Ultralow phase noise Ti:sapphire laser rivals
100 MHz crystal oscillator
R. P. Scott, C. Langrock, and B. H. Kolner
The timing stability of modelocked lasers is an impor-
tant quantity but is very difficult to measure accurately.
This is especially true for Kerr-lens modelocked Ti:sapphire
lasers  and some harmonically modelocked fiber lasers 
which demonstrate very low phase noise. We have assem-
bled a system for characterizing the phase noise of mode-
locked lasers which displays exceptionally high dynamic
range (> 170 dB) and accuracy (±2 dB). We have used
this system to characterize a femtosecond Ti:sapphire laser
and found it to have short term stability close to that of
the precision crystal oscillators used in its characteriza-
tion. This extraordinaryresult suggests the possibility that
modelocked lasers could find applications in high stability
RF and microwave sources.
The most informative measurements of timing stability
are made in the frequency domain by observing the power
spectrum of the sidebands produced when random or dis-
crete noise modulates the phase or frequency of the laser
pulse train. There are two approachesto this measurement.
In the first, a photodiode directly detects the optical pulse
train and feeds the signal to a spectrum analyzer. The
sidebands adjacent to any Fourier component can then be
related to the rms timing jitter. This “direct technique”
is of very limited utility because of the limitations of the
spectrum analyzer’s dynamic range and it’s IF and base-
band filters. Furthermore, low frequency Fourier compo-
nents can have a large amount of AM noise which can ob-
scure the true PM spectrum. Using higher harmonics can
reveal more of the phase noise spectrumbut this is also lim-
ited in accuracy beyond a few harmonics. The phase noise
amplitude (not power) grows nonlinearly with harmonic
number as the modulation drives the Bessel functions out
of their linear range (this seems to be seldom appreciated
A superior method relies on using a high stability refer-
ence oscillator which is kept in phase quadrature with re-
spect to the laser under test by a phase-locked loop (phase
detector method, Fig. 1). The output of the phase detec-
tor is a voltage proportionalto the phase deviation between
the laser and the reference oscillator. Low frequency FFT
and RF spectrum analyzers then measure the spectrum
of this phase-demodulated signal with great precision and
high dynamic range since the analyzers do not have to han-
dle the high powered carrier. A detailed treatment of this
method applied to modelocked lasers can be found in .
Clearly, the quality of the phase noise measurement by
the phase detector method depends on the spectral purity
of the reference oscillator. To date, almost all measure-
Fig. 1. Typical measurement setup for spectral analysis of laser phase
noise. LNA; low noise amplifier, PD; photodiode, LO; local oscillator
port, RF; radio frequency port, VCO; voltage controlled oscillator.
ments of modelocked lasers have been made using synthe-
sizers as reference oscillators. The internal reference os-
cillator of the synthesizer can be pulled, generally, by a
few Hertz using electronic frequency control (EFC: volt-
age tuned varactor across a crystal). However, we have
found that the spectral purity of most synthesizers is in-
sufficient to characterize a well constructed and operating
Ti:sapphire laser. Indeed, the literature contains frequent
evidence of laser phase noise being masked by the reference
II. The Measurements
To overcome the rather high phase noise floor limitations
presented by synthesizer references, we have used a very
low noise crystal oscillator to characterize our Ti:sapphire
laser (K&M Labs Model TS). The oscillator has a funda-
mental frequency of 100 MHz and a broadband phase noise
floor of <-175 dBc/Hz at ≥10 kHz offset frequencies. We
used a matched pair of these oscillators to characterize each
other and thus establish the lowest phase noise measurable
by the system (see Fig. 2, trace 2).
A critical aspect of using any reference oscillator is its
ability to track the wandering frequency of the free-running
laser. Our laser drifts less than 10 Hz in any 20 minute
measurement period. There are also occasional frequency
jumps of up to 2 Hz and the oscillator must be able to track
these as well. Alternatively, one can phaselock the repeti-
tion rate of the laser to a stable source and dramatically
reduce the requirements on the reference oscillator. We
have tried both approaches and measured approximately
the same phase noise spectrum except that within the loop
bandwidth (<200 Hz), the phase noise was reduced by
about 10 dB.
Figure 2 shows phase noise data from our measurement
© 2001 IEEE - LEOS, 11-15 November 2001, Paper: ThCC2
Offset Frequency (Hz)
L( f ) (dBc/Hz)
Trace 1: Phase noise of Ti:sapphire laser phaselocked to 100 MHz
crystal oscillator (average photocurrent 3.5 mA). Trace 2: Phase noise
of two identical 100 MHz crystal oscillators. Trace 3: Measurement
system noise floor.
