REVIEW 2 major objections 5 minor 10 references
A 50 ps resolution monolithic active pixel sensor without internal gain in SiGe BiCMOS technology
T0 review · 2 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash
Pith's one-line read A monolithic pixelated silicon sensor in SiGe BiCMOS technology achieves about 50 ps time resolution for minimum-ionizing particles without internal gain, matching avalanche-gain detectors.
desk verdict A genuine sub-100 ps result from a SiGe monolithic pixel sensor without gain, but the headline '50 ps' is the Gaussian-core width after cutting the 7-10% non-Gaussian tail, and the paper needs systematics and a tail study before I'd take the sensor-level number at face value. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The load-bearing element is the reduction of pixel capacitance. Hexagonal pixels with 65 µm and 130 µm sides are laid out with the readout electronics inside triple wells, so the capacitance seen by the preamplifier drops to roughly 70 fF and 220 fF respectively, versus the 750 fF of the group's earlier pixels. That low capacitance, together with a silicon-germanium heterojunction bipolar transistor (HBT) preamplifier and a discriminator, gives an equivalent noise charge of about 90 electrons and a total time walk below 1 ns. Time walk is corrected offline using the time-over-threshold signal, and the timing is evaluated against a reference low-gain avalanche detector with an independently verified 50 ps resolution.
What would settle it
A beam test with an external tracking telescope that selects hits by impact position would settle it: if events far from pixel edges still show the same late tail, the tail is intrinsic and the true per-event timing precision is worse than the fitted core for a substantial fraction of hits; if the tail appears only for hits near the inter-pixel boundary, the attribution is confirmed.
Extended reading notes
Core claim
The central claim is that a monolithic active pixel sensor built in a 130 nm SiGe BiCMOS process, without any internal gain mechanism, achieves a time resolution of the order of 50 ps for minimum-ionizing particles. After correcting for time walk, the Gaussian core of the measured time-of-flight distribution has a width of (46±1) ps for the 65 µm-side pixels and (55±2) ps for the 130 µm-side pixels at one operating point, with the small pixel keeping a resolution between 46 and 62 ps across thresholds. The authors attribute the route to this performance to the low pixel capacitance (about 70 fF for the small pixels), which lowers the equivalent noise charge of the SiGe HBT front-end to 90 electrons. They conclude that SiGe HBT technology can deliver both tracking and excellent timing without the avalanche gain needed by competing detectors.
Load-bearing premise
The quoted 50 ps resolution is obtained by fitting only the Gaussian core of the time-of-flight distribution and excluding all events with arrival times more than 100 ps late; the paper assumes that excluded tail (7–10% of events) comes from electrons crossing between pixels and is not an intrinsic property of the pixel's timing response.
Editorial extensions
If this is right
- Monolithic pixel sensors in standard BiCMOS processes can provide roughly 50 ps timing without adding avalanche gain layers, so tracking and timing can be combined in one thin sensor.
- The strong dependence of the resolution on pixel capacitance implies that further shrinking pixels or improving the front-end noise should push the timing below 50 ps.
- Because resolution improves with higher threshold and higher bias voltage, the current design is limited by time-walk correction and drift-field uniformity rather than by the sensor's intrinsic speed.
- The 7–10% non-Gaussian tail, attributed to charge sharing at inter-pixel boundaries, would need to be suppressed for applications that require per-event timing on every hit.
- This opens a route to time-of-flight positron emission tomography and 4D tracking in high-energy physics using one monolithic technology.
Reading between the lines
- An extension not tested here: the unexplained tail fraction should shrink or disappear when the inter-pixel boundary is placed under a high-field region; if the tail persists for hits far from pixel edges, the fitted 50 ps core resolution would not represent the full per-event timing precision.
- The authors' own data show time resolution improving with threshold, which suggests that a lower-jitter discriminator or a more precise time-walk correction could shift the operating point to even higher thresholds without losing efficiency.
- If the capacitance scaling observed here extends to even smaller pixels, reaching a few tens of picoseconds in the same process may be possible, at the cost of more readout channels and smaller charge signals.
Signed reviews
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper reports a proof-of-concept monolithic pixel detector fabricated in IHP SG13G2 130 nm SiGe BiCMOS technology, with hexagonal pixels of 65 and 130 µm side and the amplifier electronics integrated inside the pixel triple wells. The front-end uses a SiGe HBT preamplifier and a CMOS discriminator, and the pixel capacitance is reduced to about 70 fF (small) and 220 fF (large) to improve noise and timing. Lab measurements with a 90Sr source and a reference LGAD detector are used to measure the time of flight (TOF); after a time-walk correction based on time-over-threshold, the paper quotes Gaussian-core time resolutions of (46±1) ps for the small pixel and (55±2) ps for the large pixel at specific thresholds, with values between (62±2) ps and (46±1) ps over the threshold scan. The paper also reports an ENC of 90 and 160 electrons for the small and large pixels, respectively, estimated from measured gain and simulated amplifier noise. The abstract and title state a '50 ps resolution' without internal gain, claiming competitiveness with silicon technologies that use avalanche gain.
