Piano World Home Page Forums Our Most Popular Forums Digital Pianos - Electronic Pianos - Synths & Keyboards The DPBSD Project!

Spectral pan view of TP Diamond compressed layer test edited down to velocity 19 (left) and 127 (right).

OK, I agree, the layer "stretching" is probably just a filter. But the louder hammer brightness might be an extra sample overlayed on top, with the filter applied to it. I imagine this is what Yamaha calls "Spectral Component Modeling".

THose tow pictures tell us it is NOT a simple high pass or low pas filter being applied. All filters like that have a roll off. I think now that the sound is synthesised, not played from a sample at all. The synthesis may be based on samples or anlyisis of sample. It looks to me like the lower velocity just simply never have the higher order overtones, not filtered just not synthesised.

I think what is stored inside TruePianos is almost exactly the picture we see.

If I'm right this is a lot like the way Roland describes their "Super natural" technology. They record the amplitude of the varoius overtones and how the amplitude changes over time. Then when yo move some parameters (hammer hardness) it makes adjustments to how the overtones play out over time. Finally when you strike the key the sound is synthesised. Seem like TP is doing the same as Roland. Both are sample based but the samles are very, highly processed to the point where there can no longer even be called "recordings", "Measurements" is a better terms I think.

dewster
On this very close look It seems that samples are not exactly the same - but still, the differences are extremely subtle.

ChrisA
TP is not like Roland SN at all. Roland has obvious visible and audible variations between all velocities, similar to those in Vienna Imperial (I don't have picture of its layers right now, maybe dewster can post it).

TP is just a primitive limited sampler - stretched, looped, one-layered.

I think they may be using a second additive layer for the string brightness / hammer sound. But, yeah, other than that it seems rather primitive.

dewster

Maybe.

There might be reasons for that:

1) Patent license restrictions.
2) Usage of closed source third party software.
3) Maybe the work is done by more than one processor and it is technically impossible to have then both responding to MIDI.

Probably they will not tell the true reason and will not fix this in an update. It might be impossible to fix due to technical or licensing reasons.

Now for me the only conclusion is: Dont spend tons of money for that proprietary stuff, use softwarepianos ;-)

Peter

If it's 1 we need to burn down the patent office.
If it's 2 Roland needs to fire their legal department.
If it's 3 Roland needs to fire their engineering department.

;-)

I believe also that modelled pianos use a lot of processing power. So they will probably only be usable as solo instruments.

MIDI instruments are normally multitimbral, that means, multiple MIDI channels can use them simultaniously. That is probably not the case for the modelled piano. Probably this is implemented in another way than the other MIDI instruments.

So that might be another reason why there are MIDI problems.

This is also a problem with some software pianos. However, it is not a problem with Kontakt player based pianos.
Sometimes I play MIDI piano files, that have multiple tracks. Simply to solve: load one and the same piano multiply with different MIDI channels. The memory is needed only once, it is shared.
This can be easily done with the PC because it has probably 100 times more processing power and 1000 times more RAM and 100000 times more nonvolatible storage.

I uploaded a DP-BSD mp3 of a VST called "Supreme Pianos", which claims to be physically modeled, using no samples at all. I heard about it on the KVR forum http://www.kvraudio.com/forum/viewtopic.php?p=4065947; the product web site is at http://www.soundemon.com/products/spintro.htm. The mp3 is of the trial version, which inserts a few seconds of silence in its output every minute. All sliders were left at their defaults.

The file is up on mediafire: http://www.mediafire.com/file/1mxnyjro1dv/dpbsd_v1.7_sound_magic_supreme_pianos_red_wings.mp3

Despite the author's claims, this one sounds 100% sampled to me...

I think dewster is right. TP is in no way similar to SN, I even doubt they're using any kind of modeling (I don't regard using multiple components of a sampled sound as "modeling"). BUT: I still like the sound signature of TP better than that of Roland's SN (sorry Roland owners). It's not always about the best technology but about playability and quality of the basic sound, despite many technical (even audible) limitations.

Thanks! There are strange sounds at the beginnings of notes, and strange drop-outs during the decays. I'll download the demo and take a crack at doing an MP3.


Stretch groups of 3 over the whole range, and what looks and sounds like 2 unblended velocity layers.

And 1/10 the cost.

Here is spectral phase view of HP307 velocities:

I downloaded the demo and am able to get the popping noises to go away, but can't fix the short note decay thing. Decaying notes just fall off the face of the earth.

- It passes the pedal down sympathetic resonance test - the sound of this is nice.
- It passes the key down sympathetic resonance test, though the effect is subtle.
- It largely passes the pedal down silent replay test, though the vel=1 note replays.
- It fails the quick pedal partial damping test.
- No support for partial pedaling.
- It doesn't appear to be looped, though there is some evidence of a transition between note attack and note decay.
- Stretched, both visually and audibly.
- Stretch groups: 2,3(x28),2 = 30 groups.
- 2 velocity layers, the upper one blended somehow, harsh step @ vel=66.

This product strikes me as a work in progress. It might not be a strictly sampled instrument, but the end result largely boils down to the same thing. I like the fact that it isn't looped and has nice sympathetic resonance, but they need to lose the stretching, and fix that velocity switch - it just about gave me whiplash.

And, is it my imagination, or are we going backwards lately? When it comes to the subtleties of how pedals and keys work and interact, no one (except Pianoteq) is able to get everything right for some reason, and it seems like it is only getting worse.

how large is the market of people wanting to use a VST for a complete simulation of an acoustic compared to the market that just wants a 'good sounding piano in the mix' (which will be compressed/eq'd/...)? I think it is a very small niche, which is well served by mostly only pianoteq at this point, other products likely have other priorities.

Good point, though surely it can't be that difficult to do this stuff right. Most offerings seem to be the product of people who either don't understand how pianos work or don't believe in code craftsmanship (or both).

heh, given that I work as a software developer during the day, I can definitely let you know that 'code craftsmanship' is all fine & dandy, but in the end you are going to implement what is required by the feature set agreed with marketing and product management and QA.

The old adage that 90% of the program takes 90% of the time, and the other 10% takes 90% of the time comes to mind, you'd be surprised how many "can't be difficult" things take a long time & effort to get right, and if you are aiming at the 'piano in the mix' or 'film score with some piano' crowd the ROI is just not there, it's a much better business decision to spend the development time on a new/different piano rather than getting all the intricacies of the DPBSD test just right.

The pianoteq folks, given that they are operating with modeling, have it paradoxically a bit simpler, given that if your model is 'good enough' you will get a lot of these behaviours 'for free'. I can definitely see how going from a raw 4-layer sample set to something that responds to partial pedaling, harmonics and so on would not be easy at all.

I would be hoping that the market for a semblance of excellence would be expanding not contracting.

I'm wondering what could the "other priorities" be? It bothers me that mediocracy first comes to mind.

And yet when I listen to postings on another music forum where the piano isn't solo, some of the piano sounds make me cringe. And as you say, "it fits in the mix". Jeez, what would a real live group do with a real live Steinway? Kill it to make it fit in?

Glenn

Hey dewster, what's the moral of the story? Could you give us a little recap/summary of this epic thread?

probably, but given that nowadays' mixing is as far from 'high fidelity' as possible, see http://en.wikipedia.org/wiki/Loudness_war for example, it does make sense that products like Pianoteq are the exception rather than the rule...