AES50 advice w/ midas DL251 /Pro 2 + SAC

[Eric White]

Greetings and salutations...or something like that.

I have been a happy sac user for several years and I understand that someone (or several ) have had some success in using a lynx aes16-50 card to connect SAC to a midas or X32 stagebox. I am looking for advice before taking that jump. I work with national acts from time to time and am looking to add a Midas Pro2 to keep guest engineers happy. I also want to hook SAC into the mix to run AES50 to do opposite whatever the Pro2 is doing. Example -pro2 @foh,sac @ MON.

3 questions here -

#1 - Can i use the midal DL251 with SAC directly as my main go to rig stagebox setup instead of using these ada8k units and

#2 - when the DL251 is connected to the pro2 , will the generic AES50 signal from the pro2 work sufficiently on the sac rig?

#3 - latency - any ideas compared to the ada8k/adat setup we mostly use here?

Thanks a ton
-Eric

[dasbin]

Hi!

I think I'm that "someone" - or at least the first to use the DL251 with SAC several years ago. Not sure if others are doing it now.
In response to your questions:

#1 - Can i use the midal DL251 with SAC directly as my main go to rig stagebox setup instead of using these ada8k units and

Yes! I *only* use the Lynx cards with the DL251 in SAC.

#2 - when the DL251 is connected to the pro2 , will the generic AES50 signal from the pro2 work sufficiently on the sac rig?

Hmm. I don't know about that one. How would you parallel the digital signal from the DL251? I don't think it's possible. The DL251 can be placed into "mode 1" or "mode 2" depending on whether it is being controlled by PC software over the Ethernet, or by a Midas console - certainly not both at once. Maybe you can send a copy out of the Pro2 into the Lynx cards but I imagine there's probably added latency in that scenario.

#3 - latency - any ideas compared to the ada8k/adat setup we mostly use here?

Well, I can't compare directly, because I don't have ADA8ks, only the DL251 with the Lynx cards. The truth is that a 32-sample buffer always results in problems, and 64 is only usable on really light loads (no or minimal plugins) with the Lynx cards. Admittedly, the system is an i3 system from 2011, so you might be able to do better, but I was a bit surprised at not being able to totally minimize latency. It's always been acceptable for performers though, even when I had to go to 1x256 (in that case because I was using multiple IR reverbs as a design choice on a theatre production). The weird issue on this system is that I experience dropped buffers even when the CPU usage in SAC reports around 15-20%. I have to keep it under that. I don't know if it's the Lynx cards for sure, but I also tried them in an i7 with similar results unfortunately. I do know that they seem to compete with the graphics card for PCIE bandwidth somehow, at least the way SAC is coded - any scrolling through the mixer tends to result in dropped buffers if I'm on the edge. Don't have this problem with any DAW software; just SAC with its many peculiarities. I also *must* run SAC in realtime priority mode to get reliable results under 1x512 (yuck).


My main piece of advice for you is to upgrade the firmware on both the Lynx cards and the DL251. When I got the DL251, I was the very first person in the world to use it with a PC system instead of a console, I think, and I had all kinds of problems. I worked with Midas and they wrote new firmware a few times, sending it to me, and the upgrade process was a giant pain in the butt (connecting via terminal to the DL251's Linux boot process, interrupting it, and manually inputting the many memory ranges to be updated with the new version). I expect you won't have to do this with newer DL251's sold today, but send Midas an email first.
Stability got a whole lot better with Lynx card updates. Boot order used to only work if certain things powered up in a certain sequence, and the whole chain would stop working if anything went down or had to be reconnected at any time. With the newer Lynx firmware, that's all gone and it just works when it's plugged in, finally. Yay!

