i've contacted korg and arturia multiple times about various issues and bugs especially over the years... and your experience has been spot on to mine. korg never responds, and arturia is always friendly to respond but ultimately useless.

Disclosure: I design and manufacture sequencers...

I can give you some lengthy details about my design decisions and why there are limitations like bar length.
This may be too long and too detailed, but it is a topic I have lived...

So, just looking at the idea that "memory is cheap".

From the hardware side:
A hardware sequencer design will need to choose a core processor to start with. In most cases, for a sequencer, this will be a microcontroller, not a desktop microprocessor.
There are many reasons to choose one vs the other, both from the hardware perspective and the software perspective.

Things have changed/blurred with very small form factor modules that use desktop class microprocessors, along with all of the supporting components, such as memory, GPU, thermal management, etc.
Many microcontrollers today are also getting more powerful with more features, and can rival desktop processors. You can read more about that here:
https://www.totalphase.com/blog/2019...e-differences/
or here:
https://predictabledesigns.com/micro...r-new-product/

Microcontrollers in general allow programming closer to the hardware, with great control over peripherals (serial MIDI, USB, UI, etc). This can give more control over critical aspects like timing, responsiveness, priority, etc.

If I choose a desktop class processor, the next question will be "Why not just write it in software and use existing hardware?" This could be an iPad, PC, whatever and you get an advanced OS for free. It reduces cost, risk and support.
Most musical device manufacturers can't compete with the features, cost and ubiquity of something like an iPad on the hardware side.

The other factor is overall cost. Microcontrollers are cheaper in most cases than a desktop processor. You can use a rule of thumb that the cost of a component can be multiplied by a minimum 2.5x to estimate the impact to the final price of the device. Here is a good overview of that rule of thumb, if you want to learn more: https://www.youtube.com/watch?v=UwrkfHadeQQ

In most cases a device like a sequencer will use a microcontroller.

So, with more memory comes more heat, more power, more microcontroller I/O, more complex PCB layout... which creates more risk and more cost.

As an example, adding more memory could impact the following hardware:
- power supply
- power conversion
- heat dissapation, heat sinks, fans (!)
- PCB design, for traces, layers, impedance, timing, emissions, heat
- microcontroller, faster clock, more I/O for memory address and data lines
- additional supporting circuitry, muxes, caps, etc
- UI components, buttons, LEDs, display to make use of the memory
- the memory component itself, usually larger SDRAM

...and if you get any of these wrong during design, it is expensive to fix, or iterate on revisions of the hardware.

From the software side:
This may be an even bigger reason you don't see greater bar length in a step sequencer. To start with, it is important to think about the way data is stored in different types of sequencers. It was mentioned earlier, but the difference between a step sequencer and a list sequencer is crucial.

An overly simplistic way I like to think of this is the difference between Microsoft Word and Microsoft Excel.
If you think of Word as a list type sequencer, you just type as many words as you need and the memory usage is very compact. You don't really care about what happens "between the words". The sentence structure gives you the timing. In a list sequencer the timing is a timestamp on the event.

Now think of Excel as a step sequencer. In Excel you have a bunch of cells, that may or may not hold any data, but they still consume memory. The location of the cells give you the "timing".

Most of the older sequencers mentioned are list type sequencers which were limited by the total number of events, but they didn't give a lot of advanced non-linear step sequencing features.
In something like an MMT-8, two MIDI events consume only two memory "chunks", regardless of the timing. You could have one event at time zero, and another at time 3 minutes, it doesn't matter. The only difference is the timestamp values. In a step sequencer, more and more memory is consumed the farther away the two events are, there is a chunk of memory for each empty "step" between the two events.

So why would you want to make a step sequencer? As mentioned, a big reason is non-linear playback and performance features. Not just the rearrangement of large patterns, but playback directions, such as random playback, skipping of steps, etc.

At the step level, every new "capability" you add increases the memory consumed by each step, even if there is no active event data in the step. Things like per step parameter locks, step probability, step skip, ratcheting, microtiming, recalling a programmed pitch, etc. In a list based sequencer if you remove an event, it's gone. In a step sequencer, the pitch, velocity, length may still be there, they are just turned on or off. A simple detail like this has huge implications for memory usage, but may be desirable based on the focus of the device.

