Archives for : ensoniq

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.

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:

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.

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!

Demos of a few dozen hardware Reverbs

Far from any scientific or “professional” test, this is just a quick bunch of demos when a reverb is pushed a bit harder, say into the 10 second decay time and only around* -6dB below main track. Don’t use it as a reference because results will vary depending on the recording levels at – take it with a grain of salt. All tracks encoded to FLAC (lossless format). Feel free to share if you find it useful.

Regarding the Akais, both of my units are expanded with their respective FX boards (it is not the same FX board as they are many years apart) and they can be used as a regular effects processors, while Kurzweil has a sampling board which again turns it into an external effects processor (it can do way way more than just a reverb, think of it as Eventide’s little brother). I no longer remember why I recorded two reverbs from PCM-70. I guess I wanted to display it’s less chorusy side of things and more closer to the rest of the bunch.

*yeah, some reverb tails might be a bit off. I actually mixed everything on an analogue mixer during the period of a few days, so probably some are louder than others. Sorry about that!


Ensoniq TS-10 wavetable and wavesequencing monster


If i would have to choose one ROM-pler to hit the category mysterious, it would definitely be TS-10. First of all i never understood why such high second hand market price (particularly in States). You would think it is because from the impressive synthesis capabilities of having both the wavetable and wavesequencing synthesis in one machine. But i am 99% sure that is not the reason. Even today (writing this in dec/2015) and good condition TS-10 unit can set you back over $1000 USD. Which is in a way funny because in Europe you can obtain it for around 400 notes or ever less if you look long enough. Unfortunately i don’t know the secret connection of the TS-10 and US, if someone does, feel free to add a comment. Personally I suspect the secret is: 1)polyphonic aftertouch; 2)session gig players who got used to it; 3) excellent build quality; 4) excellent sequencer (again gig players territory)

With TS series, Ensoniq continued their line of transwave synths, this time introducing the sample playback in the synth engine. The first thing user would check when exploring waveform content are the transwaves. And unfortunately all those good transwaves from SD-1 are gone. In fact, this synth has a weakest set of transwaves, of all Ensoniq’s transwave series. There is total of just 8 of them. But the worst thing is, they all sound almost the same. So, on the first sight it appears this is no good synth for transwave fun, right? Well…. wrong! We got some good news.

Sample playback in TS series is not just ‘basic playback’, but it also features transwave synthesis. If you load a transwave into TS-10, you can change its properties from the basic waveform into the transwave. Now all that is left is to route a controller (LFO, env, mod wheel, etc.) on to it and your transwave is ready for fun. And we got some more good news.

Since transwave synthesis requires extreme playback precision the same can be applied for basic samples (non transwaves). You can for example use extreme short loop points, and route sample end position to mod wheel. As you move the mod wheel, new harmonics are being generated. This works best on short, white noise samples. Or instead of mod wheel you can use random LFO for some really unique effects.

Another good feature this synth has, is that you can shift the loop point and ‘browse’ through various regions of your sample. This works best on complex samples, made from small snippets, vocals for example (connected in series) merged into one large sample. Route LFO or mod wheel and you got some of the craziest vocals at the output. Believe it or not, but even some high-end professional samplers do not have this kind of loop shift feature. Now you might ask – is this all we can do with it? What would happened if we would have one sample made of 64 or 128 small short (pure waveform) samples, connected in series and then we would apply a loop shift feature onto it? Ever heard of synths such as PPG or Waldorf Microwave? Well, that is exactly what they do! Welcome to the…

Wavetable synthesis. Although not from the default state available on TS series, is possible, once you build a wavetable. Technically speaking, TS-10/12 does feature wavetable synthesis, but unfortunately there is no Ensoniq software for creating custom wavetables so one would need to make it ‘manually’ with standard waveform editing software. Considering there are total of 128 waveforms, this can be a big work. Also, every cycle must begin and end at zero amplitude. This ensures smooth playback of each individual frame, since any amplitude difference between start and end point at such short loops alters the harmonic content or totally shifts it into wrong pitch. However, once you build it, the result can be quite impressive. In fact it is possible to gain much higher quality (longer cycle waves, more hi-fi sounding) than on a standard wavetable synthesizers. This is because a single wavetable on TS can be as big as RAM size in it. For example 1 MB wavetable contains a frame with a size of 8 kB. In the days of PPG, 8 kB was the size of the whole waveform ROM!


