Hello there!
Reviving an old thread with fresh data nowhere to be found AFAIK:
According to my findings with a budget hardware USB sniffer, Steinberg dropped LTB (Linear Time Base) support in Cubase 9 (late 2016).
Testing was done under Win10 x64, from SX3 to 10.5 with a Midex-3 using the unsupported official Midex 3/8 Windows x64 drivers.
MIDI Clock and MIDI Note jitter tests were done with a MIDIGal running the MidiClock tester firmware.
When used with a Midex-3/8 interface, all previous Cubase versions starting with SX up to 8.5 are sending extra timestamp bytes with each and every timing sensitive events (Note On/Off, Clock, etc). Those are buffered into the Midex Cypress MCU memory for the embedded 8051 to work with, relative to its own more accurate and stable timebase.
When using a regular USB MIDI interface instead, those extra timestamp bytes are missing.
Same behavior with the Midex-3 being driven by any non Steinberg MIDI software sequencer.
Midex-8 is designed around a Cypress CY7C64613-128NC EZ-USB MCU combined with 2 x Exar ST16C452 UART to drive the 8 MIDI I/O.
Midex-3 uses a Cypress AN2131SC EZ-USB MCU only to drive its 1 IN/3 OUTS (with a funny electric relay that disables its by default MIDI Thru state once the interface is activated by a MIDI software).
Both USB 1.1 devices, built by BEK Systemtechnik.
NB: the Midex-3 must not be confused with the look-alike Steinberg USB-2-MIDI, which was a rebadged M-Audio MIDISport 2x2. Some people were mislead and because of this questioned the Midex line and LTB altogether, but it’s an entirely different product.
To give credit where credit is due, LTB, Midex-8 and Midex 3 were designed by Access Music, based on their Virus synthesizer adaptative MIDI Clock sync.
Here’s a quote from Christoph Kemper posting a message about this on the long gone access-list@ampfea.org back in the days (05/10/2001 02:16):
It’s time to do some promotion for the MIDEX 8 !
We have actually developed the MIDEX 8 and LTB for Steinberg.
The root idea for LTB (LinearTimeBase) was initiated by a bug in the Virus several years ago.
Some of you might remember it:
When the Virus Arpeggiator was synced to Midi Clock, the Arpeggiator slightly drifted away from the original beat; it just went a bit slower than it should.
It was a miracle to me, the only explanation was that either the computer does not send a small percentage of the Clock messages, or the Virus looses them.
But how could a system loose just a small portion of messages?
In the Virus this can only happen, when two clocks are sent immediately after another, in that case it lost the second clock.
But how can two clocks be sent right after another - they should have a distance of about 20 ms at 120 bpm? A midi timing jitter of up to 20 ms from the sequencer!!I tested this, and it was the jitter!
I needed to redesign the whole clock engine in the Virus, especially to have a smooth tempo measurement for the Delay, to sync it to the clock without producing clicks by that horrible jitter.
That new clock engine is LinearTimeBase, and it makes the arpeggiator and the delay unsensitive of clock jitter. It’s in the Virus since Virus A.Later I realised that we could design a Midi Interface, using the same technic.
I offered this concept to Steinberg, which is now the MIDEX 8.
The MIDEX 8 has a better timing than the AMT or Unitor working with Cubase.
When you drive a Midi-Interface without jitter-rejection by USB, this adds another one Millisecond jitter to the data stream, due to the USB framing.
The MIDEX 8 archieves a timing accuracy of better than 300 µs, that is the lenght of one Midi Byte - independent of the condition of the computer or operating system.
It bridges every timing information, and makes the computer working as a non-realtime machine.Try this: Everybody can measure the sequencer and interface timing using the Virus.
- Switch the Virus LED Mode to " — ". It’s the “secret” timing jitter measure mode.
- Enable MidiClock Send in your sequencer.
- Set song tempo to 140.
- Start the sequencer.
You will see one of the LFO-LEDs light up, showing that the Virus adapts to the song tempo.
After a while the LED’s should stop blinking.
- Every blink of the upper LED is a Midi Clock that came in too early
- A blink of the lower LED is a Clock that came too late.
That measurement is not very accurate, it only shows jitter greater than 2-3 ms.
Thus, they should not blink when driven by a good system.On software sequencers the timing is mostly independent of the density of the arrangement, but this must not be the case on hardware sequencers.
Test your favorite dense midi arrangement on a hardware sequencer.
Maybe the result is a surprise.Even on high midi traffic the midi clock should be stable, since it has the highest priority when transmitted by midi.
Reply to me, if you have any questions; this is an interesting topic.
Tell us your results!
Christoph Kemper
access music
So yes, when it came out in 2001, LTB worked a magnitude better than regular MIDI adapters (be it serial, parallel, ISA or PCI) under legacy or even more up to date Wintel systems.
The older, the better though.
As a matter of fact, external MIDI timing went unfortunately worse and worse at each Cubase for Windows iteration, to the point of LTB being irrelevant after being supported for 15 years. Other DAW software are no better on this aspect, some even much worse (eg. Mixcraft)
Same Win10x64 system, 120BPM 1/32 Note On jitter tests (percentage of standard deviation over the median):
Midex-3:
Cubase SX3 - 0,185% (LTB) - Clock jitter 2.753%
Cubase 6.5 - 0,937% (LTB) - Clock jitter 3,123%
Cubase 8.5 - 0,860% (LTB) - Clock jitter 3.095%
Cubase 10.5 - 2.209% (no LTB)
ESI M4U 1st gen Cypress CY7C64013-SC based with latest x64 proprietary drivers (i.e not in USB MIDI Class compliant mode):
Cubase SX3 - 1,071% (no LTB) - Clock jitter 3.816%
Cubase 6.5 - 1,285% (no LTB) - Clock jitter 3.440%
M-Audio Audiophile 2496 PCI:
Cubase SX3 - 1,496% (no LTB) - Clock jitter 5,880%
Cubase 8.5 - 1,398% (no LTB) - Clock jitter 5,585%
Cubase 10.5 - 2,261% (no LTB) - Clock jitter 6,804%
On the same system, overall figures are ~50% better under Win7 x64, with Midex still having the edge. By far.
Only beaten by pure hardware sequencers such as the Arturia BeatStep Pro or Keystep (~0,068%).
My old XP-50 Super-MRC sequencer only gets a 1,081%.
The Korg electribe 2014 (e2/e2s) are doing 1.215% (Clock jitter 1.841%).
Note that - contrary to the MIDI norm - all software sequencers are prioritizing Note On messages over MIDI Clock (there’s a setting to disable that in Cubase MIDI preferences).
Well, you get the idea. 
Pity that things goes backwards external MIDI wise, but kudos to Steinberg for supporting the USB Midex line up until today, even remotely.
Gotta love those retro-computers. 