Ooooh... Charts and Graphs

Status
Not open for further replies.
Look, I've tried to avoid saying it, but we all remember the "facts" that Cliff chucked about regarding the Kemper - like the "it runs at 22kHz and can't reproduce anything over 11kHz" one which was proven false several times over. With that in mind, it strikes me as sensible for those of us who want actual solid facts to maintain a decent level of critical thought instead of just believing a potentially-biased source who's been unreliable on the subject of other companies' products in the past.
This was clearly observable at least one point in time:
http://dallashodgson.info/articles/KemperCabAnalysis/whitepaper.htm

Again, you should perhaps doubt yourself more before you criticize someone of expertise in multiple ways... or you come across as having an agenda ;)
 
Last edited:
Yes, you simply capture the spectrum without applying any stimulus. That's then your noise floor. Then you apply a stimulus (i.e. a sine sweep over a small range so that you can discern aliasing from natural harmonics). The "noise" floor will increase. That noise is aliasing noise.

This is basic stuff and anyone arguing against the validity of the technique and results simply doesn't understand the theory and the math.
Could we perhaps get a measurement graph of the units' noise floor without the stimulus, so we can reference it vs. the stimulus graph to see the amount of aliasing?
 
...whenever I've encountered ground loops the hum has been much more apparent than any hiss introduced from multiple gain stages at high signal volume.
Correction: whenever you’ve encountered strong ground loops. Ground loops are always present. It’s just a matter of degree. The ground loop in the plot above is more than 60 dB below the signal — so quiet as to approach inaudibility.


As my history teachers always used to tell me..."Consider the source".
Indeed. Consider the source. @FractalAudio has a long history of technical excellence, rarely being wrong, and freely admitting when he actually is wrong. @digitalscream ’s history, at least on this board, is to consistently throw mud on that knowledge and integrity, and to do so in an area in which he’s clearly demonstrated gaps in his knowledge. Which source is more trustworthy?


...why, when measuring aliasing, was an analogue signal path involved at all (at least with the Fractal units, given the ground loop) when the modelling could've been directly tested via USB without passing through the converters, thus keeping everything in the digital domain and thus eliminating a source of variation?
Because aliasing is the result of sampling. If you send your test signal via USB, it’s already been sampled — by something else. That something else already introduced its own aliasing, and the Axe-fx isn’t sampling anything. The test would be invalid. You need an analog signal to perform the test.
 
Could we perhaps get a measurement graph of the units' noise floor without the stimulus, so we can reference it vs. the stimulus graph to see the amount of aliasing?

hx_noise_and_aliasing.JPG


Green is the thermal noise floor (plus some ground noise at 60 Hz). Blue is the noise floor due to aliasing.
 
Correction: whenever you’ve encountered strong ground loops. Ground loops are always present. It’s just a matter of degree. The ground loop in the plot above is more than 60 dB below the signal — so quiet as to approach inaudibility.

OK, fair enough. Why isn't it present even as a slight bump in the Helix plot, then?

Indeed. Consider the source. @FractalAudio has a long history of technical excellence, rarely being wrong, and freely admitting when he actually is wrong. @digitalscream ’s history, at least on this board, is to consistently throw mud on that knowledge and integrity, and to do so in an area in which he’s clearly demonstrated gaps in his knowledge. Which source is more trustworthy?

I'm not asserting any facts - I'm questioning his findings, and all he's done thus far is fire off a condescending comment (having been through his history on here, that's something he definitely has a long history of when questioned). I also can admit where I've got a gap in my knowledge (see directly above), because that's the whole point of what I'm trying to get to here.

Because aliasing is the result of sampling. If you send your test signal via USB, it’s already been sampled — by something else. That something else already introduced its own aliasing, and the Axe-fx isn’t sampling anything. The test would be invalid. You need an analog signal to perform the test.

TrueRTA is generating the signal in every case, digitally. Putting it through a couple more conversions between the digital and analogue domains shouldn't make any difference unless the two signals were sampled at different rates or there was wander in one of the clocks. Let's not forget, aliasing is where you can take two signals and - after processing - no longer distinguish them. That processing could be by the sampling in the converters, or it could be the modelling itself; I may have misunderstood, but there appears to be no reference to the converters in Cliff's OP and lots of references to the modelling and modellers (and "high gain", which suggests we're not just talking about converter performance here).

Personally, if my hypothesis is correct, I'd have thought you guys would be all over it. It'd certainly be a useful distinguishing point for purchasers - Fractal gear discards noisy bits of the original amp's signal, so you get a purer tone. Helix (or others) keep it for authenticity, but the trade off is the same annoyance you have to deal with.

Speaking of facts, please help keep them straight by not using quotes if you are actually not quoting someone. This is how rumors begin.

If it was a real quote and not a vague recollection, I'd have provided a link and an attribution. I thought the colloquial context of the paragraph made it pretty obvious, but I'll be more explicit to avoid confusing people if you want.
 
OK, fair enough. Why isn't it present even as a slight bump in the Helix plot, then?



I'm not asserting any facts - I'm questioning his findings, and all he's done thus far is fire off a condescending comment (having been through his history on here, that's something he definitely has a long history of when questioned). I also can admit where I've got a gap in my knowledge (see directly above), because that's the whole point of what I'm trying to get to here.



