Why Are Encoders More Expensive Than Dvrs?

Hello All, Although encoder pricing have dropped over the last few years they are still much higher than most DVRs. To me the effort that a manufacture has to go through to build a DVR is much more than an encoder. Manufactures don’t have to develop software, recording solution, iphone app, etc. for an encoder like they do for a DVR. So why do encoders cost more money?

The only thing that I can think of, is that most encoder encode each channel separately at a higher resolution/higher frame than most DVRs and it requires more encoding/chipsets that cost more money.


It's a really good question and I am not sure but I have some theories. [As background, we have a post on encoders vs hybrid DVRs that examines this comparison.]

  • Theory 1: Comparing higher end encoders to lower end DVRs. There's not as many low end / entry level encoders as there are entry level DVRs so one compares higher hardware quality encoders to lower quality DVRs (like Costco models).
  • Theory 2: Lack of scale for encoders: Low end DVRs are likely sold in 10x, maybe 100x greater quantity than encoders. This likely helps to reduce the cost.
  • Theory 3: Encoders are being used for higher end deployments where customers likely have more money to spend. As such, the sellers are less motivated to drop prices and are happy to take higher margins.

Btw, we were having another encoder discussion today where a new 16 channel Bosch encoder looks to have a street price of ~$1,000. That's extremely good as encoder pricing goes and close to even low end 16 channel DVRs.

If you stick within brands or manufacturers, you may be able to use the DVR as an encoder. Thats what we do, we dont even carry encoders as there is no point in doing so, because the 4,8, and 16 channel DVR's will encode to the NVR's. Only thing is with DVR's is most, if not all, are not Onvif compliant so it makes intercompatibility between brands difficult. But if the DVR provides an RTSP stream out, and if your NVR or software can accept an RTSP stream, then you are golden. Of course you will only get a video stream that way, no motion detection, audio, or anything like that.

But I am with you, it should not cost more than a DVR, actually should cost less IMO. All A DVR is is a glorified encoder with recording capabilities.

Sean, I am curious about the RTSP stream out functionality. What DVRs do you know that do that? And they provide an RTSP stream for every feed (i.e., a 16 channel DVR provides 16 RTSP feeds out)? It's an interesting idea, I just didn't know many DVRs provide that.

Sean, Can you clarify this "we dont even carry encoders as there is no point in doing so, because the 4,8, and 16 channel DVR's will encode to the NVR's". What DVR brand can encode to an NVR?

John, do you know of any DVR that has an RTSP stream out functionality?

We deal and distribute Dahua products rebranded with our own brand name. All Dahua DVR's have an RTSP stream so you can use with 3rd party programs such as VLC media player and NVR softwares that will take an RTSP stream. And yes you can choose which channel gets streamed out by simply adjusting the RTSP URL string. Of course, like I said, you wont get motion detection to work unless the NVR software you are using has its own on board motion detection. For example, if you used it with Exacq, you would not be able to use motion detection because it relies on the incoming IP camera or device's motion detection protocol and of course with RTSP, there is no motion detection protocol. But with a software like Milestone, you could use motion detection since it has its own on board motion detection processing.

We also carry their NVR's and with their NVR's you can choose to connect to a particular channel from a remote DVR if you want. And since they are the same manufacturer, you will get motion detection and full compatibility.

I could be wrong but I do think you can do this with Geovision, although they are more PC based while Dahua is an embedded standalone device. I would not be surprised if this option was not available for other manufacturers as well. Some of the bigger brands though, such as CNB, Panasonic, Bosch, etc. will have a Korean or Chinese manufacturer make their DVR's and NVR's and they sometimes use different manufacturers throughout these products. The key is is to just stick with the same manufacturers and I would bet that it could work.

Interesting... this must be something new, because I seem to remember a discussion elsewhere a while back with someone asking about pulling an RTSP stream from a Dahua DVR and the reply being that it wasn't possible. Or maybe I'm just "mis-remembering" things.

If that's the case though, and you can specify the channel via the URL, I bet I could make it work with Vigil as well :) Dang, that woulda made this current job a lot easier!

Of course, with Dahua, you can also use their PSS software as a minimal-yet-functional NVR/VMS to stream from and record any mix of Dahua cameras, DVRs and NVRs. Pretty nice integration all-around, really.

Sean, actually we deal with Dahua too, and there is special FW for EYEsurv ESDV-FULLD1-16 16 Channel Security DVR Real Time D1 Resolution that supports also motion detection with three independent zones

Sean, and you can RTSP stream out ALL channels simultaneously? if the other device (Exacq, Milestone, etc.) is being used as the NVR (that is recording), you would need to do this.

