KVR Audio

pj geerlings · Post by **pj geerlings** » Mon Nov 08, 2010 5:11 am

I'm still wondering about this ...

pj geerlings wrote:How many surrounding samples are used to build the oversampled image?

meloco_go · Post by **meloco_go** » Mon Nov 08, 2010 7:29 am

pj geerlings wrote:I'm still wondering about this ...
pj geerlings wrote:How many surrounding samples are used to build the oversampled image?

Hmm, well, it seems you have some image processing background, because it seems like an awkward way to ask about filter kernel size in audio=)
From latency figures I see that kernel size is somewhere around 200 samples.
However it also seems that he uses multirate filtering for higher sample rates, and I don't know to calculate effective kernel size for this.
Probably Andy can get into more specifics if he wishes.
All I can say that the passband is really flat and aliasing suppression is VERY good.

VariKusBrainZ · Post by **VariKusBrainZ** » Mon Nov 08, 2010 10:56 am

Having never used a hardware compressor or any 'pro' software compressor plugins I was beginning to think my expectations of a compressor were false, until now.

I just bought The Glue this weekend and boy am I happy..... a compressor that does what I want. Easy to tweak and no unwanted suprises or signal degradation.

pj geerlings · Post by **pj geerlings** » Mon Nov 08, 2010 3:42 pm

meloco_go wrote:
pj geerlings wrote:I'm still wondering about this ...
pj geerlings wrote:How many surrounding samples are used to build the oversampled image?
Hmm, well, it seems you have some image processing background, because it seems like an awkward way to ask about filter kernel size in audio=)
From latency figures I see that kernel size is somewhere around 200 samples.
However it also seems that he uses multirate filtering for higher sample rates, and I don't know to calculate effective kernel size for this.
Probably Andy can get into more specifics if he wishes.
All I can say that the passband is really flat and aliasing suppression is VERY good.

Maybe I should have said "new signal stream" instead of "image" - sorry about that.

My understanding of oversampling techniques is admittedly very limited - so far I have never gone beyond 2x oversampling for my stuff and in that case I used some number of points on either side of the region to generate the "center" value via band-limited sinc interpolation. It seems to work but I'm always more-than-a-little concerned about the transient smearing that goes along with it.

The glowing reviews of "The Glue" made me wonder if there is a better, less transient-destroying method of oversampling.

-pj

meloco_go · Post by **meloco_go** » Mon Nov 08, 2010 3:54 pm

pj geerlings wrote:The glowing reviews of "The Glue" made me wonder if there is a better, less transient-destroying method of oversampling.

-pj

x2 is relatively mild OS so if it is audibly destroying your transients then I would guess it might be something else. Maybe your passband is not as flat as it should be?

andy-cytomic · Post by **andy-cytomic** » Mon Nov 08, 2010 5:20 pm

News: The Glue v1.1.3 is ready for download

The last release was not good, sorry people! I have been working flat getting it right, so for everyone having trouble with v1.1.2 please download this version and give it a go, hopefully all will be well again.

The Glue v1.1.3:
* Fixed: mono oversampling now working that was broken in v1.1.2
* Fixed: hang on render in Cubase and Nuendo by not reporting latency change during render. Until Steinberg fix the issue at their end renders will not have correct automatic latency correction

The Glue v1.1.3 also has everything that was in v1.1.2:
* New: updated linear phase high quality oversampling to support up to 12 MHz (x256) - note that x2 is recommended for Realtime and x8 is recommended for Render
* New: slightly improved preset handling with next and previous buttons
* Fixed: unwanted smoothing of Mix and Makeup Gain on first process block
* Fixed: needle is now reset of when not processing audio
* Fixed: detection of offline rendering in Reaper and Sonar
* Fixed: reporting of dynamic latency in VST version in Live and possibly other hosts

djanthonyw · Post by **djanthonyw** » Mon Nov 08, 2010 5:25 pm

Amazing Andy! Thank you.

andy-cytomic · Post by **andy-cytomic** » Mon Nov 08, 2010 5:50 pm

meloco_go wrote:
pj geerlings wrote:The glowing reviews of "The Glue" made me wonder if there is a better, less transient-destroying method of oversampling.