Getting into more of the gory details of the software (but still simplified), using a microcontroller, you may not have any operating system, or in some cases the OS is just a "scheduler". This means that the microcontroller is running the sequencer code directly. It also means there is no advanced memory management being performed (usually this is taken care of by the OS, or more advanced hardware). So the impact of this is that in general it isn't a good idea to use "dynamic allocation" of memory in the embedded microcontroller world. It leads to memory fragmentation that can't easily be defragmented because pointers to dynamically allocated objects could become invalid after defragmentation and the sequencer would crash. There are ways to mitigate this, such as only allocating, but never deallocating memory, or using a very large pool of items (events) and just taking or releasing these elements as needed.

Another side effect of using a lot of these "dynamic" approaches is that you start using memory pointers in a lot of places, which typically would be a 16 or 32-bit pointer, so this pointer must also be stored and increases the memory usage even more. You can read more about these ideas here: https://barrgroup.com/embedded-syste...ory-allocation

The other (non-dynamic) approach is just to create a sufficiently large, but fixed sized array that is dedicated to a pattern, track, etc. This guarantees that you always have enough memory, but it can be quite wasteful because the memory for this array is consumed regardless of whether it is used to hold meaningful musical data.

This array based approach is a common method to handle memory in a step sequencer, and a main reason you have bar limitations.
In this case, memory may be better used to provide more tracks, more patterns instead of sitting unused in the (infrequent) case a greater number of bars is needed. The patterns are chained together by the musician if longer structures are needed, giving a sort of "dynamic" allocation of the available memory.
It is a tradeoff.

Another area I won't even get into is how the data can be permanently saved and restored into memory...

Of course the difference between list and step sequencing isn't that black and white, so there are many examples of blending some elements of both types into a single machine, both old and new. Usually there will be other tradeoffs to be made depending on what the main focus of the design is.

If you need to see examples of this in various current sequencers:
The Synthstrom Deluge seems to handle dynamic allocation of memory and provides arbitrary length patterns. I don't know what the current level of advanced step capabilities are on the Deluge.
The Cirklon gives you P3 patterns which are step type patterns and CK patterns which are list type patterns. So you have the option to choose the type of pattern.

In general, I would say trying to compete with a PC running a DAW is a losing battle for a hardware sequencer if you only look at sheer memory, display, speed, PPQN for the price, and just acting as a MIDI recorder and playback device. So the hardware sequencer needs to excel at what the PC can't do so easily, which is provide a tactile, performance oriented interface which embraces some of the limitations. It undoubtably leaves some musicians wanting more, bar length limitations is just one of the many things I get feedback on.

I didn't even cover the complexity of the other aspects of the hardware/software that have to be balanced along with something like number of bars! Think number of tracks, timing accuracy, quantization (or not)...

Appreciate the kind words about Sequential! We do our best!!

If you'd like a more verbose answer I can give you some of an idea of how the compromises are made, feel free to ask questions and I'll answer what I can.

The short of it is, if you guys have particular implementations of sequencers you like and think need further investigation and expanding on the core ideas I'm curious. I have spent the last few years dissecting and building some feature/design ideas around sequencers that I haven't been able to explore fully yet. May be with Sequential at some point, may be a side project. Either way I'm curious what you think does this stuff or other complicated features REALLY well.

The complicated part of this era of music is allowing transitions and evolving music over a longer time frame. You have different parts of the market/industry covering different amounts of time. DJs are doing something over an entire hour, producers over a much smaller scale, sound designers over an even smaller scale than that. How do you bridge what a DJ does with what a producer does with what a sound designer does?

We're smack in the middle of all those camps, with the Tempest as a surprisingly effective transition between producing/sound design into a longer time frame but we didn't do it perfectly. I know Roger Linn REALLY wanted realtime playlist but I wasn't able to get it working in the final product. That's one of my biggest regrets with that one, especially after now a decade of reflection.

I'm considering writing up a longer blog post about my experience on that instrument, trying to bridge the visions of Dave and Roger was a pretty unique and influential experience on my path and some day I have no doubt I'll find the right time/opportunity to bring my own ideas into the world. Just trying to be patient and keep drawing up designs in my spare time

The Pro3 was a new exploration in a 16 step sequencer but it really isn't our core business. Most of our customers do keyboard performance and sequencers are relatively complex for them to use, not because they don't know how, but because a synchronized live performance is still incredibly difficult to keep simple and repeatable. Lots of single points of failure and you can't have that when you play live.

Chaining patterns is effective but, in my opinion, kinda boring. Too much work figuring it out ahead of time. There are much better ways to do that and have it evolve without it straying from the foundation of the track. Not sure how much I want to talk about that for the moment since I think I may be a few years ahead of the competition in my head on that one, but I'll think about it.

