect shield to ground at a quite area of the board.

[pylo]

On the topic of a case: yeah best would be to have one too, but the options are:

• Let a case of my own be manufactured. Too expensive below 10.000pcs, so not an option.
• Design into an existing case. This would still need to be modified with cutouts for cables and windows for LEDs plus printing for labels. Also, the board is pretty small, so a standard case would make it a lot bulkier (and uglier).
• Design and 3D-print the case myself. First, I have no experience with mechanical design, and second I don't have a 3D printer.
• Ask for help in design, and share the design so that anybody can print their own case. This is not so bad, but most people still don't have a 3D printer, so only a few would be able to benefit from this solution.
• Apply heatshrink. Might not be "professional", but TBH it works, it is cheap, and is available for everybody. And as already pointed out, not so ugly as long as transparent.
• Apply conformal coating to the board. I think this could be workable as far as costs are concerned (though not so sure as it would require a 2-phase assembly), however, this is far too unreliable (abresion, thickness etc.), so I call my veto on this.

Summarizing, only heatshrinking works that I can readily ship to my backers. Having a public design for a custom 3D-printed case would be ultra-cool, but people would need to print it themselves. EDIT: Or, have a backer tier during the campaign where a printed case is included, but sold at a higher price.