Confusion on Buffer settings...

[Jeff Scott]

Hey Folks...Just got a RME Digiface with PCIMA Card for my Laptop. Hooked it up to a rented ADA8000 and started to play with the demo. I have read a lot about the different Buffer settings that various people use and have read Bob's SAC Manual Help file but I am still a little vague. Wondering if someone could elaborate on what the 2 numbers mean...how changing them effects the sound / performance ...and how do I know if I have achieved the best setting for my system?

[905shmick]

Section 12.7 of the manual.

PreLoad Buffers

This setting controls how many buffers of data are preloaded to the audio device before output streaming begins. This affects the response time of live controls such as mutes, solos, faders, etc. If this setting is too low for your system, audio stuttering or glitching during playback may occur,
especially when minimizing or sizing windows, etc. on a complex session. This setting combines with the Out Buffer Size setting to determine the final latency.

Use the PreLoad Buffers listbox to choose a setting between 1 and 6. Lower values will give you faster response times.

Buffer Size

Use the Buffer Size listbox to choose from six different settings ranging from 32 to 1024 samples.

The combination of the Buffer Size and the PreLoad Buffer value control the latency of the real-time engine. Lower latency values give live, instantaneous response and the feel of a real physical console. You eill need to acheive settings of 3 x 128 or lower for a realtime live audio console feel.

Some combinations that result in the same low latency value will cause the system to glitch while other combinations will perform comfortably. For instance, 2 buffers at 128 size might cause static or glitching on certain audio devices, but the same latency value can be obtained by using 4 buffers at 64 size, and the audio device may play back perfectly.

[DominicPerry]

Smaller buffer sizes require more CPU to process, so 32 sample buffers are hard work compared to 64 etc. This means that if you get your system to run without glitches at 4x32 and at 2x64, then you are probably better off using 2x64 as it will require less CPU.
Bob also recommends that you use at least 2x whatever buffers you try, i.e. 2x32 or 2x64 or 2x128 rather than 1x32 or 1x64 etc because there is no room for error with a single buffer.
However, IMHO, if you have a powerful machine and you can run at 1x32, then I would do so, because you will get no complaints about latency - 1x32 really is incredibly fast.

Dominic

[905shmick]

We're running 2x64 with IEMs and there's no noticeable latency.

[Trackzilla]

to explain another way, # of buffers X # of samples in a buffer = how much latency the system has from audio going in through audio coming out at a given sample rate.

The lower the latency the faster the CPU has to to crunch all of the math to keep up while simultaneously remembering it is a computer ;)

the higher the # of buffers, the more that data stream can be temporarily interrupted by another process and still recover before the information going out fails to keep up with that coming in.

The lower the number of samples, the less time it has available for each individual interruption of that stream.

Insofar as percieved latency, 1X256, 2X128, 4x64, and 8x32 will all be the same, but 1x256 will give the CPU dramatically more time to take a vacation in between chunks of data in the stream...8x32 will consume more of the CPU's attention in an attempt to keep up, but allow it to screw up 8 times before the output is affected.

Every computer has a different opinion of just how much vacation time it needs, you must experiment with various combinations to get the best result for your channel count, but in short order you will become comfortable with knowing how much load and instability is created on your rig by adding tasks to the data flow.

[RBIngraham]

I would also add that it depends on what Sample Rate you are working at. All of these buffers are the number of buffers x by the number of samples in that buffer. So if you are running at 96KHz then the same number of "samples" is actually a much smaller time value than it would be at say 48Khz. I doubt most of us are running at 96KHz anyway so it likely doesn't make all that much of a difference, but it's worth understanding what is going on there at least.

Richard

[gdougherty]

We're running 2x64 with IEMs and there's no noticeable latency.

Seconded. 1.5ms is plenty fast.

