Triggering a CoaXPress camera: What about latency and jitter? – Part 2

In our last blog we discussed the origins of latency and jitter in a triggering scheme for a CoaXPress camera. In this blog we will continue by zooming in on the actual magnitude of the latency and jitter.

What is the latency and jitter related to the CoaXPress interface?

To answer this question let us first describe how triggering over CoaXPress actually works. This information is well documented in the CoaXPress standard, but we will give a short summary below.

For CXP the trigger signal is hidden in the same data stream as other camera commands. However trigger signals are labeled in a specific way such that the interface knows that it is a trigger signal. The trigger signals are given the highest priority in the standard such that the latency is reduced to a minimum.

So what is this latency? Most CXP cameras will use the low speed up connection. Due to the CXP protocol this low speed up connection has a latency of 3.4 μs with a jitter of ± 4 ns. This fixed latency is achieved by the use of an accurate timing correction mechanism. The 4 ns jitter is actually a minimum jitter. It depends on the actual implementation of the frame grabber and camera manufacturers if this jitter can be achieved.

For more details about the origin of this 3.4 μs with a jitter of ± 4 ns you can read one of our previous blogs.

What is the latency and jitter related to an external I/O port on the frame grabber or camera?

The advantage of using the I/O port on the camera is that you do not have the additional latency of the CoaXPress interface. However this does not mean that you can state that you have no delays or jitter at all. Triggering over the I/O port on the frame grabber will in the end also encounter the CoaXPress latency and jitter but on system level it might be the most convenient way to synchronize the camera with other system elements.

Before we continue, it is important to first make a distinction between two different implementation methods for an external I/O. Often only one of these methods is available for a specific camera or frame grabber model.

1. The I/O port is directly connected to the logic input of the camera or frame grabber by a low voltage differential signal (LVDS) or transistor-transistor logic (TTL) signal.
2. The I/O port is galvanically isolated from the internal, and thus sensor, electronics.

Let us discuss these two methods one by one.

1. The I/O port is directly connected to the logic input of the camera or frame grabber by LVDS or a TTL signal.

This method of implementing the I/O access is the most direct way. For this implementation it is actually true that the delay and jitter is negligible. However it comes with disadvantages that could damage your camera irreversibly. As the I/O is directly connected to the logic input it is very sensitive to electromagnetic compatibility. All electronic systems should be close to each other to guarantee the same electrical ground. If the ground is not the same everywhere and the voltage difference between the grounds becomes too large, the sensor could be damaged. With distances too large between the source of the trigger and the device it could thus become challenging to keep the trigger voltages within the damage threshold.

2. The I/O port is galvanically isolated from the internal electronics.

Galvanic isolation solves the problem described in the first method. Galvanic isolation breaks the ground loop between the external and internal electronic circuit. It thus removes the risk of having large voltage differences due to a difference in ground potential. Galvanic isolation can, for example, be achieved by using opto-couplers. With an opto-coupler the incoming trigger signal activates a light emitting diode and on the side of the internal electronics a detecting diode is present to detect the light. There is thus no electrical contact between the two circuitries.

However, opto-couplers do have influence on latency and jitter. Opto-couplers require a certain response time to turn on or to turn off. This response time contributes directly to the latency. The actual response time, and thus latency, depends on the electronic circuit and opto-coupler that is being used. For example, Adimec has cameras in which this latency can be < 0.5 µs if the right circuitry is being used. However with a wrong circuitry this could easily increase to several microseconds. What exactly is good or bad circuitry depends on the actual implementation and is out of the scope of this article.

 

 

In terms of jitter the biggest contribution is probably from environmental conditions. The response time of the opto-coupler depends on its temperature. This might cause a different response time if you have an irregular trigger sequence. For a high density sequences the opto-coupler might increase more in temperature then with a low density sequence of triggers.

In a constant temperature environment the typical jitter of an opto-coupler circuitry is in the order of a few tens of nanoseconds. But when the temperature varies the jitter can increase to hundreds of nanoseconds or even several microseconds.

Another way to achieve galvanic isolation is using digital isolators. Digital isolators have a more controllable delay and jitter compared to opto-couplers but they require an external power source.

So as you can read above, several factors are at play in determining the latency and jitter for an external I/O port. This makes it difficult to state a single latency and jitter that is valid for all implementations. Hopefully the above information helps you in determining the latency and jitter of the external I/O port in your system.

Summary

This two part blog provides insight in the latency and jitter related to the triggering part of a vision system.

Several triggering possibilities are mentioned: triggering over CoaXPress, and triggering over the I/O port of the frame grabber or camera.

Using the CoaXPress interface adds a latency of 3.4 µs with a jitter of about 4 ns.

Using the External I/O port could add latency as well although it depends on the exact implementation. For example, galvanic isolation with an opto-coupler adds < 0.5 µs latency when an optimal electronic circuit is used. In less optimized situations this could increase to several microseconds. The jitter in a temperature stable environment is in the order of a few tens of nanoseconds but in less stable environments it can go up from hundreds of nanoseconds to several microseconds.

When the I/O port of the frame grabber is used, the latency and jitter of the I/O port and CoaXPress interface should be added together.

Do you have any questions related to latency or jitter? Don’t hesitate to contact Adimec Support.