Opinions - which sounds better

[TimStannard]

A school I do some work for had someone come and do a promo video (a teacher's son who is setting up in business doing this). I was going to stick it up on the school's website but I immediately noticed some issues with colour which I've managed to address to some extent and another teacher who is ex BBC sound noticed issues with the sound and offered to get her husband to work on the sound in Audition. I got the "improved" version back today, but I thought it was worse. As these guys were professionals in the industry, I'm not sure I trust my own ears any more so I throw this open to a "blind" test:

Which of the attached sounds "better"? Note there are three sections (the first goes on a bit) so please don't consider just teh first section.

Thanks for your input.

The files (6MB ea) are available at ftp.adhocvideos.co.uk

username: dd@adhocvideos.co.uk password: public

[Midnight Blue]

Appart from x being slightly louder, I couldn't tell the difference through the PC speakers or my reference headphones. Maybe my ears have gone.

[TimAndrews]

They have slightly different EQs; affecting the apparent loudness which MB hears. In my very humble opinion, I would:
a) Use an EQ to completely remove anything under, say 30htz (or higher). None of the audio requires deep bass sounds.
b) Gently use a compressor to remove the huge volume differences between the speakers; and it may also remove some of the spikes.

[Midnight Blue]

I couldn't help looking into this a little more. The beginning and end sections are fine, on either, to may ears. The middle section seems to have two peaks of hiss one at around 5k the other at around 10k. Using the Graphic EQ, not the track EQ tool. If you put a marker/point on 2.5k, 5k, 10k and 20k. bring the 5k and 10k to -inf and then make a W shape by putting the 7.5k to -6 db. I think this it takes a lot of the hiss noise out without taking any of the quality from the voice. The attachment show the before and after.

Tim sound test.mp3

[slimbob]

Hi Tim

I managed to eventually play these samples (quite slow with computers) I saved the links for 2 samples and played both separately using windows media player .
Couldn't tell much difference but I thought sample x was clearer and better than sample a if I had to choose Tim.

I then played both at the same time using 2 different players (ie) windows media player and nero player and then repeated vice versa to keep test as fair as I could.

Doing this my choice is sample x for definite (in my opinion) ..... sounds clearer to me Tim.

No idea which is which but that is my take on it anyway ... interesting to see what the other guys think.

[TimStannard]

Thanks all for your input. At least it confirms I'm not going mad. Only one of you has expressed a preference - that being for sample x. I can now reveal that sample x is indeed the original (so i was probably right to question it)

One of my problems when removing hum/hiss is identifying the frequencies at which they occur. I've looked at (freebie) software specturm analysers which i believe are the correct tools, but due to my incompetence I seem incapable of identifying the frequencies. It shoudl be easy, shouldn't it, to take a small section with just the hiss/hum and then identify the peaks?

Actually I think I've just realised the issue. What we have here is hiss - a sort of pink noise across a broad range of frequencies - rather than a hum which would be clearly defined by it's fundemental (and possibly harmonics)

Midnight, your sample reduces teh hiss well, but dramatically changes the sound - it doesn't sound "right" with the video (as you haven't seen the accompanying video, showing the environment, you couldn't possibly know that so this is in no way a criticism)

Tim- you've given me an idea - I haven't tried this (or parts of it) through the wonderful "levelator" app. Killing the sub 30/40Hz signal is a no-brainer as well. Thanks.

I'm going to play for a bit with all your ideas.

[TimAndrews]

I like this Download Sonic Visualiser as a one way to visualise the frequencies of an audio file. Load the WAV, and use menu to create a spectogram layer.
One may notice how the upper harmonics of the voices are still apparent around 10k.
Probably of no use to TimS in this particular case. But an interesting tool.
I made this one from 'sample a'. It shows the whole clip. Frequency against Time.

[slimbob]

Hope it goes well Tim ........