TV and video game streaming with a quantum receiver: A study on a Rydberg atom-based receiver's bandwidth and reception clarity

Rydberg states (highly excited) of atoms have been of growing interest in the past decade and have provided an avenue for making a variety of different sensors.¹ This is possible since Rydberg states are highly sensitive to electric fields and, depending on the Rydberg state used, allow for detecting fields ranging from DC to THz.^1–12 They have been used for the detection of electric fields for amplitude modulation (AM)/FM (and phase modulation) receivers,^13–20 spectrum analyzers,²¹ voltage standards,⁶ angle-of-arrival,²² and many more applications.^23–27 These sensors even allow for calibrated measurements traceable to the international system of units (SI)²⁸ for both electrical and radio frequency power.^8,29–31

A current thrust in the development of Rydberg atom-based sensors is geared toward improving the sensitivity and in understanding the limits of bandwidth for these quantum-based receivers. While the reception of various signals has been shown, the reception of live television (TV) has yet to be demonstrated. This limitation largely arises from the bandwidth to signal relationship. A recent study demonstrated this link between the bandwidth and signal strength.³² We take this a step further to optimize the signal/bandwidth and utilized optical homodyne detection to receive a TV/gaming signal that can be directly output from a photodetector to a TV.

In this demonstration, we utilized electromagnetically induced transparency (EIT) to probe the Rydberg state of interest (⁠$50D5/2$⁠), shown in Fig. 1(a) and trace in (b) trace (i). Then, we apply a radio frequency (RF) field that is resonant with the $50D5/2→51P3/2$ transition which results in the Autler–Townes (AT) splitting of the Rydberg state, as shown by another trace in Fig. 1(b) trace (ii). This 17.0434 GHz field is the carrier of our signal. The signal is in the form of analog amplitude modulation of the carrier, as shown in Fig. 1(c). Our lasers are locked to an ultra-low expansion cavity using the Pound–Drever–Hall method. We tune the lock to the atomic resonance using an additional modulation to create a tunable offset lock. By locking the laser to the center of the EIT resonance, the energy level shift translates to an amplitude modulation of the transmission signal of the laser,^13,16,17 shown in trace (iii) of Fig. 1(b). This modulation of the laser is then detected by the photo-detector and can be fed directly to a cathode-ray tube TV. Here, we used an analog to digital converter to transform the analog signal into video graphics array (VGA) format to display on a monitor, shown by Fig. 1(d) (Multimedia view). This is the demonstration of a quantum receiver being used to watch a live feed from a camera. The underlying light–atom interaction that makes this possible, EIT and AT splitting, require quantum mechanics to be described. While this is closer to so-called “quantum 1.0,” the nature of the receiver allows for the potential to be pushed into the realm of “quantum 2.0” sensors by utilizing quantum enhanced light.³³ 

In this manuscript, we present a detailed study on how the signal and bandwidth depend on the beam sizes and powers, and how these conditions affect the reception of the live video feed, which ultimately determines the clarity of the reception and if color can be received. Figure 1(d) shows a clean reception of a TV signal, with optimal beam size and optical power. In this experiment, we used the setup shown in Fig. 2. This setup allowed us to switch between homodyne detection and balanced differential detection depending on the observed signal.

We use a 780 nm external-cavity diode laser (ECDL) as our probe laser and a 960 nm ECDL amplified by a tapered amplifier that is fed into a second harmonic generation cavity to generate our 480 nm coupling laser. Both lasers' powers at the cell are stabilized at the cell location by using acoustic optic modulators. The coupling laser power is fixed to 70 mW for this study.

The polarizing beam displacer (PBD) splits the 780 nm laser into a signal and reference beam. The signal beam is overlapped with the coupling laser in the vapor cell while the reference beam is not. The vapor cell is a 70 mm quartz cell with windows at the Bragg angle of the probe laser to avoid reflections. The two windows have opposite angles. After interaction in the cell, both signal and reference beams are overlapped spatially using the second PBD. The waveplate slow axis is set to either 0° or 45°. When we want to utilize balanced differential detection, the waveplate is set to balance the signal and reference power and the second waveplate is set to be aligned with the slow axis. For homodyne detection, the first waveplate is set so that the signal beam power is 30 μW and the reference beam is $∼1.5$ mW. We stabilized the system with boxes to reduce fluctuations from air currents.

