Wireless camera has gaps in recording

[Stewy]

This may or may not be a networking forum issue, but got to start somewhere to relieve my exasperation!

I have 2 wireless cameras in my network, a Hikvision DS-2CD2542FWD & a DS-2CD2142FWD. The 2542 has worked flawlessly while the 2142 occasionally behaves like a poltergeist has sneaked into it.

It can run seemingly smoothly for weeks, then it will go though phases when there are recording blanks. This usually starts with gaps of just a few seconds (only noticeable if you happen to be watching when the clock freezes), then a few minutes and sometimes hours. I have uploaded a snapshot of how this looks on the NVR playback screen.

This has happened a couple of times before, but lots of fiddling, re-booting and talking not-so-nicely to the camera has got it running again (to be honest I never observed any cause and effect in anything I tried, but it would eventually resume continuous operation). The drops in recording just started happening again in the last couple of days, so this time I am seeking help in running the problem to ground.

It feels like a network issue, but could it be something else specific to the camera?

The erratic 2142 has line-of-sight from its position in the eaves of the garage roof to my router, through a window over a range of 8 metres (the other camera actually has to penetrate a wall, so it ought to have a weaker link). No other devices struggle for wi-fi signal in the garage.

I don't see how it can be a bandwidth issue, although the camera image gives the appearance of struggling like that's exactly what it is (during periods when the camera is working normally, the video quality only seems to struggle with bitrate sometimes when I reverse the car quickly back from the garage door, at which point the majority of the image area is changing rapidly; normally the image is smooth). Anyway, I should have heaps of bandwidth to spare. Indeed I have run streaming video to 2 smartphones and an iPad in my garage simultaneously, while the cameras continued to run without a hiccup.

I have no obvious interference on my network. I have my wi-fi fixed on channel 11, while the only other visible network (from my neighbour) is on channel 5 (he has his router on auto channel select, but I have checked when my camera has been having a bad day and I have the frequency all to myself). Unless it could be something else interfering? The garage door alarm has a wireless transmitter, but that operates on a very much higher frequency than a wi-fi network.

I run both cameras at demanding settings: max resolution and 20 fps, but with H264+ turned on (at medium quality). When the 2142 misbehaves, I try downgrading the settings, one at a time, but that doesn't seem to make an instant difference. In any case, what's the point in having a 4MP camera and not making use of all those pixels.

I appreciate I could use a PoE and powerlink solution to avoid relying on wireless. However, I'm not convinced that wireless is the problem here, tempting though it may be to blame such inconsistent behaviour on the vagaries of w-fi. What other troubleshooting might you suggest?

 

[Phil]

Thanks for the comprehensive write-up.

I'm sorry, I know it's not what you want to hear, but I think Step 1 in troubleshooting this will be to remove wireless from the equation.

Perhaps you can chuck-in a long cable over the garage roof just to prove a point?
Normally I'd recommend you connect it to your NVR with a short (manufactured) cable, but you're going for long periods without a problem, so might not want the camera out of place for as long as you need to be confident whether cable is better than wireless.

I doubt that any variation in scene, bandwidth, bitrate are significant - pretty much always going to be <7Mbps which shouldn't be too challenging.

 

[Stewy]

Thanks for the prompt response. As you point out, it is impracticable to hard wire my remote garage camera for any length of time (plus I’d have to drill a hole for an extended experiment, as the vehicle door is the only access). However, I think I have moved on in my troubleshooting and would be interested in your analysis of the results. Apologies for the eventual length of the post, but I am trying to not leave any obvious questions hanging out there.

I installed inside the garage an unused wi-fi extender that I already had. This is now only a metre away from the camera. I connected the camera to the IP address of the extender and on the web interface of the camera this now showed a signal strength of 100 (versus about 60 previously when connected directly to the main router in the house). However, the erratic performance of the camera was unchanged. The other camera remained connected directly to the router and continued to function normally. I know this whole issue feels like a comms problem, but I am less and less convinced it is anything to do with my network. The performance of the two wireless cameras is so dramatically different.

I was struck by your observation about the bitrate requirement of no more than 7Mbps versus the available network bandwidth, which suggests that there shouldn’t be a comms rate issue. This should be true even if the bitrate demand is high, as it would be in the circumstance I described in my initial note, namely a moving vehicle suddenly filling half the normally static image. Yet I had previously observed that such an event does sometimes cause the image to freeze and recording to be skipped. That had seemed a relatively minor issue compared to the occasional random total loss of video, so I had ignored the “big movement” issue for the time being, thinking the two issues were unconnected. But in light of what you said, I began to wonder….

So I investigated much more closely the recordings of this camera and discovered its image gaps were worse than I had thought. You can’t see these as “black gaps” on the playback screen because of the scale on the time axis. However, there are skips of about half-a-second to a second going on regularly. Also, when I create an intrusion alarm, the event almost always triggers a loss of video, which then takes anything from a couple of seconds to a couple of minutes to reappear. The impact seem to vary with the scale of the moving intrusion, but haven't tested that rigorously.

To further investigate the datastream from the camera, I resorted to experimenting in Live View mode. Although small movements of people (i.e. not big enough to trigger freezing of the image) looked reasonably smooth, close observation of the clock shows that while it looks continuous, it isn’t quite regular.

