also I think the change in shape at low frequencies is just because
modFM only approximates a pulse, it is not exactly one.

Victor
----- Original Message ----- 
From: "victor" 
To: 
Sent: Tuesday, November 25, 2008 6:44 PM
Subject: [Csnd] Re: new bandlimited oscillators (was: Re: UDO question)


>I had also noticed that the CSD had, for no particular reason, ksmps=1,
> so if you did not change that, I expect it would have been slow. The
> results I mentioned before were for ksmps=64.
>
> yes, the DC blocking has its issues, but that is the same with any other
> method that tries to turn a pulse into a saw or square. A few tweaks and
> it could be OK. I also had another method of removing the mean without
> actually using a DC blocker: find out the mean for a range of frequencies,
> store that on a table and just subtract that from the signal. That 
> probably
> might turn out to be better.
>
> Victor
>
> ----- Original Message ----- 
> From: "Richard Dobson" 
> To: 
> Sent: Tuesday, November 25, 2008 5:51 PM
> Subject: [Csnd] new bandlimited oscillators (was: Re: UDO question)
>
>
>> Thanks for posting these examples - I finally got the csd ones running 
>> after I twigged they needed  5.09. Somewhat slow on the G4 iMac (just 4 
>> voices before breakup), but much better on dual-core, unsurprisingly - I 
>> will need to lengthen the envelopes to be sure I am genuinely playing 
>> more than 16 voices (via my 2-octave Oxygen8 controller), but so far so 
>> good! One thing - am I right in assuming the loss of waveform shape at 
>> low frequencies  (looks like bottom partials are reduced quite a bit, 
>> from about 200Hz), and the just-noticeable latency is because of the new 
>> ultra-powerful dcblock2 opcode? Presumably for low notes (e.g. 
>> floor-shaking 50Hz) the order has to be increased pro rata. Higher up the 
>> range, the waveforms look and sound excellent!
>>
>> Richard Dobson
>>
>>
>>
>> victor wrote:
>>> No, all my stuff's GPL and there are no strings attached. If you want
>>> to have a look at the code, I have an archive with the dafx material
>>> in: http://music.nuim.ie/vlazzarini/tmp/Dafx08.zip
>>>
>>> Out of curiosity, I just checked the cost here on my computer. Takes
>>> about 2.3 secs to do 100 secs of output. Using a buzz instead, is about
>>> 2 secs per 100 secs. Not bad.
>>>
>>> Victor
>>>
>>
>>
>>
>> Send bugs reports to this list.
>> To unsubscribe, send email sympa@lists.bath.ac.uk with body "unsubscribe 
>> csound"
>
>
>
> Send bugs reports to this list.
> To unsubscribe, send email sympa@lists.bath.ac.uk with body "unsubscribe 
> csound" 

It's a factor no doubt, but increasing the order has resulted in 
substantially preserving the shape down to  a much lower frequency (not 
sure what, close to 50Hz anyway; I am watching the signal on an analog 
scope), at the obvious cost of now very noticeable latency. But I agree 
that a subtractive solution would be ideal, and looks as if it ought to 
be possible, somehow! I am a bit puzzled by the implication in the paper 
and slides, that the bipolar pulse wave also needs  DC correction - is 
it not fully symmetrical?


Richard Dobson



victor wrote:
> also I think the change in shape at low frequencies is just because
> modFM only approximates a pulse, it is not exactly one.
>