I tested the dictionary based (dict) fonts from 'bdf2font' against nominally identical modules from 'micropython-font-to-py' (binary).
For these memory tests I used a ESP8266 dev board (Wemos D1 Mini) (32k RAM, 4Mb flash).
See: import_test.py
The script loads a list of font files and records how much memory is used during the import.
It does this in 'pairs' for the matching font modules and then lists the results for these side-by-side.
There are two test sets, with 4 tests in each:
size: Four different fonts of increasing pixel size (height) are compared.set: The same font (17px high) is compared with four different character sets ranging from 'time' (15 chars) to 'full' (192 chars).
This is 'rough' testing. I'm just using gc.collect() then gc.mem_free() before and after importing the module. The imported module is then indexed to make sure it has data, and references to it a deleted with del.
Repeated runs produce slightly different results, the ones below are 'typical', ymmv.
The test was run twice; once for 'frozen' font modules, and once for 'uploaded' ones.
I followed the documentation to create a uploadable micropython firmware with the font files frozen in the modules folder. This was then flashed to the test board, the test script was then uploaded via the IDE and run.
I flashed the test board with a vanilla micropython firmware from the main download site. Then all the font files were uploaded to the filesystem via the IDE, along with the test script. I then ran the test script.
The raw results are in import_test_frozen.txt and import_test_uploaded.txt
test : bin dict
----- : ----- -----
size:6 : 416 1104
size:14 : 384 1232
size:23 : 384 1104
size:33 : 384 1136
set:time : 480 464
set:num : 384 512
set:ascii : 416 1136
set:full : 400 1872
These results show that the binary font files are very efficient, using 400'ish bytes of RAM no matter how large or long the font and character set is.
The dictionary font modules show increasing size dependent on how many characters they encode, however the size (width,height) of the font makes no practical difference to the RAM used.
test : bin dict
----- : ----- -----
size:6 : 2144 4944
size:14 : 3232 5408
size:23 : 6704 8000
size:33 : 10032 12944
set:time : 2432 1776
set:num : 1824 2432
set:ascii : 5104 8000
set:full : 11264 15232
Both font module types show increasing RAM use as the pixel size and character set increase. The binary modules seem to typically use 2/3 the ram of the dictionary based ones.
The fonts created by micropython-font-to-py show a definate advantage in terms of RAM use in all scenarios, but both font packaging methods are acceptably small for use in smaller devices in both uploaded and frozen useage.
-rw-r--r--. 1 owen owen 6627 Jul 13 12:19 mPyEZfont_u8g2_4x6_r.py
-rw-r--r--. 1 owen owen 47842 Jul 13 12:19 mPyEZfont_u8g2_courB14_e.py
-rw-r--r--. 1 owen owen 7487 Jul 13 12:19 mPyEZfont_u8g2_courB14_n.py
-rw-r--r--. 1 owen owen 21346 Jul 13 12:19 mPyEZfont_u8g2_courB14_r.py
-rw-r--r--. 1 owen owen 5031 Jul 13 12:19 mPyEZfont_u8g2_courB14_t.py
-rw-r--r--. 1 owen owen 27168 Jul 13 12:19 mPyEZfont_u8g2_spleen_12x24_r.py
-rw-r--r--. 1 owen owen 13268 Jul 13 12:19 mPyEZfont_u8g2_timB10_r.py
-rw-r--r--. 1 owen owen 42593 Jul 13 12:19 mPyEZfont_u8g2_timB24_r.py
-rw-r--r--. 1 owen owen 3971 Jul 13 13:51 ezFBfont_06_4x6_ascii.py
-rw-r--r--. 1 owen owen 4238 Jul 13 13:49 ezFBfont_12_courB14_time.py
-rw-r--r--. 1 owen owen 9504 Jul 13 15:42 ezFBfont_14_timB10_ascii.py
-rw-r--r--. 1 owen owen 5614 Jul 13 13:48 ezFBfont_16_courB14_num.py
-rw-r--r--. 1 owen owen 15962 Jul 13 13:49 ezFBfont_17_courB14_ascii.py
-rw-r--r--. 1 owen owen 33876 Jul 13 13:47 ezFBfont_20_courB14_full.py
-rw-r--r--. 1 owen owen 17010 Jul 13 13:52 ezFBfont_23_spleen_12x24_ascii.py
-rw-r--r--. 1 owen owen 34031 Jul 13 13:53 ezFBfont_33_timB24_ascii.py