E-bay usb/serial. Is this one safe ?

[ae.vasconcelos]

Hi,

Just got this adapter from e-bay, it is called "USB to serial adapter PL2303 TTL console Recovery RS232 for Raspberry Pi", but I know better than believe in everything I read in e-bay. The adapter is this one.

I removed decades of dust from my analog multimeter (duracell batteries rule ) and measured the pins. Between the GRD and TXD I have something between 3 and 4 V, probably 3.3V. However on the red pin I have 5V.

Is this normal and ok ? Is this on safe to connect to the Pi ?

Thanks in advance.

[scruss]

Yes, the red wire is supposed to be at 5V. Should be okay.

I would have preferred to choose an adaptor that used an FTDI chipset. They're a bit more reliable than a PL2303.

[rurwin]

Don't connect the red wire. Otherwise the power supply of your PC will fight with the power supply of your Pi.

[ae.vasconcelos]

Don't connect the red wire. Otherwise the power supply of your PC will fight with the power supply of your Pi.

Hummm, didn't knew that one. Thank you.

Who wins this kind of fight ?

[Douglas6]

I think it's safe to say: you lose.

[rurwin]

In the red corner is a PC power supply rated at somewhere between 300 watts and a kilowatt, capable of driving a 5V line in excess of 20 amps.

In the blue corner is a USB power brick rated at 3 to 5 watts and a Pi polyswitch "fuse" which will trigger if more than about an amp flows and a Pi power plane rated for about the same.

Between them is that red wire, rated at around an amp.

The two power supplies will have slightly different output voltages, even if that's a tenth of a volt or less. The PC supply will force the USB power supply to exactly the same voltage as itself and will supply or sink as much current as necessary to do so.

At those currents, there will be significant resistance in the red wire and the Pi power cable etc., which will tend to even out the voltage mismatch, so it might not be as violent as I make it seem. My guess is that the polyswitch "fuse" will trigger, but there may be smoke.

After the fuse triggers, instantly or after a few minutes, you will notice no difference, as the Pi is now powered by the PC. But when you unplug the serial interface, the Pi will be mysteriously dead for a few hours. Try that too many times and I believe the untriggered resistance of the polyswitch fuse will rise and you will start to get issues indicative of an underpowered supply.

[jojopi]

I used to leave the red wire disconnected because of the voltage difference issue. Then one day I realised that when the Pi is powered off, the PL2303 continues to supply 3.3V on its RX pin. Also, if the PC is powered off, the Pi's TX supplies 3.3V to the unpowered PL2303. Holding logic pins high when a chip is unpowered is not considered good, so I took to connecting the red wire.

(We now believe that the Pi's GPIO will not draw significant current at 3.3V and try to power the whole chip, but I have not checked the PL2303.)

USB ports are required to have resettable current trips of less than 5A, so there is a polyfuse in the PC as well. Between the two power supplies there are two polyfuses, the tiny red and black wires with total resistance over 2Ω, plus the power supply's cable. I would be surprised if there is enough current to trip either fuse, let alone produce smoke. I measure about 60mA.

The manufacturer's literature I have found says that resistance jumps are not cumulative after the first trip or soldering event. The only things that should cause permanent damage are overvoltage (~13.2V) and overcurrent (~100A).

[Douglas6]

Thanks to all the great sources of information here, one day I hope the knowledge will sink in to my brain. No wonder people tell me I"m not grounded.

[ae.vasconcelos]

Thanks to all the great sources of information here, one day I hope the knowledge will sink in to my brain. No wonder people tell me I"m not grounded.

Yes. I have a lot to think about.

I can see the logic of having a 5v wire on a serial connection when doing very low level stuff, like connecting to a jtag debug interface to some motherboard. But a serial terminal in a unix machine, even in a Pi is not a low level interface so I would never tough about connecting it and expecting the computer to be powered on. I guess I would have been quite surprised...

