Too good to be true

I am a bit shocked by some results I am getting out of an Arduino Mega board with a built-in ESP8266 [www.aliexpress.com] (not that these are a novelty at all, but the price point now piqued my curiosity).

I have installed an old Marlin version on the Arduino (one that I have been using for years) and the ESP-LINK firmware [github.com] on the ESP8266 (again, nothing new here).

Next, I go on streaming a sample file over USB using pronterface software and it takes around 40 minutes to complete.

The same file streamed over a TCP connection (now using the ESP as a serial-wifi bridge) takes only 24 minutes. For streaming over TCP I have used both my own Python script and Pronterface (both achieving the same results time-wise).

The board is just sitting on my desk and not connected to any 3D printer hardware at the moment, but I do not believe the speed-up achieved is possible and I am leaning towards a mistake made on my side, that somehow is eating up part of the data mid-way.

Do any of you have any experience with a similar setup? Any idea a test I could do to find the culprit? (other than attaching a printer's hardware and check the parts printed both ways :-).

Thanks a lot for your input.

How about math approach.

You know the size of the file
You know the baud rate, data bits and stop bit etc

You can calculate the minimum time It could take take under ideal conditions.

even faster if you use an sd card...though ive not done it in a while as I just print from the PC.

@Dust:

Right, but that would be less than two minutes.

I used the term streaming instead of transmission to convey the idea that actual work is done for each line of g-code received. Some of the movements taking a while to be completed.

In both cases, USB and TCP, the total time is way above the transmission time.

@MechaBits:

Very good idea, I am going to try this next (simpler than to attach a whole printer ;-). My past experience has been that SD printing uses to have an edge over USB streaming.

Thanks a lot, guys!

so they say, but it seems quite rare for me its only happened once or twice here, but obviously better to try & free up the computer.

Ok, I moved the test from one Imac running Mojave to a Win10 computer: The same file takes only 25 minutes to be streamed over USB too, more or less the same as with TCP, which makes sense.

So the cautionary tale about this issue is the USB drivers we use for both (genuine Arduino's 8U2 and CH340-based clones), at least on Mojave, exhibit significative lower performance than their Windows counterparts (at least this is one case where it happens).

Have you ever noticed the 3d printing time of a part was somehow larger if printed from Mojave?

I have found out there are several CH340 drivers available. One of them crashed my macbook and the other worked fine even though it wasn't for my system and the one that crashed my mac was meant for my system.

---

http://www.marinusdebeer.nl/