dbx166xl for webcasting

[Sgeirk]

I have a dbx166xl and onboard dsp eq on the emu soundcard we have. (a separate webcast)

Anyone have experience using the dbx166xl to process for a webcast? Any suggested settings? We're using the EMU1212's ability to roll off the highs, this is a 32kb/s live stream. I think the slope would be better in front of the dbx, though. Should we get a real EQ?

We're using existing gear (dbx and emu)...i'd rather buy an EQ than buy another computer to be the dedicated audio processing box.

I also have an old CRL audio signature that works well at smashing levels together, without clipping. Which would you use?

Suggestions welcome. Thanks.

 

[RadioMan203]

While I can't speak for which processor of yours is the better for your situation, I can tell you rolling off the high end going into your encoder will make a world of difference.

Back when I had an Orban Optimod 6200 I was able to decrease the output bandwidth to match the stream and it was extremely clean - more than likely since the encoder isn't wasting time trying to encode data that will get thrown away anyways. I'm using a different processor until I pull the plug on my webcast (Aphex 2020) and I notice the difference as far as transparency when running below all the processing thresholds. A simple EQ with the high end sliders pulled down placed right at the input of your encoder PC should help.

 

[OKCRadioGuy]

Wow.. That's an interesting concept. I'll have to try that with our 24k stream for the station. I never really thought about it, but eliminating wasting processing power on crunching higher audio frequencies might be a good thing to try! Thanks!

 

[ChiefOperator]

A while back, Cornelius Gould weighed in on this very issue and offered some advice. It was along the lines of reducing the bandwidth prior to the encoder, but I can't remember the exact info. Cornelius was on the board the other day; perhaps he can offer the guidance here again.

Also, I believe he has some articles posted on his website and a few posted on the Radio Magazine website listed, I believe, under "Streaming."

 

[Sgeirk]

Frank Foti also covers some of the challenges and basics of streaming, he's got a white paper up on Omnia's website.

 

[speakerman]

From all of the streaming sites I have listened to it is clear there are many ideas how to get it done. A few sites definitely sound better than others.

Take a look at this http://www.orban.com/products/streaming/ as aacPlus V2 sounds good at really low streaming rates if the audio is processed properly.

This is also a lower cost way of running a Flash server fed by the Orban encoder http://www.wowzamedia.com/

 

[RadioMan203]

The Orban info is right on both counts. The codec has a lot to do with it. I had been using MP3Pro since it's the preferred codec on Live365 (and the ill effects can be heard on many players that aren't equipped with MP3Pro). Those players with MP3Pro decoding sound very, very good but even with those players you could hear some of the phasey artifacts referred to in the article.

The Optimod-PC 1100 is very very similar to the 6200 I had. I'm sure you could turn down the bandwidth on it as well prior to the encoder. This works well to optimize what the codec churns on.

I'm working on a streaming station for a high school and given their limited budget I'm using the outboard solution of an equalizer before the encoder pc to limit the bandwidth. It's worked well in tests with both WMP and regular MP3 streams, so it can be done "on the cheap" or with existing external processing someone may have on the air. I suppose it would work if you picked up your off-air signal with a tuner and applied the same technique to limit the input to the encoder.

 

[littlejohn]

We have seven (at last count) of the Orban cards streaming in various uses. They do an excellent job. Orban went overboard helping us get them going properly when we first got them.

 

[radiorob2.0]

From all of the streaming sites I have listened to it is clear there are many ideas how to get it done. A few sites definitely sound better than others.

Take a look at this http://www.orban.com/products/streaming/ as aacPlus V2 sounds good at really low streaming rates if the audio is processed properly.

This is also a lower cost way of running a Flash server fed by the Orban encoder http://www.wowzamedia.com/

The aacPlus stream sounds incredible. However, Mac machines will not encode the "plus" attributes so they hear telephone quality audio.

 

[gtomas]

Wow.. That's an interesting concept. I'll have to try that with our 24k stream for the station. I never really thought about it, but eliminating wasting processing power on crunching higher audio frequencies might be a good thing to try! Thanks!