But even so, is this wise ? I understand that everything was done to keep the cost as low as possible, but maybe something like a jumper should have been included to address this kind of problem. I understand that this can happen trough the usb port too...

But what really bugs me is the fact that there is no hw flow control, and at 115200 bps/s... Who needs a console at 115200 anyway ? I think there is no provision for flashing the card trough the console, so, why the 115200 default? 9600 should be enough.

[jojopi]

But what really bugs me is the fact that there is no hw flow control, and at 115200 bps/s... Who needs a console at 115200 anyway ? I think there is no provision for flashing the card trough the console, so, why the 115200 default? 9600 should be enough.

You can change the console baud rate in /boot/cmdline.txt if you think it is too fast. However, 115200 baud is only 11.25KiB/s; about twice the speed of an analog modem. The Pi's UART has a 16 byte FIFO that interrupts when it is getting full, and the Linux kernel buffers about 4KiB for you. Even with no flow control, you only need to read the port three times a second to avoid overrun.

Hardware flow control, specifically RTS/CTS, can be enabled using GPIO31/30 on P5. But you would need a different adapter, quite possibly a USB-RS232 plus MAX3232 level converters, to make use of this. It is really not worth the trouble.

[ae.vasconcelos]

Even with no flow control, you only need to read the port three times a second to avoid overrun.
Hardware flow control, specifically RTS/CTS, can be enabled using GPIO31/30 on P5. But you would need a different adapter, quite possibly a USB-RS232 plus MAX3232 level converters, to make use of this. It is really not worth the trouble.

Yeah, I guess that times do change and do not stay the same. I remember when I had do use a FOSSIL driver in MS-DOS to write a terminal emulation program in pascal. At the time most UARTs were 8250 with no FIFO at all.

[rurwin]

I remember writing a program on an IBM PC/AT to download a program to an EPROM programmer. At 19200 baud I had to resort to assembler, because compiled C was not fast enough. (No handshaking and only 1 byte buffering in those days.)

[ae.vasconcelos]

I remember writing a program on an IBM PC/AT to download a program to an EPROM programmer. At 19200 baud I had to resort to assembler, because compiled C was not fast enough. (No handshaking and only 1 byte buffering in those days.)

Never had to use those beasts, I think that the oldest one I saw already had a parallel interface.
The 286 and specially the 386 wasn't bad machines for real-time stuff. But assembler was required to most low-level stuff.
The BIOS provided a lot of functions but I think that most (if not all) were too slow for real-time stuff.

That's a feature of the Pi that I find amusing... It doesn't not have a built in BIOS.

[catalincf]

I got today a RS232-TTL with the newer MAX3232.
In regards with the safety topic concern, I wonder if I should power the converter with the 3V3pin or 5V pin from the RPi? Maybe some one can inform me as well..
Digged on the net for weeks to understand the voltage scheme for such simple 4 wire connection usage and found a post regarding the above board/converter.
''The unit is powered through one of the connector pins and can run at any voltage i.e if the board is powered at 5V, the unit will convert RS232 to 5V TTL. Power the board at 3V3 V and the board will convert RS232 to 3V3 CMOS TTL.''
And from this quotation seems very simple. When I supply 5V from the Rpi to the board plugged in my Laptop RS232, all the signals on the pins RX TX of the Rpi should be 5V and similar when power the board with pin 3V3 of the RPi the RX and TX pins will carry 3V3 signal. But something is wrong, the pins RX and TX of RPi should only have 3V3 as 5V would be BAD. So how come the post in the link below is powering the board with 5V from RPi and does not fry the chip though the pins TX and RX that should carry 5V???
[url]http://www.davidhunt.ie/?p=3091[/url]

thanks to the GURU who will lighten up my doubt.

[rurwin]

You are correct.

One of the comments points out the same mistake. It is a mistake -- the device should be powered by 3.3V, unless it has an on-board voltage regulator, which is unlikely.