Wednesday, August 17, 2016

The CEDAR DNS 2 Portable Dialog Noise Suppressor --- The Sound Of Silence

There were a few “Wows!” and even a few fluttering hearts at NAB 2016 when Independent Audio debuted the new CEDAR DNS 2 portable dialog noise suppressor. It was followed by a stinging sensation as a result of the $4K USD price tag. CEDAR has the well-earned reputation for making expensive ($10K to $50K, USD) but effective audio restoration and forensic noise reduction software and hardware devices and winning awards for them since 1989. 

Fraser Jones and Independent Audio Staff
Fraser Jones, head of Independent Audio, a distribution company here in the USA, has always had a keen eye and ear for audio gear, (and recently some video gear) as demonstrated by the names on the Independent Audio line card; Audessence, Audio & Design, Audio Developments, Coles, DACS, Merging Technologies, Pearl, Sonifex and Thermionic Culture.  Not what you might consider mainstream, but slightly boutiquey. It was no surprise to see CEDAR on that list. 

Eric Toline
Professional Sound Services
A CEDAR DNS 2 was making its way back up the east coast from Eric Toline at the Fort Lauderdale office of Professional Sound Services to Independent Audio in Portland, Maine and landed here for a few days. I quickly put together an ad hoc gathering of Washington and Baltimore Location Sound pros; Bernie Ozol, Jim Gilchrist, Len Schmitz and Jonathan Cohen answered back quickly and the date was set. 

Len Schmitz
These are four top shelf mixers with many years of experience and incomparable ears. Some of us, myself included, had watched a YouTube video demonstrating the CEDAR DNS 2 and were concerned that either YouTube encoding had compromised the audio or that better adjustments on the mic or DNS 2 would yield better results.

Bernie Ozol brought a 12 V DC battery, cup and female, four-pin XLR powering rig, so we wouldn’t be tethered by the wall wart that is included in the DNS 2. The two-channel DNS 2 has identical controls for each of its two channels.

Jim Gilchrist and Jonathan Cohen listen to playback.
The DNS 2 XLR inputs can be easily hard switched for Mic, Mic with 48 V Phantom Power or Line Level. The DNS 2 operates in 48 Khz, 24-bit mode, however, with external AES sync, it will lock to 44.1kHz, 48khz or 96khz, The DNS 2 also has an AES3/AES11 digital input and AES3 XLR connections.

Analog output from the DNS 2 via XLRs, is at line level only. That means the DNS 2 has a pair of preamps. Our first concern was the mic pres in the DNS 2. All five of us judged that the preamps were as good as the Sound Devices MixPre D we were using during these tests. Although we liked the sound of the preamps, we liked the idea of having a mixer for better control between the DNS 2 and the camera. Again, we’re location sound guys; of course we want more control. Several folks thought a headphone jack would have been nice, but we got along just fine with the headphone jack on the camera and on the MixPre D.

CEDAR DNS 2 Analog In and AES/EBU I/O
Gain range on the DNS 2 preamps is +18dB - +78dB. Dynamic range: >102dB (at 36dB gain). Each channel has its own independently switchable Phantom Power supply. The line input nominal level is +4dBu with 20dB headroom. 

Gain range is a very wide -6dB to +54dB with a dynamic range of > 102dB. The DNS 2 runs on 8VDC to 17.5V DC (nominal 12V) and consumes 4.0 to 6.0 watts. The DNS 2 is about the size of a Cracker Jacks box and weighs just over a pound. It has strap poles on the face that allow it to be  connected to other gear or a sound bag. 

CEDAR 4-pin XLR Power and Line Level Analog Output
Our first chain was CMC641>DNS 2>MixPre D>JVC HM650. We later went CMC641>MixPre D>DNS 2>JVC HM650 but heard no appreciable difference. 

The analog input level was already set very well for the Schoeps CMC641, so we didn’t have to touch that. The input sensitivity level can be easily adjusted by depressing the SetUp button and spinning the control knob.

CEDAR DNS 2 Front Panel
I live in Baltimore Country. An arc of the I-695 beltway semi-surrounds me from about a mile away. As a result, I can easily hear a constant low level pad of distant internal combustion machines. More locally, the cicadas were singing nicely the morning we did our tests. The Light Rail was running and we also had the occasional car-by and flights to and from BWI Airport. We also ran a test with an umbrella sprinkled by water from a hose while micing from below. 

The “Learn”  procedure is very simple. Hit the “Learn” button. It can be turned off or left on. We chose to leave it on most of the time. The Noise Reduction control is variable from 0 to -20. After SetUp and Learn, you simply engage the filter and use the rotary control to vary the amount of noise reduction.

