Domesday Duplicator

Overview

The Domesday Duplicator is intended to allow high-quality back-ups of the analogue information contained on the laserdiscs by bypassing most of the 30-year-old electronics in the VP415 player. Direct RF sampling also allows all information on the laserdiscs to be stored (unlike conventional RGB sampling of the video output). Since AIV laserdiscs are a combination of video, pictures, sound and data (as well as numerous VBI streams), direct RF sampling is the preferred method of preservation.

The Domesday Duplicator is a USB 3 based DAQ capable of 32 million samples per second acquisition of analogue RF data at 10-bit resolution.

Domesday Duplicator 1_4

The hardware is a USB3 based 10-bit analogue to digital converter designed to allow the backup of Domesday AIV laserdiscs (as well as generic laserdiscs) through the direct sampling of the RF data from the optical head (laser) of a laserdisc player.

The hardware/software solution is designed to act as a sampling front-end to the ld-decode (software decode of laserdiscs) project https://github.com/happycube/ld-decode and replaces the generic TV capture card to provide high-frequency sampling with 4 times the sample resolution.  Increasing the sample resolution allows better capture of disc overall however, the primary advantage is that the Domesday Duplicator provides better performance for weaker RF signals especially at the start of a laserdisc (where the RF output has lower amplitude) and when the disc is degraded due to age and surface damage.

Domesday86 would like to thank Chad Page (the author of the ld-decode project) – without his tireless work producing ld-decode and his assistance in modifying the library to support the Domesday Duplicator project, this preservation method would not have been possible.

Image showing a comparison of the same frame captured as RGB and as RF

The following block-diagram shows the 4 high-level components of the Domesday Duplicator:

Domesday Duplicator block diagram

These components are described in more detail in the sections below.

The laserdisc player used during testing is a Sony LD-1500P modified with a BNC connector wired to the test connector CN8 exposed at the back of the player.  The RF test-point provides a 300mV peak-to-peak output from the PR-80 circuit board in the player.  This RF output is the signal generated by the optical assembly of the player and represents the raw output from the player’s laser pick-up.  The Sony LD-1500P is a PAL only laserdisc player.

Please note that the Domesday Duplicator is currently in beta and has not been fully tested.  Information on this page will be updated as the project continues.

Hardware

RF Front-end

The RF front-end design can be seen in the following schematic diagram:

Domesday Duplication – RF Front-end schematic 1_5

The RF front-end takes the raw RF output from the laserdisc player, decouples the DC offset and then applies a DC offset of 2.5V required for the ADC (note that the offset is provided by the ADC’s reference signal generator (REFT and REFB)).  A high-bandwidth OPA690 single-rail opamp is then used to add a gain of 6.45 to the RF signal to ensure the amplitude of the RF is suitable for the 2V peak-to-peak range of the ADC.  The OPA690 was chosen due to the high bandwidth and single-rail supply (allowing the Domesday Duplicator to be powered using only the 5V VBUS supply from the USB host).

The gain of 6.45 is chosen based on the assumption that the maximum input amplitude from the laserdisc player will be 300mV maximum (150mV peak-to-peak from a source impedance of 50 Ohms).  300mV at 6.45 gain provides a peak-to-peak signal of 1.935V to the ADC.  Due to the analogue nature of the circuit the real gain of the RF front-end is measured at around 6.00 providing an peak signal amplitude of around 1.8V to the ADC stage (from a generated sine-wave of 300mV).

Measurement of the RF stage shows a consistent gain across a frequency range of 60KHz to 20MHz; signals above and below this range are attenuated by the circuitry.  Given the expected frequency span of the laserdisc RF signal, the frequency response is more than adequate for the application.

The output from the RF front-end is passed through a simple RC low-pass filter with a cut-off frequency of approximately 15.4 MHz.  This LPF acts as an anti-alias filter for the ADC stage (added to board revision 1_5).

Modifying the RF front-end gain

For other laserdisc players it may be necessary to modify the gain of the RF front-end.  The gain is set by two resistors R5 and R4.  The rules for resistor selection are provided by the OPA690 datasheet:

R5 and R4 should be greater than 200 Ohms and less than 1500 Ohms.

The parallel combination of R5 and R4 should be less than 300 Ohms ( (R4 * R5) / (R4 + R5) = parallel combination resistance).  So, (220 * 1200) / (220 + 1200) = 185.92 Ohms in the configuration shown by the schematic.

