Ooooh... Charts and Graphs

Status
Not open for further replies.
@yek @Rex ok makes sense.. Though I wonder if it doesn't hide an actual weakness. In all amp sims I tried there's always that bad fizz around 10khz, can still hear it in the AX8 around 9.3kHz even though it is much less noticeable. The signal of amp sims also tend to fall down around 15kHz unlike real amps and I'm curious to know if it's linked to that 20kHz peak as well.
 
images
 
Big thing is dropping any expectation that anyone will understand anything / read past posts; and if they do, groovy

Expectations just disturb the mind...

"The great way is easy for one who has no preferences" —third patriarch of zen
 
Well, those of us that passed math and science classes in the 8th grade*: understand the charts, find them self-explanatory, and appreciate them being posted.

*The major math strands for 8th grade curriculum are number sense and operations, algebra, geometry and spatial sense, measurement, and data analysis and probability.
 
No. Aliasing is dependent on the gain. That test had obviously had less gain. I would've matched the gain to "Brand K" as on that particular device the controls do not mimic the real thing so changing the controls would introduce inaccuracy.

Hang on a minute...so you're saying that the gain was different on the units (but the aliasing depends on the gain), so the results are meaningless? The signal paths also appear to be different, hence the ground loop on the Axe tests...

Can you actually post details of how you went about this test so folk can try it for themselves? I ask because I know a few people have tried this with the Helix against the actual amp being modelled, and none of them get the results you have here. In fact, in every case the modeller follows the noise curve of the original amp within +/- 1.5dB all along the spectrum with a couple of exceptions.

In keeping with the old adage...if it's repeatable, it's science. If it's not, it's pseudo-science...or marketing material.
 
Can you actually post details of how you went about this test so folk can try it for themselves? I ask because I know a few people have tried this with the Helix against the actual amp being modelled, and none of them get the results you have here. In fact, in every case the modeller follows the noise curve of the original amp within +/- 1.5dB all along the spectrum with a couple of exceptions.
Maybe I'm not understanding your question, but how would someone test aliasing, which is a digital artifact, in an analog amp?
In keeping with the old adage...if it's repeatable, it's science. If it's not, it's pseudo-science...or marketing material.
Kind of sounds like you're insinuating Cliff is not telling the truth...
 
Looks like Jean-Luc needs someone to talk too. DIANA ...

Lesson learned: this is probably the wrong audience to discuss math.
 
Hang on a minute...so you're saying that the gain was different on the units (but the aliasing depends on the gain), so the results are meaningless? The signal paths also appear to be different, hence the ground loop on the Axe tests...
No, he's saying gain levels are different between the 2012 and 2018 tests. Perhaps consider your reading comprehension before you suggest someone of expertise is posting meaningless stuff.

Can you actually post details of how you went about this test so folk can try it for themselves? I ask because I know a few people have tried this with the Helix against the actual amp being modelled, and none of them get the results you have here. In fact, in every case the modeller follows the noise curve of the original amp within +/- 1.5dB all along the spectrum with a couple of exceptions.
Could you share the data you have? Would enjoy looking at it.
 
Last edited:
Can you actually post details of how you went about this test so folk can try it for themselves? I ask because I know a few people have tried this with the Helix against the actual amp being modelled, and none of them get the results you have here. In fact, in every case the modeller follows the noise curve of the original amp within +/- 1.5dB all along the spectrum with a couple of exceptions.

Just googled "noise curve", no reference outside of environmental studies. What are you referring to with this term?

As has been pointed out above, the alias issue refers specifically to the effect of a DIGITAL amp model and what is input into it creating aliasing noise based on the CPU capacity, converter filtering and sample rate. The comparison as I understand it has nothing to do with comparisons between the actual amp modeled by different brands, but merely used the same amp model and its gain structure and similar settings between brands as a reference, and to then input the same signal to see how much aliasing results in different modelers.

Did I miss something? Is there something else missing from the comparison test? If you send the same input into 2 different JCM 800 modeler brands with controls set basically the same and get different amounts of aliasing noise, how is this unscientific given two modelers will never be exactly the same in their implementation of both the model itself and the digital architecture?
 
Maybe I'm not understanding your question, but how would someone test aliasing, which is a digital artifact, in an analog amp?

No, I'm saying that the tests in the OP don't measure aliasing, unless something has been done to completely isolate the effect of aliasing (which it plainly hasn't) from any other effects. TrueRTA is a spectrum analyser, not an aliasing analyser.

You'd be quite surprised how closely an amp's response under identical conditions (the 9-11kHz sweep) matches that of the Helix...

Kind of sounds like you're insinuating Cliff is not telling the truth...

No, I'm explicitly saying that the conclusions he's drawn are meaningless in the absence of any context, like details on how to reproduce the test exactly.

No, he's saying gain levels are different between the 2012 and 2018 tests. Perhaps consider your reading comprehension before you suggest someone of expertise is posting meaningless stuff.