TrueRTA is generating the signal in every case, digitally. Putting it through a couple more conversions between the digital and analogue domains shouldn't make any difference unless the two signals were sampled at different rates or there was wander in one of the clocks. Let's not forget, aliasing is where you can take two signals and - after processing - no longer distinguish them. That processing could be by the sampling in the converters, or it could be the modelling itself; I may have misunderstood, but there appears to be no reference to the converters in Cliff's OP and lots of references to the modelling and modellers (and "high gain", which suggests we're not just talking about converter performance here).

Personally, if my hypothesis is correct, I'd have thought you guys would be all over it. It'd certainly be a useful distinguishing point for purchasers - Fractal gear discards noisy bits of the original amp's signal, so you get a purer tone. Helix (or others) keep it for authenticity, but the trade off is the same annoyance you have to deal with.



If it was a real quote and not a vague recollection, I'd have provided a link and an attribution. I thought the colloquial context of the paragraph made it pretty obvious, but I'll be more explicit to avoid confusing people if you want.

Arguing from a position of ignorance is not a winning strategy.
 
Arguing from a position of ignorance is not a winning strategy.

And...condescension and relying on a fan-populated echo chamber to buttress your argument? Not a winner either, IMO.

Seriously, I'm just asking for your methodology so that others can repeat it and test alternate hypotheses with a comparable method, that's all. If you're so confident that everything outside your conclusion is correct, then why the reticence?
 
hmm. been registered since 2014 but all posts are only on this topic. i see you've addressed this in the thread, but interesting that this made you speak up so intensely.
 
Seriously, I'm just asking for your methodology so that others can repeat it and test alternate hypotheses with a comparable method, that's all. If you're so confident that everything outside your conclusion is correct, then why the reticence?

I've explained my methodology. It's quite simple. I'll explain it again.

1. Input a FM modulated sine wave from 9 - 11 kHz into the DUT. The modulation frequency should be low, around 1 Hz or so.
2. Measure the output spectrum of the DUT. A spectrum analyzer with peak-hold ability should be used as the aliasing products will move around with the input modulation.

A product with no aliasing will exhibit only its thermal noise floor plus the excitation plus harmonics generated by the distortion. The harmonics will be from 18 - 22 kHz (twice the input spectrum. Turn off the input excitation to obtain the thermal noise floor. The difference between the measured noise from 0 - 9 kHz (and 11 - 18 kHz) with the source on vs. off is the noise due to aliasing.

In the comparison both the Axe-Fx III and the other product were using an SLO 100 model with all controls at noon. The Axe-Fx III gain was then adjusted to match the gain of the other product (although they were very close to start).
 
I've explained my methodology. It's quite simple. I'll explain it again.

1. Input a FM modulated sine wave from 9 - 11 kHz into the DUT. The modulation frequency should be low, around 1 Hz or so.
2. Measure the output spectrum of the DUT. A spectrum analyzer with peak-hold ability should be used as the aliasing products will move around with the input modulation.

A product with no aliasing will exhibit only its thermal noise floor plus the excitation plus harmonics generated by the distortion. The harmonics will be from 18 - 22 kHz (twice the input spectrum. Turn off the input excitation to obtain the thermal noise floor. The difference between the measured noise from 0 - 9 kHz (and 11 - 18 kHz) with the source on vs. off is the noise due to aliasing.

In the comparison both the Axe-Fx III and the other product were using an SLO 100 model with all controls at noon. The Axe-Fx III gain was then adjusted to match the gain of the other product (although they were very close to start).
but *insert argument here*!!!!!?????
 
@FractalAudio - Thank you, the extra details were what I was after.

@chris - Why would I argue? Four or fives pages later, I've finally got what I was asking for (after a lot of insults and condescension which - frankly - shout louder than any technical arguments about this community, but hey...Internets etc).
 
@chris - Why would I argue? Four or fives pages later, I've finally got what I was asking for (after a lot of insults and condescension which - frankly - shout louder than any technical arguments about this community, but hey...Internets etc).
not necessarily you. just the sentiment in general.

this community is great. perhaps take a step back to see why people reacted in that way. as you say, "internets," and some of it appeared to be the typical stuff that caused the reaction.
 
  • Like
Reactions: Rex
I’m kind of on a different course than most of the previous posts here. I think it is everybody’s right to be critical of posted results etc. That’s what science is about (and more). If stated results and facts leave room for discussion or interpretation, by all means be straight about it. This will invite involved parties to disclose applied methods. IMHO there’s no reason to question the other party’s objective or motivation, it’s all good.
 
OK, fair enough. Why isn't it present even as a slight bump in the Helix plot, then?
Because it's completely masked by the aliasing. Look at the plot again, and you'll see that.



I'm not asserting any facts...
Yes you are. Examples:

Hmmm...this strikes me as a pretty flawed "study"....

The setup was clearly not apples-to-apples for all units...

...I'm saying that the tests in the OP don't measure aliasing...

..."Consider the source". Cliff has a very obvious vested interest here...


-----


- I'm questioning his findings, and all he's done thus far is fire off a condescending comment...
Go back and read the posts again. He gave specific answers to your questions.


I also can admit where I've got a gap in my knowledge (see directly above), because that's the whole point of what I'm trying to get to here.
If you want to fill in gaps in your knowledge, the most effective method is to ask for knowledge, not to use your knowledge gaps to "prove" that someone is talking through his hat.
 
Status
Not open for further replies.
Back
Top Bottom