PDA

View Full Version : A new way of crossfading


Dreamofthenight92
10-20-2013, 10:41 AM
I'm not a soundesigner, this is just an idea. When crossfading it's common to have phasing issues (in reality, never happened in EW libraries, but chould happen). Why not to crossfade between two samples at once? For example, I have 4 samples for an instrument:
at MOD=0 first sample playing alone
0<MOD<31 first and second sample cross fading
MOD=32 second sample playing alone
32<MOD<63 second and third sample crossfading
MOD=64 third sample playing alone
64<MOD<95 third and fourth samples crossfading
MOD=96 fourth sample playing alone
96<MOD<126 fourth and fifth samples crossfading
MOD=127 fifth sample playing alone.

I tried to do this experiment with some samples of EWQL SO platinum (that are velocity based, then I crossfaded them using a Reaper tool).

What I noticed were less phasing issues and less computer resource request then crossfade all the samples all in once as modern crossfades are made.

What's your opinion?

pkm
10-20-2013, 11:29 AM
I'm curious to see where this is going...

jspencer
10-20-2013, 04:41 PM
An inexpensive flux capacitor off Ebay will fix this issue :)

JHMusic
10-21-2013, 03:06 AM
Is that not already the case? I was under the impression that the mixing coefficient was zero except during the transition and full volume areas, and that the reason for having every single voice loaded and "playing" at once was because of the nature of disk streaming.

Malys
10-21-2013, 07:46 AM
I'm seriously wondering if I'll ever be able to understand the half of the technical things that I daily read in this forum.

(Anche tu dall'Italia!)

Dreamofthenight92
10-21-2013, 02:53 PM
Sometimes I think that EW samples, that are great samples, aren't programmed as they should.
I made some examples some time ago: for example I proposed to use MOD patches also in HS short patches (as in HB), to have a continous swell also in short patches. Also I made other experiments in Hollywood strings removing the slow attack of the sustained patches, and learning a midi CC to attack envelope, and the result was...wow, let's say, better and better.
Also, I use to record rolls from various percussions, and stretch them to make the roll integrate better with the tempo of the song .

These are little things, that miss in play, and would make EW libraries be more epic than what actually are.
I'm actually esaminating crossfading (I love make slow dynamic swell), and I think that actually crossfades works very well, but there are some problems:

in HS: in niente patches CC 11 controls both volume and dynamic crossfading, and this make HS mix with Hb a real nightmare (because in HS the volume in adjusted by dynamic crossfading, so that volume decrease too much at low CC 11 values, while HB sound loud also with CC 1 set at minimum level).
The solution I proposed should solve also this problem, making the mixing process easier: using the two samples only crossfading as I proposed, in HS niente patches the niente effect will affect just the first samples, avoiding volume issues with the rest of the samples. This will make the mixing process easier.

Another benefit of using the two samples only crossfading: everybody knows that brasses "overblow" between mp and mf, the timbre of the instrument change much.
Using all samples crossfading introduce the need to impementate the crossfade with a 6 pole lowpass filter. Using the two samples crossfading, we won't need the filter anymore, because when the mp and mf samples are crossfading, the ff sample will not sound, and there won't be the need of filtering hight frequencies.

Also in trombones, in my opinion, will be avoided much flanging problem crossfading just two samples at once.

These my impressions...maybe these kind of feature are too hard to program, or there is somthing else I forget to consider, but I think someone should experiment the two sample crossfade as I'm doing

Jonathan120
10-21-2013, 02:57 PM
I'm sure they already are trying and experimenting with new and better ways to make their future products sound better and be more efficient. After all, it's in their best interest for profits to do so given how good competing companies products are sounding.

pkm
10-21-2013, 07:38 PM
Oh, my iPhone app cut off everything in your original post after the first "0", which is why I was wondering where you were going with it. I read the whole thing, and it makes much more sense now!

That is how it is done, minus any saved system resources. All the samples have to play together so they are in sync and can be crossfaded between immediately, not because they are playing at the same time. You wouldn't want to push up the mod wheel and hear the attack of a new sample.

Correct me if I'm wrong. I don't use the Hollywood Series, but every other sample library I use does it this way.

in HS: in niente patches CC 11 controls both volume and dynamic crossfading, and this make HS mix with Hb a real nightmare (because in HS the volume in adjusted by dynamic crossfading, so that volume decrease too much at low CC 11 values, while HB sound loud also with CC 1 set at minimum level).
The solution I proposed should solve also this problem, making the mixing process easier: using the two samples only crossfading as I proposed, in HS niente patches the niente effect will affect just the first samples, avoiding volume issues with the rest of the samples. This will make the mixing process easier.

