Rss

Finding the tightest hardware MIDI sequencer among a dozen (measurement tests)

Hardware MIDI sequencers have a rich history rooted in the evolution of electronic music technology. MIDI, or Musical Instrument Digital Interface, was introduced in the early 1980s as a standardized protocol for electronic musical instruments to communicate with each other.

The first hardware MIDI sequencers emerged around the mid-1980s. Devices like the Roland MC-500 and Yamaha QX1 were among the pioneering standalone sequencers. These early models allowed musicians to record, edit, and playback sequences of MIDI data, enabling them to control multiple synthesizers and drum machines in synchrony.

Throughout the late 1980s and 1990s, the market saw significant advancements in MIDI sequencing technology. Companies like Roland, Yamaha, Korg, and others introduced sequencers with improved features such as more tracks, better editing capabilities, and enhanced integration with other MIDI devices.

In the late ’90s and early 2000s, hardware MIDI sequencers experienced a shift with the emergence of computer-based DAWs (Digital Audio Workstations). These software applications offered more comprehensive recording, editing, and mixing capabilities, challenging the dominance of standalone hardware sequencers.

However, hardware sequencers persisted, appealing to musicians seeking tactile interfaces and dedicated performance tools. Companies continued to innovate, releasing units like the Elektron Machinedrum and Octatrack, Akai MPC series, and newer versions of the Roland MC series, offering unique sequencing approaches, sampling capabilities, and real-time performance features.

Fast forward to the present day, hardware MIDI sequencers remain relevant in the music production landscape. They often integrate modern features such as touchscreen interfaces, advanced MIDI capabilities, CV/Gate outputs for analog gear, and innovative sequencing methods, catering to the preferences of various musicians, producers, and live performers.

The evolution of hardware MIDI sequencers showcases a journey from the early days of MIDI technology to the present, where they continue to carve out a niche by combining the hands-on approach of hardware with the power and flexibility of modern electronic music production. But which is the tightest midi sequencer? Let’s build some custom cables and run some measurement tests to find out!

In this article the following devices will be tested

Hardware sequencers:
Akai MPC2500
Kawai Q-80EX
Roland MC-500MkII
Yamaha QX3
Yamaha RS7000

Hardware keys synths and samplers featuring a sequencer:
Ensoniq ASR-10
Ensoniq ESQ-1
Ensoniq TS-10
E-MU Emulator 4 Ultra
Korg 01/W
Kurzweil K-2600 RS
Roland XP-50

Computers featuring sequencer software:
Amiga 500
Atari 1040 ST
Mac running OSX and Windows 10 using RME UCXII and MPC Renaissance

MIDI Jitter
MIDI jitter refers to variations or deviations in the timing or regularity of digital signals. In the context of MIDI jitter can disrupt the accurate reproduction of the original signal due to timing inconsistencies. In data transmission, it can affect the timing of bits being sent across a channel. Jitter can arise from multiple sources. It might occur due to imperfections in the clock signal, signal interference, signal reflections, noise in the transmission medium and limitations in the precision of the components or the timing of the system.

In case of computers we need a perfect and dedicated USB bus that will not be interrupted by other protocols and services, which at some systems can be a difficult task to do (i.e. Windows 10 based computers). If the same USB port is being interrupted for some random reason every few seconds it will definitely have less “space to breathe”. Which brings us to the last parameter and these are drivers. So even if we are limited to i.e. the Win10 system a set of good drivers can improve things a bit. Looking at the graph with the results some might notice that Akai Renaissance is missing in the graph. The reason for that is, it would simply not fit, or if it would fit, the rest of the graph would be hard to read. This is a clear example of good vs bad drivers, in this case the RME vs Akai.

Measurements
All of the measurements were performed on a free software called MLA – MIDI Latency Analyser v2.1.1. It’s a piece of software to help us determine the effects that hardware, driver and software changes have upon MIDI latency and jitter. The program is also handy for identifying the ideal number of samples of offset to apply to MIDI tracks to compensate for round-trip latency when recording them to audio tracks. I am not the author of this software not related in any way to it, therefore I can not provide any sort of technical assistance regarding this software. A special hardware cable is required to be built before using it. If you decide to join the research, all of the details can be found on this address: http://tinyurl.com/midijitter

Offtopic: Some extra scores and some explanations
Yes, many of the keyboards feature hardware sequencers. This is why I included many of the hardware keyboards / synths into the measurement. I believe Ensoniq ESQ-1 was one of the first ones with a decent sequencer. It’s a pity it does not have some more features like setting the fixed velocity onto the recorded data or modifying the gate times. This is also the reason an additional (Features) column was included in the results table. Point is, if just because some sequencer scored high, does not mean you should try to grab it immediately, then send me 50 emails cursing me why didn’t I tell you the sequencer has nothing useful inside. This is the reason a scoring column called Features was added and it works on the following principle (how the scores are added). Please note the Features score column DOES NOT in any way relate to the MIDI jitter measurement results. If you’re curious here’s how the Features scoring column works: 

20% score = bare minimum Record, Play, Transpose, Quantize, Copy, Paste
+ 20% for advanced MIDI editing, change velocity, gate, note editing ranges
+ 20% for step recording
+ 20% for microscope edit
+ 20% for XoX style edit

So a sequencer that scored 100% has all of the above. Again this is just to make your potential shopping list easier, has nothing to do with the MIDI jitter results. And yes some of the rack synths and samplers have sequencers too! So they are included as well.

