X32 users, how to network?

[mschnell]

It's hard to believe that the USB interface of the X23/M32 generally does not work. I suppose this is some kind of incompatibility with the testing equipment.

A friend of mine (who is a computer consultant and embedded hard and software engineer and hence should know what he is talking about) does own an X32 and does work with Reaper (on PC, no Mac knowledge over there). Do you want me to ask him to do a test ?

To let me do that request: what exactly do you mean by "ASIO performance" ?

-Michael

[pipelineaudio]

This confirms exactly what Jim Roseberry found years ago, I figured they just improved it by now, but apparently not, and obviously, given the differences in Jim Roseberry PCs back then vs now, its also not system dependent.

I mean real time monitoring at the full 32x32 I/O count without dropouts. 4 years ago, and today, this tested works ok at 256 samples, but not reliably at 128, forget 64

Same system runs MOTU drivers in the same test config at 64 samples reliably and RME drivers at 32 samples reliably. Again duplicated by others, not just me, including the guru himself

The windows system audio performance issues are a silly something else entirely, but I am going to be running a separate interface for that, so no biggie I guess

[pipelineaudio]

Any idea if the behringer branded X-Dante card will work in the M32 instead of the Midas branded version of the card?

[mschnell]

Originally Posted by pipelineaudio

and today, this tested works ok at 256 samples, but not reliably at 128, forget 64

By "samples" you mean block size ? And latency set to one block ?

Same system runs MOTU drivers

Originally Posted by pipelineaudio

in the same test config at 64 samples reliably and RME drivers at 32 samples reliably. Again duplicated by others, not just me, including the guru himselfs

You tested 32 active channels in plus 32 active channels out at the same time as well on the X32, as on the Motu and RME ?

What sample rate ?

-Michael

[pipelineaudio]

44khz

And yes, 32x32 or it wouldn't be an actual comparison. Not at all my first rodeo with this sort of thing

[mschnell]

If you want to use Reaper with such interface in a "Live" situation (e.g. use the outputs as Monitors for the musicians), I suppose with 44 or 48 kHz, you do need a block size of 128 -> 3 mSec latency -> 6mSec round trip latency equivalent to 2m free air distance. Of course effects in Reaper should not add more than about 200 Samples PDC to that.

Even If Dante is perfectly stable, will it really provide much less latency ?

-Michael

[pipelineaudio]

Thats a good point, its important not to get mixed up between RTL at a given buffer, and what the actual lowest usable, safe, reliable latency is

For example if the Dante could handle 128 and was still exactly as slow as the X32 drivers, but the X32 needed to be at 256, then I would be at 9 ms vs 15ms for instance

http://kailuamusicschool.com/tech/ro...tency-roundup/

[pipelineaudio]

Well this was bizzarre. After reading a lot from Tim Gray we put a dante card in the M32 and got a new switch and ethernet cards

Dante has really strangely high reported latency in reaper, but does really weird things, like letting you run at 32 samples, with no more dropouts than 128 samples

Ideally, this was to get a lower usable, real world latency than the XUSB card was capable of, but I'm not really sure it gets there

[mschnell]

It should be possible to measure the round trip delay by means of nothing but Reaper. Maybe the easiest way to do this should be by ReaInsert, which seemingly can measure latency of external gear, but I never tested this.

-Michael

[pipelineaudio]

Dante experts: I'm getting a lot of clicks and pops, in some cases they end up on the recording, but mostly just happen during playback.

I'm running the Rednet PCIe-R card and the Midas X Dante card with a cisco SG 200-8 router. Lowering the X-Dante latency seems to make it worse, while lowering the PCIe-R latency has no adverse effect, even at 0.15. Neither DPC Latency Checker nor resplendence show any problems with the computer and Dante software shows no dropouts, yet I hear them and often they are recorded. The newer firmware menus on the cisco do not match the focusrite instructions for setting up the router optimally, is there anywhere to tell me how to check it and set it?

[Dewdman42]

I have been using an X32R with my MacPro and pretty happy with the USB DAW performance through X-Live card on the X32R. I have not tried 32 channels simultaneously at 64 sample buffer size, though, I personally don't really need that. I don't think I would expect any USB2 sound card to perform well at that spec honestly, but who knows.

Comparing the X32 USB to Motu PCI is not really a fair comparison to begin with.