Significance. If the result holds as stated, it is significant for timing detectors in particle physics and for TOF-PET, since it would demonstrate that a standard silicon process without internal gain can reach the 50 ps level. The measurement uses an independent LGAD reference whose resolution was verified in a two-detector TOF setup, and the paper is transparent about the presence of a non-Gaussian tail at the level of 7-10%. The main value is as a proof-of-concept; the sensor design and the measured timing performance are useful for the community. However, the central claim in its current unqualified form overstates the per-event timing precision of the full detector, because the quoted numbers come from a Gaussian fit to the core of the TOF distribution only.
major comments (2)
- [Section 3.2, Figure 7, abstract] The headline '50 ps resolution' is obtained from Gaussian fits restricted to TOF < +100 ps, excluding the non-Gaussian tail that amounts to 7.4% of events for the small pixel and 9.8% for the large pixel. The text in Section 3.2 says 'No event selection was applied to the data of the prototype chip under test,' but the fit range is an event selection in the analysis. The conclusions honestly state that the quoted resolutions refer to the ~90% Gaussian core, but the abstract and title do not carry this qualification. As a detector-level timing precision, the quoted value is therefore not established for hits near the inter-pixel boundary; if the tail is intrinsic to the pixel response, the per-event resolution for a substantial fraction of real hits is worse. Please either (a) quote a combined or tail-inclusive resolution (e.g., a quantile-based resolution or a two-component fit), or (b) revise the title, abstract, and conclusions to refer explicitly to the Gaussian-core resolution and list the tail fraction as a known limitation. The sensitivity of the result to the +100 ps cut should also be reported.
- [Section 3.1] The abstract states an equivalent noise charge of 90 and 160 electrons for the small and large pixels, but the amplifier RMS noise used in the ENC calculation (sigma_V = 4.0 mV and 4.7 mV) is taken from Cadence Spectre simulations, not from the measured noise hit rates (which give 2.67 mV and 2.99 mV). The authors explain that discriminator hysteresis acts as a filter, yet they do not provide a measurement of the true amplifier noise or a systematic uncertainty for the simulated value. Because the ENC is a headline performance parameter and the paper notes a 30% discrepancy with the simulation, the ENC claim needs either a direct measurement (e.g., with the discriminator hysteresis characterized) or an explicit systematic uncertainty and a discussion of how the simulated noise affects the time-resolution interpretation.
minor comments (5)
- [Section 3.2] The equation 'TOF = thexa − tLGAD' appears to contain a typo; 'thexa' should likely be 't_hexa' (or another defined symbol for the time measured by the prototype chip).
- [Section 3.2] The sentence 'No event selection was performed' is misleading because the Gaussian fit excludes TOF > +100 ps; please rephrase to specify that the data acquisition was unselected but the fit range was restricted.
- [References] Reference [2] lists 'JINST 13 (2017) P02015' but the DOI points to JINST 13 (2018) P04015; please correct the citation.
- [Figures 5 and 7] Some axis labels and text in the extracted figures contain garbled symbols (e.g., 'Threshold [e ]' and '−101'); please check the rendered math and unit superscripts in the final version.
- [Section 1.2] The word 'break-down' should be 'breakdown'.
Circularity Check
No circularity: the 50 ps timing claim is a direct measurement with an independently verified LGAD reference, not a derived prediction forced by the input data.
full rationale
The paper reports a laboratory measurement of time resolution using a 90Sr source and a reference LGAD detector. The quoted sigma values are obtained from Gaussian fits to the TOF distribution after time-walk correction, with the non-Gaussian tail excluded above +100 ps. This is a measurement choice, not a circular derivation: the time-walk correction is calibrated from the same data, but it removes a mean trend versus time-over-threshold and does not by construction set the residual width to 50 ps. The reference LGAD resolution of 50 ps is taken from the literature and independently verified by the authors in a two-LGAD TOF measurement, so the subtraction in quadrature is externally anchored. Self-citations to prior TT-PET designs provide context for the front-end electronics but are not used to prove the present timing result. The paper honestly states that the quoted resolutions refer to the Gaussian core of about 90% of events and that the tail origin requires test-beam investigation. There is no fitted parameter relabeled as a prediction, no uniqueness argument imported from self-citation, and no equation that reduces to its own input. The main caveat is the tail fraction, but that is a correctness/interpretation concern, not circularity.