Sean, will all Dahua DVR do this or only a store specific model?

Matt...... That could have been on an old firmware. Most of the newer firmwares built within the last year and a half allows this function. I will send you the string.

Robert........ As far as I know, most all of the DVR's with the newest firmwares allow you to do this. Of course I am only familiar with the firmwares that we have, but I would be doubtful if they didnt put this feature on for other companies as well

Interesting... I suppose I'll have to get another DVR in now to test this. Question: will a DVR that records at 7fps (for example) on a given channel, still stream that channel at 30fps?

Matt, Dahua has DVR's with 25 fps per channel too. If you want use it with a NVR with more professional options than Dahua NVR, you can use Digifort. Digifort has the Dahua DVR's integrated as video servers and it is possible use others fucntions like PTZ and I/O.

It will only show you your particular FPS you have your DVR set at in the encoding section. Whenever you are viewing remote video either through the RTSP stream or through the web browser or through PSS, you are viewing the encoded video, and that is the same stream that gets recorded to the hard drive as well. So for example, if you have your DVR set to record at 7 FPS, then that is what you would see from the RTSP stream or any other remote video method for that matter.

Now there is a way to make an adjustment to the RTSP string so you can view the substream if you want. You can choose to use the main stream or the substream on the RTSP stream for that particular channel if you want to. So for example, if your DVR is recording at D1 @ 7 FPS on the main stream, but you would prefer to see the substeam @ CIF at 30 FPS on the RTSP stream, you can do that as well. But just note that the substream maxes out at CIF resolution for the analog DVR's. And CIF kind of looks crappy.

I would be willing to bet the capabilities of most, if not all, DVRs are extremely crippled compared to a good encoder. For one thing, my testing has shown that there is little codec control capability in RTSP streams. GOP size and structure, bitrate control and the like are often nonexistent.

That was with an Axis and a Bosch encoder. Perhaps a DVR's RTSP is different but I've also found most DVRs lack the more sophisticated controls available on high-end encoders anyway.

"If you want use it with a NVR with more professional options than Dahua NVR, you can use Digifort."

If anything, I'm actually looking at this as a potential option with Vigil DVRs. Not that I have any immediate need for this capability, but it's an interesting prospect.

"I would be willing to bet the capabilities of most, if not all, DVRs are extremely crippled compared to a good encoder."

No doubt. But you're also paying four to ten times more for those encoders. As with so much else, it's going to come down to a balance of price vs. performance vs. features on a case-by-case basis: if you NEED to have that much control over each individual encoding channel, then it makes sense to pay the extra for a "proper" encoder. If you just need a quick-and-dirty way to "convert" analog cameras to IP streams, then it makes sense to save the money and not pay for all those features you'll never use.

"Four to ten times more"? Not sure that's the case. One of the best encoders we tested, the Axis Q7406 costs less than $267 per channel when installed in a fully-populated Q7900 chassis, according to an internet price search. That's 84 channels in 4RU - the densest solution I've seen. The cost-per-channel may be a bit higher than cheaper systems but the space savings are well worth it in our situation and it offers far more features and capabilities.

Either way, it's far less costly than our current system's encoders, which at ~$3,200 per 8 channels works out to $400/channel.

"...the Axis Q7406 costs less than $267 per channel..."

A four-channel full 30fps@D1 Dahua DVR retails for under $350. That's three times more right there. And as you point out, price for other encoders goes up sharply from there.

Density is another issue, and before you note all the configuration options available with the Axis and others, I'll reiterate my point: "it's going to come down to a balance of price vs. performance vs. features on a case-by-case basis: if you NEED to have that much control over each individual encoding channel, then it makes sense to pay the extra for a "proper" encoder. If you just need a quick-and-dirty way to "convert" analog cameras to IP streams, then it makes sense to save the money and not pay for all those features you'll never use."

"Quick-and-dirty" has no place in my vocabulary.

Well, it actually does, but not in that usage ;>)

Wish all my clients had unlimited surveillance budgets.

<like button> pressed

I have clients who were extremely impressed by the improvement in video operations of existing color analog cameras upgraded from 320x240 DVR recording (and viewing) at low frame rates to full resolution server-based recording and viewing at high frame rates, using encoders.

