I'm a little bit perplexed right now.
Originally wanted to post this in the Tonex thread, but the information is possibly as relevant here.
According to Leo Gibson, the device latency of the GT-1000 would be 0.7ms. But it's actually less, it's not even 0.6ms (0.58 to be exact).
Then, and this is where things really start to get crazy: With one (serial, obviously) loop engaged, the latency only goes up to 0.75ms.
With both loops running serially (so loop 1 returns and then instantly goes into loop 2), that value is established as latency is now at 0.97ms.
Note: There might be very slight inaccuracies due to the transients possibly be a little smeared by the triple ADDA conversion and whatever op amps in the GT, but these inaccuracies are within the 0.0x ms realm.
So, apart from the overall latency with no loops engaged being extremely low already, a little bit contrary to what often is the case, the loops add *less* than full device latency to the overall picture (very often it'd just double because there's the double amount of converters and signal processing).
And btw, as I couldn't actually believe these numbers myself, I crosschecked everything multiple times. Tried my Zoom G3 that I used as a splitting device on its own, as I was suspecting there might be some sort of offset between the left and right outs, but there's none. I even flipped the input channels on my Motu M2 to see whether something would be happening there (which was very unlikely to start with), but nada.
So the latencies are in fact *that* low. I have absolutely no idea how Boss manages to pull that off, but it's just incredible.
Anyhow, for a final test (because that might become my future setup), I inserted a Tonex One into one loop, running a clean capture, and the overall latency was 2.8ms. Without the capture loaded, it's just 2.1ms, so at least some captures add some latency (I observed that before already).
Whatever, around 3ms for a whooping 4 serial ADDA cycles is just stunning. Also, while you find people complaining about whatever hiss or other sound integrity issues when it comes to the GT loops - fuck that, sound quality is actually excellent once you use decent power supplies and check the various ground lift settings.
In the end, I guess it really shows that Roland/Boss has an enormous experience when doing digital pedals. I mean, people are chaining their pedals since decades already, and if latency was a real issue, I'm sure it would be discussed about anywhere. But looking at the GT's figures, you just know why it isn't.