Been playing with Axoloti for a long while now and having so much fun with it. And it is also very easy to learn, cause of the visual programming environment. Just they way I like it
Last weekend I was at my friends house who has got a Kyma system. I brought my Axoloti and we "battled it out". Actually Axoloti is way more low level and you can essentially build more advanced stuff on Axoloti than you can on a basic Kyma system. Crazy.
But one things springs to my mind..... The sound quality of Axoloti. It just sounds extremely digital. And Yeah, I know it is not fair to do a 1-1 comparison on Axoloti vs. Kyma, because of price and other factors.
BUT
I am curious about if they might be anyway to "update" Axolotis sound, either hardware or software vise?
My thoughts:
Software: I know Axoloti has got some upsampling techniques that upsamples Axoloti internally. This helps a bit. But not nearly enough.
Would it be possible to change some hardware components on Axoloti to "beef up the sound"? Maybe the DAC? New teorrity for me. Sorry if this is stupid question
Would be really great to be able to play sounds louder than -14db, which it the current maximum level output. Beyond this, Axoloti just saturates/compresses the signal.
I rememeber seing @johannes mentioning somewhere that Axoloti uses consumer grade components. I am thinking for example the DAC. Would it be possible to change the DAC to a better one?
In general I am curious about making Axoloti sound better. I have a hard time making it fit with my other synths(G2, Microwave XT, etc.). Any thoughts and ideas are more than welcome
Just to be clear on my intentions: I have no intention of miscrediting Axoloti in anyway. I really like Axoloti and would like to get the best quality out of it. This is my goal of writing this. I just have a hard time combining it with my other gear, because of some of the things I mentioned above.
Also really curious about if there might considered making a "pro version" later on? With "pro-grade" components.
We need to separate two things: ADC/DAC quality and algorithms.
ADC/DAC quality
On the ADC/DAC side, the design goal is clean in-to-out, I mean there is no sweetening, soft saturation, or compression before ADC or after DAC. To evaluate the performance, it's easiest to use a quantitative approach, and analyze dynamic range, noise level, total harmonic distortion and intermodulation distortion. The vocabulary sometimes used to describe sound quality like "warm", or "open" are hard to relate to actual properties.
Input and output levels are around 2Vpp, that's consumer audio level. Pro audio level often has hotter in- and outputs, but that would have required higher power supply voltages, adding a lot of components and increasing the board size. The audio ADC/DAC chip ADAU1361 on Axoloti was selected because it integrates both ADC and DAC, programmable gain amplifiers, and has a 98dB(A) SNR spec for both ADC and DAC. Many other converters with a similar level of integration claim only like 90dB(A) SNR on the input.
So I suppose the -14dB you mention comes from "consumer audio level" versus "pro audio level"? Or - do you get distortion when playing out a full scale sine wave on Axoloti? Do you use the line out jack, or the headphone out jack? The headphone out is only recommended for headphones. There is an issue with board revision 1.0, if you connect a mono-jack into the stereo output jack socket, this causes distortion. One workaround is using a stereo to dual-mono cable to connect Axoloti to other equipment - that's how it is supposed to interconnect anyhow. Or in a situation where you only have a mono jack at hand, you could insert that halfway into the output jack, that taps the right output channel without causing distortion.
I don't think substantial improvements can be made by simply swapping some components. It is possible to connect an external ADC and/or DAC, for example an evaluation board from Cirrus Logic or AKM, but this requires firmware hacking. Here is a fundamental difference with Kyma: all technical info and firmware code required to do this is available.
Could you test, looping your MicroWave XT (or any other sound material) through Axoloti running an elementary audio-in-to-audio out patch, with the output level of the source properly matched to the input level of Axoloti, and listen to the output? Does it sound like the source or like the Axoloti?
Honestly, I think, when properly interconnected, the sound quality is fine. Some audio converters and gear specifies far more than 98dB SNR. While one can't really have "too much" SNR, I start wondering which audio signal sources justify a need for 120dB SNR, perhaps capturing the full dynamic range of a microphone without any gain adjustment ever.
Algorithms
With hundreds of different objects we need to get more specific. Generally speaking, there are a lot of possible trade-offs that can be made in oscillators, filters, waveshapers etc. Again, in contrast with Kyma or the G2, all source code is available, and variations can be discussed, developed, shared, compared...
And that's before we arrive at patching.