[gdougherty]

I would also add that it depends on what Sample Rate you are working at. All of these buffers are the number of buffers x by the number of samples in that buffer. So if you are running at 96KHz then the same number of "samples" is actually a much smaller time value than it would be at say 48Khz. I doubt most of us are running at 96KHz anyway so it likely doesn't make all that much of a difference, but it's worth understanding what is going on there at least.

Richard

I couldn't hear a difference between 48Khz and 44.1Khz on my system so I stuck with 44.1Khz. It also makes recording for CD direct and simple with no dithering necessary.

[RBIngraham]

I couldn't hear a difference between 48Khz and 44.1Khz on my system so I stuck with 44.1Khz. It also makes recording for CD direct and simple with no dithering necessary.

Yep. I've never heard any difference in using those 2 sample rates. They are so close that I think it matters very little.

In fact I don't really hear much in going to 96K even. But I'm guessing that I just don't have a high enough quality monitor system or the mics or the preamps, etc.... that you really need to notice any difference in going to higher than the 44.1 or 48 K sample rates.

I was just noting that as your sample rate increases obviously your latency will decrease as the same size buffer is now a much smaller amount of time. It also means you'll use more CPU of course.

I wasn't really trying to get into the great debate of what sample rate sound best. :)

RBI

[gdougherty]

Yep. I've never heard any difference in using those 2 sample rates. They are so close that I think it matters very little.

In fact I don't really hear much in going to 96K even. But I'm guessing that I just don't have a high enough quality monitor system or the mics or the preamps, etc.... that you really need to notice any difference in going to higher than the 44.1 or 48 K sample rates.

I was just noting that as your sample rate increases obviously your latency will decrease as the same size buffer is now a much smaller amount of time. It also means you'll use more CPU of course.

I wasn't really trying to get into the great debate of what sample rate sound best. :)

RBI

Agreed and a very good point on the timing at sample rates. The quoted figures are probably all at an assumed 48Khz max since 96Khz cuts our channel counts in half. I'd much rather have 32 channels, even at a slightly lower quality if it were audible.

[RBIngraham]

Agreed and a very good point on the timing at sample rates. The quoted figures are probably all at an assumed 48Khz max since 96Khz cuts our channel counts in half. I'd much rather have 32 channels, even at a slightly lower quality if it were audible.

Sorry to be a nit pick here...

But how many channels of I/O you get at 96 compared to 48 or 44.1 is totally dependent on your sound card. Of course many are using RayDat cards here and that would cut the channel count in half since ADAT optical can only do 48K max by default, and if you use the SMUX protocol it will cut the channel count in half.

However if I had 32 channels of Echo Audiofire units I could run that system up to 96K sample rates and it would have the same number of I/O. But those units use all analog I/O. Actually the Audiofire 12 units can even do 192KHz sample rates.

Now why anyone would really want that I have no idea or good reasons to defend. :p

Just saying it is possible and the sample rates and channel counts will totally depend on what sound card you are using and what I/O format that sound card uses, etc, etc, etc....

Richard

[gdougherty]

Sorry to be a nit pick here...

But how many channels of I/O you get at 96 compared to 48 or 44.1 is totally dependent on your sound card. Of course many are using RayDat cards here and that would cut the channel count in half since ADAT optical can only do 48K max by default, and if you use the SMUX protocol it will cut the channel count in half.

However if I had 32 channels of Echo Audiofire units I could run that system up to 96K sample rates and it would have the same number of I/O. But those units use all analog I/O. Actually the Audiofire 12 units can even do 192KHz sample rates.

Now why anyone would really want that I have no idea or good reasons to defend. :p

Just saying it is possible and the sample rates and channel counts will totally depend on what sound card you are using and what I/O format that sound card uses, etc, etc, etc....

Richard

True, true. I guess I was figuring on the standard setup around here of an interface of some make with ADAT I/O making up the balance of the channels. ADAT then becomes the limiting factor.
Why they opted for multi-cable SMUX, I still haven't figured out. Optical connections handle hundreds of gigabits per second in the network world over kilometer long runs. There should be no problem for a single toslink cable to handle many channels at 96Khz or higher.

