HX3 vs Mojo61 sound engine latency

[Kurt W]

Some years ago there was a discussion about the claim that the HX3 clone engine only needs 50 microseconds to produce audio:

https://forums.musicplayer.com/ubbthreads.php/topics/2701012/Re_UHL_organ_w_HX3_inside

 

Having some spare time this Christmas I got the idea to do some amateurish testing. Not being able to measure the absolute delay I ended up comparing the HX3 against the Mojo61. Connected midi out from Mojo61 to midi in on the HX3 expander. Configured the Mojo61 to trigger at the high,fastest, sensor. Connected the left audio out from HX/Mojo to my Novation Audiohub interface recording into Logic Pro X. The Mojo61 is triggered directly from it"s keyboard while the HX3 is triggered via midi from the Mojo61. Triggered one single key and found that the HX3 actually produced audio about 2 milliseconds earlier than the Mojo61.

 

Also swapped the inputs on the audio interface to be sure the measurement not was influenced by the interface or Logic. Next step might be to trigger both units via midi. I think I have a midi hub somewhere.

 

 

 

[Theo Verelst]

At 44.1kHz or 48,000Hz sampling frequency, 50 uS is about two samples, including a potential DAC built-in filter latency (could be a few to a few hundred samples, depending). Midi will always have a standard latency of about 1 millisecond for a single note on, maybe .6, depending on some implementation. That's always going to be the determining factor for the overall latency!

 

You could wonder about the constantness of the latency, that is: is there a way to time the Note On message to the degree of accuracy that corresponds with a engine response time of 1 or two orders of magnitude faster. In principle, you can build in a Midi Out interface and in a Midi In interface (I have done that) a very accurate, transparent latency for single notes, spaced at least the normal Midi timing apart. Probably that isn't the case in these synths.

 

T

[pinkfloydcramer]

On a practical note, can anyone here hear this difference? I recall reading a comment by a prominent organist that they could.

[Jazz+]

I feel it, practically speaking. It's why I would never perform with a laptop, let alone an ipad or iPhone.

[hardware]

I feel it, practically speaking. It's why I would never perform with a laptop, let alone an ipad or iPhone.

 

Exactly why I build my own 1U PC"s.

 

I had an ADK Pro when i7"s first came out, the Express 54 ported my PCI e connector card and the i7 920 was okay.

But Zebra2, Kontakt, PLAY, PTeq and Omnisphere needed 4GHz to be able to layer my instruments to battle the drums bass and guitar live.

Doing trio and solo work the VI"s work okay, but not in an ensemble.

 

Just replaced the guts with an 8086k from Silicon Lottery I picked up a while back and it only lets you add more synths which I don"t need.

The iPad gagged on a Rhodes and a B3.

 

So I have to go with what works.

HX-3, SE-02, PC and Strymons are here to stay.

 

 

[Al Coda]

... the claim that the HX3 clone engine only needs 50 microseconds to produce audio:

https://forums.musicplayer.com/ubbthreads.php/topics/2701012/Re_UHL_organ_w_HX3_inside

 

... comparing the HX3 against the Mojo61.

Connected midi out from Mojo61 to midi in on the HX3 expander.

Configured the Mojo61 to trigger at the high,fastest, sensor.

 

Connected the left audio out from HX/Mojo to my Novation Audiohub interface recording into Logic Pro X.

The Mojo61 is triggered directly from it"s keyboard while the HX3 is triggered via midi from the Mojo61.

Triggered one single key and found that the HX3 actually produced audio about 2 milliseconds earlier than the Mojo61.

 

Also swapped the inputs on the audio interface to be sure the measurement not was influenced by the interface or Logic.

 

With HX3, the 50 microseconds come from the engine being triggered by internal manual keys inside an organ like UHL or MAG, the keyboard actions being connected to HX3 FPGA based circuit board by Fatar scan boards !

NO MIDI involved !

 

OTOH, HX3 in the Expander enclosure and using MIDI is pretty fast too, but not the fastest,- which was already mentioned on german keyboardpartner site when HX3 board and HX3 XB-2 replacement kit were introduced.

They always said using MIDI for the manuals is somewhat slower than Fatar scan boards.

 

Next step might be to trigger both units via midi.

 

That´s the only fair comparison ´cause the HX3 is triggered by MIDI only and I dunno if Mojo61 is internal MIDI or direct (Fatar scan board or whatever) trigger system.

I guess, Mojo61 being triggered by ext. MIDI will be more delayed anyway ...

 

The reason why Mojo61 is about 2ms late vs HX3 expander might be Mojo´s overall DSP technology vs FPGA technology,- IF the DSP used in Mojo16 isn´t a FPGA too.

FPGAs are extremely low latency.

A modern SHARC DSP might be on par or comes very close ...

What do they use in Mojo61 ?

 

 

A.C.