Best interface, for lowest delay between keypress and speaker output??

What latency do you currently experience with the UR824? Someone online stated 2.3mS hardware latency for it, so it seems unlikely you would find other hardware to improve too much on that component of the delay. I think the remaining contributers are going to be the combination of driver quality/stability and system performance/resulting minimum buffer size. It's going to be difficult to predict if other hardware can give you better results than you can achieve with what you've already got.

USB, when well implemented, is not at a practical disadvantage in comparison with alternative interfaces, in that it's fast enough, and the latency of any of the interfaces themselves is considerably less than other factors such as the sample buffers.

Before my eMU 0404 died, I could reliably run 3mS buffer latency, 2mS would work but the computer wasn't fast enough to sustain it, but I don't know what hardware latency the 0404 had. I never tried to use Pianoteq and my Korg's onboard sounds mixed live - there was no point as Pianoteq was so much superior to the entry level Korg, so I can't comment directly there...

Hello and thanks for your feedback!  :-)

I don't know what delay the Steinberg interface produces/adds to the chain, as I don't currently know how to measure this precisely, but it also don't feature any midi inputs, so I didn't really consider it as a option for my piano!  I mostly just mentioned it in my specs, in order let you guys know what I currently have available to me!  :-)

I guess I was perhaps not being clear enough, but I wanted to hear your opinion on how to get the shortest signal path from piano to speakers!  ;-)
Using the Steinberg(which is really a Yamaha interface), I imagine the signal chain would be something like:

Keypress -> sensorchip ->onboard DSP ->onboard USB controller -> USB controller in PC -> Steinberg soundcard ->Pianoteq/CPU processing -> Soundblaster ZxR -> speaker output!

Each step adds a delay. But how much impact in each step, I guess, is my question!  :-)

I figured, the best/shoortest way, would be:

Keypress -> sensorchip -> onboard DSP -> Midi-out -> Midi-in on soundcard(USB/PCIe?) -> Pianoteq/CPU processing -> Speaker output(on same soundcard?)  :-)

Or are there really no delay benefits of skipping the signal processing of multiple USB controllers and adding another soundcard to the signal chain?

My "bad mouthing" of USB controllers is my own general experience with(onboard)usb controllers performance in I/O data transfers, when transfering for example data to or from harddisks and USB Sticks. The performance is not very consistant is my experience and depending upon which onboard USB controller used, it's usually not very fast either.....usually a long way from what the theoretical max limits of the USB 2 or 3 standard allows!  ;-)

So, I thought maybe Firewire or PCI-express was a better option to get more consistent and closer to max theoretical speeds......or can this not be transfered to audio performance?  Yeah, I guess I have just revealed myself as being still quite new to this perticular topic!  ;-)

But if I need to shell out 500$ on a new RME soundcard interface!, I want to be sure I understand the reason behind it!  ;-)

But your telling me that software buffering/drivers is where far most of the delay stems from?  This is a CPU bottleneck right?  But seing as even pretty "lowspec laptops" can run Pianoteq, I wanted to make sure everything else besides the CPU is running as optimal as possible!  ;-)

And no, I'm actually not asking because I'm currently dealing with a issue of too much delay, but because I wanted to research AHEAD what a optimal platform for using Pianoteq with my new Roland piano would look like.....just to stay ahead of possible issues!  ;-)

Hope I explained my reasonings well enough, so everybody understands what I'm trying to achive with this topic!  :-)

I assume you're considering this from a live performance aspect?

If I was looking at using your setup, I think what I would try is using the UR824 to drive the speakers with the FP-90 line outs as inputs, both connected USB to the computer, and standalone Pianoteq using the FP-90 for MIDI input and the UR824 as its output device.

If you're not already using connecting the FP-90 via USB:

https://rolandus.zendesk.com/hc/en-us/a...a-Computer

Hey thanks for that link.....a lot of usefull information about how latency is added to a audio chain, all makes more sense now. And for those who may wonder, the link basically advertises Thunderbolt(DMA) and Presonus hardware, in order to get lowest possible latency!  But now I suddenly wonder, does PCIe hardware not offer DMA or similiar feature....article does not seem to touch on that subject?  :-)

But your recommendation for looking out for used RME hardware is noted!  ;-)

You get more bang for your buck from a Focusrite Clarett Thunderbolt interface:  https://focusrite.com/thunderbolt-audio...arett-2pre.

A Clarett 2Pre ports a 1.67 ms round trip latency without unneeded additional I/O ports amounting to overkill.

Now, I’m just getting by on my Focusrite PRO 24 DSP, as it’s now considered a legacy, discontinued FireWire interface.

Hey man, that's awesome info....gets better and better!  ;-)

As my PC's motherboard does not feature onboard Thunderbolt support, I would need a add-on TB adaptor card...which I guess is an extra added cost....wouldn't happen to know of any good bang-for-buck adaptor cards too?   :-D

But yes, your suggestien is something(not good with math) 40% lower in price than the RME, so definitely interesting.   There is only the 0,7ms vs 1,67ms performance gap....which does not seem like much but is actually nearly 2,5 times lower.  Perceiveable in real life performance? Well who knows, but lower is obviously better!  :-D

Update: Just came up with a new question....talking before about the major importance of stable and optimized drivers, what are the experiences/thoughs- on the quality of Focusrite's drivers in here, any user experiences to share?  :-)

Man, I would love to own a Roland FP-90, if one were mine to use as a MIDI controller within Pianoteq!  I’ve a Yamaha DIGITAL PIANO P-95.  And, I admit; I’m feeling a bit jealous er envious!  Maybe you, Mr.DJ, could get another of that Roland model for me, and, the Roland RD-2000 model that EvilDragon likes, right when he lets you borrow the amazing RME UFX+ top-notch audio interface  —just before he goes abruptly on his vacation.