We utilized differential detection when taking power-dependent data for long-term stability and used homodyne detection for added signal strength at lower detector gain. The detector used in these experiments has limited bandwidth at high gain, so we employed homodyne detection to increase the signal such that we could operate the photodetector with lower gain.³⁴ 

For these experiments, we varied the probe laser beam size and adjusted the coupling laser beam width to keep it slightly larger than the probe. This avoided unwanted effects from spatial intensity variations. The beams were focused at the center of the cell to achieve the desired beam waist. For each beam width, we varied the power of each beam to change the Rabi frequency. Figure 3(a) shows the EIT signal as a function of coupling laser detuning for various probe powers. The frequency axis was calibrated using the atomic structure present in the Rydberg states. The stronger EIT peak is the 5P$3/2→50D5/2$ transition. The weaker peak at the 0 MHz detuning location is from the 5P$3/2→50D3/2$ transition. The separation of these peaks is 92 MHz. We took these measurements at a beam full-width at half-maximum (FWHM) of $85 μ$m.

We also measured the rise and fall times of the atomic response by locking both laser frequencies to the EIT resonance and applying a 10 kHz square wave. This was done by locking both the probe and coupling laser to the EIT resonance and observing how the RF field effected the transmission signal. The RF field was generated by a horn antenna with a gain of 17 in our operating frequency range. The horn was placed at 33 cm from the cell location. The 17.0434 GHz RF field was generated by a signal generator and was mixed with a square wave modulation from a function generator using an RF mixer. For the case of the square wave modulation for finding rise and fall times, the RF power fed to the horn was 9.3 dBm that resulted in a field of 1.15 V/m at the location of the atoms. This modulated RF field was then fed to a horn antenna to radiate the atoms, as shown in Figs. 3(b)–3(d).

To demonstrate the effects of beam size and powers on the EIT signal and response times, we collect traces, similar to those shown in Fig. 3 for different probe beam powers and beam waists. From these measurements, we extract the EIT height, EIT width, rise times, and fall times, shown by Fig. 4. The extracted values are the average of 10 traces and the error bars are a standard deviation of the 10 measurements. Figure 4(a) shows the height of the EIT peak as a function of probe laser Rabi frequency for each beam width. The height of the EIT peak is proportional to the number of atoms that take part in the EIT interaction, which is proportional to the interaction volume. Increasing beam width increases the interaction volume, so for the same Rabi frequency a larger beam results in a stronger EIT peak. Figure 4(b) shows the EIT linewidth as a function of the probe Rabi frequency for each beam width. As opposed to the height, the linewidth does not depend on the beam size and is proportional to the Rabi frequency of the probe laser. Note that the measurements for a 55 μm beam width is not in line with the rest of the traces. This is due to the strong divergence of the probe beam used to obtain the tight beam waist at the center of the cell. For this reason, the average beam waist in the cell was 130 μm, nearly a factor of three larger than the actual beam waist. While this was the case for a 55 μm beam waist, the larger beam sizes (⁠$≥$200 μm) used in this paper had an average beam waist that was within 5% of the focused beam waist. For the case of the 85 μm beam waist, the average beam waist was roughly 25% larger. In addition to this, for the case of the 55 μm beam waist, the divergence is not simply defined. While the beam waist is changing, so is the number of atoms in the interaction. This would weight the signal such that larger areas would contribute more to the EIT signal.

We next investigated the rise and fall times for the atomic response to a square wave as shown in Figs. 4(c) and 4(d). The fall time is the time or the EIT signal to go from 90% to 10% of the maximum and the rise time is the time for the EIT to signal to go from 10% to 90% of the maximum signal. The temporal response does not significantly depend on the probe Rabi frequency, as opposed to the EIT height. However, both the rise and fall times depend strongly on the beam width. These times are proportional to the beam width, which changes the length of time an atom spends in the interaction volume. Since atoms are moving in and out of the interaction volume, smaller interaction volumes have a larger “refresh rate” for the interaction. The average velocity of room temperature Rubidium atoms is roughly 240 m/s. Hence, an atom will be able to transit through the interaction region within the transit time t_transit as given by

where ω is the beam waist [given by FWHM $/2 ln(2)$] and ν is the average atomic velocity. We compare the average rise time and fall times to the average transit times for the different beam sizes, as shown in Table I.