Some might ask how come this synth has Wavetable synthesis, yet its specs or manual don’t say anything about it – they only mention Transwaves. Well, transwaves are similar to wavetables, in many aspects identical, exept there is no interpolation calculation between to adjacent frames (waves). Single transwave is made out of 128 individual single cycle waveforms and no calculation occurs in between. In other words, what you put into is what comes out (aka garbage in – garbage out). You can’t smooth it out or change in any other way. This is what makes it different from a wavetable, along with the way the data is stored and calculated on wavetable synths. Typical stock transwave is usually made out of two major waveform frames, the first and last cycle in the transwave. However, if you have editing skills and a desire you can build any transwave you imagine, which puts this machine in the vicinity Waldorf wavetable synths and their cool wavetable banks. Unfortunately Ensoniq never provided anything remotely interesting as Waldorf’s wavetables which is probably the reason why wavetable synthesis never took off on the TS series. Kinda pity. Even the Waveboy disks and their custom wavetables aren’t much impressive (i bought them all and regretted). Still if you have patience, once you build a set of good custom transwaves, you’re in the business! And just when you though, this synth has so many cool features, we come to another chapter…

Wavesequencing – just like on the famous Korg Wavestation. Although called Hyperwave, it is basically the same thing. Offering the same methods and similar settings it has one additional and quite useful feature called crossfade volume point. As you might know, a volume loss naturally occurs in the center of a linear crossfade point and with this feature you can completely compensate it. That’s why TS produces constant volume wavesequences, making them completely undetectable – almost sounding like some kind of a morph. Of course, you can always set it to 0 dB to achieve the classic Wavestation-like wavesequence with volume loss.

Custom Transwaves (detailed procedure)
Lets now go back co custom transwaves mentioned at the beginning of the article in case you decide to put them into the TS you might encounter some problems. For example: if you want to add another layer or duplicate existing one. Once you load the sample, you can’t – for some reason. So you must do it prior to loading.Recently i found a way to transfer multi layer samples to the synth. Lets say you build few transwaves in the PC and you want to put them in TS-10 via EPS disk software. No problem, you save the sample, load it to TS and start to program it. But there is a problem. You want to add another Layer (to place the same sample there, but with different parameters for thicker sound) – TS-10 wont let you do that. So i found some really old prehistoric program called Ensoniq MIDI Disk Tools. This program requires Win98 OS, but can run on Virtual PC (Microsoft’s PC emulator for WinXP and Win7).It is a Demo version, but for some reason it will do exactly what you need (in fact, this program is for something completely different). With it, you can make a copy of existing layer and create another one (this is just a copy, so total size won’t increase!), or you can put another transwave in another layer (useful for Ensoniq Fizmo type of sounds). The trick is that this program operates directly on file. So it doesn’t matter if this is demo version, for what you need this program, will be already done even before you click exit.

Here is a procedure on how to create custom transwaves (works on ASR-10 too). This requires commercial program called Awave Studio, but if you are musician you probably already have this program as it can do 1000 other things when it comes to sample conversion:

  • First, to create a transwave use Tranzilon – nice and simple prog.
  • Then convert .wav to ensoniq .efe file via Awave Studio program.
  • Then create ASR-10 floppy with EPSDisk.exe and save this .efe file it to disk. Done!
  • In case you want multi layer, then before you use EPSDisk start that Ensoniq MIDI Disk Tools program (described above) and add 3-4 layers (or just use copy if you want the same wave, so the waveform stays in layer 1 and you don’t get unnecessary large file size). Remember, this program operates directly on file, there is no undo. So make a copy of whatever you do.

We need something to modulate all those transwaves, wavetables, etc. right? When it comes to modulations, TS offers one good feature called: modulation mixer. This is very similar to Kurzweil equation FUN’s where you can combine two controllers, apply scale and shape to one of them and get new controller at the output. With modulation mixer you can create really incredible modulators, some of them possible only on complex modular systems. Here is more info about it (from the manual):


Here are the available shapes:


Some examples (but possibilities are endless):


A couple of my patches
Originally i had idea to build a larger demo, but instead decided to build a soundset for TS-10 first, then do the proper youtube demo. However, not to leave you empty handed i found a couple of wavesequencing atmospheric demos that i did for the legendary web side (also known as Sealed’s Deep Synthesis for those of who still remember it!!). These are no ordinary sounds but mostly long evolving textures, demonstrating the Hyperwave function.

