Echo suppression?
 

I am doing an audio restoration from a cassette recording made ages ago. It is a piano music (Beethoven's Moonlight Sonata) with a man reading a script ... the piano music is fine. The problem is the man's voice ... has lots of echoes in there. Any software / technique to reduce that to an acceptable level?

The background is the piano. The man's voice comes in once in a while ... to mask just the voice itself is very hard work (assuming it can be done).

Echo or reverb? If it's reverb then I've never found any tool that can remove it - as it's a time domain problem. Discreet echos can sometimes be removed/reduced by copying and pasting a time slipped duplicate, to phase it out, but if the music is playing at the same time - I suspect you are doomed!

It is echo. But, there's background music playing at the same time. Look's like this one is doomed ...

Have you got adobe audition - I have to admit that looking at the offending track in the visual editor sometimes let's you see things that can be repaired, even when you cannot do it by listening. However, if the music gets in the way, I doubt there is much if anything, you can do.

I am using Sony Sound Forge and iZoTope RX. So I am seeing the waveform as well as the spectrum. The problem is the sheer number of places in this audio file that this is happening. I can manually fix one - that takes me about 10 - 15 minutes. But, there is no way to fix if it is a 60 minute recording with the speaker talking in about 45 minutes of it.

Secret weapon of the audio engineer: SPL Transient Designer. I use it on every mix.

It has two controls: Attack and Release. Just shorten the release time and... there you go.
There are plugins that do the same thing also.

Thanks for the great hint. I found one VST plugin that works for Sound Forge ... Schaack Audio Technologies - Transient Shaper. Also two controls only - Attack and Release. Good - reduce the Release time (like what you said) - and the echoes are much less now.

Words being eaten!

I honestly could have used this to salvage recordings many times in the past - I was frankly sceptical. So I took a piano recital, and another track of a dry after dinner speaker, mixed the two, and added some church style reverb.

Exported this to a stereo audio file, and then imported this into Cubase. First tried all sorts of dynamics processors - a tad of difference, but all succeeded in making the piano articulation suffer quite badly, when the speaking was cleaned up.

Downloaded the Shaack Transient Shaper, and installed the demo. Tried it, and it works! Fair enough the piano suffers a little, but far, far better results than I'd hoped for!

I shall download the full version - an excellent plug-in.

Thanks for this topic!
Paul

Greetings Paul,

If you are not aware of the tools out there, nobody to blame for that. I would NEVER suspect that the transient shaper / dynamics tool could be used to reduce echoes / reverbs without affecting (or minimally affecting) the background music. Thanks to Arthur, I know better now. AND added a massively useful tool / technique to my audio toolkit.

You don't have to uninstall or download the full version. That version of Transient Shaper is already the full version. You just need the license key. Cheap - for what it does. Excellent.

Glad I could be of help!

I am a recording engineer/music producer by trade. The whole video side was more of a passion/hobby that is getting more serious now.

Personally, I prefer the SPL hardware version, I use the 4 channel version as "hardware insert" on my ProTools HD3 rig.
The art is to play with the release control to achieve the optimum compromise.

Also, great for giving a somewhat 'dull' recording more 'edge' or 'bite' - not only percussive instruments, but also others.
E.g. I record a lot at Abbey Road studios. In one of the sessions there, I was recording the LSO (london symphony) and the first violin lacked a bit of 'bite' for the feeling I was going for in the mix. So I gave it a bit of 'attack', just a hint, and there it was... shiny and proud :D

Cheers
Arthur

Hi Arthur,

SPL requires a h/w DSP card - which I don't have. So it is easier for me to look for a totally software DSP, which luckily, I managed to find one. And it works the same way as SPL version too.

Abbey Road (London) studios - that really brings back fond memories for me. I was working there part-time for 6 months (1980) as an audio engineer too. That time, I was playing with Ampex 1" open reel tape recorders.

Cool!
I did about a dozen sessions there, recording and producing. So you remember Collette? She is a great studio manager!

I worked with Andrew Powell, the producer for the first two Kate Bush albums and Al Stewart's year of the cat etc. He was around a lot back in the early 80s, remember him? He was also part of Alan Parsons Project.

I only recalled Collette. Yes - she is a very nice lady to work with. I was mostly dealing with the backend - things like checking the open reel recorders are in order, whether the mixers are okay, etc ....

Thanks Arthur,

I have reached out to them to see if I can get a copy for review.

