My New Rhodes: Anyone Down for a Large Free Sampleset?

[hollowsun]

* Different velocities would be good but not essential

Steve, why is this not essential for you?
Are you saying that it could be enough to feature the velocity>amp/filter modulation with the base samples?

Ermmmm....actually... yes!!

Used in combination with functions such as velocity>sample start and velocity>amp env attack and other tricks that is!

With good programming, you'd be surprised how radically a sample set can be optimised to create realistic results with the minimum of 'bloat'. I come from that tradition - squeezing an entire grand piano onto a single floppy disk for the Akai S1000 library back in the 80s for example

That said, we don't have those archaic memory restrictions these days so some 'real' velocity variation samples would be ideal - a combination of velocity switching AND the tricks mentioned above should give a good rendition of the sound. However, as the idea is to make this sound freely downloadable (?) I would want to optimise it as thoroughly as possible with the minimum of 'bloat' to make the file size managable not just from the perspective of bandwidth but for the user's convenience.

I would also like to make an efficient instrument that does not require the earth's current supply of memory and disk storage to work properly

Steve

Steve

[Sascha Franck]

I'll gladly help in trimming, naming and looping the samples.
And yes, I agree with Steve, for many purposes velocity splits are no requirement.
For a 100% realistic sound you may need them - but in that case you'll rather need a LOT.

Having said that, quite sometimes I find 1 velocity zone to be better than 2. With 2 zones only you'll most likely notice the zones switch and applying velocity x-fades may very often result in phasing problems.

Best would be to have a fullsize patch in addition with a few "light" ones (loop points and no velocity switches).

Funkybot, why don't you just upload the raw samples at the Auditorium, along with some brief description about what they represent? That way anybody could give them a go.

[rockstar_not]

Before launching into what looks to be alot of work - you have tried the various EP VSTi from Soundfonts.it, right?

I have to say, that I probably will never use a sampled EP ever again - the Wurly and Rhodes emulations that they have there are truly amazing - and oh, by the way, they are free.

Give them a try before you fly.

-Scott

http://www.soundfonts.it/?a=cat&b=2

[Sascha Franck]

Yes, Guidos VSTis are amazing!
Still, samples would use up less CPU power (even if the soundfonts.it stuff is fairly decent), plus, they'd be crossplatform (which SE synths aren't).

[Rangtangtang]

Those Wurly and Rhodes Vsti are very good. Well, if youve got good vsti and it has the sounds you want, why bother with samples, anyway. Its just extra work? I gues it depends what your trying to do.

[fr4ncesco]

* Different velocities would be good but not essential

Steve, why is this not essential for you?
Are you saying that it could be enough to feature the velocity>amp/filter modulation with the base samples?

Ermmmm....actually... yes!!

Used in combination with functions such as velocity>sample start and velocity>amp env attack and other tricks that is!

With good programming, you'd be surprised how radically a sample set can be optimised to create realistic results with the minimum of 'bloat'. I come from that tradition - squeezing an entire grand piano onto a single floppy disk for the Akai S1000 library back in the 80s for example

That said, we don't have those archaic memory restrictions these days so some 'real' velocity variation samples would be ideal - a combination of velocity switching AND the tricks mentioned above should give a good rendition of the sound. However, as the idea is to make this sound freely downloadable (?) I would want to optimise it as thoroughly as possible with the minimum of 'bloat' to make the file size managable not just from the perspective of bandwidth but for the user's convenience.

I would also like to make an efficient instrument that does not require the earth's current supply of memory and disk storage to work properly

Steve

Steve

All valid points Steve,
we spoilt kids have a lot of tricks to learn from the old school masters (call it the art of optimization)


ciao
Francesco

[hollowsun]

I would suggest that you sample at 44.1/24 bit.

Hmmmmm!

The thing you want to minimise is noise and 24 bits will give you better quality at the quietest velocity levels and on long decays. Of course you need a clean signal to start with.

The latter sentence in your paragraph highlights the the fact that your suggestion is erroneous because it highly unlikely that a Fender Rhodes has a signal-to-noise ratio/dynamic range even approaching a 20-bit signal, let alone 24.

If Hollowsun needs 16 bit for Akai - surely a dying breed of machine in 2005 - then you could just batch convert with dither to generate the 16 bit set.

I don't need 16-bit just to service my allegedly antiquated sampler - I would prefer 16-bit because there is no need to add to yet more 24-bit bloatware when it is totally unnecessary!

Aside from the fact that the Rhodes' s/n ratio is incapable of a 24-bit signal, there aren't any convertors out there that can actually play a 24-bit signal accurately (more like 20-bits on a good day), MIDI's velocity range itself is effectively around 21-bits AND in 99% of cases, it's all gonna get mixed down (in amongst a bunch of other instruments) to a 16-bit delivery format (or worse - MP3!).

44/16 is more than adequate to accurately represent this instrument. It will also make the instrument more viable for download, will provide broader product compatibility, will use less storage space, will be quicker to load in your sampler, will place less strain on the CPU (allowing more instances of other instruments/effects to run) and so on.

Let's be practical about this!


Steve

[hollowsun]

.....you have tried the various EP VSTi from Soundfonts.it, right?

.....the Wurly and Rhodes emulations.....

I agree with your point but there's one word in your post that makes this project viable.... "emulations"

Yes - they are good, they are versatile, they are very tweakable but....

Many people want the real deal, not some digital 'emulation'.

I hear what you say but there's mileage in this project IMO.


Steve

[egbert]

The thing you want to minimise is noise and 24 bits will give you better quality at the quietest velocity levels and on long decays. Of course you need a clean signal to start with.