I'm most familiar with the Axis encoders, although I know there are some other brands of good encoders out there, and most integrators with whom I've discussed it have their own preferred brands. The encoders I'm familiar with provide multiple video streams per camera, each at appropriate settings such as for live monitoring, recording, and remote viewing.

Encoders can add all kinds of capabilities to each analog camera, including video motion detection, video loss alarm, alarm events including sending video clips or frames via FTP or email, activation of the encoders physical alarm outputs, and PTZ support including activation of PTZ presets upon motion alarm. This is without any server or application software supporte needed.

Some 4-camera encoders provide a 1536x1152 quad view. One series of encoders supports up to 64 GB memory card per 4 cameras for edge storage. Edge storage combined with the free 16-camera software make a heck of an afforable upgrade for a small analog camera system, as for some end users this means no NVR or DVR needed, and any recent model PC or laptop could be used given edge storage.

On the other hand, for one client I'm upgrading two facilities with network cameras, and moving a DVR they recently purchased and a dozen analog cameras from one facility to another. There is no network in the facility's remote building, and the analog cameras and DVR will work just fine there.

I'm a fan of reusing good analog cameras, and find that encoders can provide excellent integration into whatever VMS is being used. But also find that I can reuse existing late model DVRs, especially in circumstances where there is no live monitoring.

The product life cycles for most good analog cameras is pretty long, and they are often good at covering small indoor areas very well.

I don't find much use for new low-end or "entery-level" DVRs though.

Matt, I bet your clients do, too! Budget constraints are one reason why I like extending the use of analog cameras using edge storage for with cameras and some encoded analog cameras. But you are right about budget constraints and in some cases advanced features of network cameras and encoders are just not needed. No sense paying for what you don't need, and are not likely to need, if that's what the risk picture indicates.

"they are often good at covering small indoor areas very well." And, with proper lens and FOV selection, can give perfectly satisfactory performance for many applications.

Yes, if the FoV is 10 foot wide or less OR the user is OK with not capturing facial features or other fine details.

Carl, that is very true but I can see that I was talking to my own experience, which is with cameras that are already in place with lenses that are suited for small areas and short depths of field. Although I have to admit that a $200 or $300 lens on an existing camera is more affordable than a brand new camera plus the $200 or $300 lens!

You've now got me thinking about applying good wide-angle lenses to an existing analog camera. I'm only familiar with the Theia wide-angle-lenses, which are for megapixel cameras. Are there similar products that would be good to use with an analog camera?

I know I have to keep in mind what John is talking about, which is the fact that I'm spreading 640 or so pixels accross a wider space. I need some field testing to determine where this would really be useful. I've seen good results in small museums where a higher number of cameras at SD resolution, properly located, are more effective than a smaller number of megapixel cameras, regardless of image quality. Sometimes it's the camera's location that can be critically important.

I'm only familiar with the Theia wide-angle-lenses, which are for megapixel cameras. Are there similar products that would be good to use with an analog camera?

Cost aside, there's no reason you can't use a megapixel lens on an analog camera. May even make for a sharper, clearer picture.

That Theia lens costs the same as a shelf of analog cameras at Costco :)

Theia prices are typically $300 - $450 per lens, so it's quite a jump.

Ray, though I don't know, I am sure there are super wide angle analog lenses. However, what's the practical use of them? The pixel density is going to be so low, you wouldn't make much anything out at all more than 10 or 20 feet away. I am sure there are a few niche cases but it is likely to disappoint in most scenarios.

Member's reference: our Theia lens test results.

...I am sure there are super wide angle analog lenses. However, what's the practical use of them? The pixel density is going to be so low, you wouldn't make much anything out at all more than 10 or 20 feet away. I am sure there are a few niche cases but it is likely to disappoint in most scenarios.

I don't think it's as niche as you think. We've found it actually very common in retail to use a number of wide views (2.5-2.8mm on 1.3" sensor is not uncommon) to give overall coverage of activity, with closeups on the choke points to provide ID. When you minimize blind spots, it's not hard to follow someone through a site until the're caught in that identifying shot.

Give you an example: a current large restaurant install we're completing has a mix of Axis, CNB and Dahua domes, 40+ in total, all of them at their widest coverage. There are only three public entrances, and those are covered with Panasonic WV-CW504s with super-tight shots on the doors.

The only other factor I suspect is whether you use manual iris lens and close down the F stop. That would increase the depth of field.

True, although that wouldn't affect the "perspective" issue Carl refers to.

John,

"Yes, if the FoV is 10 foot wide or less OR the user is OK with not capturing facial features or other fine details."