Altering transients should have little effect on removing echo from poorly recorded dialog. There may be some side effect that does the job. I found that was the case with the GML noise reduction unit. It wasn't designed to reduce room reflections, but when used just a little, it helped some.

I get the impression from this forum that the people who ask about this have typically used a camera mounted mic and are too far away. I doubt that any processing can help those situations.

Regards,

Ty Ford

Hi Ford,

The SPL software requires the h/w DSP card. I found a purely s/w only implementation that sounds just as good in removing the echoes.

Sometimes, we are asked to clean up poorly recorded audio for clients ... we can't go back to do another recording. Too late. And the clients hand over a C60 cassette tape - and expect us to work miracles. Sometimes, we do perform miracles for them. No choice.

HI Ty,
You're welcome! Glad to help.

The guys at SPL (at least SPL Germany here in Europe) are pretty cool about test units. While at it, test their DynaMaxx compressor, it is excellent, as are their preamps.
There are a lot more cool units for audio for picture engineers. If you are interested, we could do a little piece about it. I remember your name from way back when, the early audio.rec.pro days I think.


Tingstern,

The unit we are talking about - The SPL Transient Designer 4 - is a 19" rack hardware unit with 4 XLR in and outs. Pure analog hardware, nothing software in it.
I think you are referring to the plugin for the Creamware Pulsar platform. While close, the hardware unit (the 19" analog rack) sounds better. A plugin is always a compromise, an imitation, and in my years experience never reached the quality of the hardware they emulate.

Hello Arthur,

Yes. i'm still hanging on at rec.audio.pro, through some fairly fierce BS, but it's still a valid watering hole if you step over the cow pies.

I reviewed the Dyna Max soon after it came out (I think).

That review and others are up on my online archive. Just for old time sake..

SPL DynaMaxx Model 9735

Ty Ford
Baltimore, Md

The SPL DynaMaxx Model 9735 ($869) is a stereo or two channel
compressor/limiter/de-compressor and gate. In keeping with SPL's tradition of
offering unique solutions for the studio. DynaMaxx provides good sounding gain
control for those who aren't rocket scientists.

The back panel is fairly familiar. There are balanced XLRs and 1/4" TRS
balanced/unbalanced jacks. Both XLR and 1/4" output jacks are hot. The
DynaMaxx also has TRS side chain jacks. There's also a Ground Lift switch, a
120/240 voltage select switch and a standard IEC mains socket. The designers
have obviously spent some time operating table top equipment and short racks,
because each jack on the back panel is labeled so that it can be read upside down
and rightside up. Nice touch.

There are no attack time, ratio, release or threshold knobs; just continuously
variable "Compress" and "Gain" rotary controls (remember the dbx 163?) and a
finely-detented noise gate knob. That means you can't dial in your favorite
settings. Instead you get two controls and two nice looking 20-digit 1dB/LED
ladder meters, one for each of the units two channels. Each channel also has a
small Signal Present LED that lights up when the signal rises above -40dB. There
is no overload indicator. Maximum input level is +24dBu. A Stereo Couple button
links the two channels together allowing control of both from the left channel.
There's also an Active button that operates a relay-operated hardwire bypass. An
illuminated power switch is mounted on the far right of the face of the unit.

The Compress knob sets the amount of compression, incorporating threshold and
ratio. The manual suggests that no peak limiter is unnecessary because, at 50
micro seconds, the transient detection is fast enough top catch the peaks and
automatically re-adjust the attack time. The input and output of the attack
detection circuit are constantly compared. If a transient does get through the first
attack control stage, a second faster stage is activated. The release time is
determined by the difference between the peak and average signal levels. A large
difference creates a faster release time. The manual reveals that a fully clockwise
Compress control yields about a 3:1 ratio.

To compress, start with the Gain and Compress knobs at zero. Turn the
Compress knob clockwise until you start to see a leftward movement of the LED
meters. They will indicate the amount of gain reduction you are applying. The
more compression you apply the lower the level of the audio. The Gain knob
controls makeup gain, so for every adjustment of the Compress knob, you have to
reset the Gain knob. The Gain knob is calibrated to the LED display, so that the
lighted part of the display moves to the right end of the display as makeup gain is
added. Basically, the Gain control lets you set the proper input level to whatever
the DynaMaxx is feeding.

