There has been quite a lot of debate recently about AAX DSP plug-in development recently with Kush Audio announcing that they would not be continuing to develop AAX DSP plug-ins and then Sknote announcing a Kickstarter project because they wanted to develop for the AAX DSP market. A lot has been said and so we thought it would be good to find out some more about the new kid on the block Sknote and Quinto Sardo has agreed to be interviewed, so what follows is our chat over email and then a response from Michael Carnes of Exponential Audio.
Mike: Tell us a bit about your company, how you started and how you got into plug-ins.
Quinto: I have a strong maths background starting from school. Also I love music and I have been very passionate about sound synthesis from the beginning. When I started it was all analog and so I started looking at Moog synths, - Guardiano del Faro, Genesis, J.M. Jarre and listening to beautiful MiniMoog solos by Italian PFM (Premiata Forneria Marconi).
In 1991, I got my degree in Electronic Engineering, especially focused on ICT, digital processing and telecommunications. With home computers making their appearance, when they became able to process audio in realtime I started experimenting with digital audio technology.
My first job was with a new company here in Italy. I was in a small team and we designed, installed and tested transmitting satellite stations dealing with parabolic antenna, power station, control systems, waveguide paths and filters.
During my time with them, I started to work with microcontrollers and DSP chips. We used Microchip, Motorola and Texas Instruments. Applications were automatic tracking of satellites, digital signal processing for telecommunications, automated control systems. In my own time because of my interest in audio and music I started testing them for musical audio applications.
After that quite intense experience, that lasted seven years, I decided to switch to a different job and join the family business started by my father in 1968. We designed and installed industrial plant including air conditioning, electrical, and control systems, and I used my experience to design monitoring and control plants that we installed in critical environments like clean rooms, operating theatres and other critical environments.
I also managed quality control for a chemical laboratory for about ten years where I got ideas about audio too! For example, our level-independent dynamics algorithms come from my work on automated chromatography analysis.
I started in audio software development when VST and AU open standards were defined and programming frameworks made available. I developed several products for about seven years making it a job four years ago.
Ten years ago I decided to setup a recording studio with some friends and run it to offer recording and production services. The experience with tracking, mixing and dealing with musicians was absolutely essential to be able to develop a taste for audio effects and tools and understanding how they are used.
Three years ago I made the decision to focus solely on audio and extended into hardware and started to develop hardware devices for audio processing in addition to my software. My main goal was to offer a complete range of products and to be able to grow the company and create more jobs, as unemployment ins high here in Italy, even for people with high levels of skill and experience.
Mike: You are clearly successful in making plug-ins in AU and VST formats, what triggered you to look at AAX plug-ins?
Quinto: The Pro Tools environment is a huge slice of the market, and it is something we should have done some years ago, if I were really focused on sales. I had promised plug-ins for Pro Tools quite a lot of time ago. However other priorities, like starting to develop hardware and structuring a real company, stole the place of porting my products for far too long.
You cannot ignore that Pro Tools is the standard at a professional level, so my company's image, both from a user's perspective and to reach different groups of professionals and audio enthusiasts is very important and offering my products in the AAX format has to be the way forward.
Mike: You develop software and also several hardware products. Was hardware development crucial for your interest in AAX-DSP and the decision to try to follow this route?"
Quinto: Yes, absolutely. Hardware development forced me to look at a different class of problems and also to interact with people used to a different workflow. This kind of development and interaction gave me a feeling for hardware-accelerated systems like HDX. Stability, power, speed are crucial in some environments, which is different from some less time-critical situations.
Mike: Why did you decide to go for a Kickstarter Project?
Quinto: Whilst I was working on AAX software, studying documents and learning its structure for native versions, I was contacted by some Pro Tools users who were using our software through wrappers or didn't use it anymore because they switched to Pro Tools, as well as potential users, asking me about offering our products as AAX DSP, too.
My first answer was "no way, its too hard and too smaller market". Then, going deeper into AAX standard, looking at Avid hardware to get a view on the overall system and being pushed more and more by a few DSP users who really wanted something from us for their HD systems, I started thinking it could actually be doable.
You may be surprised that my main concern wasn't about costs and income but if a real HDX user base, interested in our products, existed at all. This is much easier for established developers with a lot of users of their existing native products, but we didn't have any users in the Pro Tools world, so even getting seen could be a serious challenge, without any idea about the potential feedback from a range of potential users.
So, a kickstarter came to my mind. A funding project could have had several objectives....
- Help us to acquire the hardware systems needed to test our AAX-DSP new products.
