2.3 Laser Systems and their Repetition Rate Control
2.3.3 Instrumental Implementation
The present asynchronization circuit is based on that discussed by Gebs et al. [42], and is presented in Figure 2.8. The novel aspect of this general design of offset-locking circuit was the use of ‘DDS boards’; that is, components DDS1 and DDS2 in the schematic. Direct-digital-synthesis (DDS) boards are low phase noise frequency synthesizers. The devices use an external sample clock and a digital tuning word to generate a tunable digital signal that is then converted to an analog sinusoidal output using a digital-to-analog converter.
DDS boards can produce a range of frequencies up to nearly half the supplied clock frequency, with a tuning step size of∼4µHz for the particular model used herein [54]; the external sample clock in the present case is a harmonic of the master laser output.
The tracing of the circuit is perhaps best started at the master laser, as it is presently left free-running, with no active control. Small portions of power from the master laser illuminate two monitor photodiodes; these are the optical inputs to the asynchronization circuitry. PD1 is lower in bandwidth as compared to PD2 and supplies the 12th harmonic of the laser repetition rate, at approximately 960 MHz, though an amplifier and power splitter to provide the clocking signal to both of the DDS boards (DDS1 and DDS2). In this regard, when the DDS boards are initially turned on and programmed with the frequency of their reference clock, the repetition rate of the master laser, as read off of a referenced frequency counter, is used as the clocking frequency (∼960 MHz). During the course of an experiment, the repetition rate of the master laser may drift away from this initial value; perhaps a few hertz or even up to a few tens of Hz. These drifts are generally small compared to the frequency of the master laser, and thus represent a small fractional error, unlikely to significantly affect the definition of the temporal waveform timebase or subsequently, the frequencies of spectral line centers. However, with improvements in frequency performance or simply to make the system more robust, it may be worthwhile to pursue at least loose stabilization of the master laser.
Returning to PD2, this is the other photodiode that the master laser beam illuminates, and is a higher
FROM PUMP LASER FROM PROBE LASER toyFasty PZTtoySlowy PZT
PD 2 PD 1 PD 3
AMP 1 BP 3
PI PI
BP 1
BP 2AMP 2 AMP 3PS
DDS 1 DDS 2
MX 1 MX 2
MX 3 LP
BP 4 BP 5
AMP 4 AMP 5 AMP 6AMP 7 LOOP 1 LOOP 2
4.8 GHz 4.8 GHz
960 MHz
70.000 MHz 70.006 MHz
1.9 MHz G = 100G = 1
4.87 GHz 4.87 GHz
IF
LORF LORF
IF
LORF Legend:PDy=yphotodiode,yBPy=ybandpassfilter,yAMPy=yamplifier,yPSy=ypowerysplitter, DDSy=yDirectyDigitalySynthesizer,yMXy=yMixer,yLPy=ylow-passfilter,yPI/LOOPy=yProportional-IntegralyLoopyFilter Figure2.8:TheasynchronizationcircuitthatmaintainsthereprateoffsetintheTHz-TDSinstrument.
frequency model; PD2 supplies the 80th harmonic of the master laser repetition rate (4.8 GHz) through a bandpass filter and amplifier for mixing against DDS1. The slave laser similarly illuminates PD3 (identical to PD2), the signal (at 4.8 GHz) from which is band-passed and filtered to select a single comb tooth, amplified, and mixed with the output of DDS2. So overall, the master laser is clocking both of the DDS boards, and each of the master and slave lasers is having its 60th harmonic comb tooth at∼4.8 GHz mixed against the output of a DDS board.
As an aside, we point out that in a synchronization circuit, there would be no need for the DDS boards;
we could simply and directly mix the matched comb teeth selected from each laser, and use the resulting error signal to close the loop. However, for the precise asynchronization of the lasers, the DDS boards are necessary. In this regard, we point out that the DDS boards, while phase coherent, due to the common (master laser) clocking, are set to different output frequencies: DDS1 to 70.000 MHz and DDS2 to 70.006 MHz—the latter being a 6 kHz increase in frequency. And this is the key point: the asynchronization circuit is basically a synchronization circuit where we ‘give’ a small, extra amount of frequency to the slave laser, so that when its signal is synchronized against that from the master laser, the slave laser is actually running at a lower repetition rate, in this case 6 kHz lower as measured at the 60th harmonic at which the PLL circuit operates.
Dividing 6 kHz by 60 yields 100 Hz, the desired offset.
So when the loop has zero error signal, the slave laser will be operating at a repetition rate that is 100 Hz lower than that of the master. Still under this zero error condition, when the master and slave lasers’ 60th harmonic signals are received by the monitor photodiodes, the slave lasers’ signal will be 6 kHz lower than that of the master laser. However, after the slave laser is mixed with the output of DDS2, which is 6 kHz higher than that of DDS1 (and the sum frequency output of the mixer selected and amplified) the master and slave laser signals reaching MX3 will be the same and the intermediate frequency (IF) signal leading into the PID controller will be zero.
Any nonzero error voltage at the last mixer would be passed on to the PI controllers for correction to the laser. The first component in the path after the ‘last’ mixer (MX3) is a 1.9 MHz low-pass filter; this is in place to reject any high frequency components that may have leaked through the circuit to this point. The correction bandwidth of the overall circuit is<100 kHz, so any high frequency signals at this point could only serve to potentially saturate later amplifiers or otherwise interfere with circuit operation, and so should be rejected.
Following the low-pass filter, the filter is immediately run into a variable voltage amplifier, typically set to a gain of 100 during lock conditions. It should be noted that in Figure 2.8, it looks like there is quite a gap or length of cable from MX3 and the LP over to AMP6, but actually they are all connected straight together in one block. Following the mixer, any noise that is added on the circuit will be interpreted by the loop as
something for which a correction is needed. Therefore it is important to place AMP6 as close after MX3 and LP as possible. Following AMP6, the error signal is directed into the PI controlled labeled as Loop1; this is the fast loop attached to the high-speed PZT in the slave laser. This PZT or ‘tweeter’ is attached to the back of a folding mirror inside the laser cavity. The Coherent fast PZTs are known to have a relatively high capacitance, so a gain = 1 line driver, designed for high current, is used as suggested in [55].