Enhancing audio from VHS capture [Archive]

yup

6th August 2010, 08:34

Hi all!

Please advice how can enhance audio from VHS capture. I have original tape (not copy). Capture using Pinnacle Movieboard 700 in VirtualDub. Tape was recorded by Hitachi VM-2980E PAL camcorder which have only mono audio track.
I try Sounforge 9, Audacity 1.3 and Audition 3 Noise reduction, when I increase noise reduction I see artifact. Better result was using Noise Reduction with small reduction and after Adaptive noise reduction in Audacity. Please advice step by step way for enhancing audio.

With kind regards yup.

Ghitulescu

6th August 2010, 10:30

Is CoolEdit not good enough for you?

http://forum.audacityteam.org/viewtopic.php?f=16&t=33191#p89653

pandy

6th August 2010, 16:04

describe in steps what are You doing - personally i recomend Adobe Audition (this is old CoolEdit with few changes) - remember that normal noise reduction in those apps is made as a subtraction noise spectrum from signal spectrum - if in noise spectrum You have some signal it is also removed from signal thus You introduce some artifacts.

Also You can split recording for parts - sometimes noise spectrum slightly change with time so sometimes is much better to take noise spectrum from few places and denoise locally.

GrofLuigi

7th August 2010, 00:06

Very primitive and probably wrong, but I'm happy with the results (much better to my ears than any complex noise reduction):

Audition (or any other audio editor) > Spectral View > cut the top part off

I don't know if it's only my VHS recorders, but I clearly see a double distorted line where the cutoff frequency is supposed to be. No thinking, no calculations - just cut as close to the line as possible (including it).

Probably some harmonics are hurt in the process, but A LOT OF noise is too. :)

GL

Ghitulescu

7th August 2010, 08:52

It may work well for hiss, but the brumm (low freq. noise, usually interferences from mains due to bad grounding) is not so easy.
However we have yet no clue what kind of noise the OP talks about :p

yup

9th August 2010, 12:25

Hi all!
Thank You for reply.
Now I get good result with adaptive noise reduction (light noise reduction) I only set High Quality Mode.
Main problem with Noise Reduction is finding sample pure noise. I think that noise changing during recording and Adaptive Noise Reduction work better, but slower.
yup.

2Bdecided

10th August 2010, 14:02

This is far better than Cool Edit Pro's noise reduction...
http://www.sonycreativesoftware.com/noisereduction
...but I've never tried the newer NR in Adobe Audition, so don't know how it compares with that.

Cheers,
David.

yup

11th August 2010, 05:35

2Bdecided

11th August 2010, 10:43

I must try this Audition 3 adaptive NR.

The NR I linked to isn't the one built into Sound Forge, it's a plugin. In Mode 0 it's similar to CEP's NR, but in Mode 2 it seems to avoid artefacts. It's not exactly state-of-the-art though.

Mains supply hum is 50Hz + harmonics, but similar NR should be able to remove it to?

You can often get less hum on VHS mono transfers by moving the VCR away from all other equipment, or using a different VCR. Cables and earthing can also be an issue.

Cheers,
David.

pandy

11th August 2010, 13:08

2Bdecided!

I try also SoundForge NR with the same result. Now my roadmap Adobe Audition 3 : covert to mono (because original tape recording is mono), remove clip, normalize at -3dB, adaptive noise reduction (preset light noise reduction at High Quality Mode) and normalize to 0dB. I listen small artifact especially at low level sound, but result satisfactory.
Now I try find way remove mains supply hum.

yup.

recording in stereo mono track then converting to mono without reducing level before conversion but after (valid for conversion from 16/32 bit stereo to 32 bit mono) can give some small (3dB) noise reduction.

remove hum shall be quite easy - just use notch filters for 50/60Hz and multiples (50/60Hz depend from main power frequency of coz).

IMHO there is still field to use normal noise reduction - just select very short FFT and very small, signal free area to create spectrum pattern - signal free area shall be at least 2x number of samples for FFT (to avoid spectrum leakage) - to find such small area signal free You can use adaptive noise reduction or normal NR, search for this area, mark are then undo previous NR and create spectrum pattern - best is chose block at ZERO crossing points there special option in EDIT - to choose area inside or outside current selection on ZERO crossing points, also DC bias must be removed from recording before any FFT based NR.

markanini

12th August 2010, 03:42

I think Izotope RX advanced is the superior audio denoiser these days. It's the only audio denoiser I know of that has adaptive NL-means and DFT algorithms. Just like the video counterparts results are great and processing time is long.