[quaizywabbit]

True, true. I guess I was figuring on the standard setup around here of an interface of some make with ADAT I/O making up the balance of the channels. ADAT then becomes the limiting factor.
Why they opted for multi-cable SMUX, I still haven't figured out. Optical connections handle hundreds of gigabits per second in the network world over kilometer long runs. There should be no problem for a single toslink cable to handle many channels at 96Khz or higher.

if it were glass fiber optic vs. plastic then you'd be right...

[RBIngraham]

True, true. I guess I was figuring on the standard setup around here of an interface of some make with ADAT I/O making up the balance of the channels. ADAT then becomes the limiting factor.
Why they opted for multi-cable SMUX, I still haven't figured out. Optical connections handle hundreds of gigabits per second in the network world over kilometer long runs. There should be no problem for a single toslink cable to handle many channels at 96Khz or higher.

Because they wanted the equipment to be backwards compatible with Alesis' ADAT Optical format, which Alesis only wrote up to 48K in the spec, when they bought out their ADAT Multi-tracks years ago. I think it's so they didn't have to reinvent the wheel with all the data that is being shoved down the pipe. It's just using the ADAT Optical Protocol once it's actually on the lightpipe and as long as the 2 devices on each end know what's going on, then you can half the channel count and do 96K, but the upper half of you data is on another ADAT optical channel. It meant they didn't have to build a custom set of I/O chips and physical connectors and of course it's all backwards compatible with older 48K only gear.

I always wondered why they limited themselves to only 3M cable lengths really. But again that was a cost thing. Cheap plastic fiber cables, won't go long distances very well, and in order to go long distances you would also need better quality hardware ports on each end. I know for a while Apogee made some A-D-A convertors with ADAT Optical on them, but you could get the glass fiber option so they ports themselves were of better quality than your typical ADAT Optical port and you could order very long runs of glass fiber optic cables. So it was able to let you run the cables thousands of feet with no issues. I'm pretty sure the Cleveland Orch used those for quite some time for their recordings of concerts. Probably moved on to something nicer by now, but who knows...

You can even get 100' or so out of ADAT optical if you use good high quality glass fiber cables. But then the cable runs cost more than the gear you're hooking up often. :p Although there are some people out there making glass fiber TosLink cables, it's not that popular and my guess would be that is because no one can guarantee you a given result since it depend as much on the transmitter and receiver ports as it does on the glass fiber optic.

If you take a look you'll see that others have used optical cables in a much more sensible manner. Look no further than RME's MADI boxes. Of course they are not cheap, but it can do lots of channels over a single fiber optic cable or a single piece of Coax.

There are so many protocols around now that it's almost impossible to keep up with them and most of them have a fiber option available for those that need longer runs than a CAT5 can provide. And some of them will even route over a standard Ethernet network like CobraNet and EtherSound.

You can even get a fiber optic option for AudioRails if you really wanted to. :D

Richard

[Jeff Scott]

So ...how long can we run the Toslink cables between the Stage / ADA8000 and the FOH position? Could you do 100' with no problems? (the budget doesn't allow for a remote Laptop as yet...)

[905shmick]

I'd be worried about running 100' toslink cables for fear of them being destroyed.

[RBIngraham]

So ...how long can we run the Toslink cables between the Stage / ADA8000 and the FOH position? Could you do 100' with no problems? (the budget doesn't allow for a remote Laptop as yet...)

If you look around the forum there are one or two folks that talk about using 100' fiber optic cables. I can only assume they are using glass fiber optic cables.

But a quick Google search turned up these:
http://www.lifatec.com/Lifusa_patchcords.htm

Just keep in mind that no one is likely to guarantee good performance of ADAT Optical at that length, since ADAT Optical is only written to support lengths of 3 M which is about 33'.

So buyer beware.