Maybe, post a link to a specific (solo'ed) sound that you'd like to approximate, and perhaps a patch that already approximates the sound, and challenge the community to make a better approximation?
I think its mainly algorithms, @grimmreefer have you tried @johannes 4 pole filter, thats an example where its sounds much less digital.
in the same way as for my Mac using the same audio interface etc, something like Diva is orders of magnitudes less digital than say a Max patch using the the basic saw/filter. Its also noticeable with something like Reaktor, with using things like Monark oscillators/filters. ... and don't get me on the difference between reverbs
(I think Reaktor used to be known for its digital 'signature', really only since Monark, and now blocks, has that perception started to change)
of course, the analogue modelling comes at quite a cost of CPU and resources, so you not going to get much polyphony from 1 axoloti and of course, then there is the not small task, of actually developing these more advanced modules, Kyma has been developed over a really long time. (I dont have a G2, but would assume the same)
on your question, I always wondered if, in the future, there is an opportunity for plugin developers to market 'patches' for Axoloti e.g. someone like Sean from Valhalla to create a reverb. of course they wont want to share the source, but Id probably buy a black box 'firmware' for my axoloti, that allowed it to be used as a 'programmable reverb' box. of course the issue is the market size is too small, say 1000 boards, but perhaps only 25% interested users = 250 sales.
as you say, no criticism of Axoloti here, just opportunities.
EDIT: a question, did you compare more digital things with Kyma/Axoloti? say FM, was there still a big difference?
There is hardly any corner to hide for any sort of copy protection mechanism, it would require the presence of secret decryption key in Axoloti (which there isn't). A firmware upgrade can't add such a secret decryption key as it would be visible in the firmware upgrade file. And even if a secret decryption key was present, it'd require fusing off the debugging facilities, otherwise you could just read out the decrypted assembly code or decryption key. Or someone potentially develops an object that dumps the secrets. So on the practical side I don't see how I can over a reasonably safe vault for binary code
The open-sourceness differentiates Axoloti from many other products. I'm not an open-source "crusader", I do respect people and companies that sell licensed, closed source dsp algorithms or software. On the political side, in this case I'm concerned it could impede the open-source development of similar algorithms.
Yes off course, but both affect the quality of what comes out of the device you are working with. Better components = better sound... Better algorithms = better sound(That is why i am talking about both things in same thread). I think this is why a professional compressor easily cost 3000 dollars and a Behringer compressor costs 100 euro. They might have the same signal flow, but better componenets make it sound better. Probably most valid in analog domain. I am just trying to figure out what is possible and what is not What might help gettin better sound out of the litte device.
Totally understandable about the power consumption.....
About the Kyma comparison. I know there is huge difference between Axoloti and Symbolic Sounds goals. Kyma is a proffesional sound developer tool and Axoloti is something else. Kymas cheapest system comes at a price of..... 3000 dollars. And that is truely a VERY basic limited system.... so of course it is not fair to compare those two on sound quality. But I see no reason not to be inspired by it.
Kyma obviously use quality components etc., as I talked about above. That is their thing. I guess quality of the internal part of what ever device we are talking about is related to price. For example the same friend of mine with the Kyma has got a Roger Foote compressor. Compressor for uber geeks. One knob on that devide cost around 50 euro!!!! These are used because they are really precise and they last more than a life time. But this compressor also cost around 3000 dollars.
But on the patching side, Axoloti is way more approachable and appealing than Kyma. Both is open source and you can write your own custom objects/prototypes in both devices. But it is NOT recommended writing your own stuff in Kyma, but possible. This is info from my friend who owns the system, I have not checked up on this myself... and also from what I understand extremely complex to create algorithms for Kyma and not much help to retrieve from anywhere. Axoloti is a lot more welcoming in this area.
I am not saying or concluding anything, just wondering what the reason might be that there is a limit to how hot the volume can get. Maybe not -14 maybe it was -18. Sorry dont know the exact number. But I know others have noticed too:
But yes, I was thinking this might had something to do with the components. But dont hang me for this, I am just guessing. and trying to find out what the limit -18db comes from. Might be algorithm, might be components, might be a mix of both? Again no critique, just curios about it, trying to understand
What I compared in Axoloti was basicly jus a simple granular patch, so sampling. Kyma sounded in your face powerful, no clicking even with the smallest grainsizes. But couldnt do nearly as complex stuff as Axoloti. Axoloti was IMO more fun to play with than Kyma, but Kyma just sounded better.
