diff --git a/.gitignore b/.gitignore
index 2f70c28..0665a92 100644
--- a/.gitignore
+++ b/.gitignore
@@ -1,4 +1,6 @@
 *~
 __pycache__
 .ipynb_checkpoints
-*.ipynb
\ No newline at end of file
+*.ipynb
+*.h5
+*.npy
\ No newline at end of file
diff --git a/.vscode/settings.json b/.vscode/settings.json
new file mode 100644
index 0000000..38958e5
--- /dev/null
+++ b/.vscode/settings.json
@@ -0,0 +1,7 @@
+{
+    "files.watcherExclude": {
+        "**/__pycache__/**": true,
+        "**/.ipynb_checkpoints/**": true,
+        "**/*.h5": true
+    }
+}
\ No newline at end of file
diff --git a/perturbation_blk_pipeline.py b/perturbation_blk_pipeline.py
new file mode 100644
index 0000000..8cbab94
--- /dev/null
+++ b/perturbation_blk_pipeline.py
@@ -0,0 +1,25 @@
+from blk import pipeline, cache
+from perturbation_generator import Perturbations
+
+cachedir = "blk_test/"
+
+cache.set_dir("blk_test/")
+
+p = Perturbations(rng_seed=333, nsamples=500, 
+                  perturbation_filename=cachedir+"entropy_perturbations.h5")
+
+gen = pipeline.Stage(p.generate_perturbations(),
+                     arguments=[],
+                     tag = "generation",
+                     action = "manual")
+
+app = pipeline.Stage(p.save_perturbed_entropy(),
+                     arguments=[cachedir+"entropy_perturbed.npy"],
+                     tag = "application",
+                     action = "manual",
+                     depends_on = [gen])
+
+pipeline.execute(app, parallelism="none")
+
+p.load_perturbed_entropy(cachedir+"entropy_perturbed.npy")
+p.visualize()
\ No newline at end of file
diff --git a/perturbation_generator.py b/perturbation_generator.py
new file mode 100644
index 0000000..b80bd13
--- /dev/null
+++ b/perturbation_generator.py
@@ -0,0 +1,169 @@
+#!/usr/bin/env python
+# coding: utf-8
+import h5py
+import napari
+import numpy as np
+import unyt as u
+import scipy.stats as s
+from scipy.interpolate import interp1d
+from tqdm import tqdm
+
+class Perturbations():
+
+    def __init__(self, rng_seed=None, nsamples=500, 
+                 grid_min=-800, grid_max=800, grid_size=512,
+                 entropy_filename="../initial_entropy_profile.txt",
+                 pressure_filename="../initial_pressure_profile.txt",
+                 perturbation_filename="entropy_perturbations.h5"):
+        
+        self.rng_seed = rng_seed
+        self.nsamples = nsamples
+        self.grid_min = grid_min
+        self.grid_max = grid_max
+        self.grid_size = grid_size
+        self.entropy_filename = entropy_filename
+        self.pressure_filename = pressure_filename
+        self.perturbation_filename = perturbation_filename
+
+        self.rng = np.random.default_rng(self.rng_seed)
+
+        self.entropy_func = None
+        self.pressure_func = None
+
+        self.x = np.linspace(self.grid_min, self.grid_max, self.grid_size)
+        self.xx, self.yy, self.zz = np.meshgrid(self.x, self.x, self.x)
+        self.rr = np.sqrt(self.xx**2 + self.yy**2 + self.zz**2)
+
+        self.modified_entropy = None
+
+    def load_entropy(self):
+
+        if self.entropy_func is None:
+            func_r, func_K = np.genfromtxt(self.entropy_filename, unpack=True)
+            self.entropy_func = interp1d(func_r, func_K, fill_value="extrapolate")
+    
+    def load_pressure(self):
+
+        if self.pressure_func is None:
+            func_r, func_P = np.genfromtxt(self.pressure_filename, unpack=True)
+            self.pressure_func = interp1d(func_r, func_P, fill_value="extrapolate")
+
+    def generate_perturbations(self,
+                               dlnK_sigma=0.6, dlnK_loc=-0.2, dlnK_scale=0.3,
+                               tcross_mean=0.3, tcross_sigma=0.01):
+
+        # Pull in the entropy and pressure profiles K(r) and P(r)
+        self.load_entropy()
+        self.load_pressure()
+
+
+        # Randomly pick an (x,y,z) coordinate and a max dlnK for the perturbation
+        # 
+        # The use of a lognormal distribution for dlnK as well as the choice of sigma_lnK are based on Voit 2021:
+        # 
+        # >"Around each median profile, fluctuations in density,
+        # temperature, and entropy exhibit approximately log-normal
+        # distributions with long tails toward lower temperature and
+        # entropy and greater gas density"
+        # 
+
+        pick_x = self.rng.choice(self.x, self.nsamples)
+        pick_y = self.rng.choice(self.x, self.nsamples)
+        pick_z = self.rng.choice(self.x, self.nsamples)
+
+        pick_r = np.sqrt(pick_x**2 + pick_y**2 + pick_z**2)
+
+        lognorm_params = {"s":dlnK_sigma, 
+                        "loc":dlnK_loc,
+                        "scale":dlnK_scale, 
+                        "size":self.nsamples,
+                        "random_state":self.rng}
+
+        pick_dlnK = -1*s.lognorm.rvs(**lognorm_params)
+
+        # Cut off at dlnK = -1
+        while not np.all(pick_dlnK > -1): 
+            pick_dlnK = np.where(pick_dlnK > -1, pick_dlnK, -1*s.lognorm.rvs(**lognorm_params))
+
+
+        # Randomly pick a size for the perturbation using normally distributed sound crossing times.
