Ok, so I saw this post on the Helix FB group the other day that referred me to the fractal forum, where I read a post about Aliasing… WTF…. no -like seriously -WTF was the OP talking about?
Now, if he was talking about Aliens… well hell I get that, those buggers show up in my research all the time, and I can make them dance like this when I crank up the ripple Control…
Well it turns out the OP was Cliff the owner of fractal, and he wasn’t talking about aliens but, “Aliasing”… What the hell is that? Fractal has a new unit coming out, and it’s expensive, very expensive. Gotta be a marketing thing right? They’ve gotta differentiate themselves from the plebs for sure. But honestly – I have no idea. So half an hour later and half a dozen googles later I come up with a basic (very basic) idea of what foldover aliasing actually is. Apparently, sampled Audio can only record tones at 1/2 the sampling rate, and for some reason this is named the nyquist frequency. This means that standard CD sampling rates, which is 44.1khz, can reproduce frequencies up to 22khz, which is nyquist for cd quality audio. Well that’s above human hearing, so who cares right? Well apparently sounds (harmonics and the like) will “fold back” (I think of this as bounce back but whatever) around the nyquist frequency and appear back in the noise floor of the lower frequency bands, where they are called aliens… I mean aliases. Aliases tend to be particularly egregious because they fold over a frequency that is not harmonically related to the fundamental frequency, but rather related to the sample rate. Thus they appear as anharmonic peaks in the noise floor.
Ok… this exhausts my rudimentary knowledge of this subject….. Headache… Excedrin….
If you must dive deeper here are some resources.
Try Google…. I’m done with the aliens.
So here’s Cliff’s original post from the Fractal forum. I suspect it to be the new AX III vs Helix, it apparently demonstrates superior anti aliasing performance of his newest unit. Let’s Investigate shall we?
Let’s break down what the he says….
- “I believe that a target goal for a “high-gain” amp should be a minimum of 60 dB of what I call “Signal-to-Peak-Aliasing”.”
- “…so any aliasing above the threshold of audibility (60 dB) reduces clarity and definition.”
- “Analog amps don’t alias…”
- Aliasing is particularly nasty because it’s harmonically unrelated to, and can occur at frequencies well away from, the desired frequencies (and, hence, is not masked by our perceptual process).
First let’s delve into the whole “60db of signal to aliasing ratio” . I think of this as signal to noise ratio, but since we’re talking about the aliasing component of the noise floor, I’ll roll with that terminology. Now I saw some discussion in the thread about how the 60db number, and was bunk because 60 db was only relevant at 30hz and the real number at 10khz was closer to 18db… Ok so half an hour later and half a dozen googles later I came to a few conclusions….
For the record, I’m no audiologist, and I don’t care for reading doctoral dissertations on The Threshold of Audibility. Note that term, Threshold of Audibility. Not threshold of hearing which is entirely different. So let’s take a look.
Here is a chart for threshold of hearing (I like charts better than tables of numbers) And as the FB post I read indicated, it shows that it indeed is 60 db at 30 hz and approximately 18db at 10khz. There are however 2 problems with this. First, we’re talking about the frequency of the alias and not the frequency of excitation (input frequency) so at the alias frequency (lets use 5khz) it’s closer to zero… which would mean…what exactly? I have no idea… Why? mainly because of the second problem — our old friend (no not the aliens) Threshold of Audibility which refers to the human ability to hear one sound in the presence of another sound. More precisely our ability to hear an alias in the presence of the fundamental frequency. Here’s the most relevant chart I could find on the subject.
Now, I see the minus -60db on the left side, and I see the entire guitar frequency across the bottom….. I’m sure that the frequency of the alias vs the fundamental frequency makes a difference, as does the volume of the fundamental etc…
If you must dive deeper here’s some resources.
Try google …Because at this point I’m done with doctoral theses on hearing aid design.
Headache… Excedrin…. you get the picture.
So since it’s just a number someone picked to be their standard… I’ll roll with it for now. We’re just going to be looking at what is actually there and reporting it. I’ll let the hearing aid doctors debate the merits of the 60db signal to alias ratio.
Ok time to look under the hood. First up… Fuchs ods50 vs Line 6 Litigator.
Here are the settings on the Fuchs
Fuchs is just there to compare…. (’cause real amps don’t alias)
So lets see 10khz first. Fuchs (green) and Helix (blue)
Ok so this looks pretty close… Shockingly close actually. I see my Fundamental frequency at 10khz, and its first harmonic at 20khz. A couple other spikes but they all match up, and as we know real amps don’t alias. The 2 spikes at 120 and 240 on the Fuchs graph are from the power supply which I was able to dial out with the ripple control on Helix… so far so good. Up next… 11khz
Ok, so they both get messy here, but the Helix still consistently produces that spike at 5200 hz and at 3900 hz like the last graph. Now for 9khz
Ok! Well Both amps clearly like 9khz better. But what the hell are those spikes at 4.305khz and 13.28khz? No idea. Aliases? Aliens? maybe…… to be honest I used the Helix to switch in and out the Fuchs, maybe the sample rate of Helix made a difference.. Idk .