And by the way, great work @johannes & @thetechnobear I guess you have put some hours in this device. I saw someone was working on undo and colouring. Cant wait for that to be implemented(if it will).
Yes I have been trying some of these new custom objects out. They all benefit to better sound quality. Actually I use the oversampling from JHO(I think i was) just before the output of a patch. I think it is not intended to be used like that, but to me it sounds better (less flat) and thats all I need
Yes true. I guess that is down to algorithms and trying to recreate "non linearity" in a digital environment. Possible but not easy. And I think people dont give those secrets away for free as you mention. Some are more succesful with this than others I guess. And Yeah Sean from Vahalla is definatly one that needs recognition. Great plug ins they make. But sure for him it is a buisness and off course he wont give away his bread and butter. And off course this is a conflic area with open source.
Yes off course. They spend a looooong time on those devices. Nord G2 is also 24 bit 96khz internal processing. That I am pretty sure has got something to do with how "fat" it sounds.
No did not compare anything in more "scientific way". Next time Ill go visit him Ill do that. Make a few sound experiments, recreating same sounds on both systems, maybe even record and bring it back here for you to listen.
@johannes the encryption key that you are talking about, would that be for protecting algorithms from copying? Yeah that would make much sense in an open source environment.
But I am just personally thinking that there is something to gain on Axoloti. It is all ready good, but can be better. Always good to strive to become better
I wish I was a programmer and could make anything of the code in Axoloti. To contribute on that side but I cant. but maybe some ideas and thoughs can be shared anyway
Anyway, I have seen another device, small one a bit similar to Axoloti. That one has got an analog stage on the output. Maybe it was dmfx-1? Not as complex as axoloti I saw somewhere, that he has no plans in the future to sell this device alone, he needs partners to make it a product. I dont think that itll happen in near future. But everything is available on Github. If you want to, you can build it yourself. Anyway, back to the subject about the analog stage in the output. Could this also be a solution for axoloti? To have an Analog gain/drive/whatever on the output, to compensate for digital sound and maybe also to gain? I think this could be a great addition to Axoloti. Or maybe I could just build this myself and add it after. That is also a solution, I guess But it could be great to have control of this internally in axoloti.
Anyway, Axoloti is still early in development. Good things will happen, I am sure.
I love my axoloti's, and i think as far as sound quality what have you is concerned, it is as good as it needs to be. The one thing it needs, is MOAR POWAH! like, i run out of cpu/memory often when creating complex patches(and even more so if they are polyphonic), and there are a lot of more powerful processors in the same family that could do the trick i think, but im not an expert. I would totally pay an extra $100-$150+ for an axoloti with like, triple the processing power.
@debterminer, sorry dont agree wth you on that. I have many synths, Micro modular, Nord g2, Micro Q, Blofeld, microwave XT and Axoloti can in no way compete with those in quality output.
But in flexability Axoloti tops them all.
And yeah I wouldnt mind either to have a version with a lot more power either, but to me Axolotis "sound" is more important than more power. After all to get more power, you can just get another Axoloti. That doesnt help on my issue.
Quality of patching is a far bigger factor in perceived "quality" than anything listed about hardware spec, you list a bunch of VA synths that I've used or own(ed).
Taking the Blofeld for example, it's not highly specced at all and nearly all the stock patches (including the updated ones) are terrible with shockingly poor gain staging. Yet it can sound really good, digital or almost analogue (check Don Solaris' patches) when you spend some time on it. Waldorf also have such machines as the Streichfett, using a 72MHz M4 core, this box is well regarded for it's price because it has been programmed really well. It helps that it is dedicated to one function also.
I agree that the Axo can be a little quiet though, but with even the cheapest interfaces having super quiet preamps these days, I've never had issues with quiet gear (of which there is plenty). Most people record way too hot anyway considering how low our noises floors are with 24bit input/output stages and 32/64 bit internal mixing in DAWs.
Every single synth I have owned has got its own "sound", some examples again: Micro Q = punchy & raw Blofeld = soft and digital G2 = fat and powerful Virus Indigo = Clean & a little bit punchy, but nothing compared to G2.
Percieved Quality: if there was no difference in how chips, components, DAC sounds, shouldnt all synths sounds the same then? There is a reason that people buy different synth, compressor, etc. They have different qualities, they sound different. I agree with you patching goes a long way in making things sound fat. But if for example algorithms are not optimised or well written, I think all this wont matter. So a good combination of good quality components, etc. and well written algorithms, firmware, etc. goes a long way.
