blind listening test - 24 DPs and Synths

[ATW10C]

This is from a pianoworld thread.

First listen to a recording of a Steinway B.
http://oneamongmany.net/pianos.mp3

Next listen to the samples, 24 DPs and Synth and among them, the recording of the Steinway B again. 75meg, 38 mins.
http://oneamongmany.net/pianos.mp3

Note down what you like.
Then check against the list with the sample sources named. http://oneamongmany.net/list.txt Don't give away spoilers when discussing your preferences.

I got a hypothesis. Those learning piano on a non acoustic piano will like the real recording and VSTIs better than DP. And possible, a VSTI better than the real acoustic Steinway.

My preferences from the test which I just did.
I didn't give a rating like 1 to 10. I started with "it sounds ok" but noted that the tones were "isolated". Ok while lacking resonance, acceptable to listen. Wouldn't die from listening to it but not particularly oustanding. Some samples sounded above ok. Nice resonance (or is the term reverb?), wide contrast rather than a narrow spectrum of high and low pitch notes.

I like the Steinway B recording that was in the list. But I liked Pianoteq BETTER than the Steinway B. It was warm and bright, had a sparkle to it. Altogether there was 7 that were above ok and 3 were below ok.

Out of the 7, only 1 was a DP, Casio PX-200 which I liked better than Akoustik Piano. The next best VSTI after Pianoteq was TruePiano Emerald. The other 2 VSTI I liked were Art Vista and Konkakt 3.

The 3 below ok were Roland FP-7, next worse Post Musical Instruments and the worst, Korg Triton.

So are such preferences sad, an abomination or is this the future?

My background - noob, 6 months thereabouts, no prior, learning on a Casio CTK3000 (based on my field test, its about the same or just little lighter than semi weighted keys, tested Roland, Novation, CME, Yamaha), probably will go on to a Akai MPK88 on Pianoteq or TruePiano. Could care less about graded key but I do want some resistance to the keys, find semi weighted fine enough but real piano keys seem floaty. I like the Roland's stage pianos key action the best though.

 

[kongwee]

I like full weighted keys, the white and black keys are equally balance in touch. I used cheapo 4front piano quite a lot recently. I have logic piano sampler but it is very difficult to cut through mixes.

 

[illuminationexcursion]

I love this test.

3 things struck me the most as I listened to all the tracks.

1. sound stage
2. reverb
3. sustain/decay

Some sounded really fake with sustains that would not decay. Some sounded really harsh with bad reverb or too muddy with too much reverb. I think the biggest tell tale sign is exaggerated bottom end - real pianos don't sound like that.

I wrote 1 or 2 liners for each sample as I heard them. The three I liked most were:

1. Real Steinway B, I wrote this:

"sounds natural and real, no over exaggerated bass, strength used quite believable good dynamics but probably a synth"

I thought it was a synth because the sound stage was too good to be true.


2. Pianoteq3 is the bomb. I wrote this:

"flat no stereo sound stage, but real, sustain is natural, probably a real piano"

It wasn't the most impressive sounding, but it made me feel like I was sitting in front of it as it played. I play the piano myself and this made me feel most at home strangely.


3. Steinberg The Grand 2 Model 1 would be my choice of piano for studio recording. I wrote this:

"dynamics good, sounds good, i would choose this, probably a synth. love the dynamics"

It was very clean, the dynamics was outstanding - the dynamic range imparted alot of feel to the sound. Sound stage was awesome.


Others
I though the Synthogy Ivory was very natural sounding too. Acoustica Pianissimo and Art Vista Virtual Grand Classic had too much bottom end, making it muddy otherwise they would have sounded believable.


Worst
I would stay away from Korg Triton Concert Grand, Yamaha CVP 407 and Roland FP-7. A kid could tell the Triton was fake. Roland FP-7 killed my ears with piercing high ends.

 

[Cheez]

There are a few caveats to comparing piano samples/keyboards like that:

1. Not all the pianos are the same. Most are Steinway, but some aren't. E.g. the White Grand is a Malmsjo, the Kawai ES-3 sampled their own Kawai EX Concert Grand, many we don't really know (may even have been an upright piano!). Different pianos sound different, so sampling them will produce different sounds - Steinway, Bosendorfer, Yamaha C7 series etc - these are the most common sampled pianos and they are very different in character.

2. It's obvious that a single midi file is used to playback all the samples (except the real Steinway of course). Here lies the problem. Most of us pianist/keyboardist don't play exactly the same way on different keyboards. Our touch/velocity and playing nuances changes according to the touch of the keyboard and the sound of the keyboard. E.g. if the higher notes are too harsh, we adjust accordingly when playing.

3. Digital keyboards (hardware) and sampled VST (software) are very different animals. Build-in keyboard sounds are meant to be played "out of the box". Hence they tend to be more processed than VST sampled pianos. Sampled pianos may sound "thin" but they are usually dry recordings meant to be processed and mixed. We'll leave the sample/note issue - this is obviously a handicap for digital keyboards due to RAM limitations, but since we are looking for the most realistic sounding piano, the handicap is acceptable.

4. The list is not exhaustive. There are other great sampled pianos that are not included (including the Vienna Imperial, other sampletekk pianos, Galaxy II etc).