The overall gain is given by 1+(R5 / R4), so 1+(1200 / 220) = 6.45 in the configuration shown by the schematic.

10-bit ADC

The 10-bit ADC stage schematic can be seen in the following diagram:

Domesday Duplicator ADC stage 1_5

The ADC is a Texas Instruments ADS825 capable of 40 million samples per second (MSPS) at 10-bit resolution and provides a 10-bit parallel output.  The ADC data generated by the ADS825 is an unsigned binary offset with 0 representing the lowest possible amplitude, 1023 as the maximum possible amplitude and 512 representing the centre point (DC offset of 0V) of the signal.

The expected signal from the laserdisc player is shown in the following diagram taken from the Philips VP415 service manual (note that this is for a PAL specific player):

VP415 Frequency range diagram

Based on the highest expected Nyquist frequency of 8MHz the ADC must perform at least 16 MSPS in order to correctly sample the RF signal.  Due to this the Domesday Duplicator uses a sampling rate of 32 MSPS (double the required sampling rate) to ensure accurate sampling of the signal.  By doubling the maximum required sampling rate the loss of accuracy in the ADC due to ‘windowing’ is reduced.

Two line buffering ICs (74LVTH541) are used to buffer the output from the ADC towards the FPGA stage.  These line buffers reduce the capacitive loading of the databus and effectively isolate the ADC from digital noise that can cause high-frequency interference in the analogue stages of the ADC.  The design also provides power conditioning to ensure a smooth feed of both 5Vs and 3.3V to the ADC and RF front-end.

FPGA

Due to the unpredictable performance of a USB 3 interface an FPGA is used to act as a FIFO buffer between the ADC and the USB hardware.  Although a modern PC can handle many times the bandwidth required by the application the bandwidth is not guaranteed (since a general purpose PC is not a real-time system, other system activities can cause the available bandwidth to vary).

The FPGA is provided by a DE0-Nano.  The DE0-Nano is a low-cost FPGA development board from Terasic containing an Intel (Altera) Cyclone IV FPGA.  The interconnection between the Domesday Duplicator and the FPGA development board is shown in the following schematic diagram:

Domesday Duplicator module interconnection schematic 1_5

The FPGA uses a PLL function to provide a 32MHz clock to the ADC.  Data from the ADC is collected by the FPGA into a 10-bit wide FIFO buffer.  The FIFO buffer supports separate input and output clocking.  The interface between the FPGA and the USB 3.0 interface is clocked at 50MHz.  This dual-clock implementation allows the USB 3.0 to ‘catch up’ with the ADC data collection in the event of temporary loss of bandwidth.

In addition the FPGA changes the 10-bit unsigned FIFO output to scaled 16-bit signed data before passing the data to the USB 3.0 interface via a 16-bit databus.  In addition to the databus the FPGA provides several additional control signals used to provide flow-control of the data between the FPGA and the USB 3.0 interface.

The data bandwidth from the FPGA to the USB 3.0 interface is a nominal 512Mbits/sec (16 bit words at 32 million words per second) or 61Mbytes/sec.  Due to higher clock speed of the USB interface (50MHz) this can peak at  800Mbits/sec if required.

USB 3.0 interface

The USB 3.0 interface between the FPGA and the host PC is provided by a Cypress SuperSpeed Explorer development board.  This board provides a Cypress FX3 SuperSpeed USB 3.0 peripheral controller.  USB 3 is required due to the data bandwidth requirement of the Domesday Duplicator (USB 2.0 only provides a realistic bandwidth of around 280Mbits/sec which is too low for the application).  The FX3 provides a state-machine model (called GPIF II) that handles the transfer of data from the databus to the USB 3.0 interface and can run at a maximum of 100MHz (100 words of 32-bit data per second).  For the Domesday Duplicator, the FX3 is configured to use a 16-bit databus and the synchronous data clock is provided by the FPGA (at 50MHz).

The following schematic shows the interconnection to the FX3 board:

Domesday Duplicator FX3 connection 1_5

Software

Firmware

The current build of the Domesday Duplicator software is available from Github:

Domesday Duplicator Github

The Github repository contains the following items:

  • Kicad schematics and PCB design
  • FPGA Verilog HDL code for the DE0-Nano
  • GPIF II state-machine design for the FX3
  • FX3 firmware for the Cypress FX3 board

Software

The data capture software required by the host PC is currently based on the streamer example application provided by Cypress.  An open-source version of the data capture software is planned, but not yet complete.