Ground hum ids almost always more complex than a series of notch filters can handle. I let the denoiser take care that.

Whatever you do be careful not to remove too much noise or the sound will be dull. Listen to 'Output only noise' or simmilar if to make sure not not removing any relevant sounds along with the noise. Good headphones also help.

Ghitulescu

12th August 2010, 07:39

I think Izotope RX advanced is the superior audio denoiser these days.

Well, it should be, for those 1200€ it costs :p

pandy

12th August 2010, 08:29

Ground hum ids almost always more complex than a series of notch filters can handle. I let the denoiser take care that.

Notch filtering shall help and it is safer than pure spectrum substraction - substracting to much spectrum from signal introduce new artefacts... so key is remove DC bias, ground loop ad harmonics hum before FFT. Sometims is better to separate noise reduction for a few steps - starting on very mild settingd, reduction not more than 10 - 20dB and search in next steps for more fine settings.

yup

13th August 2010, 09:29

pandy!
Yes if converting from stereo to mono we decrease noncorrelated noise by factor 1/sqrt(2). It is no decreasing for mains harmonic.
Also thanks to for advice about Izotope RX, today I try denoiser, work better than other tools which I use before.
yup.

pandy

13th August 2010, 13:22

yup, obviously! - only uncorrelated noises are lowered. Hum is completely different thing, But at first You must have sureness that hum is not introduced by Your recording setup - all devices must be connected to the same line (same power bar?), all devices properly grounded (verify that socket on wall and all electrical cabling is 2 or 3 wire - 3 wire is absolutely must to avoid hums) - adding additional wire that joining both grounds (computer and VCR) can help reduce hum. Maybe galvanic isolation helps to... (i don't know Yours setup but very frequently external cable network introduce lot of hums - so... disconnect all cables, mute all sources on sound card muxer etc) - highly recommend record audio through only audio cable connected by audio transformer (IMHO not some silver special audiophile but ordinary taken from old modem)

yup

13th August 2010, 15:57

pandy!
My bench for video capture relatively good. I am using on-line UPS, isolating transformer, all cables from Ixos. I am try capture during my VHS do not send signal to capture card I see noise near -80dB. Main problem is VHS tape during recording use standard AC adapter. Also problem that speed tape not very high and we can not record with high SNR.
Now I investigate Izotope RX and think I get good result.
yup.

lordsmurf

14th August 2010, 14:01

Running everything through a mixer (hardware) between VCR and digital converter is most important. Get your gain and levels correct first. A decent mixer has some EQ, too, so you can pre-filter the high registers to top off the hiss a bit. After that, then use some software to fix what hardware generally cannot.

You might find these two links very helpful for this task:
- DVD Project Help: Restore, Filter, Improve Quality (http://www.digitalfaq.com/forum/forumdisplay.php/restore-filter-improve-17.html) / http://www.digitalFAQ.com/forum/forumdisplay.php/restore-filter-improve-17.html
- Video Guides: Restoring Audio and Video Quality (http://www.digitalfaq.com/guides/video/index-restore.htm) / http://www.digitalFAQ.com/guides/video/index-restore.htm

I restore audio almost daily, and have for many years now.
The best combo of software is Sound Forge + Audacity. For some really tough issues, Diamond Cut Live Forensics is needed. Cool Edit / Audition is never needed, although I have it (several versions worth). DC Live has proven itself more powerful than Izotope. Goldwave is good for some things, but Audacity does a lot of basic NR better (no artifacts).

You almost NEVER get everything done in just one program.

Cheap VCRs, or even some pro VCRS, add hum from the power supply. You can't fix this, aside from a new VCR (for consumer), or replacing power supply (for pro decks).

