Skip to content
David Allcock edited this page Jul 6, 2021 · 15 revisions

Sampler

Sampler is an 8-channel, 16-bit ADC EEM with an update rate of up to 1.5MSPS (all channels simultaneously). It has low-noise differential front end with a digitally programmable gain, providing full-scale input ranges between +-10mV (G=1000) and +-10V (G=1).

Design files (schematics, PCB layouts, BOMs) can be found at Sampler/releases.

Overview

  • Width: 8HP
  • Channel count: 8
  • Resolution: 16-bit
  • Sample rate: up to 1.5 MHz
  • Sustained aggregate data rate in single-EEM mode (8 channel readout): ~700 kHz
  • Sustained per-channel data rate in dual-EEM mode (SU-Servo): ~1 MHz
  • Note that the bandwidth specifications on this page are for the hardware only; ARTIQ kernel and RTIO overhead often make the effective sample rate lower.
  • Bandwidth: 200kHz -6dB bandwidth for G={1, 10, 100}, 90kHz for G=1000
  • Input ranges: +-10V (G=1), +-1V (G=10), +-100mV (G=100), +-10mV (G=1000)
  • DC input impedance:
    • Termination off: 100k from input signal and ground connections to PCB ground
    • Termination on: signal 50Ohm terminated to PCB ground, input ground shorted to PCB ground
  • ADC: LTC2320-16
  • PGIA: AD8253
  • EEM connectors: power and digital communication supplied by one or two EEM connectors.
  • Power consumption: ?

Sampler v2.0 top

Modes of operation

Sampler can be operated either in basic mode as a standard SPI EEM using a single EEM connector (EEM0), or in fast mode via a source-synchronous LVDS interface and 2 EEM connectors (EEM0-EEM1).

In basic mode, the channels must be read out sequentially, in decreasing order. The maximum SPI clock is limited by the round-trip delay in the cabling used to connect Sampler to the master. With a 20 MHz SPI read clock (25ns round trip plus setup), we get one sample one each channel per 6.9us (30ns tCONVH + 450ns tCONV + 8 channels * 16 bits * 50ns).

Fast mode allows all ADC channels to be read out simultaneously at 1.5 MHz via a source-synchronous interface. This is implemented (~ 1 MHz on each channel for all channels) in the Kasli-Sampler-Urukul "SU" laser intensity servo.

EEM connector usage:

Pin EEM0 EEM1 (fast mode only)
0 ADC-SCK ADC-CLKOUT
1 ADC-SDOA ADC-SDOA
2 ADC-SDR ADC-SDOB
3 ADC-CNV ADC-SDOC
4 PGIA-SCK ADC-SDOD
5 PGIA-MOSI NC
6 PGIA-MISO NC
7 PGIA-CS NC

"PGIA" refers to the front-end programmable-gain instrumentation amplifiers, which are controlled over SPI via a shift register. Note that on hardware v2.2 the LVDS line driver for PGIA-MISO (IC19) is not populated by default.

Tests and characterisation

To do: check power handling of termination resistors!

Tests performed using Sampler Rev 1.1 prototype hardware. See issue 226 for further details. Further tests will be done on the Rev 2 hardware.

Unless stated otherwise, all measurements use:

  • 200kHz ADC clock
  • Input termination off for signals connected directly to a voltage source, termination on for "floating" inputs
  • PGAI G=1

Bandwidth

  • Drive an input (termination off) from a signal generator.
  • Scan the signal frequency to measure the -6dB frequency on the ADC.
  • The sample rate was 125kSPS, so the signal frequency was above the Nyquist frequency for most gain settings.

Small-signal bandwidth

  • Signal was nominally -28dBFS=1Vpp/G (there was a slight error due to the ADC reference gain being incorrect in this hardware revision).
Gain Small-signal BW (kHz)
1 206
10 207
100 195
1000 87

Large-signal bandwidth

  • We made the following measurements, which confirm that the large-signal bandwidth is the same as the small-signal bandwidth to within the measurement accuracy for gains of 1 and 10.
    • G=10, 1Vpp=-8.1dBFS: 213kHz
    • G=10, 2Vpp=-1.2dBFS: 212kHz
    • G=1, 10Vpp=-8.1dBFS: 214kHz

Noise floor

  • Sample rate for this measurement was 200kHz, measurement bin width was 3Hz (65k samples)
  • We measured the noise spectrum both for a terminated input (0V input) and (using batteries as a low-noise voltage source) for signals close to full-scale. No significant difference observed between the two cases. Data below are all for terminated (0V) inputs.
  • No spurs seen above the noise floor at any frequency (e.g. no SMPS spurs seen)
  • Noise spectrum is white once DC offset was subtracted
  • Measure noise floor as a function of gain
gain LSB RMS µV RMS nV/rtHz
1 1.3 420 TBD
10 1.4 44 TBD
100 3.6 11 TBD
1000 14.3 4.4 TBD

Channel-channel cross-talk

DC cross-talk measurement:

  • Input A: apply +-17V DC, termination off. NB this voltage is clipped by the protection diodes, but was chosen to really stress-test the design.
  • Input B: termination on measure signal level
  • With input B G=1, voltage applied to input A makes <1LSB change in level measured at input B. DC cross-talk is thus <96dB.
  • To do: re-measure DC cross-talk at G=1k

AC cross-talk measurement:

  • Gain was the same on all channels.
  • Apply 50kHz, 10Vpp (-8.8dBFS) signal to one channel. NB this saturates the PGIA for G>1!
  • Measure signal on the adjacent channel as a function of PGIA gain
  • Termination is on for the driven channel, but off for all other channels
  • Note that the lack of dependence of the cross-talk on PGIA gain suggests that the cross-talk is dominated by pickup before the PGIA -- most-likely in the BNC connector.
Gain Measured signal (dBFS) Cross-talk (dBc)
1 -87 -79.2
10 -67 -78.2
100 -46 -77.2
1000 -29 -81.2
  • With the termination on for all channels, we find:
Gain Measured signal (dBFS) Cross-talk (dBc)
1 <-130 <-121
10 -111.5 -123
100 -93.7 -124.9
1000 -72.2 -123.4

Common-mode rejection ratio (CMMR)

  • Short input ground and signal together for one channel
  • Apply sinewave to input signal/ground
  • Measure the signal level seen by the ADC

G=1, signal ~2Vpp (-22dBFS)

Freq [kHz] measured signal level [dBFS] CMRR (dB)
0.01 <-120 >-98
0.1 < -120 >-98
1 <-120 >-98
10 -109 - 87
100 -87 -55
1000 -105 -83
10000 -108 -85

G=100, signal ~2Vpp (+18dBFS)

Freq [kHz] measured signal level [dBFS] CMRR (dB)
0.01 <-110 >-118
0.1 -98 -116
1 -88 - 106
10 -70 - 88
100 -50 - 68
1000 -80 - 98
10000 <-100 >-118

Harmonics

  • 25kHz signal applied to input
  • measure the level of the second harmonic (50kHz) and third harmonic (75kHz) as a function of input signal level.
Input level Second harmonic Third harmonic
0.1Vpp (-48dBFs) -51dBc
1Vpp (-28dBFs) -69dBc
10Vpp (-8dBFs) -58.8dBc

MCX connector option

If MCX connectors are preferred over BNC then the MCX_BNC_adapter adaptor can be used. This can be installed during manufacture or the conversion can be done by the user. The MCX connectors allow use of a 4HP panel instead of the 8HP required for BNCs.