For a codec, high frequency content is the most difficult part of the spectrum to deal with. That's where codecs exhibit most of their artifacts like "phasing", "squishing", "watery" and (in case of SBR codecs) metallic high-end. You'll never hear such artifacts on bass.

That is why handling high frequencies is crucial in getting coded audio (especially low bitrate one) sound good. And why audio processors employ various algorithms that try to control high-end level/density/dynamics (such as PreCode and Sensus, we have yet to think of a name for ours )

Last year I had a poster on AES convention which outlined some of the basics of audio processing for coded audio (www.gorantomas.com/articles/AES_poster.pdf). In a nutshell, you want low distortion peak control, meaning use of look-ahead limiting is preferred (with low IMD algorithm). You want low-pass filtering to be able to find best compromise between high-end response and codec artifacts (this will depend on the codec used). For example, in the latest version of DSPXmini-HD we have implemented a LP filter adjustable from 4kHz to 20kHz in 1kHz steps. Tailor your overall spectral balance and especially high-end density carefully (less is often more). If possible, control stereo image and over-enhancement of L-R content. Don't set up your processor aggressively as this can exacerbate coding artifacts. Finally, be aware of the overshoots that can occur in the codec internally and set the processor output level accordingly....


Regards,
Goran Tomas

 

[gtomas]

The aacPlus stream sounds incredible. However, Mac machines will not encode the "plus" attributes so they hear telephone quality audio.

Both RealPlayer and VLC Media Player are available for Mac and they will decode and play HE-AAC streams (at least v1)...


Regards,
Goran Tomas

 

[RolfTaylor]

Wow.. That's an interesting concept. I'll have to try that with our 24k stream for the station. I never really thought about it, but eliminating wasting processing power on crunching higher audio frequencies might be a good thing to try! Thanks!

When encoding audio with a codec the very first step is to choose an appropriate sample rate. This, along with the resolution, determine the data rate before bit rate compression.

If 24 kHz sample rate is an option on the codec you are using it is a very attractive place to start

Of course in many cases additional roll-off can and should be considered, as note in this thread.

A few more points.

As Cornelius has pointed out somewhere, some sound cards do not have sufficient filtering on their inputs. From digital audio 101 we know that if all frequencies > 1/2 the sample rate are not filtered out prior to digitization, the aliasing will sound terrible, even before you run it through a codec. Corny has found that in many cases using an external filter or a good a/d before the sound card can make a difference for this reason.

If the pre a/d filtering is good and a suitable sample rate are used prior to the codec, additional filtering can often help. At low bit rates this would be in the form of a high cut (low pass) filter to make the codec's job easier. However, a gentle roll off of the high end can make a huge difference to, particularly with dry voice coming from a high end flat microphone.

I have worked with many voiceover talents where a gentle roll-off made a huge difference in reducing artifacts while still leaving some "are" intact.

As always YMMV

 

[cgould]

Hey..this is where I chime in, I guess ;D

Forgive me if the explanation is kind of rough...feel free to ask questions if necessary...in the middle of several projects at the moment...

Anyway...

If you use high quality sound cards, you are in a good place to start, otherwise, i have found a lot of sound cards that have sufficient filtering for 44.1 kHz and 32Hz sample rates, only. Any other CODEC sample rate could result in some aliasing as the 16 or 22 kHz filtering may not be sufficient.

For example, if you choose 24 kHz sampling, the audio bandwidth will be about 12 kHz, so the sound card would use the filter for the 32 kHz sample rate, which means you'd have components above the 24 kHz Nyquist, so some aliasing will occur. By the time this aliasing rolls through the CODEC, they will manifest themselves as additional coding artifacts.

Over the years, I have found some sound cards that only work well at only ONE sample rate (44.1 kHz only, or 48kHz only). Any other sample rate will have extremely inefficient filtering. It could also be that those cards have insufficient filtering, period, but it's more noticeable at lower sample rates. It's been rare that I've seen this - usually with El cheapo OEM budget computer sound cards.

Audio processors with internal selectable lowpass filtering really help with this situation.

-Cornelius