Atari vs Amiga – the final battle, which is better for MIDI?

To answer another potential question: Why including computers in the article titled hardware sequencers? As a reference point. Nothing more. Sort of to see where you stand if you run any of the computer + audio interface combos mentioned in this test. We will also have a privilege to see the battle of two 16 bit legends, the Atari 1040 and Amiga 500. For this test Amiga was running the M.E.D. software tracker, while Atari was running the Cubase 3.0.

Results
Before looking at the graph and the table one thing to keep in mind, the smaller the number, the better the result. Ideal number in this case is 0, but we didn’t test Expert Sleepers in here, so above 0 it is. First and top of the table we have the incredible Ensoniq TS10’s. These results and numbers are ridiculous, I agree. I repeated the test several times thinking I did something wrong. Even re-soldering the cables. The number are correct, TS10 has incredibly precise sequencer. On the second place Emulator 4 stored pretty impressive, but unfortunately has a very limited sequencer, so beware. As expected the Atari as a rock solid MIDI standard still stands well, so nothing special required to be said about it. Roland XP-50’s powerful 32bit RISC processor clearly shows up, with results even slightly better than Atari if we include the Max jitter (the max amount a note will deviate from the mean value). Another interesting “battle” of the grooveboxes, Yamaha RS7000 vs MPC2500. Yamaha came out far superior. Or did it? Check out the next chapter titled Individual MIDI hardware outputs vs Jitter as the things are not always as simple as 1+1.

Continuing with the graph we see the regular Mac computer running OSX Sierra connected to a RME UCXII. The results are essentially identical to the Atari. Something that can not be said for the same computer running Windows 10. While the average jitter results are fine, in the sub 1ms range, for some reason a few of the MIDI notes will jump as far as 3ms to the front or back. Don’t worry an average listener won’t hear it, in fact no one will, however if you layer percussive sounds on top of each other then a transient jumping back and forth 3ms (6ms in total) can be very annoying at times. Now keep in mind these are RME drivers (probably the best in the world!). But to see how bad things can go with Windows 10, see the entry in the table that says Akai Renaissance (hint: it’s on the bottom). This is an example why Mac dominated the DAW all of these years, at least for people who run external gear. With MPC Renaissance having plus minus 8, that’s 16 mili-seconds combined, that’s something even a non musical person can hear, say you lay down a pattern of 1/16th hats, this kind of deviation is way too easy not to miss. So yeah, Windows 10 and external MIDI gear, not my first recommendation, or if you have to, go RME interface. I should point out MPC Renaissance was tested only as a MIDI output interface, not as a software per se, and was running Reaper DAW for the test.

I know, you can’t see a thing. Please click on the graph to enlarge it.

Continuing with the graph we see the Korg 01/W which has an excellent sequencer (actually I tested the 01/RW), packed with features almost as much as Roland XP-50 (the later is slightly superior as it has RPS realtime phrase feature which speeds up things quite a bit). 01/W is closely followed by Yamaha QX3, a super complicated sequencer, at least for me who never worked on it before, so it can be very confusing. It looks cool though and is super tight. Next surprise was ASR-10 – it’s literally on a level of QX-3 and QX is a dedicated hardware MIDI sequencer just for that. I was quite surprised as I remember having some reserved thoughts for its song mode so I tested it as well. And I was right, after measuring ASR’s sequencer in song mode, the performance unfortunately drops. I didn’t want to include the data in the table, because most of the people use it in regular pattern mode. For those interested, in song mode ASR-10 is 0.347ms average jitter and 2.7ms max jitter putting it just slightly shy of MPC2500.

Next on the graph we have the Amiga along with Atari, a cult 16 bit machine that was most of the time used for “tracker” music but had a MIDI option using the serial port interface. The results are solid, but I never expected them to be stellar as Amiga has a set of many chips inside that require a lot of coordination – it was designed as a multimedia system. And this is where the sub 1 millisecond range ends and we are entering past 1ms area starting with Akai MPC2500 and Kurzweil K2600. I was actually surprised to see Kurzweil in here, I was expecting it near the top as Kurzweil is known for its “best of everything” approach. Followed by Ensoniq ESQ-1, and Kawai Q-80EX. Last but not least of the hardware sequencers listed in here came the Roland MC500 MkII, Espen Kraft’s favourite sequencer, there’s a cool video on YT check it out. It will throw away a note or two as far as 2.5ms, but for the 80’s soundtrack scores, it will do just fine. There is something magical about those tactile switches and the fact everything is there at a press of a finger, although it can go deep in microscope edit, hence good marks on the scoring table. The graph ends here, and is missing the MPC Renaissance for the reason already mentioned.