[Eric White]

Thanks for the detailed response as it will help for reference when I make the plunge. Regarding the pro2 , the normal way this rig is supposed to work with 2 consoles sharing 1 stage box is for the stagebox to connect directly to one of the consoles and then soft patch that console to output generic aes50 data directly to the slave console ( sac i hope in this case) over a seperate set of aes50 cables. Only one console has head amp control. Its not exactly a 'Y' of the data, its a reroute out to the slave. I have no idea how clock sync will be achieved. Dasbin, have you ever measured the latency directly from in to out with any software? I have used Smaart 5.0 to measure mine and i believe i remember it being right at 5ms latency at 64x1 , but thats been a few years ago now.

Edit - how many lynx cards are necessary ?


Thanks a ton -

Eric

[Bob L]

Drivers can be extremely important... we have a SAC system using an RME Madi Fx card that can run at 1 x 32 at 90% load without dropping buffers even when F-keying all around the screens.

So, your reported performance is far under the bar... something else in the system is potentially causing the grief... or else the Lynx drivers themselves may be the culprit.

Hopefully you are using XP... if not... that is the issue... the same system described above running Win 8 will choke at 29%.

You may also improve your results with the new CPU Blocking option added to the current SAC code... this may be able to stop the slipped buffers issue for you... give it a try.

Or give me a call on my contact number and perhaps we can chase down some items in your setup that can improve your end results.

Bob L

[dbarrow]

What is the CPU Blocking feature doing, specifically? Is it just restricting the CPU to one core or something more?

[Wurst Werner]

Concerning drivers-->


Lynx Studio Technology is proud to announce a public beta for new firmware and driver for all of our PCI Express cards as well as the Aurora-TB and Hilo-TB. This new firmware and driver will provide the ultimate in low latency performance (sub 1ms round trip @ 96kHz) for both Windows and Macintosh computers.
Supported hardware list:
AES16e; AES16e-SRC; AES16e-50; E44; E22; LT-TB: Aurora-TB & Hilo-TB.
Supported operating system list:
Windows Vista; Windows 7; Windows 8; Windows 8.1; Windows 10. 32-bit and 64-bit fully supported.
OS X 10.6 to OS X 10.10 (OS X 10.11 should work as well, but since it is not yet officially shipping we will not put it on the official supported list). 32-bit and 64-bit supported on OS X 10.6 and 10.7. 64-bit supported on OS X 10.8 to 10.10.
We recommend updating to the new driver first, then flashing the new firmware.
New features for the new firmware / driver:
Best possible low latency performance. Smaller buffers can be selected and still allow for nearly 100% CPU utilization.
Windows ASIO buffer sizes selectable from 32 samples to 8192 samples (previous support was from 32 to 1024 samples).
Full 5.1 channel support on Windows WaveRT (WASAPI; MME; DirectSound).
Elimination of the safety offset for OS X resulting in 128 samples of buffering removed from the audio path.
Fixes LynxRestoreState issue with Build 57A where the wrong scene would be restored on power up.
Windows Files:
Driver Build 22f: www.lynxstudio.com/drivers/V2Setup22f.exe (http://www.lynxstudio.com/drivers/V2Setup22f.exe)
Firmware: www.lynxstudio.com/drivers/FWUpdate.Win.20150807.exe (http://www.lynxstudio.com/drivers/FWUpdate.Win.20150807.exe)
OS X Files:
Driver Build 57B: www.lynxstudio.com/drivers/Lynx_OSX_57B.zip (http://www.lynxstudio.com/drivers/Lynx_OSX_57B.zip)
Firmware: www.lynxstudio.com/drivers/FWUpdate.OSX.20150812.zip (http://www.lynxstudio.com/drivers/FWUpdate.OSX.20150812.zip)
The driver and firmware have been tested quite extensively in-house, so we would like to open the testing up to everyone so you can try it out and let us know how it works on your system. We look forward to hearing your comments.

