Rompler sample sizes

(Apologies if this has been covered before but I couldn't find anything by searching)

Reading about the new Korg SV-1 got me thinking about sample size in ROMplers: the SV-1 claims 512MB for 36 presets (about 14MB each), while my trusty Fantom Xa has 64MB for hundreds of sounds - but the SRX-11 piano board I have in it has I think only 2 (?) instruments sampled in 64MB.

I know it's not 1 sample = 1 sound, but is it always more = better? There seems to be a trend now (finally!) to bigger sample sets e.g. M3 is I think 256MB, but it's not just about size. There's Kurzweil fans who swear blind they get better sounds in smaller samples than anyone else.

Any thoughts? Is bigger always better? Will falling memory prices finally mean we get decent sized multi-samples in romplers?

nope, bigger is certainly not better

anyone can attest

Getting the most out of samples is an art. More sample memory is generally spent on velocity layers, so it's not just the volume but also the timbre that changes.

However, if those 64 megabytes sound good, what's the problem?

I guess I'm thinking more about emulated and acoustic sounds: there's a big improvement from the Fantom Xa's stock piano to the SRX-11 version.

Or does it just come down to marketing and business issues: Roland and the others have lots of nice big samples but they're trading off the cost of the flash memory, and what they can sell you later as add-on or specialist products. Which is fair enough.

The Fantom G has 256mb of ROM but the skill is how Roland puts the samples together that makes them sound good

There are 3 factors:

1) lenght of the sample

2) Quality of the sample (compression / bitrate)

3) Layers/variations needed per sample

Fantom XA has about 1000 samples. You can make as many sounds from diffrent sample as you want (example: Fantom Gs 1000s piano sounds ).

In fantom G there are I belive 1600 samples, might of been 2400 though... lets just say about 2000...

Don't forget the 4th factor - the skill and musicianship of the person creating the sample. Back in the Akai S950/Emulator 2 days we used to agonize over carefully matching samples and creating perfect loops. This seems to be a dying art, as it's better marketing to just take a zillion unlooped samples and sell 20-DVD sample libraries. Like many other things the size of Rompler sample libraries is mostly a bull{censored} marketing number. At the end of the day what counts is how the gear sounds when you play it.

In the case of acoustic instrument samples the skill and style of the musician being sampled is a huge and often overlooked factor. There are various orchestral libraries sampled in Eastern Europe that feature heavier vibrato than is usually heard in Western concert halls. I am often dismayed at the tuning issues and poorly edited sample start times found in expensive libraries on the market these days. If all your samples start a few milliseconds late because no one could be bothered to properly truncate them it doesn't matter how fast the latency on your DAW is.

As far as the SRX-11 card I can hear a big difference in quality over the compressed to death "Ultimate Grand" piano sample that is found in the Fantom XR rom. There aren't obvious loops and the sound is much cleaner.

Thanks for the answers.

So, I guess as memory prices continue to fall, we should see the trend to larger samples in ROMplers continue in the future? Which will help avoid compression issues and will improve pianos etc - but isn't necessarily a substitute for clever programming?

William Coakley actually has an interesting article on his site on this that argues against bigger being better: http://williamcoakley.com/articles.php?article=bigger.php - not least due to editing issues like the truncation at the start that TechEverlasting pointed out above.

Well I have heard the argument of good programming vs large samples several time.

And in theory , good programming wins , but in practice things are abit different.

It is simple really. It is much easier to take loads of samples and with no programming produce near acoustic results compared to taking very basic samples and with loads of programming producing the same results. In computers where size is of no concern almost all rompler are pure samples libraries with no complex programming included.

In hardware its a bit diffirent , as manufactures tend to avoid including large ROMs. But even in hardware it should come as no surprise that the best accoustic sounds are produced by Yamaha MOTIF XS boasting the largest ROM at 340 MBs.

Of course we talk strictly acoustic sounds here.

But as people said ther so many other factors that play an important role.

Examples where more ROM does equal to better, for example, my motif es with 170MB ROM has better acoustic sounds than M3 and Fantom G.

Even when compared to PC3X it can fight shoulder to shoulder (PC3 has much more rom, better programming and a vastly superior engine to support it eg . 32 layers compared to 4 of motif es etc. ) , of course in pianos and string motif es bites the dust but so do all other workstation against pc3. And when compared to MOTIF XS which has double rom , the sounds are certainly not of double quality

In any case the best judge is your own ears.

I guess it's not even just about good programming, but time invested in programming: Roland claim there's 700+ samples in their SRX-11 piano, which all need looping, truncating etc.

They'll invest the time for a flagship piano sound, but would any manufacturer put in the time to tweak hundreds of samples for any other sound in a ROMpler?

That is a very good question... Some would argue that sample size is a requirement of good programming, i think the british say "you cant polish a turd " or something along the lines in my country we say "you cannnot extract fat from a fly"

So a good sample is always a very good start in emulating any sound whether accoustic , analog , pure digital etc. but you need not to be static and that is where good programming comes, it gives life to your sample. Especially the modulation is crucial since especially accoustic sound tned to be rather naughty and modulate all the time. Take for example the overblow on a flute, you can sample loads of diffirent overblows but I feel its better to nail the programming of the whole stuff for much more realism.

Afterall this where physical modeling wins hand done in emulating acoustic instruments , it computes every single parameters , nothing is pre rendered nothing is prepared. Because the instrument is unpredictable and complex but so is the musician who plays it.

Probably why using the real thing is alot better than using an emulation.

It is a highly complex issue really , because sound synthesis is a complex issue.

Of course dont forget is not all a sound synthesis issue, is also ALOT of playing .

Never underestimate the power of playing an instrument properly. And then you have the issue of orchestration and then you have the issue of ..... blah blah blah

Which I guess also addresses some of the questions of "does the audience notice" - they won't hear the difference in a better sound, but they'll appreciate the better expressiveness and playing...

ok now you go into alot darker territory, the huge problem here is the SUBCONCIOUS which play a very important role to what we call "personal taste" .... In short just because YOU dont hear something that does not mean that YOUR BRAIN does not either!!!!

The brain is a very tricky organ and can easily play tricks on a person, many things that are felt are processed never reaching the Concious part of our brains yet they influance our decision making ability.

Take for example driving, you may drive and then later cant recall the whole process all the decisions you had to make during the process, as all was processed automatically by your brain and it decided for yourself while you were listen to the radio and were thinking about your girlfriend.

The same applies for music, you cant really say why you dont like a piece but you know that you dont. Some things are obvious(consious) and some are not (subconcious).

You can spend a lot of money and time or you can upgrade the memory. There's absolutely no shame at all in doing the latter, and it's a far more sensible approach than eternally trying to recycle those waveforms.