Having more bars does not mean it's easier to play live and musicians gravitate towards what is easy.

The path of least resistance

Artists act like a slime mold (yeah what an image eh?). Like water. We follow the creative flow towards what is easiest and, as an instrument manufacturer, we are quite aware of this. You have to make design choices that balance the cost of the instrument and the existing mental models of music production across a WIDE range of performers and artists. Not easy. More bars gives you an option but it isn't easy, it takes a lot of work. There are better tools for this (DAWs and workstations, etc).

Part of the reason I don't like workstations is I HATE looking at a screen when I'm being creative in the audio realm. I practice piano with my eyes closed because it gets me closer to the music. If I have to produce constantly switching my brain between an audio focus and a visual focus, it is infuriating. I tried to push the design of the Tempest toward button combinations I could remember easily.

You see that shift button on the upper left corner? It's there because muscle memory. The corner is one of the few places on a hardware box that is VERY easy to grab when you are in a dark situation. Because it is so essential, we added it as a second location. You have one handed operation near the main buttons, and dark/two handed operation using the corner.

So if you want to push the design of a multi-bar instrument in an interesting direction ask yourself this question: how would you design this for someone who is blind?

Isn't this fun?

Hi Pym, thanks for writing here, much appreciated!

Since you are a piano player, I'm sure you understand why that 4 bar can be a creative limitation. 1 solution would be pattern chaining. and to find a way, that notes do not get cut. another would be to expand to 16 bars (you know, 12 bar schemes and such...).

(Now linear recording for minutes is a different thing, but had be done on the MC 909 20 years ago (998 bars) and on the MPC since also decades.
It would be great to be able to do 10 minute drone recordings and such but maybe it is too complicated )

I too do not like looking at screens of synths and sequencers if I can avoid it or if they look to "mathematical", that's why I prefer some to others.

Since you are one of the designers of the TEMPEST: how big would the step have been to have 16 bars instead of 8. It's just something I like to understand (MC303 in 1996 32 bars, MC707 in 2022 8 bars). Would it make a box significantly more expensive? or, on the other side, is it so hard to program that overlapping notes do not get cut at pattern change? (I have not clue about programming, so I'm happy to be able to ask you here and I'm aware, that for drums long patterns are per se not as important as for synth lines with sustained notes)

_______



and yes, it is a very good question I also asked friends before (I know some people from my time in Berlin) how gear could be designed so even blind people can use it.

For drums an idea would be for example to have a pad like the NORD drum or the Roland hand sonic and then 8 buttons to chose a pattern, a button under it to arm it. so you can chose a pattern, arm it and then play in your pattern with a veto sensitive drum pad, maybe with 4 zones. for the sounds it could be done with macros, so you just have a few knobs to tweak the sounds. less is more here.

maybe this would also be possible for the tonal part. Like, you got 3 rings (3 octaves wich you can shift up and down ) with 12 sections (the 12 semitones) . the inner circle could be used for modulation etc.

or more conservative, the sequencer similar like for the drums, but with a keyboard jack to play into it with some midi keyboard.no need for editing. just set a loop of certain length and record. if you don't like what you recorded, record it again.

Real time play lists are awesome. Sequentix P3 has them. The list isn’t very long but you can keep changing it in realtime as it loops. Makes pattern chaining fun and is useful to build up arrangements on the fly.

Hmm so not as easy as "just add more bars" after all. Who knew.

The fact that I was able to get it to 8 bars with what we have was pretty amazing. I am so proud of myself for doing that within the constraints I had, barely losing resolution. There’s no way I could get 16 bars with the constraints I had (time/speed/memory/etc) without an amount of effort that is better spent elsewhere

Next time

Yeah the overlapping notes I really wish I could have figured out but this was my first time designing a sequencer, learned a lot, next time that should be easy to add

Perhaps adding the capability to an existing instrument but when designing a new instrument going from 4 to 16 bars is not climbing Mount Everest. It is a design trade being made. As pointed out ancient drum machines from two decades ago had larger sequencers.

Current manufactures don’t see a market for it. The OP would like more and is frustrated.

I felt the same way about keyboards that support polyphonic aftertouch. It use to be available from multiple manufacturers and then went away. Even channel aftertouch was dropped on synths like the Prologue that were billed as flagship at time of release.

