Improving Salamander Grand

[JeffLearman]

I find the Salamander Grand to be a really excellent sample set. However, even though I'm not particularly sensitive to latency, I noticed quite a bit of it in some cases. Examining a random set of about 20 samples, I saw that while most were good, a fair number had 10 ms, and one had 30 ms latency (near silence before the first significant transient.)

I plan to process the samples and remove the latency and make them more consistent.

I also noticed that the A4 sample is tinny, and that remapping the neighboring samples to cover its range made a much better sounding piano.

While I'm fiddling with it, has anyone noticed anything else about it that might want fixing?

Or has everyone found a free piano sample set that's clearly superior? I'll probably touch it up regardless.

[JeffLearman]

I found out during processing that the average latency was 15 ms and the max was 55. It'll feel better with that trimmed off.

[JeffLearman]

I posted the updated SalamanderGrand as a comparison test at

https://github.com/jlearman/SplendidGrandLatencyFix

It's keyswitched; play C8 to get the original, Bb7 to get the trimmed version, and A7 to get a merged version with trimmed on left side and original on right side. The next question is whether to replace the samples on GitHub or to provide it as an alternative.

My example here happens to be 44k/16b but the work was done on the original 48k/24b and if we replace samples of course we'd stick with that. Might be nice to also provide a 44k/16b version as well.

In any case, feedback please! Definitely let me know if I goofed anything in the sample set.

[Patrice Brousseau]

Loved this piano... but since got a few better (paid) ones. Anyway, good that you improved the original. It does a decent job if you don’t want to pay big bucks.

[PetPet33]

Thanks for your work, Jeff. A regular piano has natural latency of 15-20ms, are you aware? I remember playing the Salamander piano for the first times, and it felt clunky to me, just like a basic upright piano

That explains it.

The delay (or: latency) in pianos naturally can lend to a heavy feel. In a guitar you pick the string direct (no hammer mechanism), instant gratification

In digital pianos (keyboards) it's funny that the long down-press from mechanical pianos (etc) has remained A relic?

[PetPet33]

I found out during processing that the average latency was 15 ms and the max was 55. It'll feel better with that trimmed off.

It's a nice piano. Thanks for tinkering with it. Looking at the files now, it's a boon that the original author, A. H., left those in. Ah, good ol' Swedish m'balls.

Thanks, again. The pared down ver. of Salamander ("Little Lizard Piano") had already supplanted the original for me.

[enroe]

Thank you Jeff Learman!

The Salamander Grand Piano is certainly one of the best, very authentic sample
pianos out there.

The start of the sample can be moved back - still in the SFZ file - with the
"offset" parameter. This can be done globally - i.e. for all samples - or per sample
(within the regions).

[Propianoplayer]

Never heard of it, just checked it though. It sounds awesome for a free piano!!

[JeffLearman]

A regular piano has natural latency of 15-20ms, are you aware?

That depends on how you measure it. If you measure it from when the hammer strikes the strings, the latency is zero, and to me as a piano player, that's what matters. If you measure it from when your finger first hits the key, then you'll measure significant latency -- but it isn't real latency. It's just measuring the time between when your finger touches the key to the time the hammer strikes the string. As a player, you naturally learn that the note doesn't play when you touch it, it happens when it happens. Likewise, when playing a good hammer-action MIDI keyboard, the MIDI note doesn't get transmitted until the moment when the hammer would strike the strings, or more factually, when the internal pseudo-hammer strikes the "bottom" detector.

If a hammer-action keyboard doesn't replicate the feel and timing of a piano, that's the fault of the keyboard, and software instruments shouldn't try to compensate for that (unless it's a user adjustment.)

When should the samples begin? It depends on your mode of operation and what you want. When playing (and monitoring the sampled piano as you play), you'll usually want zero-latency so that it responds as much like a real piano as possible (assuming your MIDI keyboard also acts enough like a real piano.)

