View Full Version : USB 29 V M Frane Drop

October 7, 2014, 10:05:38

We use this camera to record videos of 10 seconds at 120 fps, the exposure is locked at 1/120 and video is recorded as an avi uncompressed Y800 format.

Using IC-Capture, setting it to record 10 seconds we get less fps than 1200 and analyzing the video with virtual dub we find the presence of dropped frames.

For this test we use an i5 processor and a 7200 rpm sata hard disk.

What can be the reason for this frame drop ?

Stefan Geissler
October 7, 2014, 11:06:14

how many frames do you miss?

October 7, 2014, 11:39:18
The frame should be 1200, but generally I get 970, 1070, 1150 in 10 seconds of video.

Instead, if set limit of recording at 1200 fps I get a video of 11 seconds.

Edit -

The OS is Windows 8.1

Stefan Geissler
October 7, 2014, 12:11:16

It is hard to say, where the frame drop happens. It can be the harddisc, so the FileWriter is not able to write all data to hard disc and drops some frames. Are there other devices on the same USB 3.0 controller, that may be interrupt the image data transfer? This is very unlikely for USB 3.0, but I want to be sure.

Is there a manufacturer's driver for your USB 3.0 controller? If so, you may try it.

October 7, 2014, 16:29:58
I did some testing, and with the hard disk at 7200 and 5400 rpm occurs frame drop, instead of using an SSD or a ramdrive there are no problems.

At this point I have questions / suggestions, from what I understand, the problem does not depend on transfer speed (data to be transferred are about 30 Mbytes per second), but should be based on response time of the disk.

I was thinking, expand the buffer dedicated to writing should fix the problem.

October 8, 2014, 18:47:55
Hello Stefan, related to this problem, you'll do some tests?

Stefan Geissler
October 9, 2014, 11:14:27

related to this problem, you'll do some tests?

Since I expect frame drops on my computer at 120fps, I think, explicit tests are not necessary to approve them. Also I have different hardware and different software installed on my computer, so tests can not be compared.

October 9, 2014, 11:55:28

Ok, I have not explained well.

I mean if you try to solve the problem of frame drop on the normal hard disk?

In my opinion, a write buffer that contains approximately 100 frames or one second of video should be sufficient to solve the problem.
I'd try, but now I can not do tests with DirectShow.

Stefan Geissler
October 9, 2014, 13:35:18
I mean if you try to solve the problem of frame drop on the normal hard disk?
No. How should we do this?

The UVC protocol does not guarantee, that all frames captured by the camera will be delivered completely to the computer. Also this does not say anything about how DirectShow will handle the frames in the filter graph.

What you can do:
- keep your hard disc defragmented
- or use an extra hard disc partition for AVI capture.
- make sure, no other processes will interrupt Windows too long
- use an efficient codec (e.g. MJPEG Compressor is not efficient)