def opt_kbbpd(lf_params,
dco_pn,
dco_pn_df,
dco_power,
temp,
tdc_steps,
div_n,
kdco,
fclk,
opt_fmax,
bbpd_tsu,
bbpd_th,
int_bits,
frac_bits,
sim_steps=10000,
max_iter=15):
cost = kbbpd_cost(lf_params, dco_pn, dco_pn_df, dco_power, temp, tdc_steps,
div_n, kdco, fclk, opt_fmax, bbpd_tsu, bbpd_th, int_bits,
frac_bits, sim_steps)
return gss(cost,
arg="kbbpd",
params={},
_min=0,
_max=0.3,
max_iter=max_iter)
def phase_margin(tf_params, fmax):
""" In degrees
"""
def cost(f):
return abs(pll_otf(f, **tf_params))
fug = gss(cost, arg="f", params={}, _min=0, _max=fmax, target=1)
return 180 + np.angle(pll_otf(fug, **tf_params), deg=True)
def bw_pi_pll(k, fz):
def h2(f):
return abs(pi_pll_tf2(f, k, fz))**2
return gss(h2,
arg="f",
params={},
_min=0,
_max=2 * np.sqrt(k),
target=0.5,
conv_tol=1e-10)
def bw_solpf(fn, damping):
def h2(f):
return abs(solpf(f, fn, damping))**2
return gss(h2,
arg="f",
params={},
_min=0,
_max=2 * fn,
target=0.5,
conv_tol=1e-10)
def opt_pll_tf_pi_controller_fast_settling(ph_margin,
max_tsettle,
tsettle_tol,
fclk,
oversamp=20):
""" Optimized PI-controller PLL for phase noise and settling time.
Subject to maximum settling time constrained by tsettle, tol.
points=1025 for Romberg integration (2**k+1)
"""
def cost(damping):
return opt_pll_tf_pi_ph_margin(damping, ph_margin, tsettle_tol, fclk)
opt_damping = gss(cost,
arg="damping",
params={},
_min=0,
_max=1.0,
target=ph_margin,
conv_tol=1e-5)
def cost(tsettle):
k = np.log(tsettle_tol)**2 / (opt_damping**2 * tsettle**2)
fz = np.sqrt(k) / (2 * opt_damping * 2 * np.pi)
return bw_pi_pll(k, fz)
opt_tsettle = gss(cost,
arg="tsettle",
params={},
_min=0,
_max=max_tsettle,
target=fclk / oversamp)
opt_bw = cost(opt_tsettle)
print(opt_damping, opt_tsettle)
#if opt_tsettle > max_tsettle:
# raise Exception("Error: It is not possible to achieve the specified phase margin and lock time. \
# Specified tsettle=%E, actual=%E. Decrease phase margin and try again."%(max_tsettle, opt_tsettle))
print("For fast settling: opt pi tsettle = %E, damping = %f, bw = %E" %
(opt_tsettle, opt_damping, opt_bw))
return pll_tf_pi_controller(opt_tsettle, tol, opt_damping)
def find_rw_k(f0, fs, samples, pn_db, df, seed=None, rw_seq=[]):
""" Fits random phase walk gain parameter k to phase noise data
args:
f0 - oscillator fundamental frequency
fs - 1/tstep of simulation
samples = length of random walk sequence
pn_db - target phase noise of oscillator
df - offset of phase noise measurement
seed - 32b value for predicable random walk sequence
rw_seq - use if random walk sequence already calculated
"""
pn = 10**(pn_db / 20)
def cost(k):
osc = osc_td(f0, fs, samples, k, seed=seed, rw=rw_seq)
print(pn, eval_model_pn(osc, f0, df))
return (pn - eval_model_pn(osc, f0, df))**2