Now ASM has released synthesizers with polyAT. I bought the original Hydrasynth and then the deluxe for that action. If they sell enough maybe others like Fatar will take note.

Similarly if a synth with longer step sequencer takes off maybe trend will change to include longer step sequencers or maybe not.

I did hear the OP has a Pyramid to use in the mean time

A very thoughtful post.

I attended a Barr Group embedded programming boot camp with my team years ago. It was a decent primer for our younger programmers and the instructor was really good.

As I stated in my other post adding a new capability to an existing product can pose a real challenge or be simply impossible. There would need to be more demand for those capabilities to become requirements for future products. Seems that demand isn’t there currently.

16 bars on a step sequencer is a crazy expectation, there's a reason almost nothing caters to that. Each step should count, a step sequencer is something that is programmed, usually for basslines and rhythms. Sounds like OP want a piano roll on all synths. Again, unrealistic. He should buy MPC Live and be set for every synth that will ever come out instead of experiencing perpetual disappointment. Not to mention the plethora of synths that have no sequencer at all - suddenly they are unlocked with the MPC's sequencer which is more than up to the task since it was actually built for this purpose.

What if you take two sequencers, and detune one of them?
Than what?

korg triton
motif XF
Roland G
Roland mc-909 - effectively unlimited step entry but with real time recording which makes it also a really good linear sequencer. really good for live performance, extremely ergonomic such that you can do everything with key strokes ( there is no touch interface but there is a very good screen).
deluge? I never used one but it is at least partially a step sequencer.

why am I even posting this, things won't ever change. I use my mc-909 every day but the era of such good HW sequencers is over.

I would LOVE for my roland 707 to allow me 16 bars. Though admit I'm some kind of outlier in the way I use it.

Geoff Mills sues you for ripping off his sound

In general I would say that the whole topic comes down to the question whether the maximum sequence length with full rhythmic resolution is limited to 1 bar, 4 bars or 8 bars.

And here, one bar is simply not enough for me, just a too severe limitation, but 8 bars cover a lot of ground for music production and song writing, with all the advantages of a hands-on sequencer. For me this sums up everything.

Some people have an anlytical understanding of music, chords, melodies. They can cope with 8 or 16 bars. They see the notes. On the other hand, an external midi sequencer over a standard midi cable will never have the precision of a build in sequencer for recording motions etc. Therefore, more bars are well justified to get the full potential.

An external sequencer will have all its interface space dedicated to sequencing, which is not true of any synth with an integrated sequencer.

Thats not true. A synth with one knob per function has all advantages in motion sequencing, see Korg Minilogue XD, Sequential Pro 3, Elektron stuff. Almost every knob can be recorded. Can there be a better interface than that? The groove does not suffer from midi limitations.

The awful old midi standard not only has problems to precisely play back midi notes with its low bandwidth, but also problems playing back cc messages etc. The required data rate increases quickly. Only solution is to use USB midi. But by far not all sequencers have that.

16 bars is no problem if you chain patterns

In my experience, despite the fact that the bandwith is huge in a USB interface, it is usually less solid than the classic one. Not sure if it is due to the USB itself or the heavily bloated modern OSes, though.

USB Jitter can be worse depending on setup. The thing I miss from classic MIDI DIN are the opto-isolators. No worry about ground loops.

What about this?
https://omg-instruments.com/wp/?page_id=52

I’m totally with you on the state of aftertouch in our world, but in the case of sequencers, things did spring up that superseded all the limits and constraints of previous sequencers and those things (specialized sequencers and workstations) still exist today. I think what the OP is asking for is that the sequencers on synthesizers get feature sets that rival sequencers and workstations.

It’s been mentioned why that’s not as easily said then done, but it also seems to miss the point of an on board sequencer. It would be like complaining about the speakers in your new TV. Yeah, they’re not so great, but to put a good system in a TV would require a lot of extra hardware and design, and ultimately some people are fine with what’s on board and for those who aren’t, there are plenty of options to go as crazy as you want.

That’s how things always are. You’re going to have some sort of baseline feature set that works for most people and specialized thing for the rest of us. I’d love to live in a world where every polyphonic synth keyboard was 61 keys or more with polyphonic aftertouch, and hell, while I’m at it, just put an Osmose style MPE system in them, for good measure, but that would increase prices for everyone to make a small number of people happy.