We found that the rise and fall times vary with the transit time, except for the $55 μ$m width. However, this is due to the average beam waist being larger than the average waist for the case of a beam focused to 85 μm. In addition to this, we note that in some cases that the rise time is faster than the transit time. This is due to the rise time being independent of the atomic decay. Establishing a Rydberg population is dependant on the effective Rabi strength of the two-photon interaction.

These results allowed us to optimize the beam size for bandwidth and achieve a sufficiently fast response to stream live video signals from a video camera and from a video game console. For this demonstration, we used an optical homodyne setup, with a signal beam of 30 μW and reference beam of 1.5 mW. The half-waveplate near the photodetector in Fig. 2 rotated the polarization of the signal and reference beams by 45° with respect to the polarizing beam cube slow-axis to mix the reference and signal on the photodetector.³⁶ 

The video format that is output by the camera and the video game console used in this study was National Television Standards Committee 480i, or standard definition. In this example, we streamed the video of a printed color test pattern, shown in Fig. 5(a). The direct signal output of the video camera was shown in Figs. 5(b)–5(e). The signal is an analog waveform that gives information on the frames, rows, and pixels. Figure 5(b) shows several fields (each is half of an interlaced frame), where we have identified the trigger marker for a given one. We also label the 240 active rows for each field. Figure 5(c) shows several rows for a given field and the trigger for each row. Figure 5(d) gives the information for each row. After each row trigger, there is a “colorburst” signal that sets the reference phase for the 3.58 MHz carrier that determines the color. Each pixel in the active area of the row is represented by one quarter-cycle of a 3.58 MHz carrier, yielding roughly 720 pixels per row. The amplitude of the cycle gives the saturation of the color, and the phase of the cycle relative to the colorburst gives “chrominance” or hue color information. The color depth for the video camera is nominally 24-bit, with eight bits of information in each of the red, green, and blue basis colors. Brightness or “luminance” is given by the signal offset, making this format backward compatible with black-and-white images when the color information is not present, as seen in Fig. 6. Figure 5(e) shows the frequency spectrum of the transmitted signal obtained by performing the FFT on two full fields of transmitted data. The data were normalized so that the differences between the 0 V trigger and voltage offset of the colorburst were the same as those for the received data. This will then allow for a fair comparison of the strength of the 3.6 MHz carrier that determines the fidelity of the signal.

To receive the video signal, both lasers were locked to the EIT resonance. The output from the camera was mixed with a 17.04 GHz carrier using an RF mixer. The strength of the carrier wave feeding the horn antenna was 14.3 dBm. This results in a field strength of 2 V/m at the location of the atoms. The signal from the video camera modulated a 17.04 GHz carrier using an RF mixer. The modulated carrier was fed to a horn antenna to direct the field to the atoms. We compare the original video signal to the down-converted signal on the probe laser received through the atoms in Figs. 6(i) and 6(ii) for each column (a)–(d). The columns show the measurement results for four different beam sizes, $800$⁠, $400$⁠, $200$⁠, and $85 μ$m, respectively. Row (iii) in Fig. 6 shows the frequency spectrum of the received signals obtained by performing the FFT. The data were normalized so that the differences between the 0 V trigger and voltage offset of the colorburst were the same for all the beam sizes. The last row (iv) in each column shows the demodulated video displayed on a screen. We see that for the larger beam size (800 μm FWHM), the received signal Fig. 6(a-ii) is heavily distorted compared to the transmitted signal Fig. 6(a-i) due to the slow rise and fall times of the atoms in this configuration. To compare the actual bandwidth received, we look at how the frequency space of the received signals changes near the 3.58 MHz carrier to asses the video quality, shown in Fig. 6(a-iii). The video from the received signal for 800 μm, Fig. 6(a-iv), is very blurry with no color present. Similarly, the 3.6 MHz carrier is not visible in the FFT. For a beam size of 400 μm [Fig. 6(b)], we see that the received signal is less distorted, and the received video Fig. 6(b-iv) is clearer but lacks color information. The FFT for this beam size does not show the carrier wave, but seems to have some sidebands starting to show. For a beam size of 200 μm [Fig. 6(c)], we see that the received signal is less distorted, and the received video Fig. 6(c-iv) is slightly blury and contains color information. However, there are splotches where the color identification fails. Figure 6(c-ii) shows the first signs of the 3.6 MHz carrier. However, if we compare the strength of the 3.6 MHz carrier here to that of the transmitted signal, it is nearly an order of magnitude weaker. For a beam size of 85 μm [Fig. 6(c)], we see that the received signal is less distorted, and the received video Fig. 6(c-iii) is nearly as sharp as the direct image in Fig. 5(a) and has all the color information with correct saturation. The FFT of the received signal also is nearly as strong as the transmitted signal and only off by less than a factor of 2.

