Nathan P
01-05-2016, 09:02 AM
Yesterday I finally actually tested the Latency of my SAC system. I wanted to know for sure the amount of time it takes from when the analog signal hits the AD to when the DA sends it back through the output. Here is what I find:
44.1K, Buffer Size: 32, Preload Buffers: 1 = 4.47ms.
(Each time I increased the Preload Buffer by one it has a result of an increased delay of: .77ms ea. time.) That value is exactly the buffer size divided by the sample rate: 32 / 44,100.
44.1K, Buffer Size: 64, Preload Buffers: 1 = 6.67ms.
(Each time I increased the Preload Buffer by one it has a result of an increased delay of: 1.45ms ea. time.) That value is exactly the buffer size divided by the sample rate: 64 / 44,100.
48K, Buffer Size: 32, Preload Buffers: 1 = 4.10ms.
(Each time I increased the Preload Buffer by one it has a result of an increased delay of: .67ms ea. time.) That value is exactly the buffer size divided by the sample rate: 32 / 48,000.
48K, Buffer Size: 64, Preload Buffers: 1 = 6.67ms.
(Each time I increased the Preload Buffer by one it has a result of an increased delay of: 1.33ms ea. time.) That value is exactly the buffer size divided by the sample rate: 64 / 48,000.
How did I come up with these measurements? I played a recording of a spoon hitting a table into the Motu, routed that input back out a SAC output, connected the output to a 2nd Motu input, then recorded both the original & the now delayed output from SAC side by side in a DAW, took the # of samples between the beginning of each wave & divided it by the sample rate (44,100 or 48,000).
This proved my suspicions that the Input & Output latency listed in some of my DAWs (Cubase) do not add up to the total system latency. It also let me know exactly what to expect from the changes to the Preload Buffer values. Before I did not know how that worked. Now I know exactly how much latency is in the system & what the results will be when if I start changing the various buffer values around.
The SAC system used here was made up of:
E8400 Core2Duo processor
G41MX-F FoxConn motherboard
4 GB of Ram
MOTU 24I/O core system running on a PCI-424 card
XP Pro 32 bit OS, SAC 4.1, (Force Single CPU & Force RealTime Priority Class)
(I'll also quickly mention some interesting lessons I've learned on this particular system in the last couple days. I have one other test to finish my conclusions. Thus far, I can say this system has more slipped buffers running Windows 7 Pro 64, or Windows 10 Pro 64 so I went back to the old XP 32 bit. I don't have a copy of XP 64 bit otherwise I could test & see if it was the 64 bit OSs or Windows 7 & 10 (Dual CPU & Force RealTime Priority Class) at their best that could not stabilize as well as the old XP 32 bit single CPU. I hope to post on this later in another thread, but time will tell if I do or not. This OS debate for me came up because I bought a new SSD & a Windows 7 64 OS thinking I'd maximize the performance of other 64 bit applications when not in use as a SAC system. As I'm learning, that new OS is only hurting my SAC system on this hardware setup. The only test left is to retest the performance now that I have installed XP Pro 32 bit on my SSD drive. All my superior XP 32 results were using my older 7,200 RPM drives. So to be fair I need to run the XP 32 on the new SSD to eliminate that variable. What a pain... to Install, Upgrade, Format re-Intall OS & software, etc.)
44.1K, Buffer Size: 32, Preload Buffers: 1 = 4.47ms.
(Each time I increased the Preload Buffer by one it has a result of an increased delay of: .77ms ea. time.) That value is exactly the buffer size divided by the sample rate: 32 / 44,100.
44.1K, Buffer Size: 64, Preload Buffers: 1 = 6.67ms.
(Each time I increased the Preload Buffer by one it has a result of an increased delay of: 1.45ms ea. time.) That value is exactly the buffer size divided by the sample rate: 64 / 44,100.
48K, Buffer Size: 32, Preload Buffers: 1 = 4.10ms.
(Each time I increased the Preload Buffer by one it has a result of an increased delay of: .67ms ea. time.) That value is exactly the buffer size divided by the sample rate: 32 / 48,000.
48K, Buffer Size: 64, Preload Buffers: 1 = 6.67ms.
(Each time I increased the Preload Buffer by one it has a result of an increased delay of: 1.33ms ea. time.) That value is exactly the buffer size divided by the sample rate: 64 / 48,000.
How did I come up with these measurements? I played a recording of a spoon hitting a table into the Motu, routed that input back out a SAC output, connected the output to a 2nd Motu input, then recorded both the original & the now delayed output from SAC side by side in a DAW, took the # of samples between the beginning of each wave & divided it by the sample rate (44,100 or 48,000).
This proved my suspicions that the Input & Output latency listed in some of my DAWs (Cubase) do not add up to the total system latency. It also let me know exactly what to expect from the changes to the Preload Buffer values. Before I did not know how that worked. Now I know exactly how much latency is in the system & what the results will be when if I start changing the various buffer values around.
The SAC system used here was made up of:
E8400 Core2Duo processor
G41MX-F FoxConn motherboard
4 GB of Ram
MOTU 24I/O core system running on a PCI-424 card
XP Pro 32 bit OS, SAC 4.1, (Force Single CPU & Force RealTime Priority Class)
(I'll also quickly mention some interesting lessons I've learned on this particular system in the last couple days. I have one other test to finish my conclusions. Thus far, I can say this system has more slipped buffers running Windows 7 Pro 64, or Windows 10 Pro 64 so I went back to the old XP 32 bit. I don't have a copy of XP 64 bit otherwise I could test & see if it was the 64 bit OSs or Windows 7 & 10 (Dual CPU & Force RealTime Priority Class) at their best that could not stabilize as well as the old XP 32 bit single CPU. I hope to post on this later in another thread, but time will tell if I do or not. This OS debate for me came up because I bought a new SSD & a Windows 7 64 OS thinking I'd maximize the performance of other 64 bit applications when not in use as a SAC system. As I'm learning, that new OS is only hurting my SAC system on this hardware setup. The only test left is to retest the performance now that I have installed XP Pro 32 bit on my SSD drive. All my superior XP 32 results were using my older 7,200 RPM drives. So to be fair I need to run the XP 32 on the new SSD to eliminate that variable. What a pain... to Install, Upgrade, Format re-Intall OS & software, etc.)