Good luck. :)

Ghitulescu

14th August 2010, 16:35

Recording audio on PC with a non pro card will always bring hum, from little noticed to obviously present. Also some other noises (you know, the interior of a PC is a hell of RF). USB solution doesn't solve this unless pricey.
The only solution to remove it (if it wasn't already there) is to use DC-decouplers (some mixers have). Sometimes the hum is the result of a bad impedance adaptation, a DI-box may help here.

2Bdecided

16th August 2010, 11:18

Running everything through a mixer (hardware) between VCR and digital converter is most important.With respect, unless the audio is clipping on the way into the sound card, or desperately quiet, what you're suggesting is a complete waste of time.

It's not like we need to set the level carefully to make use of the full 96dB dynamic range when capturing a hissy mono VHS soundtrack with approx. 48dB SNR!

I agree with using multiple software to get the best results, but as pointed out earlier in the thread, there are some new "expensive" plugins that are far better than what's gone before.

Cheers,
David.

2Bdecided

16th August 2010, 11:30

pandy!
Yes if converting from stereo to mono we decrease noncorrelated noise by factor 1/sqrt(2).For mono VHS this isn't necessarily an advantage.

Either the machine has only a mono head, and outputs two identical signals, so summing them does nothing.

Or the machine has a stereo head (each channel being half the width of the mono head), and summing them does as you say, but just get the noise down to the level you would have got from a mono-only head anyway!

The same is true for cassette tape, but no one has mono-only heads any more, so summing stereo to mono is the only choice you have. Problem is, any phase errors cause horrible high frequency loss - you need to get the azimuth spot-on, and then use something like tape restore live to iron out any inconsistencies before summing to mono.

Cheers,
David.

yup

16th August 2010, 13:24

For mono VHS this isn't necessarily an advantage.

Either the machine has only a mono head, and outputs two identical signals, so summing them does nothing.

Or the machine has a stereo head (each channel being half the width of the mono head), and summing them does as you say, but just get the noise down to the level you would have got from a mono-only head anyway!

The same is true for cassette tape, but no one has mono-only heads any more, so summing stereo to mono is the only choice you have. Problem is, any phase errors cause horrible high frequency loss - you need to get the azimuth spot-on, and then use something like tape restore live to iron out any inconsistencies before summing to mono.

Cheers,
David.

David You advice use only one channel from mono VHS tape, not using averaging?
yup.

Ghitulescu

16th August 2010, 19:13

Either the machine has only a mono head, and outputs two identical signals, so summing them does nothing.

Or the machine has a stereo head (each channel being half the width of the mono head), and summing them does as you say, but just get the noise down to the level you would have got from a mono-only head anyway!

The same is true for cassette tape, but no one has mono-only heads any more, so summing stereo to mono is the only choice you have. Problem is, any phase errors cause horrible high frequency loss - you need to get the azimuth spot-on, and then use something like tape restore live to iron out any inconsistencies before summing to mono.

First of all, the audio heads on any VHS are factory-"azimuthed" and the average joe has no possibility to change this either. The same goes for the HiFi heads (well, one can adjust them after "pickhammering" the resin out, with 99.999% chances to kill the heads). Besides, the linear track of a VHS is generally upped at 9 kHz (ok, some may go up to 11kHz).

Secondly, there's a subtle difference between a stereo track and a mono track and how these are read out using mono- bzw. stereo-heads. Exactly as it was in that example of yours. If the original track was stereo (it is not in the case of the OP), then reading it with a mono head would pick up the "band noise" of the guard track (that unused part between the L and R tracks), sorry I don't know the exact terminology.

The way I do is to use a mixer, which has the four inputs (2 stereo and 2 HiFi), well, I can't help it, all my VCRs are stereo (sold the mono long time ago), which are then mixed down to 2 tracks, using some equaliser, noise reduction and exciter (all in hardware) in between. I never experienced humm nor exaggerated noise, so the hardware way seemed to me ok (no PC needed). Also I don't use expensive scamware (like noise filtering mains they like to sell it for hundreds of €), I am confident that once the ground loop is removed, the power supplies should be able to filter all the noise out.

pandy

17th August 2010, 11:14

Either the machine has only a mono head, and outputs two identical signals, so summing them does nothing.

no - two independent amplifier channels, two ADC - uncorrelated noise will be lowered anyway - only recorded noise remains on the same level - but tape noises shall be lower (size of particles on tapes, orientation etc)

Or the machine has a stereo head (each channel being half the width of the mono head), and summing them does as you say, but just get the noise down to the level you would have got from a mono-only head anyway!

noise from mono head should be slightly lower than from stereo

pandy

17th August 2010, 11:17

David You advice use only one channel from mono VHS tape, not using averaging?