5. The different character of piano sounds sound better/worse in different genres. There are certain music which I would use a Yamaha C7, for example, instead of a Bosendorfer 290 or Steinway. Hence, it's good to have a collection of piano sound/samples for different musical genres.

My 2 cents:

1. It's not difficult to spot the real thing. The recording itself gives the clue - stage noises etc. Also, the pedaling noise is a dead giveaway. And, well, it sounds real!

2. Worst - as everybody has pointed out - the Triton. What I noted on the blind test: "thin, metallic, decay +++ etc". However, I can see that it will work well in a loud band - it will cut across well in a mix. But my goodness - the decay goes on forever!

3. It's easy to spot the Yamahas! As usual - the traditional thin and "tingly" high registers.

4. I'm surprised at the FP-7. The FP-5 was good (standard Roland sound), but I got no idea what happened to the FP-7. Overly processed (esp lower registers), thin etc.

5. The obvious ones that passed my personal test for quality - the Native Instruments samples (Akoustik Piano and the one that came with Kontakt 3). They have similar qualities, even sampling throughout the entire keyboard, althought the K3 is slightly thinner than the Akoustik.

The Garritan also passed my test, but when I found out that it was the Garritan I was listening to, I was somewhat disappointed that the sound wasn't as lush as the real Steinway. I expected more because of all the hype of the Garritan/Steinway partnership (and of course, knowing that Garritan makes really good samples).

I'm also surprised at the Nord Stage. It definitely is close on par with VST samples despite their handicap as a hardware keyboard. Not bad at all!

And to show that personal preference play a lot in sounds, I don't find Pianoteq good. It's obvious they use less sample/note, and worst of it, they have an unusual wavering sound at the decay end (ie extending their short sampled note via processing).

Background in piano samples - hardware: owned Roland ep7, RD700, various JV expansion cards, Yamaha SY77/99, Emu piano module, played and performed on Korgs, Alesis, Casios (including PX series), and other Yamaha/Roland keyboards. Currently own Yamaha S90.
Software/VST pianos: currently own White Grand Jr, Akoustik Piano, PMI Emperor and Old Lady, Kontakt Piano (stock sample) etc. Contemplating Sampletekk Seven Seas.

 

[Cheez]

Just to keep this thread alive...

I've gone back to play around with my Akoustik Piano. And it sounded very different from the blind listening test. Similarly, if you go into Garritan's website and listen to the demos for the Steinway, it also sounded very different. This is the point about sample-based pianos. How it sounded in the end depends on how it is mixed and the settings. I find in Akoustik (Concert Grand sample), the best sound to my ears will be to increase the bass setting to "heavy", hall set to "recording studio", mic at "near" position (instead of the default "close"), leave the EQ alone, maybe reduce the string resonance ever so slightly (knob to left). There are so many settings and possibilities within Kontakt alone. My settings sounded so much better than the blind test - at least to my ears.

By the way, for those new to sampling and sampled pianos - the reason why a few of the test pianos had a long decay while others don't is because of the lack of release samples. When the keys of a real piano is released, a different sound is produced - ie the release sound. In a good sampled piano, the release is sampled (ie release sample) so the keys die out very quickly just as it would decay in a real piano. When this is not sampled, the programmers have to program decay of the initial sample using things like ADSR settings.




The best pianos will include:

• Every single note of the piano sampled.
• Every single note sampled at multiple velocities with the greatest possible dynamic range (ppp to fff) - more velocities = more realism and smoother transition between dynamics, but takes up more RAM and disk space.
• Release samples for every single note.
• Initial sample/release sample etc all recorded with pedal on and pedal off (each produce very different sounds).
• Sympathetic resonance (with pedal down, the strings actually "rings" differently - most evident when you first play a key without pedal, then pedal while still holding the key down); this is mostly done not by sampling but my modelling.
• Multiple pedalling samples (including sostenuto and soft pedals; also half pedalling and the ultimate - progressive pedalling)
• A good recording with different placement of mics (very close to the strings to player's perspective to audience's perspective etc)
• Sampling a good piano
• Modelling to adjust the lid of the grand piano
• Good convolution reverb (ie actually sampling the different types of halls - from small recording room to a large cathedral) - this is useful when one wants to play the piano as a standalone. Not necessary for recording since most people would use a plugin within their DAW anyway. And in recordings, I would use a dry sampled piano because I would want to mix all instruments with a common reverb.
• Some has damper and pedal noises etc. Really, this is meant for close mic recordings simlutions. I think it's good for rendition of piano solos or quiet pieces. But in a large mix, I can give up these little nuances.

So far, the best ones will be Garritan's Steinway (really nice resonance) and Vienna's Imperial (100 velocity samples per note! Plus resonance, release sample, pedal on and off etc - giving a total of 1200 samples per note!). Sampletekk also has very large pianos (7CG and TBO - both a Yamaha C7 with 31 velocity levels per note with release sample, pedal on and off - giving close to 100 samples per note, but unfortunately no sympathetic resonance).

In general - the more sample per note, the more taxing it will be on the CPU and RAM, and it takes up disk space. Some (like pianoteq) will use more modelling and less samples - which trades realism for less tax on the CPU and RAM. In the end, it's a balance of both. The Vienna Imperial, for example, is so large that you need a separate 60GB harddrive just to store it and to stream the samples by itself. And I doubt you can use the piano with even a small chamber orchestra template, not to even mention a full orchestra.