-pj
x2 is relatively mild OS so if it is audibly destroying your transients then I would guess it might be something else. Maybe your passband is not as flat as it should be?

This post contains technical detail only relevant to someone writing their own oversampling code, so please everyone else ignore it!

If the transients are being destroyed it sounds like the oversampling filter is wrong, as meloco_go quite rightly points out you probably have a saggy passband, and / or not enough rejection in your stop band and / or a bug in your filter code.

Resampling from 44100 -> 88200 is not an easy thing to get right. It requires a linear phase fir with lots of taps. From the oversampling menu of The Glue it is easy to calculate the total taps, for 44100 -> 88200 has 105 samples of latency, which means it uses 105 taps either side of the centre sample, so that is 105*2+1 = 211 taps in total. It is a full band (ie not half band) linear phase fir filter, with a very specific shape to keep the passband 0-20 kHz within 0.01 dB, and the rejection below -120 dB in the same range. This is high quality oversampling, and since it is linear phase you can do it again and again and not touch effect the transients in your signal one iota. You can resample the same signal 10 times and still only have 0.2 dB of change at 20 kHz

Going from 48000 -> 96000 is easier, for that I use 93*2+1 = 187 taps, and for 96000 -> 192000 it is easier again, I use 27*2+1 = 55 taps.

I will leave it there for now. Please email me directly with further enquires as I feel this thread is not the best place to discuss the implementation details of linear phase fir oversampling filters

www.cytomic.com/contact

jamieblac · Post by **jamieblac** » Mon Nov 08, 2010 6:00 pm

Cytomic The Glue vs. PSP VintageWarmer2

penguinfromdeep · Post by **penguinfromdeep** » Mon Nov 08, 2010 6:07 pm

jamieblac wrote:Cytomic The Glue vs. PSP VintageWarmer2

Excuse me?

jamieblac · Post by **jamieblac** » Mon Nov 08, 2010 6:11 pm

penguinfromdeep wrote:
jamieblac wrote:Cytomic The Glue vs. PSP VintageWarmer2
Excuse me?

lol

penguinfromdeep · Post by **penguinfromdeep** » Mon Nov 08, 2010 6:16 pm

Honestly I don't know what you meant? Making a comparison? They are different beasts that's for sure ..

VariKusBrainZ · Post by **VariKusBrainZ** » Mon Nov 08, 2010 6:20 pm

jamieblac wrote:Cytomic The Glue vs. PSP VintageWarmer2

Is that a perfect combination?

Mine is The Glue v Stillwell Vibe EQ .....oh yes, my

finally wipe the floor with all your

andy-cytomic · Post by **andy-cytomic** » Mon Nov 08, 2010 6:22 pm

VariKusBrainZ wrote:Having never used a hardware compressor or any 'pro' software compressor plugins I was beginning to think my expectations of a compressor were false, until now.

I just bought The Glue this weekend and boy am I happy..... a compressor that does what I want. Easy to tweak and no unwanted suprises or signal degradation.

Hey VariKusBrainZ, thanks for the great feedback! This is exactly the kind of reaction I was hoping for in writing The Glue. I always thought compressors, both analog and digital, were something to be avoided as they mostly made things sound worse - that was until I heard a good analog compressor, then I realised how amazing they can sound

My goal with The Glue was to bring as much of that great sound as possible to a digital version for a price that your average muso can actually afford and hopefully open peoples eyes that compressors can sound good, both analog and digital - you just need the right ones!

pj geerlings · Post by **pj geerlings** » Mon Nov 08, 2010 6:37 pm

andy_cytomic wrote:... I will leave it there for now. Please email me directly with further enquires as I feel this thread is not the best place to discuss the implementation details of linear phase fir oversampling filters www.cytomic.com/contact

Thank you!

I will have to study FIR filtering a little more - per your suggestion I'll address any further questions to your contact email address.

regards
pj

Cytomic 'The Glue' Compressor