HD1 ,User samples & Ram observations after OPS v2.02

[AntonySharmman]

Yesterday I found some time to investigate new features of OPS 2 on HD1 module.

Initial expanded Ram status results seem to be :
All 13 Gb Factory Rom & EXs1-9 occupy 1141 Mb of Ram and remaining Ram will be 803 Mb.
I tested a GRD piano of 1GB size with un-looped samples of natural duration and under
virtual memory mode it occupied 45 Mb of Ram for preloaded samples !
In that best case Ram preservation ratio is 22:1 (SSD/Ram).


As you already know our X-Multisample User Bank after OPS 2 become Unknown Bank
and our first step is to convert it to new sampling section bank with pre-loading
features.
This convertion is instant using PROG/OSC-Pitch/upper screen menu selecting
Remap MS/sample Banks option and copying red high-lighted Unknown Bank to
New Sampling Section , just a couple of seconds and everything will be as before.

Therefore my prior routine was to convert my 800 Gb User sounds to new pre-load mode.
My set includes approximately 200 multisamples and all samples are precisely looped.
The results after saving modifications and loading Pre-loading KSF was a Ram save of
600 Mb and SSD/Ram ratio is 4:1 as I assume for most moderate duration User Sample sets.

Another important thing is not to delete your initially created resource files
when you keep importing user samples and save your work since always the last
KSF files have reference to these initial resource files .
I believe that the right procedure in OPS 2 save all operation (under User preloading KSF)
is to save the entire resources in every save all function.

Under my estimation there is a huge disadvantage concerning editing samples
in preloading mode since there is no edit option unless you load the other KSF
which overflows memory if our user samples contents are big , you have to preserve
some free ram space and change loading mode for specific multisamples every time you
need to edit them !

These are my initial observations on important HD1 module under OPS 2.0.2
I hope this can help you ,dig deeper and correct me where I'm wrong !

[Steve Pavao]

Under my estimation there is a huge disadvantage concerning editing samples
in preloading mode since there is no edit option unless you load the other KSF
which overflows memory if our user samples contents are big , you have to preserve
some free ram space and change loading mode for specific multisamples every time you
need to edit them !

These are my initial observations on important HD1 module under OPS 2.0.2
I hope this can help you ,dig deeper and correct me where I'm wrong !

Hi Anthony,

First off, it's great to hear about the sample memory usage reduction you're observing.

Regarding the quote above, I'm not sure I 100% understand what you're trying to express, but I wanted to point out one note in the manual which I think is possibly related to your point.:

"Data in User Sample Banks can be loaded either into Sampling Mode (for editing), or as a User Sample Bank (for playback with Virtual Memory), but not both at the same time."

- Steve Pavao
Korg R&D

[QuiRobinez]

"Data in User Sample Banks can be loaded either into Sampling Mode (for editing), or as a User Sample Bank (for playback with Virtual Memory), but not both at the same time."

This design decision was something i really liked.

I think it's very clear this way for users that the _userbanks aren't loaded in actual sample memory and that if you want to create new sets or modify sets that you have to open the original file and do your edits.

[AntonySharmman]

Regarding the quote above, I'm not sure I 100% understand what you're trying to express, but I wanted to point out one note in the manual which I think is possibly related to your point.:

"Data in User Sample Banks can be loaded either into Sampling Mode (for editing), or as a User Sample Bank (for playback with Virtual Memory), but not both at the same time."

In other words data in User Sample Banks can be loaded into Sampling Mode (for editing)
and if your actual Ram contents are 5 Gb preloaded then Ram will overflow and will be
automatically loaded again in Virtual mode ! so you have to follow a complex strategy !
For me was quite easy but what about a simple User ?

[QuiRobinez]

In other words data in User Sample Banks can be loaded into Sampling Mode (for editing)
and if your actual Ram contents are 5 Gb preloaded then Ram will overflow and will be
automatically loaded again in Virtual mode ! so you have to follow a complex strategy !
For me was quite easy but what about a simple User ?

interesting, as far as i understood from the manual the max size of a sample set that you can make is the limit of your installed memory, which is about 2.8 GB on a 3 gig expanded Kronos when you unload everything from the Kronos.

But if i read it correctly there is a way to create larger sets?

How do you do that ?

[Shakil]

when you have to edit, it will be a single sample not the whole 5gb library. I assume that KRONOS will load only one sample to ram and tben send it back to. irtual memory some how. So I see your point that there could be a scenario easily where yoj sould want edit a 50mb long sample and there is only 10 mb availae in ram. It could be rare but it is possible.

[AntonySharmman]

But if i read it correctly there is a way to create larger sets?
How do you do that ?

Available Ram with all Unloaded (except Rom Bank) is 2 Gb so that is the biggest size
of the imported multisample.
Then after final editing you save all , load only user Preloaded KSF , and 1.8 Gb of
Ram will be available for next multisample you import and so on ...

