AX8 processing latency

Hi!

I have a long history of using amp simulation software and equipment from VST plugins such as Scuffham S-Gear and BIAS to amps like Yamaha THR Head etc.. During my experiments I have discovered that soundwise the Scuffham S-Gear might be the best amp simulator available currently. However, like any other VST plugin, the round-trip latency and sound quality it is highly dependent on the audio interface / soundcard. When using a high quality PCI soundcard with proper DI box, the sound quality of Scuffham is hard to beat.

Playing without noticable latency (DAW report in-out total < 2ms) with Scuffham requires a desktop PC with PCI soundcard, and this is the reason I just got my AX8 this week. I had a wish, that I could replace the desktop PC + PCI + Mixer + DI-Box + Pedalboard -setup with single unit such as FAS AX8.

I read somewhere, that the overall latency through the AX8 should be around 2ms. However, when I started to create presets where the CPU percent was around 80% and multiple effects were enabled (delays, reverb), the processing latency of the unit was noticable. In my mind this means that the overall latency must be >> 2ms if it can be noticed with ear using headphones.

The setup I am using is AX8 --> Analog Mixer with Monitoring --> to PC. The latency that I discovered was through the analog mixer monitoring, not through PC.

Have anyone else experienced similar behaviour?
 
To clarify, the latency cannot be noticed when playing chords or slower passages. But when starting to "shred" lets say 16th notes it is noticable.
 
Any real time digital processor has no "processing latency", because all the data must be processed between a convertion cicle (AD/DA).
So the latency is unrelated to processing.
There could be two latency: the device latency (fixed, depending on sampling rate), and the "modeling/fx behaviour". Probably with 16th notes you feel sag, dynamic clip, or harmonics, loading & discarging of virtual capacitor and coil... all of the marvelous behaviuor Fractal gave us delve deep into real amp circuit. You can "fix" this behaviour stiffing the amp: try different setting of sag, negative feedback, clipping type, drive, master, and so on. Tell us if you find some parameters that affect what you hear as "latency". Hope it helps! :)
 
Actually Smilzo that makes sense, if the sample rate and buffer size is fixed, the hardware latency should be fixed. You might be right about the amp behaviour, the sag and amp dynamics can feel as latency. I have to study this a bit more. Thanks!
 
Last edited:
Actually, is there a difference in "perceived latency" when effects are put to parallel path instead of series path? I mean that the algorithms introduce software made latency (as stated with the amp model), so when stacking these blocks in series they might introduce more "latency" compared to a situation where you connect direct line from cab to output and put all effects to parallel this line.
 
Last edited:
I don't think so. You could introduce more latency using the loop block (having antoher AD/DA convertions). Otherwise the grid should not affect latency. Say you have 1000 operation at 50% CPU utilization. The CPU perform regardless of the grid 1000 operation. When you approach 100% utilization, first priority is given to audio processing (display and unit controll slow reaction), then some dropout arise in sound. Because there is no more time to process the streaming data of AD. So no latency is added, the processing is chopped. To prevent this there are warning display when you approach the limit. It's not your case, IMHO.
 
CPU in PC work in time-base frame giving each process a given time. If, say, an antivirus, a word processor, a media player, so on ask for a call with higher priority, then sound processing is freezed. So there could be latency added. In a dedicated DSP unit there is one main processor: the audio stream processing. All other process are known, and the device work at the optimal limit of harware possibilities. In any CPU the number of process depends on the user.
 
I actually ment that could the algorithm based latency be minimized by putting the effects parallel in the AX8-Edit software internally. Let's say that you have any kind of algorithm that contains any kind of filtering (almost every block has some kind of filter available to tweak), if you put these filters in series the more "attack" you lose. In contrary to situation, where you put these algorithms to a side path (meaning one grid lane upper to the main signal path after CAB block in AX8-Edit) and let the signal coming from the CAB block pass straight to output and then sum the effect lane with the dry lane in the output stage.

Of course for example the Chorus block does have a "mix" option that allows the user to define how much the dry signal is passed through, but it depends how the dry vs. wet signal is processed inside that algorithm?