- Cover a pair of months of work for the start up.
- A kind of poll among Pro Tools HDX users to see if they were interested.
- A test to see if we were able to reach them to let them know we were going to exist with those new products.
We sat down and worked out a budget to determine how much we needed to ask for....
- Avid hardware 6,000 euro (developers get special discounts)
- Advertising and promotion 1,200 euro
- Kickstarter and Paypal fees 3,000 euro
- Taxes about 12,000 euro
- The remaining should cover 500 (imagining 100 x 500 euro pledgers) plug-in licenses and 2 months of developer time.
Mike: How easy is it to get on the Kickstarter web site?
Quinto: It is very easy. I kept thinking about that website specifically for a long time but it only became available from Italy this year. So now that they welcome projects from Italy, it is all about setting text, images and possibly video to grab people's interest. Once the project is loaded onto their site, there is just a fast test passage, where they check everything for content and conformity to rules which took three days for our project.
Of course once the project goes live, the hard part is to involve and get people interested and on board. One downside was that we started getting lots of emails from people trying to sell promotional services and such things, but in our case we had a targeted focus, so I looked for the most appropriate channels and I settled on targeted Facebook ads and Pro Tools Expert.
Mike: We have seen recently that Kush Audio have decided that it is not worth their while with AAX DSP plug-ins. What makes you think you can do it when developers like Kush have decided to walk away?
Quinto: I may be wrong here, but it is my understanding from researching on the internet, that Kush Audio doesn't code in house, they define great concepts and then hire companies to make the software, collaborating with them to finely calibrate the concept and retaining the intellectual property. While I want to stress again that I don't actually know how things work there, I believe it is a completely different situation from the way we work here, where everything is done hands-on in house.
With their business model, we need to find the right programmer, with a very specific set of skills and experience. It is not just coding for Texas Instruments chips used on the DSP cards, it is about understanding Avid's framework as well. Once the developer gets the product they paid for, they get just that. They may have paid for something more, though, the acquired experience for the programmer. You would then have to start all over again for a second product and so on.
We plan to start with DSP, where the efforts will be concentrated on the first stage (porting the first set of products) but for new products it will be normal programming having acquired the know how from the porting process. Programming for DSP has an overhead of course as we have to invest in tools to code for the Texas Instruments chips as well as an HDX Pro Tools test bed rig, but costs of these will be distributed in a very different way for successive products.
I also think our company image would benefit from this too, so it is not just about sales but about overall advantages in a long term perspective.
Mike: Your comment on our story about your Kickstarter campaign that “We were able to compile one serious test product for DSP in one week. One. Don't be fooled by wrong perspectives: DSP is hard to do if you don't develop in house and you are just used to pay somebody else to do it.” How can you justify this statement when some experienced developers took a long time to produce AAX plug-ins?
Quinto: A key factor is that we plan to start with AAX DSP, however I want to stress that I cannot comment on other developer's methods but switching to DSP may require some different skills and make the option hard to manage.
Another thought is that some products may become critical on DSP, for example algorithms relying on lots of memory like some reverbs and effects based on convolution and kernels. This case may add complexity to the mission, e.g. critical performances or added programming difficulties.
Another possibility is a company who has a huge range of products and want to do it all or nothing, instead of offering a gradual porting like we are doing.
One more answer may be a simple choice to allocate these resources for another system, like the more recent Intel based options (whose performances, compared to AAX-DSP, I don't know at all.
With regard to the the quote you refer to in your question, please note that I was referring to getting the core of one effect running on the chips on a Texas Instruments development system, which is still quite a long way from getting it running in the AAX-DSP framework as a fully functioning AAX DSP plug-in.
Another issue may come from how all the code was written to manage all the different formats. Abstract code, that could also rely on third party libraries, may need to be completely rewritten. Also strongly optimised code, including lots of Intel assembly language, could be another problem.
Mike: How is the Kickstarter campaign coming along?
Quinto: I've mixed feelings. The amount of pledges is still extremely low. At the same time I have three people discussing about the $5.000 option to have one custom developed effect. I heard these projects usually tend to get pledged at the last minute. In the mean time I was also expecting a wider discussion, with more potential pledgers. This is a "all or nothing" project (they are all like this on Kickstarter but they can be different on other similar websites). If the goal isn't reached, all pledgers will get all their money back but nothing happens too! The key here is that there's no risk to invest in "nothing".
Mike: What are the 5 plug-ins that you are looking at porting over to AAX?
Quinto: The original plan is to port across STA-limit, SDC, C165a, Disto and SoundBrigade which I believe would offer a complete palette of compression and colour, with well known characteristics, and something very special, and unique from the start.