Another option you might consider, rather then Dante, is to go with the Lynx AES16e-50 card on the computer. It can stream 32 channels over the AES50 that is built into the X32...and then you just avoid USB altogether. I'm not sure the buffer size, but people reported 3ms round trip latency that way. Its a $900 card, but you don't need any special card for the X32, which is also a bonus because you then can put some other card in the X32 if you so desire, such as X-USB or X-Live for situations where you're away from your computer.

[pipelineaudio]

That card looks great too. The XUSB card wasn't able to get 128 working comfortably either. The Dante card after some tweaking seems to be working ok at 128, definitely well at 256, but again, as its an internal PCI card, it should definitely compare to the MOTU and it really doesn't sadly, but it does work.

I may try the USB card again with the same tweaks as I applied to the Dante card and see if it can at least run at 128 ok

[Dewdman42]

Personally I think that in order to get Dante to perform you would need both the x32 Dante cArd as well as a dedicated Dante pci card for your computer. Total cost $1500.

Motu pci performs very well and always has. Partly because it is using its own proprietary cable and data transmission from the converters tonthr pci card.

If you try to use Dante from x32 and going into your computers ethernet then you’re competing with the computer’s resources in various ways and while you could tune some of that there is still going to be a limit.

With the aes16e-50 card, similar as motu solution, you have an isolated aes50 cable and data transmission from the converters on x32 to the pci card which like motu would be optimized for it. Total cost $900

[pipelineaudio]

I'm running the dedicated Dante PCi-e card in the computer, the Focusrite Rednet. It is over a decade newer than the motu PCI card and can't come near the performance

[Dewdman42]

There is no point to compare them they are different tools. Dante’s main advantage is if you are going to setup a dedicated rednet network, particularly for studios and situations where a lot of cabling would be a pain. Rednet has a lot of power and flexibility and many devices and ways to connect however admin is not for the faint of heart nor is the cost.

For someone just looking to get low latency recording to their computer it’s overkill and without proper admin and networking design it could perform worse then simpler solutions. They all have pros and cons. If the desire is to use x32 for converters into a single computer, lynx aes16e-50 is probably the best solution. It’s on my Christmas list. Motu pci is of course very very old and lacks many features of newer offerings, but it was mentioned earlier as a baseline latency comparison. Even now motu’s pci stuff performs very well but the converters are old and there is much better now. Because it’s going through the pci bus rather then Ethernet it can perform quite well, certainly better then usb2 and probably better then FireWire. Maybe not thunderbolt.

The lynx card has the same pci advantages and 3ms rtl from the x32 is pretty great WITHOUT any of the Ethernet complexities and definitely lower latency then motu pci. I suspect that the Dante pcie card would also perform well at a higher cost with the advantage of being able to tap into a larger rednet network if needed

[Geoff Waddington]

FWIW, I used to use a Lynx Aurora 16VT with their PCIe card and got stupid low latency -- think I measured 2.3 round trip or something like that.

Rock solid performance down to 32.

The forum and guys from Lynx were always great about driver install questions, etc.

I'm going Lynx Card for the win

[pipelineaudio]

Originally Posted by Dewdman42

For someone just looking to get low latency recording to their computer it’s overkill and without proper admin and networking design it could perform worse then simpler solutions.

People keep making this claim, yet every time, they cannot provide numbers that even under "proper blah blah blah" that are close to other solutions, apples to apples, oranges to oranges.

Dante over rednet IS PCI

Its PCI

Not magic. It should perform similar to other PCI solutions, yet nobody has been able to show that it does, they just say that it needs proper blah blah blah

[Dewdman42]

It’s only pci if you buy both cards, one for your pc and one for the x32, total cost $1500 plus you’re using up your one and only card slot on the x32. In order to get good performance you will need an isolated and optimized Ethernet. If you only use the two devices with a cable in between and nothing else then it should be pretty straightforward to get the most out of the pci bus. If you complicate it by trying to network with more devices or share your existing home network then all bets are off.

[pipelineaudio]

Why do you keep assuming I didnt...its PCI if its PCI

Period, you cant use Dante with the X32 if the x32 doesn't have the dante card (except over VIA, but who mentioned that?)

If you use the rednet pci card, its pci

I dont understand the argument.

Please dont assume people are stupid. This setup was gone through both by Focusrite and Audinate, alongside an excellent tech who was local for a few months here. Dedicated everything.