Demodulating the color information requires a higher bandwidth (⁠$≈3$ MHz) than the black and white intensity information. By optimizing the atomic response times according to Table I with a probe beam FWHM = $85 μ$m, we were able to achieve video reception that was not only sharp but also captured color information. To determine the data rate that was received, we calculate the effective bit rate of the transmitted signal. The video camera outputs 480i video, interlacing 240 rows (scan lines) of 720 pixels at a field rate of 60 Hz to give a composite 480 × 720 pixels every 30 Hz. If each pixel carries 24 bits of color information (eight bits for each of red, green, blue) and the nominal bitrate for 480i is 249 Mbps $≈60 fieldss×240 linesfield×720 pxline×24 color bitspx$⁠, not including the information downtime during temporal alignment parts of the signal. Even though our Rydberg response rate and the detector bandwidth both fall well below this rate, we are able to recover enough phase information from the 3.58 MHz color carrier to display 480i color video using an analog-to-digital video converter. However, if we compare the frequency spectrum of the transmitted signal to the received signal for our best video quality, we can see that there is a difference in the strength of the 3.6 MHz carrier wave by roughly a factor of 2. This likely means that we are not retrieving the full information that is transmitted. This is due to the “slew rate” of the atoms that lowers the actual signal change per unit time. To be conservative, we estimate our data rate to be reduced by this factor, for an effective rate of 125 Mbps.

As a final example, we show data from streaming a game console. The video game console also outputs “standard definition” NTSC 480i video. Figure 7 shows the transmitted Fig. 7(i) and received data Fig. 7(ii) from a game console as the game is being played. Figure 7(a) is for a beam size of 800 μm and this resulted in a blurry image, Fig. 7(b) is for a beam size of 85 μm and the result is a clear color image. The stability of receiving the game signal illustrates the fact that the game could be played in real-time for several hours without losing the signal or its color clarity.

We have demonstrated the ability of the Rydberg atom receiver to receive live color video, using both a camera and a game console. The bandwidth was improved by optimizing the probe laser beam width, which tuned the average time the atoms remained in the interaction volume. We also show the dependence of EIT height, linewidth, rise time, and fall time on the probe beam width and the Rabi frequency. EIT height and width both depend on the Rabi frequency, but only the EIT height depends on the beam width. The rise and fall times of the atomic response did not depend on the Rabi frequency, but they show a proportional dependence on the beam width. This dependence is due to the transit time of the atoms through the cross section of the beam. Additionally, we used homodyne detection to overcome the limitation of our detector gain-bandwidth product by pre-amplifying the signal. The bandwidth depends on the beam sizes and powers, which ultimately determines the clarity of the reception and if color can be received. Finally, we achieved a data rate of 125 Mbps with this receiver.

This work was partially funded by the DARPA Quantum Apertures program Grant No. HR0011152628 and by the NIST-on-a-Chip (NOAC) Program Grant No. 6722231-000CHP15NOAC.

The authors have no conflicts to disclose.

Nikunjkumar Prajapati: Conceptualization (equal); Data curation (lead); Formal analysis (lead); Investigation (lead); Methodology (lead); Visualization (lead); Writing – original draft (lead); Writing – review and editing (lead). Andrew P. Rotunno: Formal analysis (equal); Investigation (equal); Methodology (equal); Visualization (equal); Writing – original draft (equal); Writing – review and editing (equal). Samuel Berweger: Formal analysis (equal); Investigation (equal); Methodology (equal); Writing – original draft (equal); Writing – review and editing (equal). Matthew T. Simons: Formal analysis (equal); Investigation (equal); Methodology (equal); Validation (equal); Visualization (equal); Writing – original draft (equal); Writing – review and editing (equal). Alexandra B. Artusio-Glimpse: Investigation (equal); Methodology (equal); Writing – original draft (equal); Writing – review and editing (equal). Stephen Voran: Formal analysis (equal); Validation (equal). Christopher L. Holloway: Funding acquisition (lead); Investigation (equal); Methodology (equal); Project administration (lead); Resources (lead); Supervision (equal); Visualization (equal); Writing – original draft (equal); Writing – review and editing (equal).

The data that support the findings of this study are available from the corresponding author upon reasonable request.