Regarding the soundset
I planned other demos but decided to put them in the Youtube video once the soundset is completed. Please don’t ask me when that will happen, though. It will be available on this same website. If it isn’t available then it means it hasn’t been made – plain and simple. 😉

And here is an excellent demo by thekyotoconnection that i found on a YouTube with TS-10 doing Hyperwaves and Pads. There is even a link below the video (on youtube) that provides access to the patches in that demo.

Which OS version to go for?
Stock version is in 90% cases 2.20. There has been some talk about OS3.10 being the better (latest) version, however from what I have heard one can not just swap the old EPROM chips. Some modification to the motherboard is needed. As soon as I find out what needs to be done I will publish it in here. Currently I have OS2.2. I don’t remember any specific bugs here and there, on this version of the system. Feel free to discuss below.

UPDATE: I have bought OS3.1 from here. Will have it installed next week. Fingers crossed everything goes well. I am also curious to check the sequencer MIDI jitter response after the upgrade.

ASR-10 synthesizing industrial sounds


I love ASR-10! Here’s a Roland TR-909 cymbal turned into a factory drone. I will provide a step by step guide here on how to create one. To get this atmosphere i used a standard TR-909 ride cymbal. First task was to loop it from around 30% until 80% (loop end). One thing i like about ASR-10 is that it features several crossfade methods for making loops smooth. Those are:

  • Crossfade Loop
  • Reverse Crossfade
  • Ensemble Crossfade
  • Bowtie Crossfade Loop
  • Bidirectional X-fade
  • Make Loop Longer
  • Synthesized Loop

For this particular purpose i used Synthesized loop since it adds its own flavor, depending on what Smoothness method you use. It almost completely removed the volume differences. After that point i copied the loop addresses and set these values to sample start and end address. After that i used Truncate Wavesample to keep just the looped part. Then i normalized sample.

To add even more flavor and bring up some harmonics from the background i then applied a function called Volume Smoothing which is a sort of a dynamics compressor that further removes dynamic changes making the sound more constant. This function has Smoothness option as well and i’ve used Fine setting. It took a while for ASR-10 to do the processing.

Now it was time to expand the sound into stereo field. I loaded the last 44.1k effect that comes on OS V3.53 floppy disk which is called “Parallel EFX” and used a preset in it called “Smaller Spaces”. It adds some industrial flavor as well.

Now it was time to resample it with this effect. I pressed the same note across three octaves (D1, D2, D3) and sampled about 5 seconds of it. Trimmed it to remove the 0.5 sec of beginning and about 1 sec of ending (to remove empty space at the end). Then i’ve normalized it. I’ve set loop mode to forward and that was it. No need to crossfade loop this time since there are a lot of harmonics and no click will be heard.

After that i’ve loaded “ROM-02 44KHZ Reverb” and used its preset called “Long Reverb”. Only thing left to do was to go to filter, set it to “3pole LP / 1pole LP” and manually open it. The result is what you hear.

ASR-10 synthesizing an organ

Here is a video tutorial I’ve made some time ago. If you use software such as Chickensys Translator it is very easy to transfer samples from a computer to an ASR-10. Sometimes a single cycle sine wave is all it takes to make a good organ sound. This video demonstrates the raw power of ASR-10. Alternative solution is to sample a sine wave, normalize it and proceed as explained in the video.

There’s a very useful thread on Gearslutz forum that details SCSI with old samplers and provides some solutions such as SCSI-CF adapters that can work with the sampler and can be used as a bridge between your computer and a sampler. Here is a link to it.

Let’s face it, we don’t use these samplers because of their specs, we use them for their sound. And so far there aren’t many alternatives on the software market. Investing into something like a CF drive is not a bad idea at all. These samplers once cost thousands for dollars. You can have them for a fraction now. Their sound didn’t degraded over the years at all.