Fair point, and no need to get snarky about it...I currently have a nasty case of con-SARS and was posting around 4am 'cos I couldn't sleep. Turns out the ol' brain ain't so cooperative under those circumstances...

Could you share the data you have? Would enjoy looking at it.

I will as soon as I can, yes. Still waiting for a full useful write up, and this forum won't let me post links at the mo anyway.

Just googled "noise curve", no reference outside of environmental studies. What are you referring to with this term?

As has been pointed out above, the alias issue refers specifically to the effect of a DIGITAL amp model and what is input into it creating aliasing noise based on the CPU capacity, converter filtering and sample rate. The comparison as I understand it has nothing to do with comparisons between the actual amp modeled by different brands, but merely used the same amp model and its gain structure and similar settings between brands as a reference, and to then input the same signal to see how much aliasing results in different modelers.

Did I miss something? Is there something else missing from the comparison test? If you send the same input into 2 different JCM 800 modeler brands with controls set basically the same and get different amounts of aliasing noise, how is this unscientific given two modelers will never be exactly the same in their implementation of both the model itself and the digital architecture?

"Noise curve" is Copyright Me, 2018. Just a way to describe the graphs in the OP, since they obviously don't just show aliasing (see above). If it did, then a real amp would show zero response apart from the original signal and the harmonics - would you like to place a bet on whether that's true or not?

(I wouldn't take the bet, 'cos I already know the answer and I'm cursed with a sense of ethics).

Oddly, the more accurate the modeller, the more noise you'd see - if every component in an amp is modelled perfectly, then the overall model will display just as much noise outside the original signal as the original amp did, and that wouldn't be aliasing.
 
Alright sorry didn't see that it was explained in the 1st page. Too much to read not enough time. Facepalm-worthy.

Read the thread. The 20k Peak has been explained.

Yes, but it doesn't actually answer my question : if it is (desirable 2nd-order distortion or not) related to that drop of signal that doesn't happen with real amps. High gain amps don't drop like that after 10kHz. I don't know if the III drops, but the I and II do.

Also, I didn't see any explanation as to why, if there was an undesirable aliasing peak at 9.3kHz, this chart would show it instead of it being hidden by the input signal. Is my wish to see a same chart with a sweep performed outside of the 9kHz range worthy of a funny facepalm picture as well? Cliff.

I've always heard so far in FAS products, like in any modelers, an undesirable fizz in the 9 kHz range, it was around 9.7 kHz in the Ultra, now 9.3kHz in the ax8. It's getting quieter and quieter though. Not talking about IRs, the amp alone (or drive block, or anything that distorts enough).
 
No, I'm saying that the tests in the OP don't measure aliasing, unless something has been done to completely isolate the effect of aliasing (which it plainly hasn't) from any other effects. TrueRTA is a spectrum analyser, not an aliasing analyser.
It does measure aliasing, just not in isolation. And that can still be useful for evaluating aliasing in a comparison provided other factors aren't as significant variables. In particular, the major difference between the Axe-Fx II vs III graphs does seem to me due to aliasing, and I don't have to know the isolated aliasing portion of each graph to make that observation. My assumption is that base-noise level of II and III are similar enough.

Basically,
A is aliasing and B is other noise,
If B1 ~= B2 or A>>B,
A1+B1 > A2+B2 ~= A1>A2

I will as soon as I can, yes. Still waiting for a full useful write up, and this forum won't let me post links at the mo anyway.
Please message me the link ;)

Oddly, the more accurate the modeller, the more noise you'd see - if every component in an amp is modelled perfectly, then the overall model will display just as much noise outside the original signal as the original amp did, and that wouldn't be aliasing.
Is this also displayed in the Helix measurement data you have? Link please~
 
Yes, but it doesn't actually answer my question : if it is (desirable 2nd-order distortion or not) related to that drop of signal that doesn't happen with real amps. High gain amps don't drop like that after 10kHz. I don't know if the III drops, but the I and II do.
What drop after 10 kHz? 10 kHz is the test signal, it's meant to drop afterwards for this test.

Also, I didn't see any explanation as to why, if there was an undesirable aliasing peak at 9.3kHz, this chart would show it instead of it being hidden by the input signal. Is my wish to see a same chart with a sweep performed outside of the 9kHz range worthy of a funny facepalm picture as well? Cliff.

I've always heard so far in FAS products, like in any modelers, an undesirable fizz in the 9 kHz range, it was around 9.7 kHz in the Ultra, now 9.3kHz in the ax8. It's getting quieter and quieter though. Not talking about IRs, the amp alone (or drive block, or anything that distorts enough).
How do you know the ~9 kHz undesirable stuff you hear is due to aliasing? You'd have to know this for sure to demand a sweep performed outside of 9 kHz range.
 
Status
Not open for further replies.
Back
Top Bottom