Frequently asked questions

When using two or more TPX3CAMs in a measurement set-up, there is a time drift between the separate TPX3CAM internal clocks. This can be solved by taking the timing signals from the LEADER detector to a FOLLOWER detector by means of daisy chaining with an HDMI cable. If you want more information please contact us for a step-by-step guide and python examples. 

Our devices require a +12 V DC supply to the readout board (max. 5 A). 


  • All data (512×512 pixel frames) can be directly saved to the hard disk of the acquisition computer. No on readout board buffering is applied.
  • For 12 bit CRW (2000 fps): The data is stored in a 16 bit frame format so 512*512*2 bytes equals 542288 bytes for the pixel data of one frame.  At 2000 fps, data will be saved at  1000 Mb/s on our fast SSDs, or 60 Gb/minute.


  • All pixel hit data (8 bytes per hit (ToT (10 bit), ToA (14+4 bit), pixel coordinate information) is saved to the hard disk of the acquisition computer in a binary data file. Or software can provide real time ‘frames” of counts or sum ToTToA/ToF per pixel.
  • For 120 Mhit/s the data stream written to the disk is 960 Mb/s or 57.6 Gb/minute. The total data stream consists of the raw pixel hits and the TDC time stamps.

In general, our products have a QE higher than 90%.

For TPX3CAM, if combined with an Intensifier, the QE depends on the wavelength you want to measure. Please request the QE and dark count plots if you would like to know more. 

For Cheetah, electron detection, the efficiency is higher than 95% above 30 keV down to almost 5 keV for EBSD/SEM/LEEM/PEEM low energy sensors. 

Yes, that is possible. The best absorption efficiency is with a CdTe sensor of 1 or 2 mm, with 17 and 32% efficiency at 200 keV, respectively.

For MeV energies, the efficiency will be ~1-2%.  

It is not easy to damage the sensor. The sensor can get physically damaged due to laser ablation or heating by focusing a high intensity laser on the sensor. 

The threshold for this is not known. 

Yes, that is possible with the timepix3 chip. However, we cannot determine the exact time. 

For example, we cannot deduce that a 50 ps pulse is actually 50 ps long, but we will detect it. 

If someone flashes light for some ps, you will see it with your eyes even though our eyes have ms time-resolution at best. A 50 ps pulse, if the detector is very well calibrated, will appear 1.56 ns long. In a less optimal calibration it will look 3-6 ns long. When no calibration takes place, the pulse might even seem 100 ns long. 

The medipix3RX chip has several modes. All frame rates are written below.

From a hardware point of view, the upper limit on the frame rate is set either by clock frequency supplied to the chips or the available data transfer bandwidth (1 Gb/s or 10 Gb/s). 

The detector’s pixels are robust and radiation hard. There hasn’t been any pixel damage of MPX3 and TPX3 detectors even for high energy and high flux electrons (e.g, 200kev, 300kev TEM, STEM, X-ray). The pixelated detector can be used for many years. The pixels unlikely become damaged in such short term if there was not exposed to any hazardous conditions, i.e. high voltage/discharge in vacuum near the detector.
We advise our customers to be very careful for any HV discharge near the detector in vacuum. That may damage pixels, even the chips!

/shutdown is the polite way. Hitting Ctrl-C or using the kill command is the rude one. 

In conventional imagers the signal is integrated in a slice of time called a frame. What you get out is an array of pixel values in an image. Data is read out frame by frame. This makes it impossible to achieve nanosecond resolution as you would need extreme amounts of data. The timepix3 chip has data-driven readout. This means that every time an “event” (or hit) reaches the sensor, the data are sent (read-out) immediately. If no events happen then there is no data transferred at all. 

With an ICCD you have frame based acquisition ad the time resolution is achieved by synchronizing a gate open time. This way one reads out events that occur only within that gate.

For timepix3, to group events that correspond to a laser pulse, we input the laser pulses to the TDC channel of the camera, which will provide time-stamps of the pulses in the data stream as well. Therefore, T0 can be obtained.

Once the laser pulse or external triggering signal in the TDC is recorded, you can choose any “gate times” you like.  You can record events continuously with time bin 1.56ns (the electronic shutter is open all the time acquiring data). Essentially, you can define any time-window to generate images with grouped data. You don’t just have a fixed frame you have to pre-define. This is explained better with this video that shows how the output of the raw data looks like. 


CRW (continuous) operates in 12, 6 and 1 bit. Not in 24. 

One trick is to sum multiple short acquisitions and provide the summed image. You can do this with our software Serval. This avoids possible counter saturation and at the same time reduces the amount of data that need to be saved. The image below shows how the reading and counting alternates. 


In CRW (Continuous Read/Write) mode there is no time gap. In other words, there is no dead time.

All modes besides SRW (Sequential Read/Write) 24 bit or 2×12 bit can operate with dead time (time between frames/time gap) less than 1 ms. That is 0.5 ms for 12 bit SRW. 

You don’t need to equalize the detector often. Generally, consider equalizing when you see that the chip does not have a uniform response/image or when there are noisy pixels appearing (e.g. due to environmental temperatures changes). Instead of equalizing you may manually mask the individual pixel, too. 

Contact us

How can we help?

Science Park 106, 1098XG, Amsterdam, The Netherlands
Tel.: +31 207235200