I couldn’t find latency figures for the DNS 2, but Cedar’s DNS1500 sports a latency figure of Ten Samples; not milliseconds, but samples. My guess is that this short a latency figure is due to the zippy SHARC processors and Cedar’s algorithm. The DNS 2 is designed with 24-bit A/D and D/A conversion with 40 bit processing resolution and 1.2 GFlops (Floating Point Operations Per Second). That may not be a lot given where computer power has gone and continues to go these days, but it’s pretty damn fast. The result; a very powerful tool.

A Schoeps CMC641 with Schoeps B5D and Rode Dead Kitty was set on a locked down boom. We shot to SD cards in my JVC HM650 HD camcorder and did a few playbacks. We didn’t have enough headphone outputs to go around, so we share shared a pair of Sony MDR 7506 and Audio-Technica ATH-M50 as we shot and took the card down to my studio to listen for more detail over my Event Opal monitors.

Our first test was determining how much cicada noise we could reduce. Even before that, engaging the noise reduction dropped the distant mostly low frequency beltway noise right out. In most of our tests, the best results were when the noise reduction knob was midscale, somewhere between three and seven. Any more than seven and we could hear artifacts or the background would drop out so completely that the voice sounded like we were recording inside. That was eerie and surprising, At some point, when reduced to excess, the voice sounded like the person was talking through a tube.

While one of us was on-camera, talking for the test, the rest of us were making side comments off mic. Those comments were louder than the background noise, but not by much. As we increased the amount of noise reduction, those voices began sounding “watery.” If you were shooting a scene in a noisy location and wanted human “walla” in the background, this would be a problem. The “walla” didn’t sound right.

There was a general consensus among the five of us that use of any noise reduction required extremely astute listening. Astuteness that might not be easy on location and being close enough to the talent to have their voice “in air” adding to what we were hearing even with the closed back Sony and Audio-Technica headphones.

Jim Gilchrist came up with a good solution if you’re recording double sound. Use a spare track on your mixer/recorder and feed the noise reduced audio to that track and perhaps feed that to the camera as well. Also record a track without the noise reduction in case post has a problem with the processed track.

We all felt that putting both unprocessed and processed tracks on a camera was asking for trouble because of the opportunity for post to simply mix them together. That not only can happen, it has happened. If you’re doing all of the work yourself, then have at it!

Next was the “Umbrella Test.” Since it was not raining, I used a garden hose. We repositioned the boom to below the talent, aimed up at the umbrella. As such the CMC641 was hearing the talent and all of the umbrella above the talent’s head quite well. Again, we were all very impressed by how well the DNS 2 pulled the low frequency background noise as well as the voice out of the rain noise. And, again, moderate settings seemed the best. 

These two tests demonstrated the ability of the CEDAR DNS 2 to learn a fairly complex but consistent noise pattern and cancel it so well without the “underwater” artifacts most noise reduction devices create when pushed too far.

In some cases, when more Noise Reduction was applied, there was a spooky kind of feeling. We knew we were outside, but the audio was so quiet that it sounded as though we were inside. Just past that there was point there was a slightly phased “tubular” sound, as though the voice was speaking through a cardboard or plastic tube in between the voice and the microphone. 

Bernie Ozol (L) and Len Schmitz (R) set up an interior shot
Our indoor test was less dramatic. We set up in my living room, the room with the most echo in the house and positioned the mic first a proper distance and then purposely two feet away to get some room ring. While not a miracle worker, we were impressed by how much room the DNS 2 could wring out of the audio. I didn’t have time to test the DNS 2 in a really large hall to see how well it might extract large room sound.

Jonathan Cohen echoed Jim Gilchrists thoughts, "This gear raises the question of how much we, as sound mixers/recordists, should be effecting the audio we record. I broached this same question many years ago to a CAS forum; where do we draw the line on 'artistic license?' Is it our job simply to record the cleanest audio tracks possible and let Post do the boosting and cutting and effecting? Lastly, and I only thought about this today, I would like to have heard how it handled clothing noise on hidden lavs. Does this magic box work well on this too?"

As we wrapped, we talked about using the CEDAR DNS 2 on the set. At $4k USD, close to the price of one high-end wireless system, how would we charge a producer for that? If we brought it along and found it got us out of a jam, we could let the producer hear the before and after. If the producer approved the noise reduction, we thought a $100 USD charge would be nominal. Would they be prepared to pay the extra hundred? If so, it would take forty uses to pay for it. If the producer already had noise reduction capabilities in post, then he/she might not want us to use it in the field, for the extra cost and that we might use it too aggressively. Our final thought was that in the right hands and in the right situations, the CEDAR DNS 2 would be a winner, even if it was in post. In the wrong hands with the wrong ears, not so much.

Incidentally, there is an eight channel, CEDAR DNS 8. The DNS 8 was originally an AC powered device, but now also has four-pin,12 V DC powering. The DNS 8 costs $10K, USD.

CEDAR DNS 8 Live Hardware Dialogue Noise Suppression

Technique, Inc. © Copyright 2016 All Rights Reserved
More from Ty Ford at