+        # The value we pick from that Gaussian distribution will be 2sigma of a Gaussian blur (roughly analogous to a diameter, but OK if underresolved)
+
+        # assume n_e = n_i s.t. n = 2n_e
+        gamma = 5/3
+        mu = 0.6
+        n = np.power(self.pressure_func(pick_r) / ( 2*self.entropy_func(pick_r) ), 1/gamma) / u.cm**3
+        c_s = np.sqrt(gamma * self.pressure_func(pick_r)*u.erg/u.cm**3 / (2*mu*u.mp*n))
+
+        pick_tcross = self.rng.normal(tcross_mean, tcross_sigma, self.nsamples) * u.Myr
+
+        assert not (pick_tcross < 0).any()
+
+        diam = (c_s * pick_tcross).to("kpc")
+
+        # Save x,y,z dlnK, and simga (diam) to HDF5, because it's easier to get into enzo
+        with h5py.File(self.perturbation_filename, "w") as f:
+            f.create_dataset("x", data=pick_x)
+            f.create_dataset("y", data=pick_y)
+            f.create_dataset("z", data=pick_z)
+            f.create_dataset("dlnK", data=pick_dlnK)
+            f.create_dataset("sigma", data=diam)
+
+    def apply_perturbations(self, overwrite=False):
+
+        if self.modified_entropy is not None and overwrite is False:
+            print("Modified entropy profile already saved to this object. Pass overwrite=True to recalculate.")
+            return
+
+        self.load_entropy()
+
+        with h5py.File(self.perturbation_filename,"r") as f:
+            x = np.array(f["x"])
+            y = np.array(f["y"])
+            z = np.array(f["z"])
+            dlnK = np.array(f["dlnK"])
+            sigma = np.array(f["sigma"])
+        
+        self.modified_entropy = self.entropy_func(self.rr)
+
+        for i in tqdm(range(self.nsamples)):
+            profile = dlnK[i] * np.exp(-(self.xx-x[i])**2/(2*(sigma[i]/2)**2) \
+                                    -(self.yy-y[i])**2/(2*(sigma[i]/2)**2) \
+                                    -(self.zz-z[i])**2/(2*(sigma[i]/2)**2) )
+            self.modified_entropy = self.modified_entropy * profile + self.modified_entropy
+
+    def save_perturbed_entropy(self, filename="entropy_perturbed.npy"):
+
+        if self.modified_entropy is None:
+            self.apply_perturbations()
+
+        np.save(filename, self.modified_entropy, allow_pickle=False)
+
+    def load_perturbed_entropy(self, filename="entropy_perturbed.npy", overwrite=False):
+
+        if self.modified_entropy is not None and overwrite is False:
+            print("Modified entropy profile already saved to this object. Pass overwrite=True to continue loading.")
+            return
+
+        temp = np.load(filename, allow_pickle=False)
+
+        if temp.shape != (self.grid_size, self.grid_size, self.grid_size):
+            raise RuntimeError(f"Loaded array with shape {temp.shape} does not match the cubic grid_size {self.grid_size} specified for this object")
+        else:
+            self.modified_entropy = temp
+
+    def visualize(self, cmap="viridis"):
+        
+        if self.modified_entropy is None:
+            print("This object has no modified entropy profile stored. Either run apply_perturbations() or load_perturbed_entropy().")
+            return
+        
+        viewer = napari.view_image(self.modified_entropy, colormap=cmap)
+        napari.run()
+
+if __name__ == "__main__":
+    
+    p = Perturbations()
+    p.generate_perturbations()
+    p.save_perturbed_entropy()
+    p.visualize()