Performance testing

RF Front-end

The following oscilloscope trace shows a 150mV peak-to-peak amplitude sine wave at 50 Ohms impedance at the BNC connector of the Domesday Duplicator (board revision 1_4):

Signal at BNC connector

The following oscilloscope trace shows the same signal on the input (pin 3) to the OPA690 opamp:

Signal input to opamp

The following oscilloscope trace shows the same signal on the input (pin 25) to the ADS825 ADC:

Signal input to ADC

ADC performance

The following FFT graphs show a spectrum analysis of the Domesday Duplicator at the key PAL frequencies for a laserdisc.  The input signal was a sine wave generated at 150mV peak-to-peak with 50Ohms impedance (board revision 1_4):

683KHz FFT – Audio 1
1066KHz FFT – Audio 2
6.76MHz FFT – Video (Sync level)
7.10MHz FFT – Video (Black level)
7.90Mhz FFT – Video (White level)

It can be seem from the FFT analysis that the peak amplitude of the signal is consistant over the required frequency range.

The following FFT shows a PAL CAV laserdisc signal from around the centre of the disc (note this disc contains only digital audio):

PAL CAV laserdisc FFT

FAQ

Q: Why not just copy the laserdisc digitally like a DVD/CD-ROM?
A: Laserdiscs are analogue, so it’s not possible to simply copy the video files from the disc as you would with digital media.

Q: Don’t the BBC already have a high-quality copy of the Domesday contents?
A: Yes and no.  The BBC have a digital betamax version of the Community and National Domesday laserdiscs which includes both the video and audio.  However, there is a lot more information on an AIV laserdisc (ADFS data, VBI data, etc.) and, more importantly, there are a number of other AIV discs not covered by this backup.  Finally, the digital betamax copy is not accessible to the public; so there’s no way to get a copy.

Q: Why not just use a video capture card on the RGB/SCART/Composite output of the VP415?
A: The video and audio produced by the VP415 has to pass through many analogue stages before it can be viewed as a PAL signal.  As the data is analogue each processing stage adds noise and distortion to the signal.  By sampling the RF directly from the player’s laser you avoid all of the lossy stages.  Furthermore, the RF sampling can be performed using any PAL compatible laserdisc player – VP415 players are quite rare, this technique allows any player to be used to read the AIV disc contents.

Q: I don’t think the picture’s that much better, what’s the big deal?
A: RF sampling preserves the contents of the laserdisc far better than ‘ordinary’ video capture.  It helps if you think of the RF sampling as ‘scanning’ the laserdisc surface with a laser; the RF sample is effectively a scan of the disc’s surface.  Once scanned the methods for decoding the disc’s contents can be improved over time and more types of information can be extracted from the image.  Right now the decoding is primarily for the video and audio content but soon it will be possible to extract VBI and data from the image too.  Furthermore, the techniques used to recover content such as video can also be improved leading to better output from the image (without the need to resample the disc).  RF sampling preserves the contents of the disc in a way that RGB capture cannot.

Q: Why didn’t you just use an off-the-shelf SDR board?
A: Although there are SDR boards available that can perform RF capture they have RF front-ends designed for radio and therefore would need significant modification for use as a laserdisc sampler.  SDR boards are also quite expensive in comparison to the ‘custom’ part of the Domesday Duplicator.  A 10-bit single-ended ADC allows the design to be a very cost-effective solution.  Domesday Duplicator is also designed (from the beginning) to be a complete open-source, open-hardware solution, so a custom design allows better control over the licensing used for the project.

Q: How big are the RF images?
A: The duplicator outputs the 10-bit samples as 16-bit signed data.  The data is created at around 62 Mbytes per second.  So a CAV disc capture is around 130 Gbytes per side (260 Gbytes per disc).  CLV discs contain more video (up to 60 minutes per side) and require around 220 Gbytes per side.

Q: This project is just the most awesome thing I’ve ever seen!  How can I help?
A: Hit the donate button on the page (above-right) and help us fund the necessary equipment needed to make Domesday86 a reality – every little helps!  All of the projects around Domesday86 are open-hardware and open-software, so donations are the only means of funding for the project.