Martin's audio & video problem-solving page
a selection of postings to various user forums
I've been a regular contributor to user groups and technical forums for the last 20 years or so, and have written hundreds of posts offering information and advice on a whole range of computer, audio & video production topics. Trouble is, I never kept copies of any of them, which is why I've created this page.
Martin Kay - May 2010
Q Sweetening an Echo?
Have a bit of audio that has turned out a bit on the echo'y side....
A Martin writes...
If this is "colouration" from unwanted reflected
sound then generally, once added, it is impossible to remove.
Sometimes it can be improved, but it depends on the nature of
the echo, how it's been generated and the frequency spectrum it
If you have what might be called "boomy" echo, where much of the energy is in one or more narrow frequency bands, as might be caused by pronounced resonances in an enclosed space (where each fundamental resonant wavelength is the distance between two parallel walls), then this can be improved by careful frequency-based filtering (eg a parametric EQ or notch filter). Reducing the energy in these narrow bands will remove some of the wanted sound, but remove a bigger overall proportion of the unwanted sound, if indeed most of it is concentrated in narrow frequency bands. However, if the echo is much "cleaner", with a nice even spectral distribution that matches the wanted sound, then frequency-based filtering is not going to help, as in every instance you remove the same proportion of wanted and unwanted sound.
To use a parametric/notch filter to reduce resonances, the technique I suggest you try is first to set the filter to boost rather then cut, and then slowly sweep it through the frequency spectrum until you hit one of these resonances in the recording, at which point the sound will get far worse than before as the resonance is exaggerated. When that happens, try and tune-in the centre frequency to where it sounds the most horrible, then switch the filter to cut, and adjust the amount of reduction to taste. If there is a Q Factor adjustment, this will control the width of the frequency band affected, either side of the main/centre frequency. In general, use as narrow a band as you can get away with so as not to affect the wanted sound too much. However, there is no magic "one size fits all" setting for this sort of work, so let your ears be the judge of what works best for a particular recording (or part thereof, if the mic was moving and catching different resonances in different parts of a room.
Q Stuck with a Red Line in Premiere Pro CS5.5
I have imported a lot of still images into my
sequence and by adjusting the duration of each I have created a flick
book animation effect.
A Martin writes...
The bottom line is:- does it play OK? If it does,
then why worry? The yellow line indicates that the playback engine
expects real-time playback. The red line means it doesn't expect
real-time playback, but that doesn't mean it won't happen (on a system
with a fast CPU). The issue is further complicated if you have an I/O
card like those from Matrox or Blackmagic, whose drivers will sometimes
cause the red line to be set in anticipation that real-time playback is
not guaranteed, but may nevertheless happen.
Based on previous versions with red/green lines (and no yellow), you may not feel comfortable unless there is no red showing, but CS5 is different, and unless you're doing a play-out to tape (where a dropped frame in a red-line segment would be an issue) there's nothing to worry about, unless there's so many dropped frames that the preview is un-watchable. Your exports will be fine, as everything's effectively getting rendered at that stage anyway.
Q Desk Microphone - Recommendations?
I'm currently recording a voiceover in Final Cut
Pro using a lapel microphone. Obviously this isn't an ideal situation.
A Martin writes...
You could always plug it into your camera?
Assuming it will take an external input. If not, budget in a very basic
mixer like a Behringer Xenyx 502 which can feed into the line-in of your
PC, via a long cable.
Here's a check-list for recording good speech.
1) A reasonably "dead" room, acoustically, so it doesn't sound like you're in a room when the pictures might be of an exterior scene. Avoid small, empty rooms with bare walls and lots of hard surfaces. The more soft furniture, thick curtains, cushions and general clutter, the better. If needs be, temporarily hang something like a duvet across the room to damp down any resonances.
2) A reasonably quiet environment. It's pretty obvious you don't want extraneous sounds.
3) Any half-decent microphone. In my experience, a mediocre microphone in a good environment will produce a more usable recording than the world's best microphone in a poor environment (ie one with bad acoustics). Also, in that context, avoid placing the mic on a desk stand, as the desk is a hard, reflective surface. Better to stand up and use a floor stand (or even hand-hold the mic, with care) so that it's in "free air", about a foot in front of you. I don't know what your budget is, but there's a lot of relatively low-cost large-diaphram mics that are intended for studio vocals recording, like the Behringer C1/C3 and Samson C01 models, JoeMeek JM47, Rode NT1a, etc. Buy the best you can afford, and it'll probably last you a lifetime. Most of my best mics are over 25 years old.
4) Record at as high a level as possible without clipping! If it clips (and distorts) you can't fix it in the edit - just record it again. If it's way too low in recorded level then you compromise the signal to noise ratio, and you can get quantisation distortion from not using enough bits in the digital range.
Q What is the File size / Fps / Resolution dependency?
The problem I've encountered goes beyond my logic.
A Martin writes...
The answer is pretty much in the question, where
it's stated that "The only factor having much to say was bitrate".
