Suggestions to replace on board serial port (DB9)?

[videobruce]

I need to get a different serial port card for two different pieces of test equipment. The on board port I have is buggy.

I did a quick search and found this (also in 4 other places);
Serial HOWTO: What Are UARTs? How Do They Affect Performance?

It mostly talks about speed, which really isn't a issue here. I'm mostly interested in compatibility. The test equipment designs are around 10 years old thought one is still being sold. I was told to watch what I buy, to stay away from 'cheaper' chips.

The on board I have now is what is on a Foxconn nVidea N570 chipset MB that uses the stock M$ driver. One device I'm having a issue with corrupted files either being sent or being received. I can't tell. I also don't know if it is the serial port on the PC end or the device end.

Anyway, any suggestions? I see PCI cards as low as $9, while I also see ones for over $100 which I find hard to believe.

What is the difference between these 16C550 that are the cheapos and the more expensive 16550, 16650 & 16950 for over $30 other than the buffer size?

I am using XP Pro w/sp2.

[Gizmo]

AKAIK, the 16550 and the 16C550 are the same chip, except that the 16550 uses older N-MOS process technology, whereas the 16C550 uses CMOS process technology. Because of this, the 16C550 doesn't have the ability to drive the same kinds of loads that the 16550 does.

As far as the difference between the cheapo boards and the expensive ones, I couldn't tell you specifically, but I DO know that cheap designs tend to play rather fast-and-loose with the RS-232 spec. The generally don't properly buffer the TTL-level outputs from the UARTS with good RS-232 transceivers. RS-232 requires that the signal swing between +3 and -3 volts IIRC. Many cheap boards (and PC mobos, too) actually use regular TTL outputs, which means that they swing between about +3.5 and +0.7 volts. Although most RS-232 buffers will accept this input level, it's technically a violation of spec, and can result in unreliable communications, particularly over longer distances.

In addition to that, cheaper boards that don't outright violate the spec will still only swing between +3 and -3v or so, by using the +5 and -5v rails as supplies, or by using very, very weak outputs, or a combination of the two. The net result is still poor RS-232 performance in demanding applications. A good board will provide outputs that swing between +12v and -12v (or very close to that), with proper I/O buffering of the UART.

[videobruce]

The older N-MOS is better than the newer CMOS?
Two of the boards I was looking at use a MCS9835CV chipset;



Then how does one know what's what with a certain board?

[videobruce]

The older N-MOS is better than the newer CMOS?
Two of the boards I was looking at use a MCS9835CV chipset;
Syba
Rosewill

Then how does one know what's what with a certain board?

[Gizmo]

Originally Posted by videobruce

The older N-MOS is better than the newer CMOS?

Not really. If the board is properly designed, the drive capabilities of the chips won't come into play, and the CMOS chip is more power efficient.

Originally Posted by videobruce

Two of the boards I was looking at use a MCS9835CV chipset;
Syba
Rosewill

Then how does one know what's what with a certain board?

Sometimes the only way to know is to ask the manufacturer. Other times you can get what you need to know off the spec sheet.

The Rosewill board looks like it has a pair of RS-232 transceiver chips on the pins, although without talking the manufacturer or looking at the board I can't say with certainty.

The Syba board looks like it may also have transceivers, but I'm even less certain with it than the Rosewill board.

The NM9835 is a single-chip PCI interface solution providing 2 16C550 compatible uarts and a IEEE 1284 compatible parallel port. With a little wiring and the proper driver, you could probably get yourself a free parallel port out of one of those cards, in addition to the 2 serial ports.

[videobruce]

What's the difference between a 16C550 compatible UART and a 16550? The cards w/ a 16550 are 2x the price.

[Gizmo]

Compatible means that they have the same software interface. Basically, if you have software that wants to talk directly to the hardware, and that software knows how to talk to the 16550, all you have to do is give it the I/O address of the port, and the software should be able to work with it.

[videobruce]

How about hardware differences?

[Gizmo]

Well, obviously there are going to be some hardware differences. It's not pin compatible, after all. But that's where the buffering comes in, and that's part of the reason it is important. If the thing is properly buffered, then it will interface with anything that adheres to the RS-232 specification, regardless of hardware.

[Áedán]

One of the issues with the 16550 (and immediate family) is that they don't interface with the PCI bus. In fact, their interface is rather more like the old ISA bus. Hence, you need a bunch of interface electronics to connect them to the PCI bus. They're also years old - the current PC16550D chip was introduced in 1993.

The original 16550 had a 16byte buffer for storing received characters, whereas the 16650 and 16950 both have a 128byte buffer. That's not necessarily an issue, depending on baud rates. The 16650 tops out around 1Mbit/sec, whereas the 16950 can manage up to 60Mbit/sec. At very high baud rates, a larger buffer is a good idea. However, if you're at standard baud rates (IE 115,200 baud and below), then a large buffer isn't likely to be much use.

The interface to the outside word, as Gizmo has already indicated, is of prime importance. Poor quality buffering will generally lead to poor quality results, even if the UART chip is the best in the world.