I then set up an experiment with two monitors side-by side. One monitor was connected directly to the NVR, while the other was my main desktop computer monitor. My desktop PC established a connection to the problem camera via the ethernet cable to my router and using the Hikvision web interface. All other wireless devices on the network were switched off (including the other camera). So the only network connections were the one wireless camera being tested and two cabled devices (NVR and desktop computer).

I then proceeded to experiment with the bitrate settings for the camera. In all cases, this was done at 20 FPS, for both mainstream (2688x1520) and substream (640x360) feeds. For a consistent test movement I made use of the reflection on my car bonnet (off the bright overcast cloudbase) of aircraft 2000 feet above on final approach to Heathrow. This is really useful because the distinctive shape moves across the bonnet at constant speed, while only occupying a few pixels of the image. Any skipping or even slight hesitation in the image is readily evident. There was no movement in the rest of the image.
I made the following observations:

1. The image seemed to have equally smooth motion on both monitors when using the same stream simultaneously, on both mainstream and substream feeds.

2. When comparing mainstream (with H264+ on as usual) to substream, it was noticeable that substream motion was definitely smoother. This reduction in mainstream fluidity had not been evident, at least to my eye, when just viewing the mainstream feed on its own. As a check I switched the two streams between monitors to eliminate any monitor performance impact, but this made no difference.

3. Switching H264+ off for the mainstream feed made a huge difference. The image moved very erratically and the entire video feed would sometimes freeze for no apparent reason.

4. Finally I switched to constant bitrate (8192 Mbps mainstream and 1024 Mbps substream). The substream performance appeared no different than it had on variable bitrate, i.e. motion was still apparently smooth. The mainstream performance was, if anything, slightly worse than it had been on variable (but hard to tell as performance on both settings was randomly erratic).

It appears to me that bitrate transmission from this particular camera is indeed a problem. Even when I allocated a large bitrate allowance manually, the image was unstable. However, it surely cannot be a lack of bandwidth on the network, can it? I’m no expert though. Am I missing something in my logic?

As it seems the problem is not as intermittent as previously thought (although more blatantly obvious sometimes than others), I can do some additional testing via cable. I plan to hook up a laptop to the camera and see how different degrees of motion in the image affect the video stability. However, the laptop isn’t at home at the moment, so that will have to wait.

Is there any specific testing I should do with the cabled connection when I get the opportunity? Can I track the bitrate live? (the NVR interface allows you to see this and FPS “live” on the screen, but they jump about so much I’m not convinced how reflective that onscreen display is of what is actually going on in genuine real time).

regards
Graham

 

[Phil]

Test it with a cable connection (to your laptop) and see how it runs.
The h.264+ CODEC is amongst the most efficient, so will result in a lower bandwidth stream.
A 20fps video stream may be excessive, most people run at 10fps or less - most of our YouTube samples are captured and played at 6fps.
It's very hard to know exactly what is going on / going wrong with a wireless connection.
Even having a transmitter and receiver close together isn't always a dead cert.
We receive remarkably few complaints about HikVision's wireless cameras (honestly, I would expect more).
We've never found a suitable bandwith monitoring tool - they must exist, I've seen them in YouTube videos, but we have not yet identified a suitable tool (there are some expensive / Pro network monitoring software solutions).
Your router may be able to tell you more - obviously it depends on the make, model, level of device.

 

[Stewy]

Thanks. I'll give that a try to see what effect it has, while waiting for the laptop.

However, I'm struggling to see how halving the FPS helps me to answer the key question, namely is the problem with the network or is the camera faulty? Going down to 10 FPS will halve the bitrate but presumably your calculation of <7Mbps was done at 20FPS ? Given that the problematic camera was the only wireless device on my network during the tests, then there would have been heaps of network bandwidth to spare. As for the camera, its spec says it should be able to deliver 20 FPS at max resolution. Certainly the other wireless camera (the 2542), which has the same image spec, delivers that performance reliably.

Unfortunately my router only provides performance data for the cabled input connection from the ISP, not for the wireless signal.

 

[Stewy]

I finally got my laptop back from its summer holiday and was able to test the 2142 camera by directly connecting it to the laptop by cable, thereby removing my wi-fi network as a variable. I then watched the Live View image on the web interface while we conducted various tests in the driveway (the camera is mounted in the eaves of the garage, looking down the driveway).

I was able to reproduce the freezing of the image, although it did seem like the camera was generally more reliable than it had been on wi-fi (hard to be sure because its wi-fi performance is not consistent). However, I still don't think all is well. Basically, driving the little hatchback up the drive was fine but switching to the Land Rover resulted in the image freezing whenever the vehicle got close to the garage. The obvious difference in these tests is the size of the moving image. I could also get the image to freeze by having the kids run towards the camera. Whenever they got close enough (to get a good snapshot of a recognisable face), the image would usually freeze. So given the objective of the camera is to monitor the driveway, I have a problem in that it fails in just the circumstances it is there to monitor.

The difference between the vehicle and the people experiments is that while the kids are smaller, they were moving faster.
So in both cases it seems like this is indeed a transmission speed issue. But it can't be the fault (at least solely) of my wi-fi network. Moreover, changing the bitrate to a constant 8Mbps did not cure the problem. Halving the FPS made no significant difference, at least within the variables of the experiments.

I'm tempted to conclude that the camera is definitely not performing as it should, or maybe the peak bitrate is actually well in excess of 8Mbps at times. But is my logic sound? Anything else I can try?