In this question:

Why usb to serial port converter can’t program avr microcontroller?

Cornelius claims:

"DTR and RTS are the only pins of the serial port that can be bitbanged."

But is that claim true? Is there some problem bitbanging the TxD pin that I am not aware of?

How would I bitbang TxD? Simple:

First: Make sure that there are no characters queued for transfer on TxD.

Then: In Delphi, I would do it like this:

  Com1Port : TBlockSerial; //  TBlockSerial comes from SynaSer.

// SynaSer is part of Synapse communication library:
// http://www.ararat.cz/synapse/doku.php/start

procedure TfrmMyCom1Test.SetTxD(const Value: Boolean);
  EscCommands : Array[Boolean] of DWORD = (SETBREAK, CLRBREAK);

  if NOT Com1Port.InstanceActive
    then Exit;

  // Break:
  if EscapeCommFunction(Com1Port.Handle, EscCommands[Value])
    then FTxD := Value;

  // Value = False -> SETBREAK ; Sets TxD to Logical 0.
  // Value = True  -> CLRBREAK ; Sets TxD to Logical 1.

When a call is made:

SetTxD(False); // This makes TxD a Logical 0.
SetTxD(True);  // This makes TxD a Logical 1.

So to me it seems it is possible to bitbang TxD also (on COM1, for example) as long as one takes care that there are no queued characters queued for transfer on TxD.

Actually, there may be another way to bitbang TxD:

If you make your circuit (AVR serial programmer for PC serial port) so, that a logical 1 (Mark) on TxD translates to 0 V on the AVR SCK pin (on the ATmega328 this is PB5), and a logical 0 (Space) on TxD translates to about 5 V on SCK, that makes a 2nd way of bitbanging TxD possible, but adds an extra programming difficulty: Either one must find a way to know, when transmitting a byte/character is ready (HOW?), or alternatively one must wait long enough so that the delay is enough for transmitting one byte to always be finished, which may waste some time.

This way of bitbanging simply sends 1 to 5 bits to the AVR (either all bits ZERO or all bits ONE), and regardless of sending just 1 bit or 5 bits, one batch of send always uses the time of 10 line bits to send.

Like, at 115200 bits per second, each bit consumes 8.681 µs, which makes 10 line bits to consume 86.81 µs.

I thought the idea of my 2nd way of bitbanging would be obvious - but maybe not, so let me explain:

The 2nd way allows to bitbang 1 to 5 bits to the AVR (but either all bits sent must be ZERO or all bits sent must be ONE).

First: set the MOSI pin (could use either DTR or RTS for this, and use the other one of these to control the RESET pin of AVR) to match the bit value you want to send.

Next: send one of the following bytes (Hexadecimal values):

FF, FD, F5, D5, 55

regarding if you want to send 1 (0FFh) to 5 (55h) copies of the same value bit.

Since the start bit is 0 - translates to 5V at SCK and the stop bit is 1 - translates to 0V at SCK.

and each 1 data bit translates to 0V at SCK, so the data byte must contain one less 0 bits (which translate to 5V at SCK) as how many copies of the same bit you want to send, but there must not be two (or more) consequent 0 bits in the byte, but each two 0 bits must be separated by a "1" bit.

The ATMega328 datasheet: https://ww1.microchip.com/downloads/aemDocuments/documents/MCU08/ProductDocuments/DataSheets/ATmega48A-PA-88A-PA-168A-PA-328-P-DS-DS40002061B.pdf

The general idea here is to use the built-in COM1 port of a windows PC to use bitbang-mode of the COM1 port to do serial programming of the ATMEGA328 microcontroller.

Of course, the same idea would work to program any microcontroller that accepts SPI as the mode of serial programming.

  • 1
    \$\begingroup\$ Yes, your first way will work: you can set break condition or clear it; the difficult is how to get good timing. I confess I don't understand your second suggestion: if you send something out of the UART you'll get at least some SPACE and at least some MARK (because of the start and stop bits). So how do you propose to get solely your signal? \$\endgroup\$
    – jonathanjo
    Jan 7 at 16:52
  • \$\begingroup\$ Depends on which COM port it is and how it is implemented. I've done it so answer is yes, but I also know it might be impossible if the hardware or software interface does not support it so the aswer is it depends. \$\endgroup\$
    – Justme
    Jan 7 at 16:55
  • \$\begingroup\$ What specific chip are you referring to and can you link the data sheet? \$\endgroup\$
    – Andy aka
    Jan 7 at 17:08
  • 2
    \$\begingroup\$ Your second method isn't really "bitbanging" as most people use the word: meaning "software control of each output pin". It certainly is an inventive way of doing SPI and similar clocked serial. But can I ask What is your question for EE.SE? \$\endgroup\$
    – jonathanjo
    Jan 7 at 18:22
  • 1
    \$\begingroup\$ the answer to your question is entirely dependent on the COM1 hardware ... why would you need to bitbang TxD specifically? ... any if the other pins could be used to transmit data \$\endgroup\$
    – jsotola
    Jan 7 at 19:17

1 Answer 1


I understand the question as "Can software set the level of TxD to high and low, respectively?"

This kind of bit-banging TxD is possible, but only without exact timing.

Desktop operating systems like Windows or Linux are not real time operating systems. Therefore, you cannot fulfil timing requirements, not even in the range of tens or hundreds of milliseconds.

So the complete answer is "it depends" on your timing requirements.

Concerning your second suggestion, I would not call this "bit-banging". It is an interesting implementation of SPI, and you would not even need the inverter, as SPI knows different modes.

Actually, you can "abuse" the asynchronous serial interface for many different things. I once used the standard COM port to receive and send signals compatible to an old home computer of the 80s, which only knows to save and load its program from cassette tapes. The PC emulated the cassette drive in this case. However, this is not "bit banging" for me.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.