This is a total -non- problem. Just a curiosity. GP5 on my M4 running smooth as silk using a Motu M2.
I plugged my Cables in to do a real world RTL Check and got 2 x Different Figures.
96k / 32 Bit RTL
→ Oblique RTL 2.4905ms
→ GP5 Internal Latency Checker RTL 2.1ms
→ Both tested the same exact way.
F.w.i.w .. I repeated these tests several times including several machine and Motu M2 re-boots and the number for each RTL Test never changed in any way.
Whilst the difference isminiscule / imperceptible [0.3905ms] it is actually quite a large difference in % terms of %19
→ is there a reason for the difference ?
→ which one is considered the more correct / accurate RTL ?
Like I said - totally does not matter either way - I am just curious as to why the difference and which one is most accurate ?
So suppose I have two thermometers and no other devices. I put them both into a cup of hot water, wait a few minutes and then check the readings.
One of them says 110F and the other says 112F
How exactly would you decide which one is correct?
Ah … I don’t think the analogy of using 2 x imprecise Analog Tools like thermometers placed into an organic material in different parts of the glass is comparable in any way
Measuring the fixed electronic and digital structure of a hardware device with a very precise measuring tool that can report to the 4th decimal point is totally different … not to mention it is completely consistent and repeatable in its test reporting.
Being repeatable and consistent is irrelevant if the internal calculation for measurement is incorrect—it could be giving you the exact same incorrect reading every time.
The point @dhj is trying to make is that you have two measurements and you’re asking which one is correct. How can you possibly know which one is correct with just those two data sets? There is of course the third possibility that neither one is correct!
The analogy is perfect. The issue has nothing to do with the tools. I could have replaced those analog thermometers with digital thermometers displaying 5 digits of precision and if one of them displayed 1.2345 and the other displayed 1.3356, how would you know which one was correct?
This has gotten a bit out of hand - I am not talking about or interested in subjective qualitative perspectives.
All I am querying is why -factually and technically- the same “box” delivers different results when both measuring tools are measuring the same thing the same way
ie: a stimulation signal is sent in to the Input Stage - the signal travels through “box” AD/DA and other electronics - the signal then hits the Output Stage … and each program records how many ms this took.
Nothing more, nothing less.
Subjectively and qualitatively two or more things can be true at the same time.
However … technically and mathematically only one can be “right” at any given point in time.
So … technically and mathematically …assuming Qblique and GP5 RTL measure the same thing the same way ie: stimulation signal sent in to the Input Stage - signal travels through “box” - hits the Output Stage … then each program records how many ms this took … only one can be accurate or the “most” accurate at any point in time (?)
You would need a certified measurement, but, as you said, it’s just out of curiosity so spending money for a certified lab to confirm your results is not really useful.
In this case: If you want to boast about it: use the lowest outcome, if you need it for designing your audio setup: go for worst case.
If they were both measuring the same thing the same way, then they would both produce exactly the same result (whether that result is correct or not would be a different question) but since they produce different results, that assumption is clearly invalid.
I think David is suggesting they are (probably) technically measuring slightly different things (or doing the measurement differently). So, I suppose someone (not me, I’m clueless) could spend time digging in to exactly how the respective measurements are done using the different applications.
It looks like that time is better spend elsewhere.
“A difference that makes no difference is no difference”
I have never measured the latency of my own system - it works, it sounds great, it responds properly to my playing so I am really not concerned with that measurement for myself
I assume you are using the same buffer size, but I don’t think you mentioned what that is.
If you were running one test at a 32 sample buffer and the other at 48 you’d get a difference around that size.
I’m a little surprised you get exactly the same measurement every time. I’ve done this kind of experiment in the past and generally got some modest variability in the measurements. I don’t know why and never really cared.
I would think some part of the difference is what audio signal is being sent out and how the detection of the return signal is being defined. In a perfect world when there is no intentional output and no noise at the input you would be reading straight zeros all the time coming back in. Except there is always going to be some noise present, so what level is the software choosing to define as “no signal” when there is always noise present? If one algorithm is waiting longer to define “that’s definitely my signal and not just noise” then they would measure differently.
In the absence of analog noise, to get the most precision you’d probably want to output a perfect square wave, where the signal goes from zero on one sample to something significantly higher on the next sample. In the digital world that’s easy. But then it gets converted to analog, goes through some kind of cable, probably some analog gain manipulation, then converted back to digital. That’s not going to look like a perfect square wave because perfect square waves don’t exist in analog circuits.
So now we’re back to the question of at what point does the measurement algorithm decide the level is clearly signal and not noise.
Let’s say the testing algorithm is using a 1khz sin wave instead of a square wave. That wave has a period of 1ms. If one algorithm is looking for a signal near peak then that’s not detected until a quarter of the period, or 0.25ms.
I’m just making this up on the fly. I have no idea how any of them actually work.
Looks like a really detailed tool !
] to try and find out:-
