Become a design partner

AnalogLab by Coherence Engine

AnalogLab: follow the signal to the qubit.

Models the analogue control path, room temperature to the qubit and back.

Skip the story
1

Every part has its own place in the fridge, at its own temperature.

Example readout chain: the six plates of our demo design. Other designs use other plates.

2

The chip being read. Its faint signal starts here, at the resonator.

3

The signal climbs the chain, and each part does one job.

ISO1
Lets the signal out, stops noise getting back in.
CBL1
Carries the signal up to the 4 K amplifier.
HEMT1
Makes the faint signal louder.
ADC1
Reads the result, at room temperature.
4

Noise picked up on the way makes the readout mistake a 0 for a 1.

5

Most of that noise has one source. The amplifier HEMT1 adds 89.6 % of it, the cable CBL1 7.3 %. But HEMT1’s noise reaches the input magnified by the loss of CBL1 in front of it, so the cable is the fastest fix.

Shares: HEMT1 89.6 %, CBL1 7.3 %, RTAMP1 3.0 %, ISO1 0.1 %.

6

The same chain with a NbTi cable and the LNF-LNC4_8F amplifier: system noise falls from 58.9 K to 2.46 K, and the chain passes.

Fig. 1An example readout chain, plate by plate: our demo design, an invented design. Noise shares and totals from our demo design.

2

Where the noise comes from.

Probe anywhere along the chain. HoverTap a part’s bar and the part lights wherever this page names it.

Fig. 2Each part’s share of the system noise of the demo readout chain, and the system noise of the whole chain in four versions; the one that passes is in teal. HEMT1’s noise reaches the input magnified by the loss of CBL1 in front of it, which is why the NbTi cable lowers the total far more than a better amplifier. From a worked example on our demo design. Illustrative, from an invented design.
Data for Fig. 2
PartShare of system noise
HEMT189.6 %
CBL17.3 %
RTAMP13.0 %
ISO10.1 %
VersionSystem noise
As drawn58.9 K
Better amplifier44.7 K
NbTi cable3.28 K
Both, passes2.46 K
Fig. 3The chain’s response across frequency, as a shape. Past the bend it is set by the low-pass filter, which lights when the cursor passes the bend. Illustrative, from an invented design.
Fig. 4A drive pulse before the chain (thin line, envelope dashed) and as it arrives at the qubit (teal), normalised to compare shape: stretched, with ringing in its tail. Illustrative, from an invented design.

3

AnalogLab, in four views.

Compare parts, build the chain, probe the signal, budget the noise.

AnalogLab parts browser searched for Low Noise Factory and filtered to HEMT amplifiers: four LNF HEMT amplifiers listed, with gain, measured-at and data-covers facets on the left and a preview of the LNF-LNC2_6B, its |S21| curve and its facts, on the right.AnalogLab parts browser searched for Low Noise Factory and filtered to HEMT amplifiers: four LNF HEMT amplifiers listed, with gain, measured-at and data-covers facets on the left and a preview of the LNF-LNC2_6B, its |S21| curve and its facts, on the right.
AnalogLab parts browser filtered to HEMT amplifiers, with two Low Noise Factory parts, LNF-LNC2_6B and LNF-LNC4_8F, ticked and compared side by side: their transmission curves and a table of their facts.AnalogLab parts browser filtered to HEMT amplifiers, with two Low Noise Factory parts, LNF-LNC2_6B and LNF-LNC4_8F, ticked and compared side by side: their transmission curves and a table of their facts.

Swipe the screenshot sideways to see all of it.

