Hi and welcome!
Happy to hear you are enjoying the lizard synth.
So far I've heard from 3 other people about the same problem. In one case it turned out to be too slow / not genuine card, bought on eBay. In other cases the card seems good enough but for some reason still does not perform well - the problem is that until I get hold of the same problematic card I can't really debug the issue.
Your 2GB card is likely to be just a little bit too slow. Is it Class 4? If yes, that would explain it. I've been using SanDisk Ultra Class 10 (U1) 8GB and 16GB cards without any problems. The other guy who has encountered this problem has used the same type of card, only it was 32GB (which should not make any difference at all so it is quite a mystery, unless the sector size is different and it plays some role - I'll need to get one of those and try, as currently I only have a 32GB Extreme which works well).
If you want to see how your card performs, there is channel #3142 - SD Card speed test, it outputs the result over serial console and
you can use PuTTY to see the result. The test writes and reads 2MB of data, with my cards this takes cca 3-3.5 sec to write and 1.5-1.7 sec to read (the read info will show up only when running second time as the file needs to exist on the card first, things are done in this order to rule out any possible cache effect).
The sampling rate of 50.780 kHz is what the loopsynth uses internally. This odd value is simply a result of architecture, the ESP32 chip is driven from 26MHz crystal and the master audio clock is normally derived from it using a PLL (phase locked loop) circuit that gives very stable and low jitter signal.
But then not all values are possible, for example when requesting 48kHz from the I2S controller, 50781.250Hz comes out - this is a consequence of I2S bus timing, where MCLK (master clock) frequency is a nice number, precisely 13MHz (half of the xtal's 26MHz). Without PLL it is possible to obtain exactly 48kHz but that requires MCLK to be an odd / fractional value where the I2S controller uses some tricks to achieve that, and the clock is not smooth. This results in additional "quantization" noise added to the output signal (
more theory here if you are interested).
This of course results in whatever is recorded to SD also getting encoded in 50.78 kHz. I personally did not find this to be much of a problem as the software I used to edit audio or video was able to resample the tracks behind the scenes. PC has a lot more power so the algorithms to manipulate audio give a lot better results than if this was attempted in the synth's own small CPU (which would also need to be done in real time).
Some 3rd party sound engines - like Clouds - use 44.1kHz as they require more CPU power and need to squeeze the last bit of it, but where it was possible I decided to keep the ood sampling rate as it sounds noticeably better.
About the card formst - the default format with which the card comes from the shop is usually good. We use
FatFS library that supports FAT16, FAT32 and exFAT too - but there is still a limitation in Gecho because of the older ESP-IDF version used, which limits the card size to 32GB - I believe this is because larger cards are SDXC and use exFAT (this will likely be fixed in future firmware once I get time to port the code base over to the new version of the framework).
For best results and especially if you plan to use the material later, it is not advisable to record to SD, rather just take raw audio and record with external device, as from time to time glitch may occur and ruin the recording, for example when the card times out while erasing sectors, or something like that. Of course for the channel #222 to run well, this problem needs to be eliminated.
Show Posts
There is a simple trick to it though, check this out. 
And you're right, this is how "off" is currently indicated. The manual has been printed a while ago and then I probably changed the indication, because seeing no LEDs was confusing. And forgot that it results in this inconsistency. I'll probably fix it online, even though it will become inconsistent with the paper version.
, I mean 2111) that play slot 0.