Perhaps you’re still somewhat overwhelmed by the newness of that brand new keyboard of yours or plainly unaware of the satisfactory performance results people report about USB when used as MIDI connections from keyboards to computer processors, and, consequently, computers to audio interfaces, thusly monitors or speakers!

Take a deep breath.  Put anxiety to rest, as you may rest assured; latency is inconsequential as with USB you consider and compare Pianoteq default internal buffer performance, that occurs from MIDI messages sent from your keyboard, to, alternatively, the FP-90’s own internal sound generation performance.  Understandably it is inconsequential so long as your operating system has been aptly configured: https://support.focusrite.com/hc/en-gb/...Windows-10.

Matter-of-factly, the Focusrite Clarett 2Pre USB (a model offered without Thunderbolt) is also an economical option.

However, if you’re insistent and won’t budge about Thunderbolt, a card priced around $40 (US) is available from eBay: https://www.ebay.com/itm/ASUS-Thunderbo...%7Ciid%3A1.

Check if it’s even compatible: https://support.focusrite.com/hc/en-gb/...lt-3-ports.

I’m on a Mac and unfamiliar with your OS.

I’ve my own question, scherbakov.al and Mr.DJ, “Is the shown 0.7 ms latency a reference made to indicate the available round trip latencies or something just one way?”

Have you tried what you already have?

Too expensive for what it is. RME also has very good jitter figures.

I was questioning whether or not 1.67 ms round trip latency advertised by Focusrite is an approximate equivalent (short of 0.27 ms) to 0.7 ms stipulated by RME by it possibly avoiding any round trip references altogether.  When you consider 0.7 ms doubled at least hypothetically for a round trip figure (1.4 ms), Focusrite with its lower priced Clarett 2Pre seems an inexpensive alternative to the comparable RME product.

Added to the Focusrite Clarett's price-point-advantage, it comes with included DIN MIDI connectors built-in, unlike the RME which makes you pay extra just for breakout cables, if you intend to purchase it and use it as a MIDI interface or even for analog audio run through cables to your audio-in connected speakers and monitors!

Yes, but they're entry level and performance is at entry level too. Definitely not RME level.

Modern UAC2 compliant USB-Audiointerfaces transmit data in 125 µs microframes. This needs two buffers at the startpoint and two buffers at the endpoint of the USB-connection and results in a "Streaming latency" of 500 µs.

In other words 0.5 ms latency for crossing the packed-based USB-connection to the audiointerface.

Described in http://docplayer.net/37591176-Why-do-yo...ass-2.html

... or if you like it visualized in a 3-minute-clip:

https://www.xmos.com/developer/video/wh...io-class-2

You wrote "likely", but is there something exactly known?

Serial MIDI has 31.25 kbit/s, but even USB 1.1 has 12 Mbit/s and so can transmit data 384 times faster. How big is the latency of a MIDI-event, when a keyboard is directly connected via USB-cable to the PC?

And what is the MIDI-latency, when the keyboard is connected via 5-pole-MIDI-cable and a midi-to-usb adapter to the PC?

I did some tests of latency and comparison between an RME Babyface Pro and a Focusrite Scarlett on Windows 10 using the RTL utility:

With the Focusrite, I could achieve +/- 8,3 msec of real roundtrip latency at 44,1kHz, but I had to set the buffer at 32 samples to achieve that.

With the Babyface Pro, I have 7,3 msec of real roundtrip latency at 44,1kHz, but with the buffer set to 128 samples (2 levels higher).

So, if your aim is to have the smaller latency, then the RME is the best choice because it gives even better latency than the other popular interfaces but at a higher buffer size, which increase the stability of the system.

I usually use the RME at 44,1/64 samples. The feel and the connection with the sound is perfect, I just don't feel the latency. I can say that even at 44,1/128 samples I don't feel it.
Every attempt to go down is a non sense for me for that reason.

But. . .

A lot of people here are using USB to access PianoTeq on a computer, so they are USB packet-based, regardless.
(e.g. playing either directly with a USB MIDI digital piano OR via DIN MIDI cables to a USB interface).

For playing large chords fast, low 31,25 kbit/s DIN MIDI can get bottlenecked, creating timing issues that are different than systematic jitter. In theory, there is no MIDI speed constraint on USB other than what the USB protocol and systems bring.
___

1. In practice, who knows how different digital pianos implement USB MIDI vs DIN MIDI. Some USB MIDI digital pianos may use old boards to process sensor data for DIN MIDI then pass that data to a USB board.

2. Alternatively, DIN MIDI digital piano system is:   digital piano > DIN MIDI cable > Audio Interface > USB cable > computer

** Unfortunately, many systems likely suffer from BOTH slow DIN MIDI bottlenecks & packet-based USB protocol.

Good point. For a Mac it's among the best, I believe. And they even added a free rack of FPGA effects, some officially endorsed by Pultec.

From the VPC-1, are you using MIDI out via DIN cable or USB cable?

Thanks Scherbakov.

Did you ever try running the VPC-1 with DIN MIDI cables? If so, did you feel any difference in latency / performance?

I suppose the serial DIN MIDI cables are much slower, particularly for lots of fast notes. But that second USB computer input could have its own glitches that the RME drivers don't have control over. I've seen very little documentation for potential real-world differences.

No, I did not have the opportunity to try.

I tried running a 5 pin din from Kawai VPC1 to a Zoom uac2 interface (midi in) but found no improvement (or difference), though this is still running into a USB audio interface so not the best test. I do though use the external power supply with the VPC1 but I think powered over USB seems fine also.