If you want a reliable solution I would use AudioRails or the one other ADAT CAT5 extender that was mentioned in the forum recently. Just look around the forum, it's been talked about several times what the various options are for running sound at FOH including wireless networks, wired networks, CAT5 extenders for your KVM, ADAT Extenders so you only leave the A-D-A at the stage but your SAC rig is at FOH, etc....

Richard

[gdougherty]

So ...how long can we run the Toslink cables between the Stage / ADA8000 and the FOH position? Could you do 100' with no problems? (the budget doesn't allow for a remote Laptop as yet...)

I've thought about it, but at ~$30/cable I'd only consider it for install use. Even then I'd pickup one or a pair and test to see if it works with your hardware before springing on a full set. For live traveling use, I'd rather replace $30 or less of CAT5 than $240 of optical cabling.

[Jeff Scott]

So if running 100' Toslink is a pricey/Bad idea ...and using an Audio Rails / Cat 5 extender is also a $$$$ option....Do most of you run a standard snake to FOH? Or are most of you using a remote/wireless laptop at FOH?

[Trackzilla]

I do remote. sometimes I run a 4ch snake as well so I can use it for local monitor & TB, but sometimes I run those wireless with IEM & a wireless mic, just depends. Most shows it's just Lappy at FOH.
Occasionally I'll hook a second monitor to Lappy & use one monitor for SACRemote & leave TightVNC open on the other one, again, just depends on my mood & the show.

[Bob L]

There's no reason not to run standard snakes and keep SAC at FOH position just like how you do things now with your current consoles... later as you get the bidget... you can pick up laptops and Netbooks and AudioRails and whatnot if you want to change from the standard configuration.

No need to change everything at once... start slow... get your SAC rig working from FOH and get comfortable with virtual mixing... then expand on the idea.

Bob L

[RBIngraham]

So if running 100' Toslink is a pricey/Bad idea ...and using an Audio Rails / Cat 5 extender is also a $$$$ option....Do most of you run a standard snake to FOH? Or are most of you using a remote/wireless laptop at FOH?

The most logical and cost effective way to achieve this depends a lot on what you do and the typical venue type you work in. For example I work mostly in theatre and of lot of them have analog mic lines run already to FOH. So you might as well just use those lines, and put your SAC rig at FOH. For someone doing sound for a wedding band that doesn't make all that much sense, because it's unlikely you'll play in many (if any at all) venues with mic lines and a FOH position put in place.

If you have a spare laptop already, clearly a wireless laptop remote at FOH is a fairly cost effective solution, you probably just need a wireless router and you're in business. The issue with that solution is if you want headphones for a pfl at FOH or a talkback mic at FOH, then you still have to figure out how to get that to FOH. Some folks just use wireless everything, so they use a wireless network to remote SAC, a wireless handheld for talkback and wireless headphones or IEMs for their Solo/PFL feed. If you already have all those things, then it makes sense, if not then you might be better off with a small 8 channel snake so you can have a few hard lines between FOH and the stage. At that point you might as well just bundle a CAT5 cable with the small snake and run a wired Ethernet connection for your SAC remote.

Another option is to use a CAT5 KVM Extender. With this you don't need to have a second laptop or computer at FOH. You're still running the computer on stage it's just that the computer has a really long cable to hook up the keyboard, monitor and mouse. This of course still has the problem of getting talkback and Solo to FOH, but it's a tiny bit cheaper (not a lot cheaper!) than buying a laptop just for that purpose, and it also has the advantages of being a hell of a lot easier to set up than dealing with all the network hassles you need to work out when running a SAC remote, plus you can control all your plug-ins at FOH, which at least at this moment you can not do with the SAC remote.

There are a lot of ways to make this all work. It's part of the very cool flexibility of using a SAC rig. It depends on what gear you already have and what type of venues you work in and what parts you would like to improve. As Bob pointed out, if you already have a snake, you can just put SAC in the place of your current mixer and then move into more complex and high tech set ups as time and money allow.

Richard