Green Hippo

Testing HD-SDI latency with Creative Technology

Testing HD-SDI latency with Creative Technology

Following on from the announcement that Green Hippo have a new range of input cards, including a cutting edge implementation of HD-SDI, the company’s head of Software Development, Sean Westgate gives an insight into the somewhat tricky techniques for testing the developments.

July 2009: So Green Hippo spent months developing this new HD-SDI input card, getting sample cards, optimising the drivers and implementing it at a really low level to make it fly. Was it all worth it?

I, as Technical Director of Green Hippo, was just as curious as many of our customers to find out how fast these new input cards in Hippotizer really were, so we contacted Scott Burges from Creative Technology London to see if they could help us with some latency tests. The problem is, how do we reliably measure the time it takes for a camera image to travel from the point it enters the Hippotizer until it is displayed on the output? Many applications where live camera is used require this delay to be as short as possible for obvious reasons: If you are filming a singer or a presenter and you hear the sound before you see the image on a large screen it looks “out of sync”. Unfortunately many pieces of modern equipment introduce a delay in the video processing pipeline; there are converters, mixers, obviously Hippotizers and finally display processing equipment like LED wall controllers all potentially adding delay to the video image. Why don’t we have zero latency on the Hippotizer, surely that would be best?  Yes, of course, but unfortunately this is not possible with current PC hardware as the image needs to be converted from one format (HD-SDI serial data stream) to another (internal texture, which is a progressive image) and finally presented at the right time to the output device. So even in an ideal world you would get one frame input delay as you need to store the whole image before you can process it as a whole, and up to one frame on the output as you need to wait for the next image to refresh before you can present the processed image. But how fast is the processing in between?

When designing the test case it became very obvious that our previous test methods using oscilloscopes wouldn’t work as we do not own one that can read both the HD-SDI input and the DVI output. So we thought if we had 2 monitors side by side, one showing the input and the other one the output and we then film the 2 side by side we would see the difference and delay between the 2 images. To measure the delay we could use a camera, so that we can record the results and then review them. We tried this with equipment available in our workshop, but it soon became obvious that although a CRT monitor could provide a zero latency image of the output of the Hippotizer, we did not have suitable equipment for displaying the HD-SDI input signal without any delay. Plus we did not have a camera that would allow us to record at higher refresh rates, such as 50 or 60 frames per second which would enable to measure the delay in 17-20ms intervals.

Creative Technology came to the rescue and with the assistance of Scott Burges a test setup was rigged with a zero latency HD-SDI monitor, a zero latency CRT monitor and a HD-SDI camera connected to a hard disc recorder.

Here is what the camera recording looked like:

On the left is a CRT monitor, one of the old variety – newer LCD displays often introduce a delay, but this old analogue fellow has no delay. On the right is the (very expensive) HD-SDI monitor of broadcast quality again with no delay. A Mac with Final Cut Pro is used to generate the HD-SDI signal with timecode. The above image shows the system displaying 1080i @ 60Hz and as you can see, the output of the Hippotizer is 2 frames late which equates to a delay of 66ms (2 frames at 33ms each).

We ran through all commonly used formats and came up with the following results:

Latency:

As you can see above we also tested with different loads on the Hippotizer as we wanted to see if loading up the unit with more processing would introduce an additional delay in the HD-SDI input. Strangely enough the opposite seems to be the case: when the unit is working harder the delay goes down! But there is an explanation for this: as the load increases there is more time to catch the next refresh on the output so the chance of displaying the image earlier increases. However we wouldn’t recommend doing this intentionally as it can have a negative effect on the smoothness of playing video clips.

All in all we were delighted to see that our hard work has paid off and we are very close to the theoretical maximum latency reduction. This combined with the cards automatically detecting the input format makes them very easy to setup and use.

Many thanks for all the guys from CT for their help and support on this especially, Scott Burges for making it all happen, Tim Volkner for designing the setup and giving us access to the equipment and Antony Bezencon for doing all the tests on the day.

Many thanks also go to Takeshi Hayakawa at Japanese Hippotizer Distributors TSJ and Japanese Hippotizer users Hibino for Beta testing the card and verifying these results in an independent test. 

For a list of supported formats please check our knowledgebase article: http://www.green-hippo.com/support/index.php?_m=knowledgebase&_a=viewarticle&kbarticleid=54&nav=0,4

For more information and how to purchase this great new card please contact your local dealer.