Downloads and FAQ


Attention: In general, software packages must be completed by the user specific initialization files belonging to the appropriate delay line detector or TDC. The specific files are TDC_GPX3.INI and the TDC firmware file, which is named "*.BIT". These files have been provided as part of your detector or TDC package at delivery. You may either copy the initially delivered files into the updated software folder or alter an existing INI for the needs of your individual hardware. Be aware, that any INI alteration may come with the risk of wrong configurations and may lead to uncertainties in the behavior of your hardware. The software support and compatibility for TDCs before delivery dates of 2017 is only minimal. If your TDC contains no USB 3 port, the packages which can be downloaded here may not suit your hardware. Please contact us directly in order to get information about hardware upgrades.

Windows® is a registered trademark of Microsoft Corporation in the United States and other countries.
Labview™ is a trademark of National Instruments. Neither Surface Concept GmbH, nor any software programs or other goods or services offered by Surface Concept GmbH, are affiliated with, endorsed by, or sponsored by National Instruments.


  • DLDGUI Software v7.4 (PDF)
  • DLD3030
    Manual of a Delayline Detector DLD3030 adapted to a Scienta analyzer.
    Customer specific layout with integrated micro electrostatic lens system
    for parallel operation of the DLD together with a SPLEED.
  • DLD4444
    Manual of a Delayline Detector DLD4444 in a customer specific layout for
    a direct readout of MCP pulses in addition to the normal DLD operation.
  • DLD4040-4Q
    Manual of a 4 Quadrant Delayline Detector DLD4040-4Q with advanced
    readout system. The customer specific layout includes a translation and
    rotation stage with integrated position readback sensor.
  • SPLEED-DLD Combination
    Manual of a High Voltage SPLEED-DLD Combination adapted to a High
    Voltage Scienta analyzer.
  • DLD8080
    Manual of the standard DLD8080 for X-ray, ion, and electron detection
  • DLD120120
    Manual of the standard DLD120120 for X-ray, ion, and electron detection

Product Information and Application Notes


  • Delay Line Detectors - Principle of Operation (PDF)
  • Guidelines Surface Concept Delay Line Detectors (PPT)
  • Survey of Particle MCP Detectors with Delay Line Anode (PDF)
  • Development Roadmap (PDF)
  • General Terms and Conditions (AGBs) (PDF, English) (PDF, deutsch)

Frequently Asked Questions

What is the mechanism of coupling pulses into the new meander DLD through the insulating layer?

The detection is based on a pure capacitive coupling. The insulation layer is only some ten µm thick with an appropriate epsilon combined with a low conducing layer on the MCP bound side, where the electron clouds are accelerated towards. The meanders below are the counter electrode of that capacity. They locally interact with the image charge density gradient from the high voltage layer.

What is the length of the delay line, what is the delay? Does that mainly set the limit of the maximum count rate?

The lengths and the delays e.g. in a DLD3030 are about 1.2 m and 9 ns and scale with the active area of the detector. It is rather short for the DLD3030 and would enable 100 MHz if the events would appear equally distributed in time. Unfortunately, most experiments probe random particle generation processes, so that an average rate of 100 MHz causes a highly significant appearance of multiple events within a period of even < 3 ns. This is already below the typical FWHM of DLD output pulses. Following the meanders, the entire system is very fast. The amplifiers and CFDs can operate at 200 MHz pulse frequencies, the TDC can handle about 186 MHz in a short burst, while it reads in with 80 MHz. The main bottleneck is the algorithm that identifies pulse quadruples from the delay line ends that represent a single event. Currently, this sets a limit of a maximum rate for random hits of some MHz. Improved conditions are achievable for intense bursts with long delay times. For this case we developed an adapted system with multi-hit capability that can acquire 100 MHz equivalent rate in stroboscopic short bursts in the µs range. Also a permanent random count rate of 8 million counts per second (depending on the particular detector layout) is achieved. More...

Why does the standard detector system not offer higher count rates, is the limitation caused by the used TDC or by the USB 2.0?

