Hardware testing with an AWG¶
This page describes how to test a completed Domesday Duplicator using an arbitrary waveform generator (AWG) rather than a LaserDisc player. Feeding the Duplicator a pure sine-wave of a known amplitude and frequency tests the whole capture solution end-to-end — input stage, amplifier gain, anti-alias filter, ADC, FPGA and USB — against a signal whose expected result can be calculated in advance.
The test is worth running on any newly built board, after any repair, and whenever a capture looks wrong in a way that could be either the player or the Duplicator.
For the theory behind the circuit under test, see the Hardware Guide.
What you need¶
- An AWG able to produce a clean sine-wave to at least 16 MHz from a 50 Ohm output, with the amplitude settable in millivolts peak-to-peak. Amplitude flatness matters — see the note under Setting the AWG.
- A short 50 Ohm BNC coaxial lead to connect the AWG to the Duplicator's BNC input. Keep it as short as practical, as you would for a real capture.
- A fully programmed Duplicator — DE0-Nano bitstream and FX3 firmware both loaded — with the capture application running on the host.
- Optionally an oscilloscope and a BNC T-adaptor, to confirm what the AWG is actually delivering.
The principle of the test¶
The ADS825 ADC accepts a 2 V peak-to-peak input centred on a 2.5 V DC offset, so 2 V p-p at the ADC is digital full-scale. The test targets 75% of full scale, which is 1.5 V p-p at the ADC. That is the same amplitude a correctly configured LaserDisc capture should produce: high enough to use most of the ADC's range, with enough headroom left that signal peaks do not clip.
Everything between the BNC input and the ADC is a linear amplifier whose gain is set by the 4-way DIP switch on the Duplicator PCB. So the input amplitude needed to reach 75% of full scale is simply:
The capture application's amplitude readout reports the peak amplitude of the captured signal as a fraction of full scale, assuming a sine-like crest factor of √2. A pure sine therefore reads 0.750 when the test is set up correctly — the readout is the measurement instrument for this test, and no separate analysis of the captured file is needed.
Preparing the capture application¶
In the capture application's preferences, enable:
- Show RMS amplitude as number — this is the 0.000 to 1.000 readout the test uses. The chart version can be enabled as well, but the number is what should be recorded.
- Show advanced capture statistics — this adds the
RecentMin:andRecentMax:readouts and the clipped-sample counts, which give the same measurement in raw 10-bit ADC codes.
Use the 10-bit .lds capture mode. The amplitude readout is derived from the same data
in every mode, but the raw code values quoted below are 10-bit values.
In 10-bit terms the ADC's range is 0 to 1023, centred on 512. A correct 75% test signal therefore swings between approximately code 128 and code 895, and both clipped-sample counts should stay at zero throughout.
Setting the AWG¶
Set the generator to:
| Setting | Value |
|---|---|
| Waveform | Sine |
| Output load / termination | 50 Ohm |
| DC offset | 0 mV |
| Modulation, sweep, burst | Off |
| Amplitude | See the table below |
Check the output load setting first
An AWG displays its amplitude according to the load it has been told to expect. The Duplicator's input is approximately 50 Ohms, so the generator must be set for a 50 Ohm load. If it is left on High-Z it will deliver twice the displayed amplitude, and every reading in this test will come out at 1.5 rather than 0.75 — with the signal clipping hard against both ends of the ADC range. The right-hand column of the table below gives the equivalent High-Z setting for generators that cannot be switched.
Set the DC offset to 0 mV
The generator's DC offset must be 0 mV so that its output is centred around 0 V. The
Duplicator's input stage adds its own 2.5 V bias to place the signal at the centre of the
ADC's range, so any offset the AWG adds is passed on and shifts the captured signal away
from code 512. That eats into the headroom on one side and can clip the peaks on that
side while the amplitude readout still looks close to correct. A symmetrical pair of
RecentMin: and RecentMax: readouts about 512 is the check that the offset is right.
The amplitude required depends only on the gain configuration set on the DIP switch. Switch positions are given in the order 1-2-3-4, where up is 1 and down is 0.
| Configuration | Switch position | Gain | AWG amplitude (50 Ohm) | AWG amplitude (High-Z) |
|---|---|---|---|---|
| 15 | 1111 | 2.02 | 743 mV p-p | 1486 mV p-p |
| 7 | 0111 | 2.17 | 691 mV p-p | 1382 mV p-p |
| 11 | 1011 | 2.27 | 661 mV p-p | 1322 mV p-p |
| 13 | 1101 | 2.45 | 612 mV p-p | 1224 mV p-p |
| 3 | 0011 | 2.54 | 591 mV p-p | 1182 mV p-p |
| 14 | 1110 | 2.59 | 579 mV p-p | 1158 mV p-p |
| 5 | 0101 | 2.79 | 538 mV p-p | 1076 mV p-p |
| 6 | 0110 | 3.02 | 497 mV p-p | 994 mV p-p |
| 9 | 1001 | 3.04 | 493 mV p-p | 986 mV p-p |
| 10 | 1010 | 3.34 | 449 mV p-p | 898 mV p-p |
| 1 | 0001 | 3.80 | 395 mV p-p | 790 mV p-p |
| 12 | 1100 | 4.00 | 375 mV p-p | 750 mV p-p |
| 2 | 0010 | 4.40 | 341 mV p-p | 682 mV p-p |
| 4 | 0100 | 6.00 | 250 mV p-p | 500 mV p-p |
| 8 | 1000 | 8.50 | 176 mV p-p | 352 mV p-p |
Testing at a single gain setting is enough to verify the frequency response. Testing at the two extremes — configuration 15 (2.02) and configuration 8 (8.5) — additionally confirms that the gain switch and its resistor network are behaving, since both should produce the same 0.750 reading from very different input amplitudes.
