Talk:RS-232/Archive 5

Reduced voltage issue

The History section talks about "reduced voltage" circuits using +5 and -5 volts and implies those don't conform. But the article later on states that valid signal levels go as low as 3 volt absolute value. So a circuit that actually generates +5 and -5 would appear to conform (provided it's tolerant of significantly higher input voltages). Perhaps the article meant "TTL RS232" [sic] found in some older PCs, where the signaling uses TTL levels. That indeed isn't conforming, because one of the two signals is around 0 volts rather than negative, and because the input doesn't accept negative voltages either. Paul Koning (talk) 12:21, 24 May 2017 (UTC)

The article is incomplete in its coverage of voltage requirements and may be a bit confused in the description of the problem.

The spec actually covers both the voltages that an output pin should provide (+5 to +15 and -5 to -15) and the voltages that an input pin should accept (+3 to +15 and -3 to -15). (And any pin, whether input or output, should tolerate an applied voltage of up to +/- 25V without damage.) You won't be able to review the actual TIA-232-F document without buying it (last time I did it was close to USD 100, and my DRM'd copy got lost in a system upgrade) but you can find a pretty authoritative summary here. See Figure 3.

The difference between 5 and 3 volts is supposed to provide noise immunity, and the higher the output voltage swing, the more noise immunity you have. A circuit that runs from a +/-5 volt supply will commonly emit only +/- 4.5 V due to forward voltage drop, and there are other things that can further reduce the output voltage swing.

The above paragraph addresses only the DC domain. Things get more complicated when you consider signal rise time, inter-circuit noise radiation and pickup, etc. A wider voltage swing is not necessarily better, especially at higher speeds. The TI paper I linked above covers some of this.

But to make a long story short, an RS232 interface that claims to provide only a +/- 5V voltage swing is often just too close to the edge of the spec, esp at higher speeds. You can argue that per the spec it ought to work, and you'd be right, but it often doesn't. Jeh (talk) 18:05, 24 May 2017 (UTC)

• I'm a bit rusty on this, but RS-423 is basically the 5V version of RS-232. If memory serves from about 30+ years ago, it was used on systems which were predominately standard 5V TTL circuits, and essentially a cheaper alternative to implementing full 232. As far as I can remember, you could normally directly connect 232 and 423 interfaces, as long as they were both properly to spec (since the lower and upper limits, respectively, overlapped). Top speeds of the 1980s 423 devices were 9600 or 19200 bps. Murph9000 (talk) 03:22, 27 May 2017 (UTC)

RS-423 describes the electrical characteristics of one signal without saying what that one signal is used for (e.g., for sending data one direction, for sending data the other direction, for flow control, or for turning on the light in the garage). 72.22.12.62 (talk) 17:43, 10 November 2017 (UTC)

Vulnerability to Electromagnetic Interference (EMI)

Based on what I've read in the book "USB Complete" by Jan Axelson, if anything, USB should be less vulnerable to EMI than RS-232. Beyond "low speed" operation, USB is required to have high quality shielding. RS-232 does not require RS-232 cables to be shielded at all (though many cables are shielded slightly, anyway). In all cases, USB devices must use differential signaling. That fact alone should make USB less prone to EMI, and it is in practice. RS-232 almost never uses differential signaling, and if it operated at a higher speed than it does, it almost certainly would corrupt data easily.

Does anyone else have something to say about this?

74.178.57.187 (talk) 23:12, 24 February 2018 (UTC)

It isn't really straightforward to compare the two except as complete system, checking feature one against the other doesn't really give any sense of how vulnerable the link is. USB was designed with economic factors very firmly in mind - it had to be cheap to implement - whereas this was much less of a concern for RS232 when it was designed. Therefore when features such as shielding are added which add appreciably to the cost ask why they are needed as opposed to simply counting them as a good thing.

Sure, USB works over a twisted pair and has screening, those are both things that help with noise immunity, although the screening is arguably less useful than it was now laptops (lacking a ground connection to anchor the screen to) are so prevalent. RS232 has neither baked into the spec. However compared to USB RS232 has much higher working voltages (needing a bigger disturbance to alter the interpretation), much greater currents (the more current the less interference can pull the voltage levels from where they should be), has much lower slew rates (less mutual interference between connections) and the lower speeds allow brief glitches to be discarded, both through active techniques such as majority voting and simple passive properties such as the capacitance of the receiver. USB needs the features you cite to get an acceptably reliable transmission at the levels, currents and speeds that it operates: it doesn't make it "better" than RS232 in this regard, that is simply needed for USB to work at all: the higher speeds tend to make the error margins much tighter.