Fig. 5.1aFind a maker's parts and narrow them down. A search for Low Noise Factory, filtered to HEMT amplifiers, lists the four LNF HEMT amplifiers in the library by maker, with facets for gain, where each part was measured and the band its data covers. The preview shows the LNF-LNC2_6B: its |S21| curve, its facts and the datasheet each one comes from. Screenshot of AnalogLab v0.4.37; every number on it is the app's own reading.Fig. 5.1bTick two parts and they sit side by side. Filtered to HEMT amplifiers, two Low Noise Factory parts are ticked, LNF-LNC2_6B and LNF-LNC4_8F: their |S21| curves, each over its own band at its own temperature, and a table of their facts, with the facts that differ marked ≠. Screenshot of AnalogLab v0.4.37; every number on it is the app's own reading.
AnalogLab analysis pop-out window with four views in a grid, on a four-qubit multiplexed readout line: the chain’s |S21| from 3 to 9.5 GHz, with the filter band edges, the amplifiers’ gain, a step where the TWPA's gain drops, and four resonator notches, each labelled with its frequency, depth and loaded Q; a noiseless pulse view with its S21, phase, group delay and distortion at ADC1; the noise budget across frequency, 2.95 K system noise, with a bump between about 6.3 and 6.9 GHz; and the noiseless ADC1 record with its figures.AnalogLab analysis pop-out window with four views in a grid, on a four-qubit multiplexed readout line: the chain’s |S21| from 3 to 9.5 GHz, with the filter band edges, the amplifiers’ gain, a step where the TWPA's gain drops, and four resonator notches, each labelled with its frequency, depth and loaded Q; a noiseless pulse view with its S21, phase, group delay and distortion at ADC1; the noise budget across frequency, 2.95 K system noise, with a bump between about 6.3 and 6.9 GHz; and the noiseless ADC1 record with its figures.
Fig. 6Four analyses of a four-qubit multiplexed readout line at once, in AnalogLab’s pop-out window. The chain’s |S21| from 3 to 9.5 GHz (−46.7 to 26.5 dB) shows the filter band edges, the gain of the TWPA and HEMT, a step where the TWPA’s gain drops, and four resonator notches the app finds and labels: 5.41960, 5.77960, 7.15960 and 7.46960 GHz, each 37.6 dB deep, with a loaded Q of 4.0k or 3.9k. Beside it are a noiseless pulse view (S21 at the carrier 25.83 dB, group delay 9.708 ns), the noise budget across frequency (2.95 K system noise, with a bump between about 6.3 and 6.9 GHz) and the noiseless record ADC1 acquires (480 samples, peak 19.4 a.u.). Illustrative, from an invented design. Screenshot of AnalogLab v0.4.37; every number on it is the app’s own reading. Select it to open it full size.

4

The Components Library.
Real manufacturer data.

It models real parts from manufacturer datasheets and published measurements, so the chain you simulate is made of parts you can buy.

Our growing list includes

  • Low Noise Factory (LNF)
  • Zurich Instruments
  • IntelliConnect
  • Kelvin Quantum
  • QuantCore

5

Four disciplines in one shared space.

Physicists, RF and microwave engineers, cryogenic engineers and control engineers each own part of a control chain, and each usually works in their own tools and files. AnalogLab gives them one centralised space: one design, in which each of them sees the parts, the plates and the readings the others are working on.

Table 1One design, four disciplines: what each brings to the chain, and where in AnalogLab they read the result. The figures named are on this page.
SymbolDisciplineBrings to the designReads in AnalogLab
PhysicistsWhat the qubit needs from its drive and its readout.The pulse as it reaches the qubit (Fig. 4), and the readout chain’s system noise (Fig. 2).
RF and microwave engineersThe parts and how they are wired: gain, loss, filtering, isolation.Each part’s |S21| curve, side by side with another’s (Fig. 5.1), and the chain’s response across frequency (Fig. 3).
Cryogenic engineersWhich part sits on which plate, at what temperature.Every part on its plate (Fig. 5.2), and each part’s share of the noise (Fig. 5.4).
Control engineersThe pulses the room-temperature electronics send down the line.The pulse as it travels, probed anywhere along the chain (Fig. 5.3).

6

Questions about AnalogLab, the chain and the fridge.

  1. 1What does Coherence Engine do?

    We make system-level design and validation tools for quantum control. Hardware teams use them to design, simulate and check their control systems before they build them, so there are fewer hardware iterations and bring-up is faster.

  2. 2What does AnalogLab simulate?

    The analogue control path of a superconducting quantum computer: from the room-temperature electronics, down the cryogenic chain to the qubit, and back. It models noise, distortion and the pulse envelope, and what the chain does to the qubit’s readout: its signal-to-noise ratio and the Stark shift at the qubit.

  3. 3Is it tied to one hardware vendor?

    No. It is vendor-neutral. You can model parts from different suppliers side by side and compare them before you buy.

  4. 4Who is it for?

    Everyone who designs or buys the control path: physicists, RF and microwave engineers, cryogenic engineers and control engineers, and the people who sign off the hardware. How they work together in it

  5. 5Why simulate before building?

    Every try on real hardware means warming the fridge, changing a part and cooling it again: days each time. Simulating first lets you compare the choices and spend fridge time on the ones that matter.

  6. 6What parts are in the library?

    55 real parts with datasheet or measured data, including 23 CryoCoax cables, across 53 component types: sources, lumped, attenuators, cables, filters, splitters & switches, isolators & circulators, instruments & loads, amplifiers, mixers, qubits. Our growing list of manufacturers includes Low Noise Factory (LNF), Zurich Instruments, IntelliConnect, Kelvin Quantum and QuantCore. See the Components Library

  7. 7How do I become a design partner?

    Tell us about your system through the contact form. We work with a small number of teams at a time. Go to the form

Know a group building quantum hardware?

Introduce us, and they can try their next wiring change on screen before it goes in the fridge. They will thank you!