If you're doing a MIDI score and the piano part is already created, and you want as natural a sound as possible, you might want to hear the mechanical "pre-note" noise of the action. In this case, you'd want the samples to start earlier, before the hammer strikes the strings, and you'd adjust the MIDI timing to play the notes earlier to compensate.

I can't think of any case where variable latency is appropriate.

[JeffLearman]

The start of the sample can be moved back - still in the SFZ file - with the
"offset" parameter. This can be done globally - i.e. for all samples - or per sample
(within the regions).

Right, and the version here:
https://github.com/sfzinstruments/SalamanderGrandPiano
does it that way.

Since the latency varies per sample, it has to be done per region, and I haven't looked at how Kinwie coded it (using my offsets) to see how he provides the adjustment. But Kinwie is a gifted mapper, and I bet he got it right.

[JeffLearman]

In digital pianos (keyboards) it's funny that the long down-press from mechanical pianos (etc) has remained A relic?

Not sure what you mean by this, but if you mean to say that digital pianos feel like real pianos, then that's a feature, not a bug.

Maybe someday we'll learn that the ideal control mechanism for a digital piano is NOT to try to replicate the feel of the world's finest grand pianos, but to do something different. That would be likely to catch on like wildfire, just like the Dvorak computer keyboard.

[JeffLearman]

... and, you're welcome, folks! I'm just nerdy enough that I enjoyed doing it, but I mostly appreciated having a sampled piano that felt right.

[PetPet33]

In digital pianos (keyboards) it's funny that the long down-press from mechanical pianos (etc) has remained A relic?

Not sure what you mean by this, but if you mean to say that digital pianos feel like real pianos, then that's a feature, not a bug.

Maybe someday we'll learn that the ideal control mechanism for a digital piano is NOT to try to replicate the feel of the world's finest grand pianos, but to do something different. That would be likely to catch on like wildfire, just like the Dvorak computer keyboard.

Potato, potató?

Speaking of a "piano" is a lazy Latin (Italo) relic in the first place, so sorry for that (as a linguist). Let's speak of mass-produced digital MIDI keyboards that are used to digitally trigger all manner of (hammerless) instruments with no consideration for the "feel of the world's finest grand pianos". The long down-press is futile. The MIDI keyboard manufacturers are probably trying to replicate something, yes. A monkey can replicate, true instrument evolution is different.

[JeffLearman]

The long down-press is futile.

I disagree; I think it gives more control over velocity. I've played instruments with shorter key depth of travel and it reduces my ability to control dynamics, even when playing sounds other than pianos. Of course, I grew up playing piano, and perhaps if I grew up using some sort of surface that didn't move but registered velocity perhaps I'd prefer that.

But you're certainly right that a lot of design is silly emulation of historical artifacts. For example, I have zero use for "graded hammer action," where the low keys are heavier than high keys, similar to a real piano. It never bothered me that this detail wasn't recreated on most digital pianos up until round 2005, when it became all the rage. Sure, it might matter for a concert pianist who sadly has to use a digital for practice (but then again, that kind of person should be using a digital with a real piano action.)

I also get a kick out of all the trouble some go to, to replicate annoying aspects of the original instrument, like the release noise: something instrument designers tried hard to make silent but failed. Of course, lots of people love this, and I do admit it makes a digital instrument sound less fake when fiddling -- I just haven't ever found it musically significant. Then again, contrast that with things like Hammond keyclick, which probably was not intended, but really adds important character. So, it's a grab bag. Some attempts at faithfulness are a waste of time (IMHO) whereas others are crucial. And everyone has different opinions on which are which.

Should the pedal squeak on a digital?

[JeffLearman]

BTW, I have two digital keyboards; one has hammer action keys and one has "semi-weighted" keys. I prefer the hammer action for most things except for Hammond, clav, and wurli. I prefer the weighted keyboard to play a wind instrument or strings due to better dynamics control. And I never did like the feeling of diving-board synth keys, though my first three electronic keyboards were that type and I used them for over a decade.