You can argue against some hypothetical strawman or you can argue against a system that was setup the best way it possibly could according to the companies who create them themselves. I'm telling you, it doesn't compare to other PCI solutions. The dante drivers are iffy.

The scary part is the dongle version (the PCI card) when compared to the software version, performs about the same once you take out the 8 msec they add to make you buy the dongle. Buffer size for buffer size, the PCI is just as glitchy as the software version

I'm not someone who doesn;t do EXTENSIVE testing of this stuff, its almost insulting how many times we've had to restate the system and keep being accused of using something eolse or some siliness. sheesh

[Gerrit]

Have you considered the possibility that the problem is on the X32 side?

I use a Focusrite REDNET PCIe card in my vintage Mac Pro and latency is very low:
Focusrite REDNET 1 over single switch Dante network to/from REDNET PCIe. Analog to analog true round trip latency measured using a two channel oscilloscope.

Sample rate is 96kHz

Buffer    Round trip latency
128        4.7ms
64         2.5ms
32         2.1ms

This on a mid 2010 Mac Pro (3.33 GHz 6-Core Intel Xeon, 24 GB 1333 MHz DDR3). A buffer setting of 128 samples will run fine on this machine even with lots of tracks.
BTW I never configured anything on the network switch and it's working just fine.

[Dewdman42]

Whoa relax pipeline, I was not accusing you of anything nor stating what equipment I think you have or don’t have, just commenting generally on the pros and cons of different approaches including cost.

Btw it’s not JUST pci. It’s pci plus whatever is happening to transmit the data over the cable to/from the x32.

@Gerrit, your performance is very good at 2.1ms RTL!! You have a lot of hardware how much did it cost? Partly you are getting such low latency because of high sample rate, the x32 only does 48k so how are you dealing with that?

Having your isolated rednet network is definitely crucial. Any other rednet devices or just the two connected to the switch?

[pipelineaudio]

The other devices could be a problem, but there most definitely are signs that the Dante PCI driver itself is the problem, especially the difference in performance across multiple systems until the ASIO system was switched to MMCSS Time Critical

that made a HUGE difference in the amount of dropouts and glitches, as it does on many crappy drivers like TheSycon for instance

I really do want to see network audio as the future, especially for a school like mine, but having it all in the hands of one manufacturer who already has an arguably shady dongle pushing practice is frightening.

I've been trying to keep up on all the network audio talk going into NAMM this year

[pipelineaudio]

I'm trying to figure a way to check if its something besides the Dante PCI card's ASIO driver that is the problem here. The Dante software shows no packet loss or errors, is there anything else I can check?

[cyrano]

Maybe it's time to try another computer?

You had problems with USB, you have problems with ethernet.

I've had no problems with either in, admittedly short tests. But that was on a Mac. Maybe something in the hardware of the PC you use isn't really up to par for audio?

[pipelineaudio]

I guess I didnt say it upstream, but this has been a test across a few systems, including two Jim Roseberry ones.

What I have yet to see is anyone plainly say that they are able to run 32x32 at 128. I keep hearing stuff about me...so really, has anyone been able to do this safely with Dante?

[SonicAxiom]

I'm exclusively using Dante hardware, all devices set to a Dante latency of 0.25 ms and yet near zero-latency is still no requirement for me because I can use my digital console (that gets parallel splits of all recording signals via Dante) to provide a latency-free cue mix while signals also get recorded directly. I'm simply not using live-monitoring in Reaper. This circumvents the entire low-latency issue alltogether. Reaper's ASIO latency could be anything, it simply wouldn't matter. I worked with DVS instead of the Dante PCIe card for almost two years. ASIO latency was 52 ms (!) and still no problem because the console delivered the live-monitoring for all tracking signals and mixed it with a cue mix (previously recorded tracks) coming from Reaper.

Only occasions where I actually need low latency and live-monitoring is when I have to record and monitor VST instruments. My ASIO buffer setting of 512 samples normally allows for this but I can also switch to 256 samples (6 ms) for this special task without ever getting audio glitches. Having a networked, digital console at hand eliminates the need for real-time monitoring through the DAW. The X32 should be more than capable of providing that. Who knows, requirements in pipeline's case may be very particular and negate offering parallel cue mixes through the X32.

.

[pipelineaudio]

Talking to Audinate again, asking flat out if this should be possible or if its asking too much