Currently, there is no faster TDC on the market than the device we use for our packages. This device can permanently read at 80 MHz which lets us process 20 MCPS average event rate in a single DLD system. It is always necessary to measure 4 pulses accurate in their relative arrival time at the four ends of the two meanders for each event. The main limit is the efficiency of the quadruple finder algorithm. The processing time of this algorithm dramatically increases when multiple hits appear on the detector, due to the mixed arrival of pulse quadruple sets at the read-out electronics. The processing time of this algorithm is in competition with its error probability and currently the used compromise realizes an average random rate in a standard detector (2D/3D) of 2-5 Million/sec for continuous operation still with an acceptable error rate. It is possible to transfer > 28 Mbyte per second or more than 7 Million 3D event results (x,y,t) per second permanently via the current USB 2.0 layout. Thus, the data transfer protocol is not the limiting factor but the algorithms complexity is the main bottleneck. Also, at high permanent rates, the MCP stack starts out-gassing because of local heating by high displacement currents. Due to their finite recovery time in single pores, the pulse heights decrease at high permanent count rates what proves to be another challenge using the standard 4 times coincidence counting principle of DLDs. This is based on the fact that all four pulse recognitions are working with thresholds and only one missed pulse of four might be sufficient to lose the entire event.

Wouldn't a hexanode solve most of the problems with incomplete or ambiguous multiple hits in the high count rate regime and enable higher data redundancy?

A hexanode device definitely is a very good choice for low average event rates, but multiple hit applications. Still, it has significant multi hit blind areas. An event rate improvement is only possible for multi hits in bursts, while the average rate is even smaller than for a standard DLD. This is because the multi hit algorithms complexity raises with the number of measurement results that has to be handled per event. Therefore, the 4fold DLD is the faster multiple hit system and the concept minimizes blind areas while keeping the high average count-rate. It avoids by design the appearance of quadruple intermixing and the correlated redundancy problems using very short delays.

Why is the specified spatial resolution worse than the pixel resolution?

Like in any other pixel detector, two neighbored, peaked intensity distributions cannot be resolved from each other, if not at least some kind of intensity minimum can be recognized between them. So, at least one or two pixels between two peaks are needed to recognize a local minimum significantly.

What is the difference between absolute and relative time resolution?

The absolute time resolution is the smallest distance of two neighbored, peaked time distributions of events at a constant detector position. To resolve them, one needs to identify a minimum between the peaks in a time histogram. If time distribution peaks appear at different detector positions, their relative time distance can be determined with higher precision.

How can you measure so many pixels when the pitch size of the meanders is as large as 0.8 mm?

Initially, very short pulses of some ten ps FWHM are coupled into all parallel arranged meander line segments. The largest pulse is coupled next to the center of mass of the hitting cloud, decreasing in amplitude to left and right. This pulse group travels with strong dispersion and when it reaches the electronics, the pulse group is appearing as its envelope. Pixel assignments appear by measuring the center maximum of the resulting envelope very accurate. This enables the determination of up to 20 pixels between neighboring physical meander lines.

It seems that the spatial resolution corresponds to a better time resolution than specified for the best absolute time resolution?

That is correct. This is due to the canceling of systematic errors of the TDC and the pulse processing units when determining the time difference which corresponds to the position. For instance, pulse shape and height dependent time shift errors will be removed by calculating the time difference for two measured pulse peak times from the same event.

In the flyer you mentioned the count rate of DLDs is 3MHz. Does that mean that the dead time between two events is 300ns in 2D+1t mode?

The 3 MHz (precisely: 3 million counts per second) is a truly random rate, so the dead time is much lower than 300 ns. If you ask for the smallest possible time the DLD accepts between two subsequent events, the answer is that it can handle multiple hit delays as low as 10 ns using the current readout system. That holds for short bursts and multi-hit recognition mode. The delay length of a delay line detector is also important to consider in that context. It is about 15 ns for a DLD4040 and about 75 ns for a DLD8080. When the signals of an event are still travelling the delay line (that needs 15 ns for DLD4040) while a next event hits, there might be conditions in which the pulses overlap too much and so the event recognition is impossible or delivers wrong results. This means, that within these 15 ns both hits have to occur clearly separated from each other in the three dimensional x,y,t-space to be recognized correctly. Without using any multi-hit mode, we determined the statistical dead time values in a random hit regime to be in the range between 20 ns and 80 ns, depending on the detector size, which is slightly larger than the delay lengths.

Is the standard DLD system capable of handling multi-hit events?

