General 2D Information

The length of a 2D experiment is approximately a function of the number of scans (nt) for every point acquired, the number of points acquired in the second dimension (ni), and the length of time to acquire the data (at) and the delay to return to steady state (d1). There is also a small contribution from the mixing time parameter (mix) in TOCSY, ROESY, and NOESY.

Why would you want to increase/decrease each of these parameters:

nt - number of transients. most importantly nt must be set to a specific multiple of some number that is different for different experiments (nt should be a mutliple of 2 for gCOSY, gHSQC, gHMQC, gHMBC, nt = 1 is acceptable but nt = 2 is preferred; nt should be a multiple of 8 for TOCSY but 4 is acceptable; nt should be a multiple of 16 for NOESY, ROESY but 8 is acceptable). Also, the greater the nt, the better the signal to noise, but also the greater the length of time to acquire data, in many samples, the minimum nt is probably enough signal to noise.

ni - number of increments. should be an even number for gHMBC and gCOSY, else any number is fine. Most important is that greater ni gives greater resolution in second dimension, and will increase signal to noise for a period of time in some experiments (gHSQC, gHMQC, gHMBC, gCOSY). Normally, an absolute minimum of 32 real ni points is necessary for correlations with carbon (HMQC, HSQC, HMBC- but note that HMBC is a magnitude spectrum so an ni of 64 in gHMBC = 32 in an gHMQC), and 64 for proton-proton correlations (gCOSY is magnitude also so that is 256 for COSY). In general, the minimum standard for the basic experiments:
gCOSY: minimum = 64, ok = 128, suggested = 256, high resolution = 512
gHSQC/gHMQC: minimum = 32, ok = 54, suggested = 128, high resolution = 256
gHMBC: minimum = 128, ok = 256, suggested = 400, high resolution = 600
TOCSY: minimum = 64, ok = 128, suggested = 256, high resolution = 512
NOESY/ROESY: minimum = 128, ok = 200, suggested = 350, high resolution = 700

at - acquisition time. Normally, at is not set; the computer sets at based upon sw (sweep width) and np (number of points). Longer acquisition time leads to better resolution in the dimension that you are detecting, but also leads to larger file size, slower processing, slower data transfer, slower display. Also, in experiments with decoupling (gHMQC, gHSQC), decoupling is on during acquisition and sample will heat. It is generally suggested to have np = 2048 unless the sw is very large >6000 Hz.

d1 - delay 1, allows for return to steady state magnetization. The more steady state magnetization, the more magnetization that will be detected, but also the fewer the nt and ni that you can acquire in a length of time. Theoretically, 3-5 times the longest T1 time is correct; reality is that would require days to weeks to acquire. A reasonable rule is: if you want to compare intensities of crosspeaks (NOESY, ROESY), a longer d1 is necessary, then choose ~3*(the longest T1 you care about), else choose ~1.3*(the longest T1 you care about). Also, d1 allows for the probe to cool after experiment (so no shorter than 1 second for HMQC, HSQC).