- From: Ehsan Akhgari <ehsan.akhgari@gmail.com>
- Date: Wed, 28 Aug 2013 12:34:41 -0400
- To: Karl Tomlinson <karlt+public-audio@karlt.net>
- Cc: "public-audio@w3.org" <public-audio@w3.org>
- Message-ID: <CANTur_4A4Cmd-dyr7AoUEDc=z+onep3xxk=Geo9S3bjOn=dAYw@mail.gmail.com>
(2) is what Gecko, Blink and WebKit implement, so we should clarify the spec here. I'm not sure if this adequately answers your question here. If it doesn't, please let me know! -- Ehsan <http://ehsanakhgari.org/> On Fri, Aug 23, 2013 at 12:56 AM, Karl Tomlinson < karlt+public-audio@karlt.net> wrote: > When I first read that the DelayNode "delays the incoming audio > signal by a certain amount", I imagined something like > > (1) y(t + d(t)) = x(t) > > for input x(t), output y(t) and delay d(t). > > However, I get the impression that delay lines are often > implemented as > > (2) y(t) = x(t - d(t)). > > So I think I may have been reading more into the statement above > than was intended. > > Can we clarify, please, which effect is intended as the effects > differ for variable delays of noticeable length, and for larger > rates of change in delay? > > There is some discussion on these two approaches in [1]. > > Both use cases sound to me like they could be useful, so may I > propose specifying equation (2)? > > I propose (2) because, for small rates of change in the delay, (1) > can be approximated by applying changes in the delay at a later > time. It may be much harder to apply changes in the delay at an > earlier time, if one were to try to use (1) to approximate (2). > > If the rate of change in delay becomes large, then (1) introduces > amplitude changes and possibly overlaying behaviour, both of which > may be not easily emulated using (2). However, I suspect most > implementations are not likely nor expected to handle large rates > of change in delay well anyway. > > "When the delay time is changed, the implementation must make the > transition smoothly, without introducing noticeable clicks or > glitches to the audio stream" and the easiest way to do this is to > limit the rate of change in delay. > > [1] > https://ccrma.stanford.edu/~jos/doppler/Doppler_Simulation_Delay_Lines.html > > >
Received on Wednesday, 28 August 2013 16:36:11 UTC