autodetect baud rate?

[spamiam]

Discussions here of serial LCD adapter problems got me thinking about construction one from a dedicated ZX device. Certainly there are cheaper serial adapters out there, but maybe they are not as reliable as we would like.

Maybe I could do better myself.

So, I would like to try making a dedicated serial-input, parallel-output device.

I want to be able to have the ZX AUTOMATICALLY synchronize itself with the incoming serial data. IOW, I do not want to have any specific required baud rate. I want the interface to look a an incoming synchronization stream of characters, and then realize what that baud rate is.

How would do that? I suppose I would look at the timing between rising edges on the data stream, and that will be an integral number of bits of data. The shortest time should be one bit wide. Longer times may 2 or 3 or more bits wide.

I would suppose that after several bits, it would be possible to deduce the baud rate (baud rate == bits per second == 1/bit time, right?)

What do you think would be the most accurate way to measure the bit times?

-Tony

[spamiam]

follow-up

I did a little checking. My first idea is to use PulseIn() on the serial input pin.

Have Register.TimerSpeed2 = 1.

I think that 2400 baud would measure 614.4 timer periods, and 115200 baud would measure 12.8.

After sampling about 20 or 30 pulses it probably would be possible to take the shortest duration, and

Baud = 1/(duration * 10)

assume N-8-1. Assume standard speeds 1200, 2400, 48000, 9600, 14.4K, 28.8k, 33.6K, 57.6K, 115.2K. Other data speed could be used as well, I guess.

I could verify bit durations easily enough easily enough on the ZX.

-Tony

[dkinzer]

In order to determine the baud rate by examining the receive line you have to know something about the data being transmitted. Often, the detection algorithm requires a specific character to be transmitted, e.g. a carriage return. If the transmitted character has a 1 as its LSB, you can measure the duration of the start bit and deduce the rate from that measurement.

If you search the Internet using the keywords "automatic baud rate detection" you should get some useful hits.

[spamiam]

[Genesis]

&H55 is a common one. So is carriage return. There IS a serial backpack out there SOMEWHERE that uses &H55 - I found it once and now have lost the link to it

The carriage return can detect on no more than two characters with a simple UART, so doing so when you can bit-bang should be simpler. I have code that does this in "C" for Unix machines (a custom "getty" program that handled it.)

The only "gotcha" is that you have to make some assumptions about parity and stop bits. If you allow all possible combinations then you must get much more clever.

Mu only caution on something like this is that there's a severe issue with power consumption using the ZX for this. For line-powered things this is not a big deal, but for battery-powered devices it definitely is. The power issue is what drove me away from the Parallax backpack (which works reasonably well - theirs is a power pig) in the first place.

[spamiam]

[Genesis]

Most so-called "Serial" LCDs are really like the Parallax - just a standard HD-based LCD with a serial backpack grafted on it......

At least this is true for the smaller ones (e.g. 16x2, 20x4, etc)

[stevech]

The 40 x 4 LCD I have uses that "backpack" which is a PIC chip with just two possible baud rate choices as I recall.

[spamiam]

[stevech]

yes, my 4 x 40 LCD has a "Backpack 443" module; PIC chip. Up to 9600 baud.