To get an idea of the amount of gain reduction for a particular setting, I stopped
the music and waited for the LED to come to rest. In one case it stopped at +6dB,
which was where I had the Gain knob set. With the music on, the display pushed
to the left and ranged between -2dB and +2dB. That meant a gain reduction of
somewhere between 4dB and 8dB. The meter goes from -10dB to +8db, implying
the possibility of up to 18dB of gain reduction.

The Soft Limit mode is much more sensitive to peaks. When I kicked in the Soft
Limit switch using the same music, the display started showing a gain reduction
range from +6dB down to -10dB (16dB), with a noticeable compaction of sound.
When I reduced the Compress Setting (from 6 to 5), the track got louder so I
reduced the gain from 6 to 3. With Soft Limit in I was now seeing gain reduction
excursions from a baseline of +3dB down to -6dB, a total of 9dB of gain reduction.
Further experiments with some studio recordings of a Yamaha grand piano
yielded similar results. With Soft Limit In, I could hear some "crunch" and a bit
of pumping at some of the more percussive moments. Without Soft Limit, the
piano just sounded tight.

The Effect Compression button sets the compressor release time to 60ms, allowing
for some grungier settings, but I found that more aggressive settings were
possible with Soft Limit. The Noise Gate is rather ticklish. With or without
compression, even in a quiet studio, the gate fluttered quite a bit, opened abruptly
with a small noise each time it opened. If you're gating loud sound sources with
percussive attacks, you won't run into this problem.

Recording bare voice tracks with moderate amounts of gain reduction caused the
compressor to pull up some noise at the ends of paragraphs, but the effect was
hardly noticeable within sentences and phrases. The Noise Gate and was too slow
to open or close for spoken word recording. I could never find a setting that didn't
clip off the first moment of sound or pop open unexpectedly. Music with longer
decay trails and more continuous tones fared a lot better.

ACcentUate the poSitive
In addition to its unconventional approach to compression, the DynaMaxx also
de-compresses in a way that makes any sound that crosses the threshold jump
out even louder. Use the Soft Limit circuit and the sound jumps out even more.
This is a new subjective area for you to wander around in. I put some recordings
from the new Larksong madrigals Xmas CD I'm just finishing and set the
threshold for the early part of a verse. When the group hit the chorus they got a bit
louder, crossed the threshold and got bumped up in volume. The effect sounded
uneven and jerky. Taking Soft Limit out smoothed things out. I popped in the
amazingly compressed KIX "Hotwire" CD and found it was so compressed that
setting the de-compress control was a bit difficult because the mix was so dense
and compressed. In the verses and solo break where the tracks were less dense,
the kick drum popped out above the mix. On the chorus, where everything was
turned up to eleven, the wall of guitars jumped out followed by everything else.
The manual suggests using the de-compressor on samples that have been overly
squished and on other sources. I can hear it being used on soloed parts more than
I can on a full mix.

The manual spends a noticeable amount of words in praise of the THAT 2181
VCA chips used in the DynaMaxx; offering that they are responsible for the lack
of dulling of the processed sound and reduced distortion and noise. No question
about it. The DynaMaxx does seem to exhibit less destructive "crunching". If
you've become addicted to dynamic range control byproducts, you'll probably end
up using the Soft Limit and Effect Compress all the time.

Although the DynaMaxx can be used on individual inputs or tracks, it should also
do well hung across the main two-mix ouput of a console to give raw sounds a
more polished surface. For loading into workstations, you can use DynaMaxx to
bring the audio right up to Digital Zero without mangling the sound. Because of
its ease of operation, it's also a very good choice for people with limited audio
experience who are processing audio for the Internet.

Several words of caution for anyone starting to use compression; use less than you
think you need. It's always better to crunch more later than to over-crunch. Also,
as with many compressers the DynaMaxx will suck up the noise floor with low or
no audio present. For example, the space between the opening chords of the
Rolling Stones' "Start Me Up" was large enough to make the DynaMaxx reach
down for the analog noise at the bottom of the CD track. It did this with 8dB of gain
reduction with Soft Limit IN and Effect Compress IN or OUT. The Noise gate was
simply not fast enough to close down gracefully between the guitar chops.

IN CONCLUSION
If you can get past the idea of not having all the control parameters to play with,
there is value to a gain reduction device that can provide somewhere around 18dB
of gain reduction in Soft Limit Mode without dulling or ducking or pumping, and
get a bit nasty sounding as well.

Application: single, dual channel or stereo, analog, dynamic range control

Plus: easy to use, flexible I/O interface

Minus: somewhat slow and jittery noise gate, slightly more expensive.

Regards,

Ty