The latter sentence in your paragraph highlights the the fact that your suggestion is erroneous because it highly unlikely that a Fender Rhodes has a signal-to-noise ratio/dynamic range even approaching a 20-bit signal, let alone 24.

Steve - I don't know exactly how clued up about sampling theory you are but that fact that converters are not noiseless down to the least significant bit when sampling at 24 bit is not a convincing argument against using the extra bit depth.

Consider the situation where you have 20 bits of true resolution available (converters with S/N ratio of 120 dB). When you are recording a signal that is low in amplitude - so that it is only using something like the last 6 or 4 bits of a sixteen bit sample - you are getting very low resolution 16 to 64 discrete levels. I am thinking of ppp samples OR more importantly perhaps the end of the decay tail of a long note.

If you record the same signal at 24 bits with the same converter (which has 20 true bits of resolution above the noise floor) you will have 14 to 12 bits of resolution for the signal - 10 to 8 of which are above the noise floor. This is a very significant increase in resolution.

In the real world, people playing these samples through a sampler with a compressor will get reduced noise on the decay tails if the noise levels from the sound source (Rhodes) itself have been minimized. By this I mean a clean signal path - removing noisy components or circuitry if necessary.

I have the Scarbee Rhodes and Wurli and they are very clean 24 bit samples - Scarbee has done whatever was necessary to clean up the instrument output/preamp. You can read about what he did to prepare the instruments on his site.

[hollowsun]

we spoilt kids have a lot of tricks to learn from the old school masters (call it the art of optimization)

ciao
Francesco

Ha

Ciao

Steve

[hollowsun]

I'll gladly help in trimming, naming and looping the samples.

Thanks

And yes, I agree with Steve, for many purposes velocity splits are no requirement.
For a 100% realistic sound you may need them - but in that case you'll rather need a LOT.

Yep!

Funkybot, why don't you just upload the raw samples at the Auditorium, along with some brief description about what they represent? That way anybody could give them a go.

Hmmmm.... my only thought on that would be that there might be lots of different versions floating around which could be confusing.

Might be better for just one common, 'master' version of the samples to be available once I've sorted them out and then have experts such as yourself step in to do the appropriate tweaking for the different sampler formats.

Steve

[kevink]

.....you have tried the various EP VSTi from Soundfonts.it, right?

.....the Wurly and Rhodes emulations.....

I agree with your point but there's one word in your post that makes this project viable.... "emulations"

Yes - they are good, they are versatile, they are very tweakable but....

Many people want the real deal, not some digital 'emulation'.
Steve

But a sample set isn't the real deal, either. The "real deal" is a wood-and-metal box that weighs 40kg, with hammers, tines, and funky little singl-coil pickups.

Samples and physical models get different aspects of an EP "right". Samples record what it really sounds like - at set of specific key pressure curves. Interpolation between velocity layers is a gross oversimplification of the actual non-linear processes that produce Rhodes signal. Conversely, PM synthesis gets closer to the expressiveness, particuarly if you're playing with a weighted keyboard, but there's no point at which it sounds, exactly, *exactly* like the real thing. I much prefer *playing* Mr. Ray (or, I would imagine, Lounge Lizard, though I don't own it), but I would imagine that if some producer told me that he had to have a "real Fender Rhodes sound" on a track, he'd be more readily fooled by samples.

[hollowsun]

Steve - I don't know exactly how clued up about sampling theory you are

Ermmmm... being doing this professionally since 1986.... so yes - I am fairly familiar with the principles!

Listen - I am not disputing your claims as they appear in theory or on paper but in practice, there is little (if any) audible benefit to be had from recording individual samples in isolation from this Fender Rhodes at 24-bit.

You talk of "real-world" and "if the noise levels from the sound source (Rhodes) itself have been minimized". And there's the rub! In the real world, the noise levels in the Rhodes are likely to be such that 24-bit is redundant - few sound sources are, such is the quiescent noise in mics and/or the rooms in which they are recorded.

Now..... if we were talking about a rugged and highly dynamic 'performance' from the same Rhodes, I may well concur that recording at 24-bits might be worthwhile. But for individual samples, I believe that 16-bit is sufficient and will make for a more efficient sampled instrument.


Steve

[hollowsun]

But a sample set isn't the real deal, either. The "real deal" is a wood-and-metal box that weighs 40kg, with hammers, tines, and funky little singl-coil pickups.

Point taken

Samples and physical models get different aspects of an EP "right". Samples record what it really sounds like - at set of specific key pressure curves. Interpolation between velocity layers is a gross oversimplification of the actual non-linear processes that produce Rhodes signal. Conversely, PM synthesis gets closer to the expressiveness, particuarly if you're playing with a weighted keyboard, but there's no point at which it sounds, exactly, *exactly* like the real thing. I much prefer *playing* Mr. Ray (or, I would imagine, Lounge Lizard, though I don't own it), but I would imagine that if some producer told me that he had to have a "real Fender Rhodes sound" on a track, he'd be more readily fooled by samples.

Don't get me wrong - these modelled products are fine but some/many people would feel more comfortable playing samples from 'the real thing' and Funkybot (and I) are offering this for free. Can't see the problem myself!


Steve

[Stupid American Pig]

Just buy a rhodes, use Mr ray for composition and then record the real thing if you feel it isnt 100% accurate. TBH, I prefer Mr Ray in many situations, as its just "easier" than hooking up my rhodes and tweaking the jstation and or mic'ing an amp.


Though I was playing my rhodes last night, and there is just something about it that cant be beat.