Yes, you do notice the occasional error on an RS232 link, mostly on longer links that have more scope to pick up interference. You can't really hold that against it compared to USB when USB would be unable to span such distances in the first place. Even with the twisted pair and the shielding in place the standards acknowledge that errors are still going to be detected - checksumming and retransmission are an integral part of the standard and have been since day one: if the immunity was that good this complexity wouldn't be needed. 3142 (talk) 11:51, 19 February 2019 (UTC)

Common ground

Since my edit yesterday was reverted maybe we need to be a little more explicit. From the article:

Use of a common ground limits RS-232 to applications with relatively short cables. If the two devices are far enough apart or on separate power systems, the local ground connections at either end of the cable will have differing voltages; this difference will reduce the noise margin of the signals. Balanced, differential serial connections such as RS-422, RS-485, and USB can tolerate larger ground voltage differences because of the differential signaling.

Using USB as an example here is completely misleading. Sure the signal is transmitted on a differential pair, but the reference is to the bus as a whole otherwise the comparison is meaningless. USB (prior to version 3 for simplicity, although the same issues hold there too) consists of four wires - the differential pair, +5V, and, erm... a common ground. It is also limited in terms of the length to a quite hard 5m (~16ft) by propagation delays. RS232 has no specified length limit - the preamble to the standard mentions 50ft as a target but 1000ft+ is easy enough in practice.

So, we are comparing the two and saying USB is superior because it lacks something that it has, and this missing thing that is present makes it better because it is worse at the very metric we are using as justification.

No, that is not a valid example. 3142 (talk) 20:22, 18 February 2019 (UTC)

But the USB bits aren't "read" by reference to the common ground. As in any good differential scheme, they are compared against each other, so ground differences at the end don't matter. RS-422 and -485 have a "ground" pin in their connectors too, you know. Jeh (talk) 12:48, 19 February 2019 (UTC)

Once the link is up and running and transferring real data as opposed to bus enumeration metadata. The bus as a whole is dependent on a common ground, as is the initial connection even on the differential pair. Since I discussed USB2.0 in the first instance we'll stick with that, since it's still current and keeps things compared to USB3.0. Read through section 7.1 of the specification and take a good look at figure 7.1, there are a couple of other places this is covered but will stick with there. You see a pair of pull down resistors to ground terminating the pair, fine, that's standard differential stuff. However we also have a switchable pull-up resistor on one of the lines used for signalling the device presence and ultimately triggering the enumeration process. We also have the single-ended receivers necessary to detect those unbalanced conditions. The text also makes this explicit:

This state is achieved by using the low-/full-speed driver to assert a single ended zero...

transceiver activates an internal current source which is derived from its positive supply voltage and directs this current into one of the two data lines...

The link isn't balanced in those conditions, the common ground is necessary. Without it the link can't be established and no data can transfer. It is absolutely essential to the protocol even without the requirement to supply power.

I would suggest referring to the RS-422 and RS-485 standards. Neither specifies a common ground. One is often included for a few possible reasons - I've seen shielding, power supply and unbalanced handshaking signals all done on the same connector as both of them, but the standards make no mention of one. Both are fairly barebone standards (e.g. lack of any pinouts or even connectors) so a practical implementation has to build on them by necessity.

Ultimately though this is getting far more technical than it needs. Remember, the article makes two assertions by citing USB as an inappropriate example:

• USB does not use a common ground.
• USB can span greater distances than RS232.

Neither assertion is true. How can it be a valid example? Yes, it is a shit one and there is nothing wrong with calling it as such. Refer to WP:CIVIL, that is not a personal attack but simply calling a spade a spade. 3142 (talk) 15:06, 19 February 2019 (UTC)

Extra DB-25 pins

The standard defines a whole heap of extra connections on DB-25. These obviously cannot be used on a DE-9 connector.

Perhaps a description of when these extra DB-25 connections are used and what sort of systems use them?

See: pinouts.ru/SerialPorts/RS232_pinout.shtml for a description of the extra connections.

Also, if somebody knows why the DE-9 connector has never been added to the recent revisions (revision F as of writing this), please add it to the text.

Chris Fletcher (talk) 14:24, 10 March 2019 (UTC)

Questions of "why did they do it this way" are often difficult to answer while meeting WP's WP:RS requirement. My guess, but it is only that, is that it was easier to simply define the DE-9 connector in a new standard, TIA-574 than to go through the protocol to make a new version of an existing standard.

Most All of the additional pins are described in the article, in the "Seldom-used features" section. They're just not all in a table. Jeh (talk) 10:11, 16 March 2019 (UTC)

V.24 Article Request

Request for V.24 article added to Wikipedia:Requested_articles/Applied_arts_and_sciences#R–Z. Squideshi (talk) 20:32, 3 April 2019 (UTC)