SSB phase noise as a function of carrier offset frequency.
system, a modified HP 3047A. Trace 1 is the phase noise
of our Ti:sapphire laser at 100 MHz as measured against a
low noise 100 MHz crystal oscillator. Trace 2 is the phase
noise of a pair of identical 100 MHz crystal oscillators and
trace 3 is the system noise floor established by driving both
ports of the phase detector from a common crystal oscilla-
tor with bandpass filters and a 90 degree phase difference
between the two ports. (All phase noise is thus correlated,
simulating a pair of ideal oscillators). The phase noise of
the laser is seen to be quite good and close to that of the
crystal oscillator except in the audio region from 30 Hz to
20 kHz. Note; no special efforts were made to acoustically
isolate the laser from room noise such as fans, flowing wa-
ter, pumps, etc. We believe that with modest effort, the
noise in the audio range could be substantially reduced by
improving the phaselocked loop performance and acousti-
cally isolating the laser cavity.
The spur at 2.5 kHz is due to a resonancein the piezoelec-
tric translator used to control the cavity length for phase-
locking. The step down in noise at 25 kHz occurs at the
transition between the FFT and the analog RF spectrum
analyzer. Notice that for the data below 25 kHz, the FFT
analyzer is subject to signals which are simultaneously > 80
dB apart. In order to avoid overloading the front end of
the analyzer and exceeding the dynamic range of the A/D
converter, the instrument downranges to keep the peak at
200 Hz within linear operational limits. Up at 25 kHz, the
measured phase noise is actually below the front end noise
of the FFT analyzer by a few decibels. Above 25 kHz, the
RF spectrum analyzer takes over with a standard mixer-
type front end which has greater dynamic range and can
measure the weak noise power in the presence of the strong
noise signals at low frequencies.
The spurs at multiples of 100 kHz are due to the switch-
ing power supply in the diode-pumped solid state laser
used to pump the Ti:sapphire laser. And, finally, the spurs
above 10 MHz are due to RF sources elsewhere in the build-
If we wish to convert the phase noise data to rms timing
jitter, we must integrate the total double-sideband phase
noise power spectral density. Since the spectrum of our
laser is dropping rapidly from 1 Hz (i.e. the slope= −40
dB/decade), almost all of the timing jitter is contained in
the first few Hz of offset frequency. Thus, upon integration,
we find that ∆trms = 9.8 ps. On the other hand, if we
start the integration higher in frequency, say from 1 kHz
to 40 MHz, we find that ∆trms= 53 fs. Thus, timing jitter
is seen to be a matter of perspective.
The data indicate that fairly ordinary Kerr-lens mode-
locked lasers have the potential to rival high stability crys-
tal oscillators in terms of short term frequency stability
(phase noise). Long term drift can be easily controlled
by phaselocking to a primary or secondary frequency stan-
dard. This suggests that designing lasers with “technical
noise” considerations in mind may lead to applications in
high purity RF/microwave sources .
In conclusion, we have demonstrated remarkably low
phase noise from a conventional Kerr-lens modelocked
Ti:sapphire laser using a carefully assembled and charac-
terized phase noise system. The ultimate phase noise floor
of -170 dBc/Hz is, we believe, the lowest laser phase noise
reported to date, and approaches that of high performance
100 MHz crystal oscillators.
 A. Poppe, L. Xu, F. Krausz, and C. Spielmann, IEEE J. Select.
Topics Quantum Electron., vol. 4, no. 2, pp. 179–184, 1998.
 T. R. Clark, T. F. Carruthers, P. J. Matthews, and I. N. D. III,
Electron. Lett., vol. 35, pp. 720–721, Apr. 1999.
 R. P. Scott, C. Langrock, and B. H. Kolner, IEEE J. Select. Top-
ics Quantum Electron., vol. 7, July/Aug. 2001. To be published.
 H. Shi, I. Nitta, G. Alphonse, J. Connolly, and P. J. Delfyett,
Electron. Lett., vol. 34, pp. 2250–2252, Nov. 1998.
 X. S. Yao, L. Davis, and L. Maleki, IEEE J. Lightwave Technol.,
vol. 18, pp. 73–78, Jan. 2000.
© 2001 IEEE - LEOS, 11-15 November 2001, Paper: ThCC2