Simulating from an Epanechnikov kernel density estimate / exact functional form of the Epanechnikov kernel?

It's my first time posting, so apologies if I'm breaking any etiquette. I've used the ksdensity function to estimate a density using the Epanechnikov kernel and would now like to make repeated random draws from this density. I'm doing so according to the suggestion at How can I draw a value randomly from a kernel density estimate? while sampling from the kernel using the scheme described at Generating a sample from Epanechnikov's kernel. On simulated data, this ends up looking something like this:

reps = 1000;

weight1 = 0.4;

mu1 = 10;

sigma1 = 5;

mu2 = -7;

sigma2 = 3;

v1 = rand(reps,1);

v2 = mu1 + sigma1*randn(reps,1);

v3 = mu2 + sigma2*randn(reps,1);

data = (v1 <= weight1).*v2 + (v1 > weight1).*v3;

clear v1 v2 v3;

[density,points,bandwidth] = ksdensity(data,'Kernel','epanechnikov');

u1 = (-1)+(1-(-1))*rand(reps,1);

u2 = (-1)+(1-(-1))*rand(reps,1);

u3 = (-1)+(1-(-1))*rand(reps,1);

u = (abs(u3) >= max(abs(u1),abs(u2))).*u2 + (abs(u3) < max(abs(u1),abs(u2))).*u3;

clear u1 u2 u3;

sim = randsample(data,reps,true) + bandwidth*sqrt(5)*u;

My concern has to do with the last line of this sample and that multiplier sqrt(5). I'm inferring this term is necessary based on the discussion at Different definitions of Epanechnikov-Kernel; i also find that it seems to deliver densities that better line up with my data than when i omit it. however, to be certain that it belongs i'd need to know the precise functional form of the kernel that ksdensity is using. (if i understand the terminology correctly, i think it would also suffice to figure out the "canonical bandwidth" being used). Does anyone know these details or where they can be found? Thanks very much!