- Not necessarily. Someday you should try placing a camera a good distance away from a scene, then zoom in with a 5-50mm lens. You'll be surprised at the results.

Ray,

I sometimes disagree with John's focus on megapixel cameras as a be-all, end-all. Analog (and maybe 4CIF IP) cameras have many suitable applications:

  • Elevator cars
  • Offices
  • Area overviews where other cameras actually capture faces and incident closeups
  • Slot machine bank overviews
  • Hallways (one camera at each end zoomed in so that midway fills the screen width) (I call that "criss-crossed")
  • "211" shots
  • Any application where it is not necessary to see fine details, just a general view of what is happening.

Anyway, for many applications requiring wide FOV on analog cameras we use either Computar T2Z1816CS (manual iris), TG2Z1816FCS (auto-iris) (1.8-3.6mm) or Pelco 13VA1-3 (manual iris) or 13VD1-3 (auto-iris) (1.6-3.4mm) lenses.

John, I'm thinking in particular about one small museum application where many of the cameras are (or would be) 6 to 10 feet away from the target areas, which include showcases and interactive exhibits. Also, small entry areas.

Carl, thank you for expanding my thinking. And I'll check out the lenses that you mentioned. I appreciate the specific references.

Carl, you probably misread my statement. You quoted me as saying, "Yes, if the FoV is 10 foot wide or less." To which you responded, "try placing a camera a good distance away from a scene, then zoom in with a 5-50mm lens."

Please re-read. My point is about width. Sure, you can get a 50mm or 80mm lens and see far away from the camera but the FoV width limitation will remain.

I don't necessarily agree with a 10' width limit. More importantly, zooming in effectively shortens the perceived distance between objects that are actually quite distant from each other in the plane of the lens. Hence the hallway "criss-cross" cameras (though it is a bit disconcerting at first to watch people walking at normal speed but either shrinking or growing in size at a relatively slow rate).

That's fine, Carl. You can choose 12 or 15 feet, etc. but there's a real limit to how wide you can go and get details with SD.

The other issue with long length lense is DoF limitations, which are essentially irrelevant at 5mm but becomes a big deal at 50mm.

We haven't noticed. The applications where we use the "criss-cross" effect apparently don't have any issues with depth-of-field. Perhaps we have enough light?

There are other useful applications for the effect, including supermarket aisles.

Carl, it's primarily a factor of the length of the lens. The longer the lens and the deeper you need to view, the more appearant it becomes. The only other factor I suspect is whether you use manual iris lens and close down the F stop. That would increase the depth of field.

  • We have plenty of light
  • The light level doesn't vary
  • We almost always use manual-iris lenses and close the iris until the picture looks its best
  • We focus for best picture across the entire length (yeah, it might be slightly soft at the extremes but not really noticeable)
  • We use #5 welding glass when focusing outdoor cameras

The biggest problem we encounter with long focal length lenses is bubble distortion where the dome acts as part of the lens. It makes focusing difficult.

The biggest problem we encounter with long focal length lenses is bubble distortion where the dome acts as part of the lens. It makes focusing difficult.

That's one kudo I have to give to Arecont: at least some of their cameras we installed came with a small plastic "bubble" that matched the curve of the dome and could be snapped on the end of the lens for focusing, specifically to compensate for this problem.

Something with AF or ABF is a good way around it as well - the Panasonic domes with ABF have about a five-second delay between hitting the ABF button and the ABF process starting, which gives plenty of time to slip the dome in place.

Carl, yep that combination will do it - deep field of view plus long lens. The challenge is many users don't have 24/7 high light levels in their hallways (plus many/most cameras today are auto-iris).

Yes, that is, and will continue to be, a problem. Personally I don't care for auto-iris lenses in an indoor casino environment. That's one reason we like the Ganz ZCD-5000-series cameras. One of the few integrated domes that use CS-mount lenses. We also often set auto-iris cameras to manual in the menus, then set shutter to 1/30 and close the iris until the picture looks its best.

Matt, I guess you are right about a megapixel lens working with an analog camera - I can't think of any reason why it wouldn't work as long as I use a lens that's a workable match for the sensor size.

Yup, and of course, matching the mount type. As John notes, they are more expensive, but then, ultra-wide analog lenses are pretty spendy too. In such an instance, it may pay to think of lenses the way many DSLR users do: as investments - pay more for the MP lens to go on your analog camera, with the thinking that it could be re-used later should you upgrade that location to a MP camera.