k = gss(cost, arg="k", params={}, _min=0, _max=f0 / fs, conv_tol=1e-2)
return k
def opt_pll_tf_pi_controller_tsettle(damping,
tsettle,
tol,
pn_dco,
pn_dco_df,
m,
n,
kdco,
fclk,
fmax,
points=1025,
mode="tdc",
sigma_ph=0.1,
delay=0.0):
""" Optimize tsettle of PI-controller PLL for phase noise with fixed
damping
"""
def cost(tsettle):
k = np.log(tol)**2 / (damping**2 * tsettle**2)
fz = np.sqrt(k) / (2 * damping * 2 * np.pi)
tf_params = dict(k=k, fz=fz, delay=delay)
return pll_pn_power_est(pi_pll_tf,
tf_params,
pn_dco,
pn_dco_df,
m,
n,
kdco,
fclk,
fmax,
points=points,
mode=mode,
sigma_ph=sigma_ph)
return gss(cost,
arg="tsettle",
params={},
_min=0.01 * tsettle,
_max=tsettle,
target=0.0,
conv_tol=1e-10)
rms = np.std(pn_sig.td)
print("\nSimulated RMS PN:\t\t%.2E rad -> %.2f dB" %
(rms, 20 * np.log10(rms)))
est_rms_pn_design = np.sqrt(lfs["bbpd"]["int_pn"])
print("Filter design RMS PN Est:\t%.2E rad -> %.2f dB" %
(est_rms_pn_design, 20 * np.log10(est_rms_pn_design)))
SIM_PN_RMS[n] = rms
EST_PN_RMS[n] = est_rms_pn_design
def f(beta, target_pn):
lf = design_filters(DCO_FOM, 0, FCLK, FOSC, MAX_ALPHA, DCO_POWER,
KDCO1, beta)
return abs(target_pn - lf["bbpd"]["int_pn"])
beta_opt = gss(f, "beta", {"target_pn": rms**2}, _min=0.5,
_max=1.5) #1.0088433280171367 #
alpha = alpha_opt(beta_opt)
print("ITER %d RESULT:\tbeta = %.3E,\talpha = %.3E,\trms_pn = %.3E" %
(n, beta_opt, alpha, rms))
BETAS[n] = beta_opt
ALPHAS[n] = alpha
print("\nITER\tBETA\tALPHA\trms")
for n, v in enumerate(BETAS):
print("%d\t%.8f\t%.8f\t%E" % (n, v, ALPHAS[n], SIM_PN_RMS[n]))
# plt.plot(ALPHAS, SIM_PN_RMS**2)
# plt.plot(ALPHAS, EST_PN_RMS**2)
# plt.show()
foo()
plt.figure(1)
def opt_pll_tf_pi_controller_bbpd(tsettle,
tol,
pn_dco,
pn_dco_df,
n,
kdco,
fclk,
fmax,
delay=0,
points=1025,
max_iter=15):
""" This does not work yet
"""
sigma_ph = 0.01
def cost(sigma_ph):
m = 2 * np.pi
tf_params = opt_pll_tf_pi_controller(tsettle,
tol,
pn_dco,
pn_dco_df,
m,
n,
kdco,
fclk,
fmax,
points=points,
mode="bbpd",
sigma_ph=sigma_ph,
delay=delay)
_sigma_ph2 = pll_pn_power_est(pi_pll_tf,
tf_params,
pn_dco,
pn_dco_df,
m,
n,
kdco,
fclk,
fmax,
mode="bbpd",
sigma_ph=sigma_ph,
points=1025)
print(sigma_ph,
np.sqrt(_sigma_ph2) / n, (sigma_ph - np.sqrt(_sigma_ph2) / n)**2)
return (sigma_ph - np.sqrt(_sigma_ph2) / n)**2
sigma_ph = gss(cost,
arg="sigma_ph",
params={},
_min=0.0,
_max=1 / n,
max_iter=max_iter)
print("opt sigma_ph", sigma_ph)
m = 2 * np.pi
tf = opt_pll_tf_pi_controller(tsettle,
tol,
pn_dco,
pn_dco_df,
m,
n,
kdco,
fclk,
fmax,
points=points,
mode="bbpd",
sigma_ph=sigma_ph,
delay=delay)
return tf, sigma_ph
def opt_pll_tf_pi_controller(tsettle,
tol,
pn_dco,
pn_dco_df,
m,
n,
kdco,
fclk,
fmax,
delay=0.0,
points=1025,
mode="tdc",
sigma_ph=0.1):
""" Optimized PI-controller PLL for phase noise and settling time.