Ok, Blofeld was a bad example, haha Blofeld is a cool synth, but it is a "budget synth", created to bring Waldorf back in buisness. They were almost out of buisness and had to build a synth that was appealing, but with low production costs. Older Blofelds are extremely buggy, the pots fail over time, when connected via USB you can hear ground loop from computer in Blofelds output. From what I understand the newer versions of Blofeld doesnt have these issues, which is great. But mine, I think I have been through every single issues descriped in Blofeld user groups. Anyway, Blofeld also has some of the digtialness mentioned above, so it was not the best example. I like it but it has been a bit of a rollercoaster. But some of the other synths are good quality and doesnt have those "digital artifacts".
But having quiet preamps doesnt mean anything if the input synth has got noise and on top of that is low. If you gain the preamp 15 db, you also gain the noise floor 15 db. On G2 for example I dont need to use preamp, cause all gaining can be done internally. So no pre amp gain = noisefloor not raised. I think it would be great to have gaining internally all the way to 0db. I just dont understand what might cause this -15 db limit. There must be a reason for it, if it is either algorithms or components etc. I dont know.
Therefor I ask if anyone else might have knowledge of why this limit is there
I think you miss my point entirely, maybe I should have been clearer too though.
In this context by patching I mean combination of code/patching, whether that is on an axo with objects and their underlying code. On a dedicated VA synth the dsp code for the oscillators and filters are magnitudes more important than the DAC and output circuits (within reason ofc, we're not talking about 80s digital vintage here). Sound design is based around how these components sound, you could put some basic sawtooth / low pass code on a micro Q and it would likely sound bad too.
You can only get good sound if you put in effort, there is no point stacking extra output hardware on a device if the raw waveforms are not there.
Johannes explained why the output is low compared to more dedicated hardware. I've not needed anywhere near 15db gain to bring the axo into a very acceptable recording level, but I'm not trying to peak/peg the meter at 0. 98dB SNR isn't "noisy" at all.
For the record I just scoped the difference in noise floor between the Axo connected and just the same connecting cable terminated. A whole 1.1dB, I get more variance than that between some of the channels on the UAD audio interfaces at the studio.
I am not qualified to say what is most important, but I do know both things plays a role. And it was really not my point what matters the most. One might play a bigger role than another, but that I do not know about. There are reasons to why companies chooses specific DACs, converters, etc. for their products, that is my argument that choice of components also matter. Combined with algorithms of course, but both matter.
From my perspective is it not really about gaining it to zero db. I know I mentioned that. My perspective is that Axolotis does have a little bit thin & low output and I just want to figure out how to get the best output from Axoloti as possible. If there is nothing that can be done with the -15 db limits I guess that is just how it is. I'll just continue focusing on the software side of it. If you say it is mostly the code, then lets see what comes up. I do believe things can be improved with code and algorithms too and also on patcing level. There are all ready many things that can be done to make it sound better and trying new things out all the time.
A custom CNC lathe manufactured knob may justify 50 euros and look cool, but that does not affect the sound quality. On the other extreme end, the "mini" trend in the industry, like the volca's that have a 6mm diameter knob for filter cutoff, that impedes playability in my opinion. A 3000 euro system is like a privilege to own, I wanted to create an affordable device. Compromises are made in the design, and the maximum output level is the result of this. I don't believe the compromise made degrades Axoloti Core to a toy.
Sure. The analog part of the circuitry uses a single (non-bipolar) 3.3V supply voltage. It is not possible for such a circuit to generate a higher peak-to-peak output level than the supply voltage. No algorithm or swapping some components can boost the output level. But the non-pro output level is simply "fixed" by raising the input gain of the mixer it is connected to. Honestly I believe the audio quality of Axoloti is not really limiting, and if you do, I think you'd mention background noise level specifically, not general character. The sound quality comes from algorithms.
That's really guitar oriented, integrating a set of analog distortion/overdrive circuits.
