Designing a microbolometer ROIC means balancing sensitivity, dynamic range, pixel size and power consumption. In uncooled infrared imagers, the detector’s resistance changes with temperature, but extracting that small signal can require sensitive analog circuitry and long interconnects. A time-mode readout architecture offers another approach: convert the detector response into a time interval, then digitize that interval with a time-to-digital converter (TDC).
The architecture described here combines event-based readout, two-step integration and a column-parallel TDC. Its purpose is to reduce dependence on power-consuming analog signal paths while addressing noise, readout dynamic range and detector self-heating.
Why rethink the microbolometer readout chain?
An uncooled microbolometer absorbs infrared energy, warms slightly and changes electrical resistance. The readout integrated circuit (ROIC) biases the detector and measures the resulting electrical change. For a focal plane array, this process must work across many small pixels while meeting system requirements for sensitivity, scene range and power.
Conventional readout chains commonly amplify a detector signal, manage offset and noise, and sample the resulting voltage before digitization. When sensitive analog signals travel along lengthy metal connections, buffers may be needed to drive the loads. Noise-cancellation circuitry can add further area and power. These demands make it difficult to improve noise performance without increasing the readout’s complexity.
Self-heating adds another complication. The bias used to measure a microbolometer also dissipates energy in it, changing its temperature and resistance. That shift can consume measurement range and complicate signal processing. Readout design therefore needs to consider both the detector’s thermal behavior and the electrical path used to measure it.
Microbolometer ROIC time-mode processing
In time-mode processing, the detector response is represented by a time interval rather than being carried through an analog voltage chain all the way to conversion. A TDC then translates the interval into a digital value. This approach moves more of the signal path into digital circuitry and avoids an intermediate voltage conversion for the time-based signal.
The proposed system uses a column-parallel architecture for frame-synchronous imaging. Each column’s readout circuitry serves its pixels, while digitization takes place outside the individual pixel. This can preserve more pixel area than placing a complete digitizer in every pixel, an important consideration for arrays with small pixel pitch.
At the detector interface, a bridge network includes active and reference microbolometers. The readout uses a capacitive transimpedance amplifier (CTIA) with a modified reset sequence to create the time-based measurement. Rather than relying only on a conventional integration ramp, the circuit establishes a forward ramp at the input, followed by a backward-integrated ramp at the output. The resulting interval is measured digitally.
How event-based readout and two-step integration work
In an event-based scheme, circuit transitions mark the beginning and end of a measurement interval. The detector-dependent response affects the interval between these events, allowing the readout to encode the signal in time. A synchronous counter-based TDC measures that interval and produces a digital representation.
The two-step integration sequence extends the measurement interval compared with a single conventional integration phase. A longer interval can improve the signal-to-noise ratio (SNR), while the time-domain representation allows the design to reduce reliance on a high-performance analog chain driving long interconnects.
The TDC uses a coarse-to-fine conversion structure:
- Coarse measurement: A synchronous counter records the larger portion of the interval.
- Fine measurement: An inverter-based delay line resolves the remaining fraction of a clock period.
In the reported implementation, the counter is 12 bits wide and the delay line contributes four least-significant bits. The TDC is located in the column and serves a single imager column. This arrangement combines counter-based measurement with finer timing resolution from the delay line.
Managing self-heating and readout trade-offs
The modified CTIA reset scheme also provides a means of compensating for self-heating. Since active and reference detectors can experience bias-induced temperature changes, the timing and sequence of bias and reset operations matter. The architecture uses the two-step measurement behavior to account for these effects, helping limit the influence of self-heating on the readout result.
This is significant for dynamic range. If self-heating shifts the detector’s operating point, the readout may need to accommodate a larger change than the scene signal alone would require. Addressing that effect within the measurement sequence can reduce its impact and make the readout less dependent on circuit imperfections. The approach does, however, require additional array-control complexity.
Time-mode processing is not a universal replacement for design trade-offs. The ROIC still has to balance noise, integration time, frame timing, pixel area and power. Longer measurement intervals can support improved noise performance, but the available interval is constrained by system timing. The architecture’s value lies in changing how those requirements are met: using timing and digital conversion rather than relying entirely on analog amplification and voltage-domain sampling.
Reported implementation results
The prototype was implemented in 130-nm bulk CMOS. The reported readout achieved a per-channel power consumption of 66 µW and readout NETD below 10 mK. These figures describe the reported implementation and should be understood in the context of its circuit, detector and measurement conditions; they are not a guarantee for other designs.
The results demonstrate the potential of combining a highly digital signal path with a specialized detector-readout sequence. For engineers evaluating infrared focal plane arrays, the central design lesson is architectural: noise and power are affected not only by individual circuit blocks, but also by where the signal is represented and how it is converted.
Practical implications for infrared system designers
For system integrators and ROIC designers, an event-based time-mode approach is worth considering when a design must reduce analog overhead while retaining sensitivity. Its main elements are:
- A bridge-type detector network with active and reference microbolometers.
- A CTIA front end using a modified reset and two-step integration sequence.
- Column-level time-to-digital conversion combining coarse counter measurement and fine delay-line resolution.
- Array timing and control designed to account for detector self-heating.
These elements need to be evaluated together. TDC placement, timing control, detector bias behavior and pixel-pitch constraints all influence the final architecture. A highly digital readout may simplify parts of the analog signal path, but it shifts design attention toward timing, control and the behavior of the complete detector-readout system.
Conclusion
A time-mode microbolometer ROIC converts detector response into a measurable interval, using event-based readout and a two-step TDC to produce digital output. The two-step integration sequence extends measurement time and supports SNR improvement, while the modified reset scheme offers a way to address self-heating. The reported prototype demonstrates low per-channel power and readout NETD below 10 mK. For industrial infrared imaging applications, the practical takeaway is to treat readout architecture as a system-level lever for balancing sensitivity, power, dynamic range and pixel-area limits.