Bit-rate is exactly that, the rate at which the data (in bits) is
generated (per second) by the codec. For example, if the bit-rate is
480Kb/s (four hundred and eighty thousand bits per second), and the
video was 20 seconds long, then the file size would be 20x480Kb (=
9600Kb). There are eight bits (small 'b') to a byte (big 'B'), and file
sizes are normally expressed in bytes (or kilo-bytes KB, or mega-bytes
MB), so the value of 9600 kilobits is divided by eight to get a value in
bytes. 9600 / 8 = 1,200KB or 1.2MB.
It is the bit-rate and duration that dictate the file size, nothing else. So what effect do you get from changing the frame rate or pixel resolution? Well, there is another equation which links those things to bit-rate, along the lines of:- Pixel Resolution x Frame-rate x Relative Quality = Bit-rate. If you want to increase the values of any of the items on the left (Pixel Resolution, Frame-rate or Relative Quality), you should increase the Bit-rate by the corresponding percentage. If the bit-rate stays the same, then increasing one of the other values (Pixel Resolution or Frame-rate) will cause a decrease in quality from the codec, and visa-versa.
Of course some codecs are better than others, particularly newer ones like H264, so simply using a different compression codec and keeping the other values the same will change the relative quality of the encoded video. Strictly speaking, the "Relative Quality" element is really "Relative Quality/Codec Efficiency", such that a more efficient codec will produce a particular subjective quality at a lower bit-rate than a less-efficient codec. However, if you're already using the best codec available for the job, it makes the equation simpler to leave it as it stands.
Q How can I improve the quality, particularly volume, of a recorded piece of audio??
Hi All, I have a problem that I need some
'sound advice' with!
A Martin writes...
Yes, I specialise in audio and, from what you
describe, the phrase that comes to mind includes the words "cat" &
"hell" and general negative connotations.
You don't describe in what way your attempt to lift the volume was unsatisfactory, but I can guess. If the sound you're left with is from the ME64 which was set up for the reading, then it was in the wrong place and probably not pointing in the direction of the vows. That gives you three problems.
1) The recording is low level. If that was your only problem, then you could simply lift the volume to the right level. It might bring up some electonic noise with it, but you might be able to reduce that with software noise filtering and EQ.
2) If the mic is "off-axis", then it will have an uneven frequency response to sounds coming from that direction, which can be difficult to correct, given the complex EQ that would be required, and the fact that any EQ you apply will affect the noise as well as the wanted sound.
3) If the mic is "in the wrong place" and further away from the source, it will pick up a lot of diffused (reflected) sound, particularly from the direction which is "on axis", to which the microphone is most sensitive. These combined effects are known as "colouration" (if it was light rather than sound, it would be the equivalent of colour fringing and colour casts on the image). To remove it, you're up against the physical laws of entropy, and in my experience it's a non-starter. EQ won't help, and there's no "un-reverb" effect that I know of. It is probably possible to enhance it for forensic purposes (ie simply being able to hear what's being said), but it still won't be a nice sounding recording - more like the comms sound on the Apollo space mission.
If it was possible to get a good recording from putting a microphone in the wrong place and fixing it in software, no-one would bother to do it properly and boom operators would no longer be required on dramas, for example, but I don't see that happening yet.
Those, sadly, are the audio facts of life, in my experience.
Q How to overlay a minutes & seconds counter?
On one of my projects, the customer has requested
a running time showing just minutes and seconds on the bottom left of
the screen as it is a training video.
Although it is our plans to change over to Final
Cut Studio in the next few months, we are still running three suites
with Avid Liquid Edition 7 and they are still working very well.
A Martin writes...
Could you not just crop it to include the part of the timecode display that you require?
Or you if can't apply the crop to the timecode effect (and I'm assuming you do have a "crop" effect?), make two copies of the footage, one on top of the other, apply the timecode to the top one, and then crop away everything apart from the desired digits to reveal the other copy underneath. I've done that in Premiere with two versions of the same sequence.
Just for the record,
this procedure will work in real-time in Premiere CS5 - no intermediate
rendering or "fusing" required - so I would be hopeful of the latest Final
Cut having similar capabilities.
Premiere allows whole timelines (or as they call them, Sequences) to be interactively nested into other sequences, hence it's very easy to add an effect to an entire sequence, and to have a second version of it (with a different effect plus a crop/mask) on that same timeline.
Q Sound balance with two mics into my camera?
As a 'newbie' I have a Sony HVR HD100E camera
which has a 3.5mm mic input that the on-board mic connects to (No
other)). I also have a Sony UPW V1 lavaliere mic receiver on the camera.
A Martin writes...
Two thoughts on this.
1) Unless you're using the most rudimentary editing software, you can adjust the individual levels of each channel in post. I think even Premiere Elements can do this through the Channel Volume effect, which is certainly present in every version of Premiere Pro.
2) If the track with the lower volume (the on-board mic) is really too low to boost, then you need to attenuate the other input so that both can be recorded at a decent level. This can be done with two resistors, costing pence, some wire and and a soldering iron. You may have to pay someone to do this, but it's not going to cost a fortune. As to what values the resistors should be (ie how much attenuation should be produced), you should look at the audio meters in your (audio or video) editing software and see what the actual difference is (in dBs) between the channels, or how much gain you have to add to the lower one to match the higher one. From that an engineer can easily work out what values to use. (This is a service that I can provide)
3D - a view from the sidelines
posted to the IOV Knowledge Base 3D forum, May 2011
|Elsewhere on this web site:-|