Individual MIDI hardware outputs vs Jitter
Akai MPC2500 has 4 MIDI outputs while Ensoniq TS10 has single MIDI output. So if you want to run 4 external devices with the MPC2500 you will still get results that are shown in the table, which is something that can not be said for a TS10 despite being far superior. Speaking about MIDI chaining, first of all, each additional device in the MIDI chain will add 1ms of delay, some might add even more and some might add totally useless data to their MIDI thru port. For example if you have a Yamaha TG-33 never place it as the first device in the chain, it will make the rest of your day pretty miserable. To avoid MIDI chaining problems you will need a MIDI patchbay. But then keep in mind the second part, which is that all of the output data still has to pass through one single MIDI port of our main sequencer on TS10. While 31.25 kilobits-per-second (Kbps) seems enough for a couple of MIDI notes, the moment you start sending control CC messages for several external moduiles you will soon reach the bottleneck of your MIDI interface. This is why a MIDI device with 4 hardware outputs, will in many cases or always be superior to a single MIDI port connected to a patchbay. My point: don’t dismiss the MPS2500 because of its position in the table, or think that TS10 will solve all your sequencing needs just because it is first on the list. Increasing the number devices chained to the single MIDI output will increase the MIDI jitter related issues, as the data will be more and more packed where there will be no more space left, and jitter will literally take over at one point.

The comment section welcomes any extra infos, anecdotes and stories related to this subject. So feel free to comment!

Comments (8)

  1. Stéphane Bertoli

    I know it’s expensive, but don’t forget Sequentix’s Cirklon, a wonderful machine (https://www.sequentix.com/).
    Regards

  2. Jack

    Thanks for the effort. I also have several devices with built-in sequencers and had a thought: If one were to sequence, say, the ESQ-1 using the internal sequencer, but clock it using the TS-10 or another device, how would that affect the ESQ-1 internal timing and any device further down the chain? Would the ESQ-1 thru be more prone to jitter due to CPU usage or less prone to jitter because the ESQ would be ignoring all data except the clock?

    • Don Solaris

      As far as I remember reading these old manuals, the internal sequencer still runs it’s clock but is corrected and guided through MIDI data which itself if very limited. I am very skeptical the jitter can be improved this way. I would be glad someone corrects me! 😀

  3. analogholic

    Oh wow…time to dig up the TS-10 then… 🙂 So cool when finding another way to use a machime you already have
    Thanks for doing the test

  4. Nemanja Smiljanic

    Very comprehensive tests, thank you for this!
    I am glad that someone like you had finally put a nail to the “Amiga can’t do MIDI”. Can’t do my a**. I wish I had these results back in the day when we had a war on Atari vs Amiga. I am not sure why we never got Notator or Cubase on the Amiga, as it was far superior music computer as it was basically an 8bit sampler.

  5. Poumtschak

    Hey Don!
    Nice job you did here.
    I take you tested MIDI ON messages and not MIDI CLOCK (which are less accurate)?
    480 attacks per minute means 1/4 note @120BPM, is that correct?
    I did some tests of my own with the MIDIGal Clocktst firmware: https://midisizer.com/midigal/midiclock-tester/ @120 then @175BPM to stress out the MIDI chain.
    The MIDIGal can probe MIDI Clock or 1/4, 1/8, 1/16 and 1/32 notes per bar.
    As a matter of fact, jitter increases with the number of events per bar, and the BPM value.
    Some sequencers also have sweetspot @certain BPM value, multiple of the 1MHz scan rate of their engine. The electribe 2014 are one of those (near perfect timing values @120BPM, much worse at odd BPM).
    In those tests, the XP-50 is good enough but nothing to write home about.
    Best figure I’ve had so far are with the Arturia Keystep and Beatstep Pro (latest firmare).
    When it comes to PC computers, best results were obtained under Win7x64 + Midex3 + Cubase SX3 (LTB), which are actually better than the XP-50. On the same computer and software combination, the PCI Audiophile 2496 is 4 to 5th time worse timing wise.
    Win10x64 or each Cubase version adds some latency too.
    Gotta test Cubase and Notator on the MiST box.
    I have a spreadsheet of values, will post it somewhere someday.
    Anyways, Cubase with LTB being tight at last, I’ve been reworking my old Super-MRC XP-50 sequences in VST5, with much fun!
    (FYI Awave Studio 10.2 can batch convert the XP-50 SVQ files to MID)
    Keep up the good work!

  6. Poumtschak

    http://hinton-instruments.co.uk/reference/midi/promidi/pg05.htm

    “Scheduling Delays: most computer based sequencers operate on an internal time interval, usually around 2.5ms. If the tempo is accurate to several decimal places this technique is undoubtably used and any MIDI event that misses one interval has to wait until the next causing a quantisation dependent on how busy the sequencer is.”

    This also applies to hardware sequencers.

Leave a Reply

Your email address will not be published. Required fields are marked *