[Bob L]

From the Help file:

Force CPU Blocking
This option is designed to help reduce or eliminate slipped buffer problems on Windows 7 and 8. It can be used on a troubled XP system as well, but is mainly designed for Windows 8. Win 8 has changed the main thread scheduling system to attempt to even out cpu scheduling time across all cpus and all thread priorities. This allows other background processes to step on the main engine loop which requires as close to realtime priority as can be gotten inside an already non-realtime OS. When other cpu cores run code that accesses memory, all priorities are ignored and the main engine thread can be interrupted, causing it not to finish processing the audio buffers for all mixers within its designated looptime when under heavy loads. This results in missed data buffers causing audio glitches and distortion. The CPU Blocking code attempts to block the other cpu cores (up to 4 cores) from accessing memory during the realtime portion of the engine loop threads by forcing them into tight internal loops which only operate within the cpu registers themselves. This can have a significant effect in keeping the main engine threads from being interrupted as often, thereby allowing the audio processing to complete in time, and eliminate slipped buffers. Each system will react differently to these performance options. Test them individually and in combination to find what works best for your setup. This option saves with the preferences. If your system has more than 4 cores, it may be to your advantage to shut some down, in the motherboard bios, to use only 4 cores or less.


Bob L

[Naturally Digital]

Drivers can be extremely important... we have a SAC system using an RME Madi Fx card that can run at 1 x 32 at 90% load without dropping buffers even when F-keying all around the screens.

So, your reported performance is far under the bar... something else in the system is potentially causing the grief... or else the Lynx drivers themselves may be the culprit.

Hopefully you are using XP... if not... that is the issue... the same system described above running Win 8 will choke at 29%.Dasbin,

To add to this, FWIW I have run SAC on an HP xw4600 (X38 chipset) with 3 x LynxTwo cards and I was able to run all day long at 2x32 without slipped buffers.

CPU was E6850 (3Ghz) with lightly tweaked WinXP. Video card was Nvidia QuadroNVS290 with driver only installed.

Yes, different Lynx cards and the LynxTwo are PCI vs PCIe for the AES50 but maybe this info will inspire you to get to the bottom of what's stepping on that card.

[dbarrow]

Bob, what is your MADI FX card connected to downstream?

[Bob L]

The test system in my message was connected to the new Montarbo prototype I/O boxes.

The new Elton John system we just fired up at Caesar's Forum is using 96 chans of ADA8200s connected thru two RME Madi to ADAT boxes into a Madi Fx card as well. Everything is humming along nicely so far.

Bob L

[dasbin]

Regarding the pro2 , the normal way this rig is supposed to work with 2 consoles sharing 1 stage box is for the stagebox to connect directly to one of the consoles and then soft patch that console to output generic aes50 data directly to the slave console ( sac i hope in this case) over a seperate set of aes50 cables. Only one console has head amp control.

Ok, I haven't done that, and I imagine it would work. Will it add latency too? Probably some, but I'm not sure. I have my DL251 set to be the clock master. In your case you probably want the Pro2 to be the clock master (being as it is in between the two items connected to it) and slave the PC and DL251 to the Pro2. AES50 carries the clock alongside its signal.



Dasbin, have you ever measured the latency directly from in to out with any software? I have used Smaart 5.0 to measure mine and i believe i remember it being right at 5ms latency at 64x1 , but thats been a few years ago now.
I haven't, sorry.



Edit - how many lynx cards are necessary ?
If you just use one card, you'll only get up to 32 inputs from the DL251. Actually it's a little goofier than that and the DL251 might change how it divvies up the channels per AES50 port dependent on the current samplerate... hard to remember right now but I think the first port/card gets fewer channels when operating above 48Khz (they are moved to the second port/card). You get all the outputs from the first card in every case, though.
If you have two cards (like I do) you can make full use of all the inputs on the DL251 at any sample rate - again, they shift around where they appear a bit depending on the samplerate, which can make setting up the Lynx mixer software (already not a trivial task!) tough to figure out.

[Eric White]

Well, Ive bought the pro2 and looking to already include SAC in the mix (quite literally) and need advice again before i go the next steps. Dasbin, do i get to go up to the 96k clock rate with the lynx cards or am I stuck at something lower?

I also cannot decide whether to do a split snake and leave SAC as its own system with preamps or do the LYNX cards. Thoughts anyone??

Thanks -Eric