Subject to maximum settling time constrained by tsettle, tol.
points=1025 for Romberg integration (2**k+1)
"""
tsettle_min = 0.01 * tsettle # have to constrain, 0 will cause div-by-0
def cost(tsettle):
opt = opt_pll_tf_pi_controller_damping(tsettle,
tol=tol,
pn_dco=pn_dco,
pn_dco_df=pn_dco_df,
m=m,
n=n,
kdco=kdco,
fclk=fclk,
fmax=fmax,
points=points,
mode=mode,
sigma_ph=sigma_ph,
delay=delay)
k = np.log(tol)**2 / (opt**2 * tsettle**2)
fz = np.sqrt(k) / (2 * opt * 2 * np.pi)
tf_params = dict(k=k, fz=fz, delay=delay)
if fz > fmax:
raise Exception(
"Please increase fmax of loop filter optimization, frequency of TF zero in optimization exceeded fmax."
)
return pll_pn_power_est(pi_pll_tf,
tf_params,
pn_dco,
pn_dco_df,
m,
n,
kdco,
fclk,
fmax,
points=points,
mode=mode,
sigma_ph=sigma_ph)
opt_tsettle = gss(cost,
arg="tsettle",
params={},
_min=tsettle_min,
_max=tsettle,
target=0.0,
conv_tol=1e-5)
opt_damping = opt_pll_tf_pi_controller_damping(opt_tsettle,
tol,
pn_dco,
pn_dco_df,
m,
n,
kdco,
fclk,
fmax,
points=points,
mode=mode,
sigma_ph=sigma_ph,
delay=delay)
print("opt pi tsettle = %E, damping = %f" % (opt_tsettle, opt_damping))
return pll_tf_pi_controller(opt_tsettle, tol, opt_damping, delay=delay)
print("kbbpd = %f" % kbbpd)
print("pow ideal = %f" % np.var(M2PI * ph))
print("pow bbpd = %f" % np.var(x * kbbpd))
print("diff pow = %f" % (np.var(x * kbbpd) - np.var(M2PI * ph)))
print("Excess noise factor = %f" % (np.var(x * kbbpd) / np.var(M2PI * ph)))
error = np.var(ph * M2PI - kbbpd * x)
print("error = %f" % error)
def cost(kbbpd):
return np.mean((ph * M2PI - kbbpd * x)**2)
opt_kbbpd = gss(cost, "kbbpd", {}, _min=kbbpd * 0.1, _max=kbbpd * 10)
print("opt kbbpd = %f" % opt_kbbpd)
print("\n*****************************************")
print("MSE optimization")
print("pow ideal = %f" % np.var(M2PI * ph))
print("pow bbpd = %f" % np.var(x * opt_kbbpd))
print("diff pow = %f" % (np.var(x * opt_kbbpd) - np.var(M2PI * ph)))
print("Excess noise factor = %f" % (np.var(x * opt_kbbpd) / np.var(M2PI * ph)))
error = np.var(ph * M2PI - opt_kbbpd * x)
print("error = %f" % error)
# plt.plot(x*kbbpd, label="BBPD")
# plt.plot(M2PI*ph, label="ideal")
main_pn_data = pllsim_int_n(verbose=False, **main_sim_params)
pn_sig = pn_signal(main_pn_data, DIV_N)
# pow_pn = np.mean(pn_sig.td**2)
# print(pow_pn)
pow_pn = 20 * np.mean((pn_sig.td)**2)
print(KBBPD, pow_pn)
# error.append(np.abs(pow_pn - est_pow_pn))
return pow_pn
return cost
cost = kbbpd_cost(est_pow_pn)
print(gss(cost, arg="KBBPD", params={}, _min=0, _max=0.3, max_iter=15))
for x, y in zip(np.geomspace(0.01, 0.1, 11), error):
print(x, y)
plt.subplot(2, 1, 1)
plot_td(pn_sig)
plt.subplot(2, 1, 2)
plt.plot(main_pn_data["osc"].td - DIV_N * main_pn_data["clk"].td)
plt.show()
plot_pn_ssb2(pn_sig, dfmax=8e8, line_fit=False)
plot_pn_ar_model(pn_sig, p=200, tmin=0)
plot_lf_ideal_pn(DCO_PN, DCO_PN_DF, **lf_params)
print(noise_power_ar_model(pn_sig, fmax=FCLK / 2, p=100))
plt.show()
