On Sun, Aug 31, 2014 at 7:29 PM, Len Ovens <len@ovenwerks.net> wrote:
On Sun, 31 Aug 2014, Moshe Werner wrote:

all the recent talk about less audio interfaces being supported for Linux got me
thinking.
First I want to mention that I'm neither a programmer nor an electrical engineer, but
what if we could develop our own AI?

This is not the first time for this idea. There are one or two people working on it. The idea that seems to be the best is an ethernet connected AI because this seems to be the digital interface that stays around and is best supported. The idea is to use an arm based board with a netjack master and built in audio IF. The only project I know of is to at first provide stereo i/o as a proof of concept.

Interesting, I didn't know this. Can you send a link to it?


I imagine a modular approach with options to customize and add different features like
building blocks.
One user needs a lot of preamps, the other works mostly over ADAT or Madi...

A few comments about ADAT, while the format of the messages is well known, it is not open. Licencing may be a problem. True MADI uses obsolete HW. There are a number of audio over ethernet protocols with various openness or not.
 
But ADAT is one of the most important and widespread interfaces, so we can't just ignore it

Just a quick note about DIY HW. It is not generally cheaper than buying the same capability made by someone else. There are premade solutions out there, but they are not in the range of a lot of linux musicians.

I didn't say cheaper. I am willing to pay for decent hardware, and also software. I think I've donated more money to the Ardour project than I paid for Protools (I admit, I've also got a Mac running Protools, there are certain paying customers that still insist on that).
DIY gives you a certain amount of freedom, and quite some satisfaction. I'm just about to finish a Gyraf G7 tube mic, which was a lot of fun building.
But really my point is not so much to build the AI by myself, but to support a project that reflects the ideology that we stand behind.
I did pay a lot for my hardware, I would be much happier to pay it to someone that invests in our community and also gives real support for paying customers.
For now RME are the closest to this description but who knows for how long.



We could even implement a DSP chip with LV2 effects to lessen the CPU load.

Almost any arm based board would be able to add some DSP capabilities. It may also be possible to use a DSP board instead of an arm board, but the learning curve for programing it may be higher... like having a running up to date linux kernel and jackd running on it for example.


Maybe this is just wishful thinking, but it would solve one of the biggest problems for
Linux audio musicians.

The biggest limit to audio interfaces in basement studio is getting over the fact that a good audio interface is going to cost more than the computer it works with. This is first of all a psycological problem not technical.

I can see your point, but I don't totally agree. Cost is one matter but IMHO the main factor of frustration is the lack of support, and generally cold shoulder, we as a community get.
What will happen if RME decides do drop the support for Linux (whats left of it)? How long can we chase after laptops with express cards just to run the RME box.
Another thing we're missing is the diversity that exists in the "pro" world.
If RME would support their whole range of products under Linux that would be slightly different, but today we're pretty much bound to a handful of interfaces.
Now someone that has interest in Linux audio will be on the run at least when he finds out that his dear AI is not supported under Linux. (Happened to all my audio engineer buddies that saw Ardour, got excited about the idea, then asked if it will work with "name your favorite AI", I said not yet, that's it...)

Most of us are looking for those ~$500 solutions. That is, we expect to get 4 to 8 i/os with mic pres for $500 or so. That is why ADAT looks good to us. The Audio Science 8 i/o PCIe card is just over $1k but just has line in (my D66 is line level too BTW) so then you need mic pres. Most of us have supplied these with a mixer of some sort.

Really, an audio interface is two parts, The adc/dac, and the preamps. It is easy enough to do the line level input, even balanced.

My system consists of a 32 channel inline console->Alesis HDR24x AD->RME hdsp9652 so tell me about it :)

Going one step up, there are a lot of mic preamps around that provide a lot more control than a mixer or 8 input audio IF. Most of them have s/pdif or aes3 out. So maybe an audio interface that provides that makes more sense. However, the reason the generic mic pres work for most people is that they are using a set of dynamic mics with a few condeser mics (probably low end like mine) and the pre doesn't matter as much as it might with a ribbon mic for example.

So the problem is less technical than it might seem. (on the technical end, using netjack as an audio interface does use more CPU than the same alsa card used locally) To make such a project worth while it needs to have a wide appeal, this means cheap... I can't think of any other way to put it. The goal ends up being 8 i/os with 8 mic pre for around $500. Having something more stable and lower latency than USB2 might give you another $200 (maybe more) to play with, but writing OSx/win drivers would give a lot more room (netjack does this OOTB).

Totally agree, we're witnessing genius on a daily basis in this list, so I don't see a technical problem at all.
But I think the goal would be to bring something real new to the world, like a "stackable" or modular interface that every user could more or less configure using standard building blocks. That would cater the whole range of users in our community.
The problem is that I'm neither a programmer nor a hardware designer, I'm a user, so I'm just thinking out loud.

Best

Moshe