An analog "sweetening" stage between DAC and output can add character but would prevent it impossible to sound neutral. I don't really believe there is a universal sound sweetening effect that is always desirable. My argument is, if this would exist, why only apply it once and not N times for even better sound? If I really need to select one, it 'd be probably the addition of an output transformer. Transformers are perfect unbalanced to balanced converters, killing all ground loop noise, and while transformer specs are far from ideal they - I believe - play an important part of the sound character of upper class preamps, compressors, mixing desks and eq's. But good transformers like jensen or sowter are expensive and bulky components... If you want to explore this road without DIY, perhaps try a passive DI box like Radial. Adding a +15dB output circuit to Axoloti Core would raise the noise floor too, the difference between this and +15 dB gain on a mixer input is that the cable in the former case carries a +15 dB hotter signal, making it less sensitive to interference. If this is a concern, converting to balanced with a DI box is the solution. Upsampling followed by downsampling before going into the DAC is not going to remove any aliasing or improve the sound, the benefits of upsampling/downsampling come from processing between up and downsampling compared to the same processing without up/downsampling. Otherwise it is only a subtle lowpass filter. Can't remove salt from soup with a sieve, so to speak.
That 'd be the difference in the algorithms at work. For instance, my latest granular128 uses triangular windows, those are more click-y by their nature than hamming windows, adding other window types to the code is possible. One day this will happen, whether I do it or someone else does. Different VA synths generally get their different characters from differences in the implementation of the oscillators, filters, envelopes etc.
Of course hardware choices matter, but at this point, we're at the stage of vastly diminishing returns in performance gains. You can buy an ADC/DAC for a few dollars from Digikey that outperforms most of everything from a decade ago. There are reasons that high-end studio gear manufacturers are focusing less on performance numbers and more on things like added DSP and routing/integration.
Descriptions like "thin" really are not helpful though and are verging on Hi-Fi enthusiast terms. List measured frequency response etc and issues can be looked at objectively.
Reasonable discussion though, just remember that this is an open system that is fairly new, we're all welcome to dig in and make what we can of it. Even if that means learning a few things.
No but the point is quality components do matter in their respective area. Better DAC = better sound. Better knob = more precise more durable.
Neither do I
About the analog stage output; I guess I can build it myself. Been looking on a few different solutions. Id just like to have the ability to gain on the instrument, so I am not depended on an external preamp. So an analog gainstage after Axoloti it'll be. Yes this it will raise noisefloor but I guess that is just how it is.
Just saw it yesterday, but havent tested it yet. Window type do matter a lot. At least from my experience.
Yes. I like a good discussion and open one too. Again, I am not trying to critize anyone or anyting. Just curious about what we/I can do I do like Axoloti. Fun and very deep tool. And I am sure it will develop over time. Better oscialltor algorithms etc. But that also uses more DSP I guess. Maybe sometime more Axolotis can also be combined to compensate for more DSP heavy objects. Just dreaming....
I think its hard to come to a real conclusion at the moment, since there is room for improvements in the algorithms etc, and still a lot for many of us to discover in the patching/coding side...
one possibility eventually, that perhaps will allow us to determine this in the future, is if axoloti supports digital outputs ( e.g. via USB) (actually I guess you could already record the digital output to a raw file) then you would be able to play this through your own DAC/amps and determine for sure if an 'upgraded' DAC actually makes a big difference.
I do think it makes a difference - Ive a Virus TI, and this (to my ears) sounds considerably better when passed through its DAC/amp to analogue output vs hearing the digital output. (which it can send directly to a DAW)
There is obviously some colouring involved on the output stage.
Of course, the key here though, is Access has decided on a 'signature sound' which is perhaps easier since its a VA, but would the same 'output stage signature' also be suitable for a DX7? and theres the rub, Axoloti is many things to many people, so I understand the idea of keep it 'neutral'
perhaps off-topic, I bought a HiFiBerry DAC+ for my PI2, which makes a point of stating it uses a Burr-Brown DAC... I might at some point, see if i can run a side-by-side comparison.
Yes true. there is still lots to explore as is now.
Yes there are reports of people not being to happy about Virus TI's USB audio. But a lot of people still use it. It is an option.
Yeah. My conclusion is that I need to add something after Axoloti. I thought about using an old Micro Modular I am barely using anyway, and use that as an post Axoloti gain/fx/compressor/whatever... And build a case for them both. Also dreaming about adding a micro computer that can run the editor for the Micro modular in the same small system. Micro modular editor is really old and only runs on older operating systems, so the micro computer has to be backward compatible with older OS's. But I havent really looked into if this yet, just an idea for now.
Regarding MM/NM. It has FilterBank module wich colours sound significantly in 'warm' way even in default state. I tried it with sonic potions lxr drum machine and got results that was even too 'warm' to me