Ray,

The Computar manual-iris 1.8-3.6mm lenses are much cheaper than Theia - on the order of $120 each last time we bought them.

Matt, though I don't condone the practice, I understand at least some casinos use SD lenses on their MP cameras. That's an experiment I want to try sometime down the road. I feel it's a good bet at least some SD lenses may be suitable.

as far as i know all dahua dvrs support motion detection with multi box zones. i was just referring above that if u used the rtsp stream on a 3rd party nvr

Carl, most of my wide-angle and telephoto lens work is outside, so I need the auto-iris capability. But it is good to know about the Computar lenses as I have some potential projects that are mostly indoor camera applications.

Ray, why do you need auto-iris capability outdoors? If the camera digitally and dynamically controls the shutter (like most modern cameras do), you should be fine.

John, first of all I should have said P-iris, because that's what we're using for our parking lot applications, which is what I was thinking of when I wrote my comment. With P-iris we're able to set the preferred f-number and then the P-iris functionality goes from there. So to put it one way it's automatic iris control over which we have some influence. I don't know the engineering details of how it all works. As long as it works in the field, clients are happy and so am I.

When the P-iris technology first came out with the Kowa lenses, we had a few long days in the field with multiple cameras (two of each camera, one with auto-iris and one with p-iris for comparison) driving vehicles around the parking lot at various speeds, also with people walking, doing our best with manual settings and varying light conditions. P-iris lens won out always against the auto-iris lens, and we couldn’t find manual settings combinations that performed as well as the automatic settings did.

I don’t have my notes handy on the two sets of testing we did, but in total we tried 1.3, 3 and 5 MP cameras using the camera default lenses and the Kowa lenses but not on all cameras. In the low light conditions we had challenges relating to focus and sometimes motion blur (cars speeds were up to 45 MPH in the entry and exit lanes), but I don’t remember what the camera setting and lens combinations were where we had issues. The main think I remember is that we had better performance in the low light conditions with the P-iris lens.

[Note: I edited this to clarify because it was too long and rambling.]

Ray, interesting feedback. I am surprised there was motion blur. If it was daytime, certainly the shutter would be fast simply because of the high light levels. If it was night time, both cameras (P and auto iris) would have to be fairly slow simply deal with the lack of light. I wonder if there was an issue with the specific auto iris cameras being used because motion blur is not typically a problem.

Also, if the lens was anything under 10mm, the depth of field should be nearly infinite even if the lens was all the way open.

I can believe there were some benefits with P iris in a parking lot but motion blur and depth of field seem to indicate some other issue.

Focal length was 20mm and the focus issues were at the near side of the depth of field, during low light conditions in color mode. I don’t remember where the motion blur came in except that it was also during low light conditions. The tests were two years ago, and we’ve done additional testing since, and I think I was mixing up some of the later testing (at about 35mm focal length) with the earlier tests.

We have more tests coming up in a couple of months at another site (mostly for wide-angle lens and WDR funcionality), and if time permits we’ll play around with this more and see if I can get more data on the original challenges and solution details. It bugs me that I can't remember the details enough and didn't pay attention to where the field notes went.

20mm makes sense to see Depth of Field problems.

Btw, if there was motion blur at night, it's probably the shutter speed and not the iris, since all controllable iris cameras go to 100% open in the dark.

Let us know how future tests go!

It was dawn and dusk rather than full night, and could well have been a shutter speed issue. We also added white LED lighting to the entry/exit lanes, because traffic lights at those lanes were causing havoc with camera automatically controlled settings in IR mode at night. Red light was very intense to the camera vs. yellow and green. We added white LED lighting and went to color mode at night. So it's a complex picture and I can't tell you how glad we were to have a successfulll deployment and forget about it (until this post!!!).

The price is a combination of monetary effects.

At my previous employer, I was in the chip design world. I know for a fact that it takes a lot of $$ (millions) to make a single ASIC chip device. If one considers the cost of the specialty ASICs used to grab the incoming analog stream and intelligently digitize it and then be able to pass it downstream to the application, it becomes a volume calculation as mentioned in John's #2 theory.

The vendor that put up the $$$ to make the speciaty ASICs needs to recoup that cost and make a profit.

Of course I am assuming that the ASICs used in the encoder devices have more intelligence and capability that what we see on the basic capture cards.

I think I will take one of ours apart in my lab to see what is different inside.

Add to this that encoders are using custom packaging and PCB integration compared to being able to use inexpensive PC chassis, one can see additional costs here.