Shakil
I Know how for sure but this option could be avoided , for many reasons

[Ksynth]

This convertion is instant using PROG/OSC-Pitch/upper screen menu selecting
Remap MS/sample Banks option and copying red high-lighted Unknown Bank to
New Sampling Section , just a couple of seconds and everything will be as before.

Can you explain exactly in which area you are to carry out this change?

Sampling mode?

HD1? How do we bring that up?

etc

Thanks.

[QuiRobinez]

But if i read it correctly there is a way to create larger sets?
How do you do that ?

Available Ram with all Unloaded (except Rom Bank) is 2 Gb so that is the biggest size
of the imported multisample.
Then after final editing you save all , load only user Preloaded KSF , and 1.8 Gb of
Ram will be available for next multisample you import and so on ...

thanks, i understand that.

But when you talk about a 5 GB library you actually mean 3 (or more) files that you put in autoload? or do you mean that you have created one 5 GB file?

I still have to look into it more, but my impression after the tests yesterday was that you can create soundsets with multisamples in that 2 GB memory and save that as one file. Now whenever i want to modify one of the samples in that set i need to load the complete 2 GB set to modify that one. Of course i could create a KSF file for every multisample, but that would be quite complex to maintain.

[AntonySharmman]

Ksynth
While in HD1 PROG/OSC-Pitch/upper screen menu selecting
Remap MS/sample Banks option and copying red high-lighted Unknown Bank to New Sampling Section.

QuiRobinez
Of course I mean more than 3 multisamples , and these 5Gb contents was just my test
when I finish my setup I let you know for User Samples GBs.

Try to explore it via manuals , and sorry but I have to leave and get prepared 'cause I leave for vacations today !

Greetings

[Steve Pavao]

In other words data in User Sample Banks can be loaded into Sampling Mode (for editing)
and if your actual Ram contents are 5 Gb preloaded then Ram will overflow and will be
automatically loaded again in Virtual mode ! so you have to follow a complex strategy !
For me was quite easy but what about a simple User ?

Here is the workflow I use:

Let's say I have a sample set I saved. There are 2 files:
Test.KSC
Test_UserBank.KSC

A. If I want to edit the samples, then I load Test.KSC with setting "Clear Sampling Mode Data". It will load into Sampling mode. I can edit what I need to, then resave Test.KSC if I want to keep my changes. (Test_UserBank will also be re-saved as a result of re-saving Test.KSC).

or

B. If I want to load my sounds for VMT playback, I load Test_UserBank.KSC w/ Default Settings. It will use VMT.


I don't do both A. and B. at the same time. It's not allowed. I do one or the other. If I happen to do A, then if I want to do B right afterward, in Step B I load with setting "Clear Sampling Mode Data".

This is all pretty straightforward. It is described in the manual.

For larger sample sets that span multiple KSC, it's a matter of "lather, rinse, repeat" the steps above. I think the "simple User" can follow these steps and harness the power of the feature with no problem.

- Steve P.
Korg R&D

[QuiRobinez]

QuiRobinez
Of course I mean more than 3 multisamples , and these 5Gb contents was just my test
when I finish my setup I let you know for User Samples GBs.

Try to explore it via manuals , and sorry but I have to leave and get prepared 'cause I leave for vacations today !

Greetings

no problem, thanks for your help, i'm going to dive some deeper in the manuals

have a nice vacation!

[QuiRobinez]

...

For larger sample sets that span multiple KSC, it's a matter of "lather, rinse, repeat" the steps above. I think the "simple User" can follow these steps and harness the power of the feature with no problem.

yes, that was basicly how i did it. The only thing i missed was the area above. I didn't know that you could combine multiple large KSC's into one new sampleset. I thought that you always had to load the multiple KSC's in autoload to complete your soundset.

But i still need to spend some more time with the manual and my test projects, so it's just a matter of time before i know how to do it

[danatkorg]

yes, that was basicly how i did it. The only thing i missed was the area above. I didn't know that you could combine multiple large KSC's into one new sampleset. I thought that you always had to load the multiple KSC's in autoload to complete your soundset.

Yes, you can absolutely combine them together! See the "PRELOAD.KSC" file, for instance, which includes data from most of the factory EXs.

This is a really important point, actually, since it's critical for being able to easily store the entire state of the machine with Save All. When you save a KSC file, you have the option to save just the Sampling Mode data, just the links to EXs and User Sample Banks, or both. The "links to EXs and User Sample Banks" gathers up all of your sample references into a single KSC file, regardless of the source of those samples. So, one KSC can load a set of converted SoundFont2 data and several Korg EXs (or selected samples thereof) for a cover-band gig; another KSC could load some third-party KRONOS sounds, two of your own custom User Sample Banks, and EXs7, for a recording project; etc.

[jimknopf]

Dan, I'm certainly not lazy figuring this all out, and I'm sure I will just by trying what you write.

But I also think it would be really helpful if there was a short video instruction covering the whole area of sampling import, conversion (user streaming) and management of sample sets including autostart options, to have a picture of the whole context before going into single steps. I got by well so far, but I'm not sure if I understood everything perfectly well.