Assumptions & free parameters
assumptions (4)
- domain assumption The reference LGAD detector used for the TOF measurement has a time resolution of 50 ps RMS in the custom setup.
- domain assumption The non-Gaussian tail in the TOF distribution is an artifact of electrons crossing between pixels, not an intrinsic response of the pixel.
- domain assumption Cadence Spectre simulation provides the true amplifier output noise (sigma_V = 4.0/4.7 mV) used to compute the ENC, despite the direct noise-rate measurement giving lower values.
- domain assumption The most probable charge deposited by a MIP in the 26 um depleted layer is approximately 1600 electrons, estimated from sensor geometry and resistivity.
Cite this review
Pith. "Pith review of A 50 ps resolution monolithic active pixel sensor without internal gain in SiGe BiCMOS technology." pith.science (2026). https://pith.science/paper/3VSJ5GIB
@misc{pith2026190809709,
author = {Pith},
title = {Pith review of: A 50 ps resolution monolithic active pixel sensor without internal gain in SiGe BiCMOS technology},
year = {2026},
howpublished = {\url{https://pith.science/paper/3VSJ5GIB}},
note = {Machine review of arXiv:1908.09709}
}
read the original abstract
A monolithic pixelated silicon detector designed for high time resolution has been produced in the SG13G2 130 nm SiGe BiCMOS technology of IHP Mikroelektronik. This proof-of-concept chip contains hexagonal pixels of 65 {\mu}m and 130 {\mu}m side. The SiGe front-end electronics implemented provides an equivalent noise charge of 90 and 160 electrons for a pixel capacitance of 70 and 220 fF, respectively, and a total time walk of less than 1 ns. Lab measurements with a 90Sr source show a time resolution of the order of 50 ps. This result is competitive with silicon technologies that integrate an avalanche gain mechanism.
Reference graph
Works this paper leans on
-
[1]
M. Benoit et al., 100 ps time resolution with thin silicon pixel detectors and a SiGe HBT amplifier, JINST 11 (2016) P03011, http://dx.doi.org/10.1088/1748-0221/11/03/P03011
-
[2]
L. Paolozzi et al., Test beam measurement of the first prototype of the fast silicon pixel monolithic detector for the TT-PET project, JINST 13 (2017) P02015, https://doi.org/10.1088/1748-0221/13/04/P04015
-
[3]
A monolithic ASIC demonstrator for the Thin Time-of-Flight PET scanner
P. Valerio et al., A monolithic ASIC demonstrator for the Thin Time-of-Flight PET scanner, arxiv:1811.10246, accepted by JINST, https://arxiv.org/abs/1811.10246
-
[4]
L. Paolozzi et al., Characterization of the demonstrator of the fast silicon monolithic ASIC for the TT-PET project., JINST 14 (2019) P02009
work page 2019
-
[5]
The TT-PET: Thin Time-of-Flight PET project, SNSF grant CRSII2-160808, http://p3.snf.ch/project-160808
-
[6]
Ruecker et al., Half-Terahertz SiGe BiCMOS technology, IEEE SiRF Symp
H. Ruecker et al., Half-Terahertz SiGe BiCMOS technology, IEEE SiRF Symp. Dig., pp. 129-132, DOI: 10.1109/SiRF.2012.6160164, 2012
-
[7]
Sola et al., First FBK production of 50 m oltra-fast silicon detectors , NIM A 924 (2019) 360-368
V. Sola et al., First FBK production of 50 m oltra-fast silicon detectors , NIM A 924 (2019) 360-368
work page 2019
-
[8]
L. Paolozzi, Silicon monolithic pixel detectors in a SiGe Bi-CMOS process for sub-100ps time resolution., 12th "Trento" Workshop on Advanced Silicon Radiation Detectors Oral Presentation (2017) Direct link: https://indico.cern.ch/event/587631/contributions/
work page 2017
Show all 10 references
-
[9]
Allaire et al., Beam test measurements of Low Gain Avalanche Detector single pads and arrays for the ATLAS High Granularity Timing Detector, JINST 13 (2018) P06017
C. Allaire et al., Beam test measurements of Low Gain Avalanche Detector single pads and arrays for the ATLAS High Granularity Timing Detector, JINST 13 (2018) P06017
2018
-
[10]
write newline
" write newline "" before.all 'output.state := FUNCTION blank.sep after.quote 'output.state := FUNCTION fin.entry output.state after.quoted.block = 'skip 'add.period if write newline FUNCTION new.block output.state before.all = 'skip output.state after.quote = after.quoted.blo...
Reviewed August 14, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.