Multi-hit mode is available as a special software configuration setup. It needs a higher data stream density for finally finding the multi-hits and so it currently limits the maximum average processing rate to about 1.7 MCPS. This limit is only due to the USB transfer speed limit, because for multi-hit four times more data has to be send into the PC. The software multi-hit finder can handle up to fivefold hits, but this number is only an arbitrary limit that was defined in order to keep the processing speed on the PC high enough. This option is available for all normal DLDs. Further, we developed a 4 quadrant delay line detector for applications in which many multi-hits are expected in a very short time (short burst operation). More information on this device can be found here: There is a note on this page concerning the development of an improved fourfold TDC, which was completed in the meantime and reaches about 8 MCPS in permanent random operation. The disadvantages of this 4Q-DLD solution with respect to normal DLDs are the lower spatial resolution and a crosswise blind area between the 4 single DLD segments.

What are the advantages and limits of the 4Q-DLD systems?

The fourfold DLD is able to achieve a burst “equivalent” count rate of more than 100 MCPS but only for remarkably shorter burst durations because only 32 memory positions are available for each of the 16 TDC channels. For a short burst, the 32 memory cells buffer the results which are subsequently coming in with small average time distances. The smallest possible time distance for a single channel is 10 ns (currently limited by the CFDs, not by the TDC); between different channels it can drop to 0! Because of the 4 independent quadrants design, results for a maximum of 128 particle hits can be stored into these 16 very fast queues with a smallest average time between subsequent events of 2.5 ns corresponding to a minimum burst lengths of 320 ns. While the average read-out rate from the single queues is slow in comparison (80 MHz for all 16), one needs some suitable silent time for the readout on the interface when the burst has filled the fifos completely. The main advantage here is that 2.5 ns minimum time distance between subsequent events is far below each normal delay line time length as well as any TDC systems ability on the market today. It is even smaller than the typical FWHM of pulses that can be read out from any single DLD.

How many particles can be detected with the 4Q-DLD in a burst longer than 30 ms duration when a particle hits every 30 ns?

To understand this we must consider how fast the queues are filled during operation. 2 event quadruple sets are filled into the queue, while 1 quadruple set is read-out by the interface. This leads to loss of data after about 6 µs burst length (that is 196 events x 30 ns), unless we would give it then 6.4 µs silent time for reading the low level fifos. If we could do so, about 500,000 particles can be detected within the 30 ms time window. This is the case of a typical micro bunch group in free electron laser experiments what that detector was developed for, initially.

So during 6.4µs silent time the anode mustn't be hit by any event or can it be hit but the event will be lost?

The requirement is that no data should get into the TDC FIFOs during this period. So either, the experiment/anode mustn’t detect any events or we can switch the TDC inputs blind.

What happens when the multi-hit queues are filled in the new 4Q-DLD 4fold-TDC, how many events per time can then still be processed into the 128 MByte buffer memory?

Each particle on the detector must be determined by 4 measurements due to the DLD’s operation principle. Currently four 80 MHz TDC interfaces are used with 4 channels on the four DLD quadrants each. This means that the readings per quadrant can be handled with 20 MHz average read-out rate. Four results per event lead to 5 Million quadruple data-sets per second per quadrant. Thus, the system is characterized by an average reading offset time of 200 ns between subsequent events on a single quadrant. If all events are equally distributed over all 4 quadrants, events with an average reading time gap of 50 ns can permanently be streamed into the TDC memory.

If we wouldn't use the 4fold DLD, but the 80mm diameter standard DLD, would the limitation of the detection rate be due to the electronics?

No, the limits are given by the length of the delays of the detector anode and the speed and reliability of the algorithms that find the quadruple assignments for each event. The larger the detector, the longer the delay will be, so for 80mm, the delay will be above 60 ns, which sets the shortest average single hit time distance for “random” event distributions to 180 ns (about a factor of three is due to statistics). Even that limit is hard to reach because the quadruple finder algorithms are not perfect. It is not a limitation by the electronics at all. If one improves data processing by better multihit recognition, the system needs even more time for finding the correct result assignments and many cases of possible ambiguous assignments lead to data losses. Therefore, multihit recognition helps only for short bursts with sufficient silent time and if the detector design is optimized to avoid ambiguous cases (e.g. overlaps of pulses for one or even both delay directions x, y). This design optimization is what we did for the fourfold DLD, short delays and 4 parallel, almost independent read-outs. On the other hand this downgrades the spatial resolution, but that problem can in principle be corrected by a more complex parallel read-out electronics with even higher resolution. This will be realized by the upcoming read-out with a 4fold TDC that is under development.

« back to top