Based on Hartfield & Otto 2011 which cite Haldane 1924:
Once established, the frequency of A1B1 can be modeled by the standard deterministic equation for haploid selection (Haldane, 1924): $p(t) = \frac{p_0}{p_0 + (1 - p_0)(1-s)^t }$
p = lambda t: (p0)/(p0 + (1 - p0)*(1-s)**t)
N = 1e6
p0 = 1./N
p1 = 1-p1
s = 0.01
f = [array([p0, p1])]
tick = 0
S = array([1,1-s])
while f[-1][0] < 0.99:
tick += 1
f.append(f[-1] * S)
f[-1] /= f[-1].sum()
if tick%100 == 0: print tick,f[-1][0]
f = array(f)
100 2.73199429462e-06 200 7.46377043595e-06 300 2.03907499683e-05 400 5.57055417224e-05 500 0.000152172803824 600 0.000415626408093 700 0.0011346741306 800 0.00309384842062 900 0.00840733978591 1000 0.0226391835409 1100 0.0595165224224 1200 0.147404271787 1300 0.320805405939 1400 0.563397103729 1500 0.779025113177 1600 0.905938943823 1700 0.963387337251 1800 0.986280125989
t = arange(0,tick+1)
y = np.vectorize(p)(t)
plot(t, y, 'b-', lw=1)
plot(t, f[:,0], 'r--', lw=3)
[<matplotlib.lines.Line2D at 0x6373310>]
Tfix = lambda N,s: -2*log(N)/log(1-s)
N = logspace(3,8,50)
T = Tfix(N,s)
plot(N,T)
xscale('log')
grid(False,'minor')