So You loose energy stored in audio track (using less than half of total recording energy stored in magnetic domains)

2Bdecided

17th August 2010, 12:32

David You advice use only one channel from mono VHS tape, not using averaging?If your machine has a linear stereo head, you should average. If your machine has linear mono head but stereo connectors, is makes absolute no difference.

Note: linear stereo heads are very very very rare in Europe. There are mono only machines, and machines with VHS hi-fi stereo + legacy mono head. Linear stereo heads were more common in the USA, but still hi-fi stereo + legacy mono is more common.

linear stereo = original VHS mono audio track, split into two even thinner tracks, giving hissy stereo. About 40dB SNR. Often with Dolby NR to improve it a little.

hi-fi stereo = FM and companded with pre+de-emphasis, recorded "under" video signal across almost full width of tape, decent quality (unless there are tracking errors). About 90dB SNR. No Dolby NR needed!

Cheers,
David.

2Bdecided

17th August 2010, 12:38

First of all, the audio heads on any VHS are factory-"azimuthed" and the average joe has no possibility to change this either.Which is why I mentioned azimuth only in the context of audio cassette ;)

The same goes for the HiFi heads (well, one can adjust them after "pickhammering" the resin out, with 99.999% chances to kill the heads).Azimuth of Hi-fi stereo heads is a different issue altogether.

Secondly, there's a subtle difference between a stereo track and a mono track and how these are read out using mono- bzw. stereo-heads. Exactly as it was in that example of yours. If the original track was stereo (it is not in the case of the OP), then reading it with a mono head would pick up the "band noise" of the guard track (that unused part between the L and R tracks), sorry I don't know the exact terminology.Yes, agreed, though this noise is far less than background noise of actual recording, so is unlikely to contribute audibly. Wouldn't choose to play stereo in mono anyway (unless tape had linear stereo sound only, and no linear stereo heads to play it back).

The way I do is to use a mixer, which has the four inputs (2 stereo and 2 HiFi), well, I can't help it, all my VCRs are stereo (sold the mono long time ago), which are then mixed down to 2 tracks, using some equaliser, noise reduction and exciter (all in hardware) in between. I never experienced humm nor exaggerated noise, so the hardware way seemed to me ok (no PC needed). Also I don't use expensive scamware (like noise filtering mains they like to sell it for hundreds of €), I am confident that once the ground loop is removed, the power supplies should be able to filter all the noise out.Agree with all of that - but unless you have some very expensive hardware, good software noise reduction is the most effective way of reducing VHS mono (or linear stereo) hiss IMO. For mains hum, best not to pick it up in the first place, like you said.

Cheers,
David.

2Bdecided

17th August 2010, 12:40

no - two independent amplifier channels, two ADC - uncorrelated noise will be lowered anyway - only recorded noise remains on the same levelTape noise is typically 40dB higher than amplifier and ADC noise. Not sure how the 100x less noise from the amplifier and ADC is in anyway relevant.

Cheers,
David.

Ghitulescu

17th August 2010, 15:37

Azimuth of Hi-fi stereo heads is a different issue altogether.
Yes, one can adjust the tracking only. With a decalibrated HiFi head drum, one needs two passes, one for video and one for audio (as the tracking may be different).
Yes, agreed, though this noise is far less than background noise of actual recording, so is unlikely to contribute audibly. Wouldn't choose to play stereo in mono anyway (unless tape had linear stereo sound only, and no linear stereo heads to play it back).
That is also true.
but unless you have some very expensive hardware, good software noise reduction is the most effective way of reducing VHS mono (or linear stereo) hiss IMO. For mains hum, best not to pick it up in the first place, like you said.
Good points anyway. I don't exclude apriori the possibility of having a noisy tape, just I haven't seen one*. For normal hiss (no hum) the hardware way is ok to me, maybe not perfect, but it works in real time (yes, a bit expensive).