AWG amplitude flatness
An AWG's own output amplitude falls off towards the top of its range, and a generator with ±0.5 dB of flatness error will produce ±6% of apparent error in this test that has nothing to do with the Duplicator. If the results are marginal, put an oscilloscope on a BNC T-adaptor at the Duplicator's input and confirm the delivered amplitude at each test frequency before blaming the board. Note that the ADC injects clock noise back onto the input, which will be visible on the oscilloscope while the DE0-Nano is fitted and running; measure the peak-to-peak of the underlying sine, not the noise.
Test frequencies¶
The Duplicator's anti-alias filter is a 2-pole elliptic low-pass filter with a 13.2 MHz stop-band, protecting the 20 MHz Nyquist limit of the 40 MSPS ADC. Four frequencies inside the pass-band and one above the cut-off are enough to characterise it.
The expected readings below are relative to the measured filter response shown in the Hardware Guide, converted to the linear amplitude readout. Leave the AWG amplitude unchanged as the frequency is stepped.
| # | Frequency | Purpose | Expected reading | Pass |
|---|---|---|---|---|
| 1 | 1 MHz | Reference — flat part of the pass-band | 0.750 | 0.72 to 0.78 |
| 2 | 4 MHz | Lower video band | 0.74 | 0.68 or above |
| 3 | 8 MHz | Upper video band | 0.72 | 0.65 or above |
| 4 | 12 MHz | Just below the filter cut-off | 0.69 | 0.60 or above |
| 5 | 16 MHz | Above the cut-off — should be rejected | 0.05 to 0.10 | 0.15 or below |
Test 1 is the reference: the amplitude is set here, and every other frequency is judged against it. Within the pass-band the filter has some designed-in ripple, so tests 2 to 4 are expected to sit slightly below the reference — around 1 to 2 dB down at worst — but never above it.
Test 5 is the one test that must not read 0.75. At 16 MHz the signal is well into the filter's stop-band and should be attenuated by at least 15 dB, so the readout should collapse to a small fraction of the reference. A 16 MHz signal that survives at anything like full amplitude means the filter is not working, and any capture made with that board will contain aliasing that cannot be removed afterwards.
Procedure¶
- Power the Duplicator, connect it to the host over USB 3.0 and start the capture application. Confirm the application reports the device as connected.
- Note the DIP switch setting on the PCB and look up its gain and required AWG amplitude in the table above.
- Set the AWG to sine, 50 Ohm output load, 0 mV DC offset so the output is centred around 0 V, the amplitude from the table and 1 MHz. Leave the output off for now.
- Connect the AWG to the Duplicator's BNC input with the coaxial lead, then enable the AWG output.
- Start a capture and let the amplitude readout settle — it is a rolling average, so give it
several seconds. Record the amplitude reading,
RecentMin,RecentMaxand both clipped counts. - If the reading is not 0.750 at 1 MHz, resolve that before going any further; see Interpreting the results.
- Step the AWG through 4 MHz, 8 MHz, 12 MHz and 16 MHz without changing the amplitude, recording the same readings at each frequency.
- Stop the capture and turn the AWG output off before disconnecting anything.
Captures of a few seconds are sufficient. The files themselves can be discarded — the readings are the result — though keeping the 1 MHz capture is useful if the data ever needs to be examined in a decoder or spectrum tool.
Optional: alias check¶
For a stricter test of the filter, set the AWG to 24 MHz at the same amplitude. This is above the 20 MHz Nyquist limit, so anything that reaches the ADC will be reflected back into the capture as a 16 MHz alias. The amplitude readout should stay near zero. Any significant reading means unfiltered energy is aliasing into the sample and the filter stage needs investigating.
Interpreting the results¶
| Observation | Likely cause |
|---|---|
| All readings roughly double the expected value, clipped counts rising | AWG set for a High-Z load rather than 50 Ohms |
| All readings low or high by the same proportion, at every frequency | Wrong gain assumed — check the DIP switch against the table; check the AWG amplitude with an oscilloscope |
| 1 MHz correct, high frequencies far too low | Filter component value or solder-joint fault in the LPF stage; or the AWG's own output rolling off |
| 16 MHz reading close to the reference | Filter not attenuating — check the LPF inductors and capacitors. Do not use the board for captures until this is resolved |
RecentMin and RecentMax not symmetrical about 512 |
AWG DC offset not set to 0 mV; if the generator is correct, a DC offset fault — check the input stage divider (R402/R403) and the ADC reference divider (R301/R302) |
| Clipped counts rising with the reading near 0.75 | Excessive noise or distortion riding on the signal; check grounding, cable length and shielding |
| Reading unstable or drifting at all frequencies | Poor connection, unterminated cable, or interference reaching the input |
A board that reads within tolerance at all five frequencies has a verified analogue front end, a verified anti-alias filter and a verified digital path from the ADC to the capture file, and any remaining capture problem lies with the player, the RF cabling or the host.
Recording the results¶
| Test | Frequency | AWG amplitude | Amplitude reading | RecentMin | RecentMax | Clipped | Pass |
|---|---|---|---|---|---|---|---|
| 1 | 1 MHz | ||||||
| 2 | 4 MHz | ||||||
| 3 | 8 MHz | ||||||
| 4 | 12 MHz | ||||||
| 5 | 16 MHz |
Record the gain configuration, the DIP switch position and the AWG model alongside the readings — the expected values are meaningless without the gain setting, and the AWG's own flatness is part of the measurement.