The No.1 Website for Pro Audio
 All  This Thread  Reviews  Gear Database  Gear for sale     Latest  Trending
Okay, so how much does hardware vs software compression really matter?
Old 17th May 2019
  #61
Lives for gear
 
Conundra's Avatar
 

Verified Member
The gap is closing all the time. It might not even be as long a wait as for ordering a Sontec or a piece of Stam hardware
Old 22nd May 2019
  #62
Lives for gear
 
Jerry Tubb's Avatar
 

Verified Member
digital ITB for band specific compression like de-essing, analog hardware for overall program compression.

best, JT
Old 24th May 2019
  #63
Gear Head
 

Quote:
Originally Posted by FabienTDR View Post
This is a bit like saying: The biggest problem of modern rockets is the shape of their sails. It sounds cool to laymen, but of course it doesn't make any sense.

A dynamics processor is something taking a value, and returning one. Instantly.

Here's a dynamic range compressor:

y = tanh(x)

What exactly is the problem of finding out the value of y?!


A compressor doesn't have to "wait" for results. It simply looks at the input, and returns an output. Maybe the best counter example, perfectly showing how ridiculous this idea of a "delay problem" is: Digital clipping.




No. There is no general problem. Zero difference to analogue circuits. Takes an input, returns an output. Simultaneously.
Y=tanh(x) is a kind of function called a sigmoid function. It has a smooth nonlinearity, and is used for a few specific mathematical purposes (a soft nonlinear function that approaches a limiting value.) These comments are general WRT analog & digital, except when 'aliasing' or 'Nyquist' is mentioned -- that is sampled digital only.

The key to the statement above is 'nonlinearity', and as such produces harmnoics (and will also produce intermod.) The intermod comes in when multiple signals are applied to the nonlinear function, summed together, thereby producing sum+difference frequencies.
With all of this nonlinearty, it is likley that the various distortions will exceed the Nyquist frequency -- and then things get uglier than just the normal nonlinearity.

A gain control system IS nonlinear, but the nonlinear aspect is not on the signal -- other than a gain contorl signal. Gain control (compression/limitiing) is not done by applying a nonlinear function directly to the audio unless clipping or distortion is desired (at least, on high fidelity audio.)

Gain control is done like this: newsignal = originalsignal * gain control.
The hard work happens in the calculation of the gain, and then what to do to the signal if the gain calculation takes a long time. This delay in the gain calculation is the reason why the signal sometimes needs to be delayed (even sometimes in the analog world, but less often than in the digital world.)

There are all kinds of signal processing advantages if the gain calculation can be allowed to take a long time -- but it is a bad thing in cases where the input and output audio must have a minimal latency delay between them.

Why let the gain calculation take a long time? Well, certain kinds of averaging can be done, in some cases to both 'sound better', be able to filter out 'ripple', and also to help avoid the Nyquist frequency in the output signal result.

The 'ripple' happens because of the signal level detection techque and the subsequent filtering of the 'level' representation that eventually becomes the gain control signal. The side effect of ripple is a form of intermod siginal & gain control that mostly just sounds bad. Also, the attack/release times (mostly the attack times) often AM modulate the audio signal in a more egregious way -- thereby producing a lot of sidebands which can very easily wrap around the Nyquist frequency -- and that is a bad thing. The shape of the gain attack especially needs to be considered in the design to avoid the aliasing issue from wrapping Nyquist. (There are techniques that are *really good* at minimizing ripple from the start, but those methods create a long delay in the control signal.)

The above is a long description why 'compression' isn't simply a nonlinear sigmoid function like tanh(x), but the sigmoid might sometimes be used in a nonlinear limiter with a soft characteristic.

John
Topic:
Post Reply

Welcome to the Gearslutz Pro Audio Community!

Registration benefits include:
  • The ability to reply to and create new discussions
  • Access to members-only giveaways & competitions
  • Interact with VIP industry experts in our guest Q&As
  • Access to members-only sub forum discussions
  • Access to members-only Chat Room
  • Get INSTANT ACCESS to the world's best private pro audio Classifieds for only USD $20/year
  • Promote your eBay auctions and Reverb.com listings for free
  • Remove this message!
You need an account to post a reply. Create a username and password below and an account will be created and your post entered.


 
 
Slide to join now Processing…
Thread Tools
Search this Thread
Search this Thread:

Advanced Search
Forum Jump
Forum Jump