The OP still hasn't explained what kind of noise he has, nor posted a sample.

*I've seen distorted sounds on tape (overmodulated), sound that was AGCed, decalibrated HiFi sound, but no noisy tapes. A noisy tape may have been subjected to multi-generation copy, in this case the video has itself problems. Should this be the situation then a processing on PC is badly needed, because only the software can be so largely customised.

**I know that I was (still am) a bit picky concerning the hardware I buy, but I never realized that stereo/stereo VCRs are so rare. All my VCRs (5) are stereo/stereo.

2Bdecided

18th August 2010, 12:57

**I know that I was (still am) a bit picky concerning the hardware I buy, but I never realized that stereo/stereo VCRs are so rare. All my VCRs (5) are stereo/stereo.Maybe not in Germany. In the UK they are. But then we have few S-VHS machines too compared with Germany.

Cheers,
David.

pandy

18th August 2010, 17:13

Tape noise is typically 40dB higher than amplifier and ADC noise. Not sure how the 100x less noise from the amplifier and ADC is in anyway relevant.

Yes - however we are interested in highest possible quality of recording then with modern hardware record in stereo even 40dB SNR mono is reasonable. Also using highest possible sampling ratio even if useful bandwidth is max 10 - 11 kHz

BR
A.P.

Ghitulescu

18th August 2010, 17:52

Yes - however we are interested in highest possible quality of recording then with modern hardware record in stereo even 40dB SNR mono is reasonable. Also using highest possible sampling ratio even if useful bandwidth is max 10 - 11 kHz

Definitely.

Even if the integrated mic (of the camcorder) is not able to capture a higher sound spectrum and the AGC reduces the dynamic quite substantially.

Due to processing in software, one must ensure that there's plenty of headroom on both sides, to avoid clipping when applying the effects (-18dBfs is not only an urban myth here).

pandy

19th August 2010, 11:08

-9 - -12dBFS should be safe and sufficient - especially if we intend to add dither at the final stage, careful choose for dBFS is vital to quality.

yup

23rd August 2010, 12:31

2Bdecided!
From manual, my VCR have one non-movable head for audio.
Using Audacity I plot spectrum for left channel, right channel and mono (50% Right and 50% Left)
see picture
http://forum.doom9.org/attachment.php?attachmentid=11419&stc=1&d=1282562874
We can not see difference between 3 spectrum.
Now I try investigate DC Forensic also. Both DC Forensic and Izotope give better results than other program, Izotope also use multi-threaded CPU (100% dual core, 67% quad core).
yup.

Ghitulescu

23rd August 2010, 14:09

Mono track read with a mono head and therefore transferred using a mono cable. Splitting the signal into 2 (for left and right channel, I'm not currently aware of single channel recording cards, other than in the pro domain) would inherently increase both the noise and the artefacts. Use one channel only and discard the other one (duplicate track). You may test which channel might work better (in terms of noise and/or artefacts, including distortions, hum, interferences) but for VHS mono sound it would play no role..

lordsmurf

11th September 2010, 12:30

With respect, unless the audio is clipping on the way into the sound card, or desperately quiet, what you're suggesting is a complete waste of time.
It's not like we need to set the level carefully to make use of the full 96dB dynamic range when capturing a hissy mono VHS soundtrack with approx. 48dB SNR!

No. Don't agree. The mixer lets you manage the values of the audio.

If it were distorted, for example, a mixer would at least reduce the over-amplified signal to more sane levels. Otherwise the soundcard would continue the over-amplified signal, possibly (and likely) increase the gain along the way. Once the levels are dropped, then you can work on removing distortions.

A good mixer can help with hiss, muffling, over-bass, over-treble, etc. Help, not cure. Help.

External audio hardware is very important in the overall processing and restoration of audio. Only having software puts you at a disadvantage overall.

Ghitulescu

11th September 2010, 13:35

A good mixer can help with hiss, muffling, over-bass, over-treble, etc. Help, not cure. Help.

