Softsynths vs hardware more mem. but not better

Sirbellogs made a comment in another post that brought out something I have been wondering. I have just within the last few weeks finally made the step from keyboards to trying softsynths. After demoing many I bought Kontakt as a full sampler for midi in Cubase SX3 and Bandstand to use a GM module in Band in a Box. Although I am very happy with each of the products so far because it is handy having the sounds stored with the song as long as you have the software installed instead of having to worry if you have your sound modules still hooked up through your midi interface the same when you open a song from a year ago, but I am wondering like Sirbellogs why companies have not been able to make softsynths who can use as in these two products 15g of samples that come with Kontakt and 2g in Bandstand, sound unbelievably better than any keyboard whose limits are still around 256mb of sample memory. When in my opnion the sounds in my FantomX6 sound better than the sounds that come in these softsynths. Anybody out there with technical knowledge and can explain this to me

 

I am not technically proficient either, trying to make the same jump you are. I believe I can answer part of it at least.

I have a Technics KN7000 with very believable sounds for a keyboard, but with a few parameters available to tweak, what you hear is what you get. I have Colossus, B4II, Kontakt2, and soon will have Bandstand.

At first, I was not overly impressed until I started to experiment with effects, modulation, and different articulations. I think many of the fine libraries available, lean toward the sequencer. Eventhough I haven't learned it yet, it is the combination of articulations(scripting e.g) written as a series of commands, that will give you a finished product vastly superior to any keyboard, along with the right effects.

Now, where does that leave us live players ? Well, not with an interface as easy to manipulate as a keyboard, but as we continue to learn to manipulate these samples, and with the constant refinements from the programmers, I am confident that we are on the right path. It might take a while, however.

 

i think also there is a mistaken understanding that somehow larger samples equals better quality. It doesn't.

Our monster computer based samplers are trying to emulate what hardware machines often do naturally.

the electronic technology in a rhodes is much inferior to whatever we're using kontakt on... but it's just not comparable because the rhodes produced those sounds... whereas our softsamplers are using massive samples of what the rhodes did in an effort to emulate it...

beyond that, engineers have often been able to produce more with less... and a well engineered 16MB sample can sound superior to a massive but poorly engineered sample from a softsampler...

 

I think the notion of 'better sound' is quite vague and subjective. A more objective benchmark would be 'exactness of emulation' of natural instruments. Take a grand piano as an example, which sounds completely different when played pianissimo vs. fortissimo. It's not just a matter of volume but a rather complex dependency of timbre (overtones), key noise and resonance. Hence, to emulate the complete dynamic range of a grand piano you need multiple samples for each note, ie. about 10 samples per note to ensure smooth transitions. Also, the duration of the samples must be quite long to capture the natural decay (about 10sec).

Hence the sum of all samples would be around 88*10*10sec=8800sec. Since memory for CD quality stereo sound is 176.4kB/sec, we need for all our samples 8800*176.4kB=1552320kB, ie. about 1.5GB memory.

In fact, that's the minimum requirement if you want the illusion of playing a real Steinway or Boesendorfer, impossible to achieve with conventional hardware based sound generators.

 

Hi all,

I believe you have to also take into account time. the keyboard manufacturers had many man years to develop their sounds. they were also forced to use small sample sizes and be very creative. often times they are still adding on to the same sample set they used 15 years ago.

so I think it is just a matter of time to finetune the current sample libraries. Kontakt certainly has no technical limits to getting the most realistic instrument emulations. however, it is also way more complex than most keyboards so sound designers need time to learn about the possibilities and best practices in regards to sound vs. performance.


best,
Markus

 

Markus

Thanks for the input. I, and perhaps others tend to think that immediate improvement can now be made without the constraints of size. I can now see that it is not so simple as picking up where keyboard programmers left off.

Taking this for a fact, I bet that refinements will come along a lot faster than in the past, judging by the relative warp speed the software industry is moving now..

 

Thank you for all the excellent information. I guess I am more used to presets on keyboards already having tons of time by proffesional programers put into things like filters, envelopes, efx, etc. which greatly affect the final sound, but also which I have no desire to get into doing myself. I am not into programing sounds at all, so I may struggle with getting the type of sounds I am used to out of a softsynth like Kontakt without putting in a lot of work. But I still feel it is where the future of using sample based playback of sounds in recording is going are really is already upon us, so I better get busy.