I guess same kind of problem would exist in effect loops in traditional amps also, if you have a series effect loop with many effects versus parallel effect loop. However, there the effect lane actually passes through multiple AD/DA conversions in addition to the software processing latency introduced inside those effect pedals.
 
Last edited:
CPU in PC work in time-base frame giving each process a given time. If, say, an antivirus, a word processor, a media player, so on ask for a call with higher priority, then sound processing is freezed. So there could be latency added. In a dedicated DSP unit there is one main processor: the audio stream processing. All other process are known, and the device work at the optimal limit of harware possibilities. In any CPU the number of process depends on the user.

No, if there was not enough CPU power to process audio you would here clicks or gaps in the middle but no constant latency. The CPU is scheduled to work on blocks of audio data, namely your sample buffer. The slower your CPU is or the more it has to do besides audio processing the bigger this buffer needs to be and the higher the perceived latency will be. But this latency is always constant. Having a fast CPU and no disturbing hardware like USB 3 interfaces or Nvidia graphics cards is key to good audio performance / small buffers / low latency.
 
I actually ment that could the algorithm based latency be minimized by putting the effects parallel in the AX8-Edit software internally.

The editor has nothing to do in audio processing. The processing is done in AX8/AXEFX/FX8...

No, if there was not enough CPU power to process audio you would here clicks or gaps in the middle but no constant latency.
Thanks, it was what I hoped to tell!
 
The editor determines how the algorithms are aligned inside the AX8. You mean there is no computational difference whether a delay block is set in series to the "main signal path" or in parallel with a "shunt block"?
 
The editor determines how the algorithms are aligned inside the AX8. You mean there is no computational difference whether a delay block is set in series to the "main signal path" or in parallel with a "shunt block"?

Correct. Whether you put a delay in parallel or in series doesn't make a difference as long as you mix the FX part with the dry part. The dry signal part will not suffer any noticeable latency.
 
I have a DI box, so I ran my guitar into my DI and recorded the output from DI. The DI also has a thru (or "To amp" output), so I ran that output to my Ax8 and recorded my AX8.

On a preset with 84% CPU usage, the signal through the AX8 was delayed by ca. 150 samples at 48000 samples/second, which is roughly 3.5 ms. On a much lower CPU preset, I got 105 samples, which is roughly 2 ms.

As the signal through the AX8 has been processed, the two waveforms look nothing alike, so my latency estimates are nothing more than rough estimates. A 3.5 ms latency corresponds to having your ear 1 meter away from the speaker.
 
A few ms of latency is completely negligible. Sound takes time to travel through the air. It moves at roughly 1 foot per ms. If you are standing 8 feet away from your speakers, you are hearing the sound with about 8 ms of latency. So unless you always play through headphones or stand ridiculously close to your speaker when playing, you're not going to notice a couple of ms of latency at all. Have you ever heard anyone say that standing a couple feet further from their amp make the latency noticeable? At 30 feet, maybe. At 2, nope.
 
Hi!

I have a long history of using amp simulation software and equipment from VST plugins such as Scuffham S-Gear and BIAS to amps like Yamaha THR Head etc.. During my experiments I have discovered that soundwise the Scuffham S-Gear might be the best amp simulator available currently. However, like any other VST plugin, the round-trip latency and sound quality it is highly dependent on the audio interface / soundcard. When using a high quality PCI soundcard with proper DI box, the sound quality of Scuffham is hard to beat.

Playing without noticable latency (DAW report in-out total < 2ms) with Scuffham requires a desktop PC with PCI soundcard, and this is the reason I just got my AX8 this week. I had a wish, that I could replace the desktop PC + PCI + Mixer + DI-Box + Pedalboard -setup with single unit such as FAS AX8.

I read somewhere, that the overall latency through the AX8 should be around 2ms. However, when I started to create presets where the CPU percent was around 80% and multiple effects were enabled (delays, reverb), the processing latency of the unit was noticable. In my mind this means that the overall latency must be >> 2ms if it can be noticed with ear using headphones.

The setup I am using is AX8 --> Analog Mixer with Monitoring --> to PC. The latency that I discovered was through the analog mixer monitoring, not through PC.

Have anyone else experienced similar behaviour?
Post the preset that its giving you this problem please.
 
Back
Top Bottom