So you want the curve to match between HS and HB. That's fine if you're copy/pasting parts back and forth, but the better way is to program each one individually. Become the horn player, not the violinist controlling the horn player's dynamics. Why ask for a screwdriver to be redesigned to hit nails easier when you could use a hammer?

If you matched the curve, but brought only the lowest dynamic layer down to zero, the curve wouldn't be smooth anymore.

Another benefit of using the two samples only crossfading: everybody knows that brasses "overblow" between mp and mf, the timbre of the instrument change much.
Using all samples crossfading introduce the need to impementate the crossfade with a 6 pole lowpass filter. Using the two samples crossfading, we won't need the filter anymore, because when the mp and mf samples are crossfading, the ff sample will not sound, and there won't be the need of filtering hight frequencies.

Also in trombones, in my opinion, will be avoided much flanging problem crossfading just two samples at once.

Again, this is the way that it already is in every library I have used. Correct me if I'm wrong, but are you asserting that HB plays the highest dynamic sample at the lowest mod wheel level and uses a combination of a long crossfade and a LPF?

Here's how the curves in Nils Liberg's crossfade script for Kontakt looks. This is with 4 velocity layers:
http://i.imgur.com/25WSdGq.png (http://imgur.com/25WSdGq)
Notice only two samples are playing together, but all are there, just not always audible.

Dreamofthenight92
10-21-2013, 11:10 PM
So that it seems that samples already crossfades two per once, but they stream all together to reserve phase locking. Understood!

About HB and HS: they already match, if you but CC 11 at medium level on HS, and CC 1 on medium livel on HB, both libraries play on mezzoforte or so and are belanced. The problem is only with niente patches, where not only the dynamic in affected by CC11, but also the volume, so that I have volume issues. For now I found more confortable to use the "legacy" patches, where the CC 11 only controls crossfading, and creating the niente effect using CC 7 when needed.

In HB, a lowpass filter opens when the modwheel is at medium level (and it works very well), don't know if also the highest dynamic level sample is playing or not.

I thought that the cross fading process used to combine all samples because I saw all voices streamed by the play engine at all the dynamic levels.

What do you think about using crossfades on short string patches?

JHMusic
10-21-2013, 11:45 PM
They stream together because the current portion of any of those sample layers might be needed at any given moment, and only the first portion is loaded into RAM and is ready to be played at a moment's notice. Theoretically, if you loaded every sample entirely into memory, you could avoid this, but I suspect PLAY doesn't take that into account.

trumpoz
10-25-2013, 02:11 AM
In HB, a lowpass filter opens when the modwheel is at medium level

To my ears, as a brass player that statement is plain wrong. That is the sound of a brass instrument playing at approx mezzo-forte (ish). If there was any process of the sound like a LPF, us brass players would be able to pick it as it sounds un-natural. The highest dynamic layers are streaming, but they are silent.

Jonathan120
10-25-2013, 03:15 AM
To my ears, as a brass player that statement is plain wrong. That is the sound of a brass instrument playing at approx mezzo-forte (ish). If there was any process of the sound like a LPF, us brass players would be able to pick it as it sounds un-natural. The highest dynamic layers are streaming, but they are silent.

It even tells you they are streaming but silent in the manual LOL.

Dreamofthenight92
10-26-2013, 12:23 AM
Lowpass filters are common used in crossfades: I'm sure that EWQL SO orchestra and choir use 1 pole lowpass filter for crossfades, while I red somewhere that HB use a six pole filter

trumpoz
10-27-2013, 03:06 AM
Lowpass filters are common used in crossfades: I'm sure that EWQL SO orchestra and choir use 1 pole lowpass filter for crossfades, while I red somewhere that HB use a six pole filter

Just had a quick look at the HB manual - there is a low-pass filter for Marc-Sus patches which controls the harshness of the attack, and a LPF at the very highest dynamics of the leg slur patches.

Nothing in the HS/HOW manuals about any sort of filter.

Just musing..... the issue with your suggestion for crossfading is that on solo instrument sustains is that 2 samples will sound simultaneously giving a chorus effect. I know there are a couple of libraries that have managed to crack this one, but they are very very few.

Dreamofthenight92
10-27-2013, 11:27 AM
Thanks for your explanation. Sometimes I say something of nonsense, but please apologize me, I try to learn:)

JHMusic
10-27-2013, 01:18 PM
If I remember correctly, the use of LP/HP filters when crossfading is to do a filter sweep in the same direction on both input signals, one with a LP and one with a HP, so that you crossfade in the frequency domain rather than an amplitude weighting between the two.