The god-damned HV20 cellphone trick

Among HV20 users there is this magical method to lock+set exposure: the cellphone trick. You place a brightly lit white/gray cellphone screen in front of the HV20 and you follow a convoluted set of actions afterwards. And this is a “trick” that many people use, doing it because they think it works. They think that the HV20 has a “magical point” of f2.8 where no gain is introduced. Well, what they actually end-up doing is under-exposing. A sensor of a camera is what it is, it can’t be “overclocked” to perform better magically.

This trick doesn’t work. It’s a fucking myth. Stop using this stupid method. This is how this guy’s video ended up looking by using this method. Here’s another test video doing the same trick. Yes, the grainy artifacts are gone on the second half of the video using the “magical method”, but this is not because the method is magical, but because the camera UNDER-exposes. You can do the same thing without going to all of these weird steps, just under-expose. Here’s another test, again, it has no grain simply because the scene is under-exposed and it’s only saved because the candles are bright enough. But you don’t need this “trick” to get there.

Instead, do either of the following as long as your camera supports setting a custom white balance (all Canons do):
1. Buy a gray card (check the link on how to use it). As long as you can get close to your shooting subject, there is no substitute for the gray card, sans a light metering device. It costs $4 and you have no excuse of not owning one.

2. If you really want to use a similar method, at least use a method that works more precisely. For example, zoom all the way in to the same level as your shooting subject and fill the frame with a white or brightly lit point (e.g. a white shirt of a person, or a wall). Then, set the custom white balance at that point, and then zoom out. Lock exposure after zooming in, and compensate based on the brightness of what you zoomed on. As a plus, you will have an “ok” white balance too.

That’s the same method used by many photographers, and it’s one that works best without extra equipment. No cellphones, no extra steps. That’s it. Amusingly, the common wisdom in still photography (which is called “expose to the right”) says exactly the opposite: never underexpose, even if you have to increase the sensitivity/gain to get a proper exposure.

Update: The cellphone trick only makes sense if the exposure compensation doesn’t follow the same program curve as the exposure. Now, if indeed the exposure compensation follows a different program, *and* if that program doesn’t change the gain, *and* if the regular program changes the gain before going wide-open, it makes some sense, but none of those assumptions are documented anywhere.

7 Comments »

Jorge wrote on March 10th, 2008 at 4:18 PM PST:

Assuming you’re on a dark location (with no means to add light) you really have 2 options:

1- shoot a “better exposed” image WITH NOISE
or
2- shoot an underexposed image WITHOUT NOISE.

I prefer the last option and I understand that the cellphone trick helps you MAKE SURE that the HV20 won’t crank up the gain (and with the trick you even have more than 10 different EV values to choose from).

I don’t think the method is stupid at all.


This is the admin speaking...
Eugenia wrote on March 10th, 2008 at 4:24 PM PST:

The “magic” method is simply not what you should be doing. An under-exposed image has loss of information. A properly exposed image doesn’t. Also, read carefully. With my suggestions you can STILL shoot SLIGHTLY under-exposed, until you get a good combination of noise/information. And you achieve that with much fewer steps than the magic method, PLUS, you get a better white balance, PLUS it does not require you to shoot in AV mode to change the aperture size. So why not do it the right way, methods used for years by photographers, and instead use a cellphone method that someone invented out of his ass, without even having the tech specs of the HV20’s internals?

When you use one of the two methods I suggest, you might end up with a bit of noise. But noise can be removed on post using a plugin like this one, or simply using color grading. My video did have some noise originally, but most of it was removed when graded (except in 1-2 scenes). Look at the scene at 1:27, the second method was specifically used.

If I had under-exposed using the cellphone trick, my video would not have had as much visual information to work with. The editor plugins would have had a hard time locking in to the visual information, as it would mostly be a blob of black stuff on screen.

The point remains: if you need a camera that doesn’t have much noise, get an $8000 camera. The HV20’s sensor can do only what it can do, no matter what tricks you do with it. Which is why is best to have a bit of noise on dark scenes (in order to secure visual information), and then remove it on post.


Dave Rosky wrote on March 10th, 2008 at 5:36 PM PST:

It seems there are two different topics intertwined here. Underexposing just to reduce noise is indeed a fallacy. If you “gain-up”, you loose dynamic range by adding noise, if you underexpose, you loose dynamic range by reducing the signal – either way you loose dynamic range. You can’t get around the laws of physics (and the laws of economics that make consumer camcorder imagers shrink every year).

But underexposing because you specifically *want* the scene dark is fine.

Consumer camcorders gain-up because Joe Sixpack wants to be able to shoot video in his living room with no lighting other than his 60-watt table lamp. The problem is that gaining up makes the room in the video look brighter than it looks in reality, as well as being noisy. That’s where I think there is value underexposing (however you might choose to do it) – I might want a dark room to actually *look* dark in my film.


Dave Rosky wrote on March 10th, 2008 at 7:03 PM PST:

“Now, if indeed the exposure compensation follows a different program, *and* if that program doesn’t change the gain, *and* if the regular program changes the gain before going wide-open, it makes some sense, but none of those assumptions are documented anywhere.”

It’s not documented, per se, but from the various descriptions on the web, it seemed clear that people had figured out that’s what’s happening – that the gain gets locked when you use the trick.

What would have been nicer (and would avoid the kludgey trick) would be for Canon to simply include a setting to not allow automatic gaining up – similar to the setting they already have for not allowing automatic slow shutter.

Of course, if they put in every possible setting anybody could want, we would have 27 levels of menus 😉


Dave Rosky wrote on March 10th, 2008 at 8:09 PM PST:

Sorry to post here so many times, but this topic got me thinking about the fact that the Hi Def camcorder market seems ripe for a migration from high-end consumer models like the HV20/30 to a new level of pro-sumer models, something similar to what is happening now in digital still cameras. Not so many years ago, high-end “prosumer” digital cameras were mostly still fixed lens cameras with small sensors. Digital SLR’s were still very expensive and cost $2500 or more. In the last few years that has been shifting. There are now many entry and mid-level SLRs with APS size sensors and extremely good image quality for around $1,000.

In terms of video cameras, I would gladly pay, say, $1500 for a camcorder similar to an HV30, but with a larger sensor (or 3 larger sensors), more pro-level manual controls (like gain, for example), and possibly a few other upgrades like a real focus ring, etc. The problem, is there is not yet such a middle ground, like there now is with digital SLRs. To take the next step up from the HV20/30, you have to spend thousands more and be burdened with a physically much larger camera.

I wonder if we will see a market transformation in video similar to what DSLR’s have done in photography. Until there is, we’ll be stuck with high low-light noise and having to use “tricks” to set gain.


Tim wrote on March 10th, 2008 at 8:16 PM PST:

I think if it’s a low light situation the best you can do with the HV20 is shoot in 24p and in cinemode. Zoomed in shots are easier to compose and thus easier to light. A little relfector could take care of some shots. Something like this camera-mounted reflector might come in handy.

It’s the wide angle shots in low light that are the HV20’s downfall. Especially if there is a lot of untextured background. My apartment with it’s plain sheetrock walls and 100 watt bulbs does not look good on video. You either have to bring in more light are choose your shots very carefully.


Jim wrote on March 10th, 2008 at 9:11 PM PST:

Do you have any experience with Neat video?
It tells you to find a blank wall in a frame to test for noise, but I never seem to find one big enough to properly sample, have you found this to be true for you or any advice?


Comments are closed as this blog post is now archived.

Lines, paragraphs break automatically. HTML allowed: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>

The URI to TrackBack this blog entry is this. And here is the RSS 2.0 for comments on this post.