However these plans are not cast in stone, I would really value users ideas and thoughts even whilst the kickstarter project is running. For example, if it looks like a lot of people are missing some special delay, we coud even replace one in the list with a brand new delay effect.
Mike: Can out explain the issues with plug-in development and why it seems to be easier for some than others (For example Softube compared to Slate Digital) in ways our community can understand?
Quinto: I simply think the difference of costs between developing it directly and hiring somebody could be huge in this case, because it is really far from "standard". If you consider the startup and investment, or if you just balance on a single product. It can also be just how each company allocate their resources, they may prefer to develop something new in the existing range of formats instead of doing this.
Mike: How long did it take you to get Disto running as AAX Native?
Quinto: I'm currently testing it on Windows and it took me about one week. However it may be that the testing stages may need be quite long because it is our first AAX product.
The good news is that the Avid AAX framework is very clean and well documented and the fact that it was born to be compatible with DSP brings a very clean structure (the DSP core, the part of the code running at samplerate, is separated from its parameters (for example the position of controls) and from its internal state, for example the memory for a delay that helps to keep code clean also for native.
Mike: How long do you think it will take to get Disto to AAX DSP format?
Quinto: We planned two months from the start to the first products. From then on with new products native and DSP should go in parallel. I do expect that for some specific effects it will be harder to program because of special features like the use of lots of memory.
Mike: In plug-in development are there cases where its much harder to produce AAX versions of plug-ins because of how the code development had been undertaken up to that point?
Quinto: With regard to native that shouldn't be a problem. But when you move onto AAX DSP then yes, you may have to re-think a lot of the code or, even worse, face some limits you simply hadn't on Intel.
Mike: What is the next stage for you?
Quinto: We'll put some more effort to be sure HDX users are seeing this and have the opportunity to choose. I'm also sure that if we'll be able to release an AAX Native plug-in before the end of the kickstarter that would help raise our profile. Also we will be exhibiting at the upcoming AES will be very helpful in the long time but it is distracting us from the immediate work in hand.
Response From Michael Carnes From Exponential Audio
I’m uncomfortable with the crowd-funding aspect, but not for the reasons you might think. I fear there’s tremendous room for misunderstanding. Perhaps the project is later than hoped. Perhaps a funder was expecting a different sound or workflow and finds that the solution isn’t implemented in the manner that was expected. These are the sort of things that can really crimp a developer’s independence. A schedule might be artificially compressed as a result of this pressure. An appealing feature might be removed. An appealing feature might not be added. I fear that there could be problems even with goodwill all around. I much prefer to adopt the risk and take my chances that I got it right.
It’s a good thing that Quinto has done some surveying of the potential market, but we all know that it’s easy to vote on a survey. The real vote is when someone puts down their money. A promise to do so isn’t the same as it actually happening. And I wonder if he’s seriously investigated the alternatives. Avid DSP is priced for the very high end of the market and the trend seems to be that many users find the DSP no longer necessary. UAD on the other hand has a number of well-supported products targeted at different segments, with a number of attractive price points. A given plugin can run on anything from a Duo to an Apollo 16—on virtually any DAW. If I had intentions of writing for specialty DSP chips, it would be hard to ignore them.
Finally I’ll make a comment on the technical side—at some risk of repeating myself. The ease in implementing an algorithm on specialty chips is directly related to the memory requirement. A DSP chip has a small amount of very fast on-chip memory. In most cases it will also be attached to off-chip memory which is quite a bit slower. The real trick is in making sure that you use the on-chip memory. Classic compression/distortion/EQ algorithms don’t need a lot of RAM, and can usually be squeezed into on-chip memory. This is even true at high sample rates (remember that memory requirement is directly proportional to sample rate). On the other hand, it’s very hard to get a modern reverb or delay algorithm into on-chip memory. And at high-enough sample rates there may not even be enough memory on the card! The programmer’s effort becomes one of managing memory, and it can be a huge effort. In cases (this was true at Lexicon), I had to sacrifice a feature or two just to make things fit. Think of throwing a passenger out of a balloon so you can make it over the mountain. Not a pleasant alternative. So it really matters what sort of algorithm you’re considering.
I do wish Quinto the best of luck and I hope that my cautions are overblown. Wouldn’t be the first time I’ve been wrong about something.
So there you go, what do think of what Quinto has said here, do you agree, or disagree. What do you think about Michael Carnes' response? Are you more or less inclined to support Quinto's Kickstarter project now you have all read this?