I understand the creative limitation of 4 bars and I'm ok with it. I don't just do classical piano, I actually am VERY experienced with electronic production and will be releasing a debut album soon enough from my explorations in this space (https://soundcloud.com/solkata will be updated soon enough). Combination of classical and EDM along with a lot of cultural influences from my travels. This is my primary focus outside of work right now as I wait for the Focusrite/Sequential stuff to mellow out and show me what direction makes sense career wise

I'm not interested in 10 minutes of drone music. What interests me is INTENTION.

I'll repeat that: What interests me is INTENTION

The music that impacts me the greatest is when every portion of it is intentional. This does not have a timeframe around it. Sound design impacts me just as much as a track as much as a DJ set on different time frames.

If I cannot find a way to encourage intentional use of the sonic space in a larger period of time, I will gravitate towards design ideas that move slowly in that direction, because it is a need in the industry, but mostly I try to design however I find helps you do things easier and more intentional in realtime. I don't like producing music. I like being in a flow state. It's a combination of discipline and training so I try to limit my design time to areas where I see a significant efficiency improvement from my own experiences.

The Tempest maxed out the memory because it was very important to allow all beats to be accessible in realtime. On the older MPCs you had to STOP THE DEVICE to do a lot of the things you can do on the Tempest. It's hard to see how amazing this improvement is on a hardware device if you haven't used the predecessors... you can't compare it to a PC.

It was very important to allow resolution up to 32 Triplets, and because we didn't have a variable memory solution we found it FASTEST (remember realtime is important) to just use a memory heavy implementation because we were running right up against the limitations of the speed of the chip.

To allow memory improvements I created a scheme that reduced resolution of the steps in a way that 99% of our users will never notice and it doubled the amount of memory we could use for the bars. Could I do that further and get 16 bars? Sure, but it would take a tremendous amount of work, far more than it took me to double it from 4 to 8 bars and we have to keep making new products to survive as a company.

All of the designs you are thinking about involve trade offs. You have to take into account your current user base and what they are comfortable with, then you have to gently guide them toward a different way of doing things

Note: I did NOT say a BETTER way of doing things. I said DIFFERENT for a reason. I cannot know whether or not it is better for you, I only know it feels better for me.

We all are learning and evolving this discipline together, which is why I decide to engage on forums on topics like this occasionally. It helps me clarify my thought process.

The best music is something that you feel is familiar at a core level but makes you feel or think about something in a way you never have before.

Same with an instrument.

I could not agree more. The best is for me simple ui with maximal artistic freedom.

And just to make it clear: I did not ask for more bars on the Tempest or blame you (who am i to do so) for "only" 8 bars. I took the Tempest just as an example / opportunity (since you were so nice to give us insights here) to ask you (so I can understand better), why bars are limited. And I hope for all users, that - if possible - gear to come will allow more bars again or at least that overlapping notes do not get cut at pattern change. but I guess I've said this 1 or 2 times before in this thread

and all the best for your album, looking forward to hearing it !

Yep, I didn't take any blame from your writing, all good, we're all doing the best we can! Happy to help you understand a bit better, it also helps me understand better.

Or maybe they know that most pros using their gear actually use DAWs for sequencing.

Which is true for Yamaha as well.
Remember that Yamaha initially put out the Montage without any sequencer at all, based on their discussion with a lot of pros. After a lot of internet noise (which I'd say was made mostly by weekend warriors rather than pros), they relented, but I'm still wondering what my MODX didn't get by having the dev team (and the computing resources onboard) busy with a sequencer I don't care about.

I am sure it's not true. It was a joke in response to the humorous post before it.




Both are no longer the same companies. The Kurzweil we love had Stevie Wonder , Ray Kurzweil and different ownership. Maybe Ray can upload his consciousness to the cloud and knock out another classic workstation with Stevie?
Roland also had different owner, staff , facilities, etc.

I usually find it more useful to look at the engineers and design team rather than the company to find stuff I like.



That's because the pros moved onto other gear, changing the market.


I think the market has split and changed. There's jus as much pro gear as novice gear.




That's happening with most goods and property. It's not just synths.

If the old stuff is so good why do you need anything else anyway ?

I like the TORAIZ Squid, for me, it's the perfect sequencer.
It's 4 bars. but can do down to 16! BARS (1/8th resolution then though)
The only thing it doesn't do is transpose whole clips with the harmonizer, only the active track, further it's non lineair heaven
This is sequenced live on the Squid (RD-9, Prophet-10, SE-02 and TD-3)
Tomorrow my Oxi One arrives, it has 384 step pattern length, let's how that goes.