A good mixer is like a good TBC, both won't alter the signal. What you're talking about is "signal processing" and normally is done outside the mixer (that's the reason for INSERT I/O). It's like processing colour with a TBC, it's not its job but it may come handy. ;)

Nevertheless I still wonder what kind of noise and distortions our OP has that brought us into such an academic debate.

lordsmurf

11th September 2010, 13:50

Well, yes, I do refer to the added functions found in the higher end (and more costly) mixers.
EQ adjustments, for example.

Emulgator

12th September 2010, 17:03

yup, if it is hum only, not noise as you mentioned, a suggestion for any audio editor:

http://dvd-manufactur.de/files/2010 09 12 Mains Hum Filter.png

yup

13th September 2010, 08:45

lordsmurf!
Advice hardware equalizer cost effective. For video I use TBC CTB-100. Before I mainly think about video quality.
Emulgator!
I have not only mains harmonic noise, but also buzz and continuous random noise.
yup.

NoX1911

13th September 2010, 16:57

@yup:
What about posting a small part of your untouched audio sample?

yup

14th September 2010, 06:42

@NoX1911!
No problem. Please!
http://www.sendspace.com/file/a1oq44
This small part at begin tape. Lecture on Russian.
yup.

NoX1911

14th September 2010, 16:44

Not sure where you want to go but i think the standard methods are quite passable. NR artifacts are low/moderate. Frequency cutoff is at ~11kHz so you won't notice a big treble loss as well.

select1 - Audition noise pattern reduction 100%.7z (http://rghost.net/2626510/private/e31660deda761503b2a947a5926e8f83)
(every 2 secs alternating org/nr)
http://www.abload.de/img/selectfjt3.png

yup

14th September 2010, 18:10

NoX1911!

I get error 400 Bad Request when try download 7z file.
yup.

NoX1911

14th September 2010, 18:35

Try this:
select1 - Audition noise pattern reduction.7z (http://rghost.net/2626510/private/e31660deda761503b2a947a5926e8f83) 1.48MB
(Mirror: http://www.sendspace.com/file/n5rfad)

The '%' char in the filename corrupted the link.

2Bdecided

24th September 2010, 10:55

No. Don't agree. The mixer lets you manage the values of the audio.

If it were distorted, for example, a mixer would at least reduce the over-amplified signal to more sane levels. Otherwise the soundcard would continue the over-amplified signal, possibly (and likely) increase the gain along the way. Once the levels are dropped, then you can work on removing distortions.

A good mixer can help with hiss, muffling, over-bass, over-treble, etc. Help, not cure. Help.

External audio hardware is very important in the overall processing and restoration of audio. Only having software puts you at a disadvantage overall.Sorry, but this is utter nonsense.

The input level needs to be high enough to sit above the noise floor, and low enough not to clip. That's the only external processing you need - getting the level right.

Everything else can be done in software. Lots of things can only be done in software.

You are totally deluding yourself if you think you have some audio restoration hardware that can't exist in software.

Cheers,
David.

2Bdecided

24th September 2010, 11:11

Also using highest possible sampling ratio even if useful bandwidth is max 10 - 11 kHzUnless you have dodgy hardware, save time and trouble and record at the rate you need for the delivery medium: 48kHz for DVD or 44.1kHz for CD.

It would be utterly pointless using 192kHz or 384kHz (or whatever the "highest possible" is on your setup) for mono VHS!!!!!

Cheers,
David.

Ghitulescu

24th September 2010, 13:12

Except for VCD all other media need 48 kHz (as contrasting to 44.1).

lordsmurf

25th September 2010, 01:58

2B, you state that you disagree with what I wrote...
Sorry, but this is utter nonsense.
... then said the same thing I did with different words as your rebuttle. Seriously?
The input level needs to be high enough to sit above the noise floor, and low enough not to clip. That's the only external processing you need - getting the level right.
I don't think you read what I wrote. Don't skim.

Again, you cannot fix some errors in software. Correcting levels is necessary BEFORE it's turned into a digital file.
Or, in your own words:
The input level needs to be high enough to sit above the noise floor, and low enough not to clip. That's the only external processing you need - getting the level right.
... which is what I've been writing here all along.

But you can also further take advantage of a good EQ in a mixer, to pre-process audio. It DOES helps, and will often exceed software-only work. There is a reason massive audio boards are still in use -- this is it.