jay mitchell
Roadie
- Messages
- 620
Me, too. If people won't accept the basics, it's very difficult to have a discussion of the finer points.
The Latency Thread?
- Thread starter pipelineaudio
- Start date
pipelineaudio
Shredder
- Messages
- 1,931
What basics aren't people accepting? This is all a derail. I started the thread to measure latency. Take it up with DPA.
www.dpamicrophones.com
The basics about comb filtering (and how to avoid it)
Comb filtering is a well-known problem in audio engineering. When you hear it, you immediately want to get rid of it. As a trained sound engineer, even the slightest hint of comb filtering causes your neck hairs raise.
fretworn
Rock Star
- Messages
- 4,158
jay mitchell
Roadie
- Messages
- 620
Then download Room EQ Wizard and acquire IRs of the device using a loopback of the sweep as the absolute time reference. That way, you're eliminating the RTL of the soundcard/computer/drivers from your measurement.
There's nothing to "take up with DPA." Comb filtering is caused by arrival time differential, not latency. Note the word "differential." Two or more signal paths are required to create comb filtering, which occurs independent of latency. Similarly, interaural time differences ("ITD") come from arrival time differences of a single sound at each ear. Your ear/brain can detect and process ITDs of less than a millisecond. That's how you localize sounds laterally.
The ability to hear multi-source interference ("comb filtering") is independent of your ability to detect latency.
- Messages
- 6,508
How so??? At the very least, you're going to be getting AD/DA latency, which can be significant depending on the number of stages - it's usually quite negligible though. If you set REW up to use ASIO, then you're definitely going to have some amount of latency imparted by the ASIO driver, sample-rate and buffer settings.
And even if you cut those things out, that breaks the very definition of RTL - roundtrip latency.
A typical use-case:
Signal in > AD conversion > sample-rate and buffersize result in driver latency factor > DA conversion which has it's own latency > speakers > ears
For an accurate RTL figure, you want to measure everything in the chain between the signal and speakers.
It doesn't make sense to me to weight a test towards the very fastest possible processing chain, when it isn't representative of actual use-cases.
I agree with you entirely on comb filtering not being caused by 'latency' - it is the wrong word for it.
Last edited:
pipelineaudio
Shredder
- Messages
- 1,931
I have roomEQ wizard, this has nothing to do with anything I posted
This has nothing to do with anything I said, I think you are thinking of someone else
This has nothing to do with anything I said, I think you are thinking of someone else
pipelineaudio
Shredder
- Messages
- 1,931
Comb filtering can and does most definitely happen when two of the same signal arrive at different times
- Messages
- 6,508
The word 'latency' isn't a catch all term for any time a signal doesn't arrive instantaneously.
Sascha Franck
Rock Star
- Messages
- 6,079
So, if you record a singer, monitoring through headphones and he/she is complaining about "weird warbly FX" (or whatever they may say), in other words; comb filtering happening because of the misalignment of the signal heard through their skulls and their headphones, that comb filtering is *not* caused by latency in your book?
pipelineaudio
Shredder
- Messages
- 1,931
Was someone saying it was? In the context of this thread's OP, round trip latency, it is certainly a term for a delayed signal. This is like 666 layers of pedantic hell, I don;t even understand what is going on here
pipelineaudio
Shredder
- Messages
- 1,931
Ok, someone is on my wavelength, thank you. @Sascha Franck do you have any idea what this rabbit hole they are going down is even about in the context of this thread? I'm genuinely totally confused by this obfuscation
jay mitchell
Roadie
- Messages
- 620
To be sure we're talking about the same thing, the procedure I cited is for measuring the latency due to an outboard device or system, not system RTL. The procedure cancels the effect of RTL. However, RTL is easily measured within a DAW using the same hookup.
The loopback is physical: it takes the sound card's analog output, and, via a cable, applies it back to its analog input. The sweep sent to the external device goes through exactly the same process (the sound card has to have two inputs and two outputs). The analyzer uses as its inputs the loopback and the return from the DUT, both of which have exactly the same RTL. and uses the direct loopback as the zero-time reference. That means that any apparent remaining latency is due to the DUT.
Last edited:
pipelineaudio
Shredder
- Messages
- 1,931
Here lets at least agree that in this thread in this context we are talking about: https://en.wikipedia.org/wiki/Latency_(audio) which can and does cause comb filtering if there is a parallel path involved with a different latency (yes, air is a system)
Lysander
Shredder
- Messages
- 2,235
I should know better and stay out of this one, but...
For the n-th time, no. The fact that you have a "parallel path" with a delay will cause comb-like interference, but has jack shit to do with latency. Both sound signals propagate at the exact same speed.
Your own link explains how latency is the time between an audio signal entering, and exiting, a system. There's no such thing as "parallel latency".
pipelineaudio
Shredder
- Messages
- 1,931
Yes, for the n-th time there is parallel latency. Look this is just a derail anyway. We hopefully will continue to measure and add to the RTL numbers of different devices in this thread, and I'll try to update the first post with figures as they can come in
Lysander
Shredder
- Messages
- 2,235
Here's the classic diagram for a comb filter arrangement. You know, a system.
See that straight line below? That's you original, unaltered signal. Now ask yourself, how long does a change in X takes to be reflected in Y? And can that figure ever be affected by delay K? Because that's the actual, textbook definition of latency.
I have to agree with others mentioning the same sentiment, it's really hard to discuss something when there's such a basic misunderstanding of what it even means.
See that straight line below? That's you original, unaltered signal. Now ask yourself, how long does a change in X takes to be reflected in Y? And can that figure ever be affected by delay K? Because that's the actual, textbook definition of latency.
I have to agree with others mentioning the same sentiment, it's really hard to discuss something when there's such a basic misunderstanding of what it even means.
Last edited:
pipelineaudio
Shredder
- Messages
- 1,931
the comb filtereing is detected where it is summed. Why is this such a hard concept for you to understand? You are mixing two signals together. One delayed, one not.
I'll put this in super simple terms, hopefully, it will stick. Get a glass of water. Put some dirt in it and stir it. Now it no longer looks like the original glass of water OR the dirt by itself right?
I'll put this in super simple terms, hopefully, it will stick. Get a glass of water. Put some dirt in it and stir it. Now it no longer looks like the original glass of water OR the dirt by itself right?