Headache… Excedrin…. Rinse… Repeat….
New test gear arrives tomorrow… More soon.
New Gear Day!!!!
My dentist loves sweetwater as much as I do.
Anyway, I spent the better part of 2 days getting this rig dialed in. I’m not a Windows guy, plus the laptop I was using was in as far over its head as I was… anyway a quick bribe to my youngest son (who just got a gaming pc for christmas) and I had a totally workable laptop. It just needed a total clean wipe a windows reinstall, and all my software moved over… Thus the better part of 2 days. But in the interim I still had my old unit going. so I continued to conduct tests, figuring I’ll just update the screenshots…
Here are some posts from those tests.
This is a line 6 helix Cali IV Lead vs my Mesa Boogie Mark V in mark 4 mode.
And this is the Line 6 Litigator vs my Fuchs ODS 50 HRM.
Everything looks ok. I’m not loving the shape of the graph, it was a sweep file I just chopped up in logic and looped, and I wasn’t particularly happy with the result, but both amps received the same signal, and since we’re looking at the differences, not the shape of the graph, it seemed relevant to me. So I put em up on facebook to show how close the line 6 was to a “real” amp.
Right about now The laptop has finished loading windows. Time for my interrogation by Cortana.
Headache… Excedrin…. It’s becoming a theme…
Cortana seems satisfied with my answers… So let’s get back to it. Load up the software and away we go!
First up lets look back at that litigator and see if there’s any differences from the old gear.
Here’s that 9k test tone. This time no aliases. It turns out all digital gear creates aliases, and my old (cheap) test gear was no exception. So either my test gear wasn’t able to accurately produce the desired frequency, or it interpreted incorrectly on the return. Either way, the Aliens are gone… for now. Now lets run that sweep test from logic and see what we get.
Ok so the presumption here is that those lines going up and down to the left of the sweep frequency are aliases, but there’s so much noise. It is at this point I realize that almost anything I do with the computer causes a noise spike. Even as simple a thing as hitting Printscreen to take a screenshot, so runnung an audio file is killing the performance. So I’m betting that it’s processor noise, or video card noise, but it’s apparently coming from within my pc and not elsewhere. Maybe a signal generator will work. But TrueRTA doesn’t support sweeps, and plus it’s noisy when I first turn on a test tone. So a couple googles later I find an online tone generator. Ok lets see what we get. First up . Helix with no blocks installed… a blank path. Lets see how just the Hardware performs.
Looks good. No sign of Aliens here… Just a steady sweep from 9-12k and a nice clean graph. Now lets put a litigator in the path and see what we get.
Holy crap! WTF is that??? …Unexpected to say the least. Is this the rare and seldom seen sweeping Aliens.. Uh I mean Aliasing caused by harmonics folding back over the Nyquist frequency and appearing in the noise floor of the amp model…?? Unexpected, but only because I’ve seen no sign of it up till now. (It’s amazing what a bit of new gear will do) …Anyway, assuming it is aliasing, let’s see what the vaunted “signal to aliasing” number is. At the 4-5 second range we have the loudest harmonic fold thru at -64db at that point the fundamental frequency is at -10db so that’s 44db “signal-to-peak-aliasing” or 16 db shy of Cliff’s stated minimum. Well now I guess we may as well try my amp again, and lets see what we get. Cause remember real amps don’t alias…
Holy Crap! THE ALIENS HAVE INFESTED MY FUCHS!!!!
Well that was even more unexpected than the last test that came up weird and stuff… with harmonic waves sweeping all over the place… Real amps cant alias, so what is this I’m seeing? Is my new rig incapable of measuring this stuff?
Headache… Excedrin…. walk away… get some sleep.
So the next day I’m thinking about this test, and reading up on antialiasing filters, (yes thats actually a thing), and I started to realize that every A/D converter is susceptible to aliasing, and better ones will have some sort of anti-aliasing filter. Since anti aliasing filters are just basically low/high pass filters, maybe I overloaded my A/D converter, and compressed the signal so much the filter couldn’t differentiate between aliases and non aliases. Ok so I redo the test on the Fuchs, carefully calibrating my test rig, and boom.
The aliens are gone! A bit of further testing reveals that if I push that input signal over +31dB of gain thats where the filter starts to break down. So now that I know the limitations of my gear I can finally run that test and see what the heck Cliff is talking about.
The graph above is that same sweep as I subjected My Fuchs to, 9-12k sine wave. This time the test was conducted with the “peak hold” function on, the same way Cliff apparently performed his when he compared his model of a high gain amp to the helix model.
Placater (green), SLO lead OD (purple), Cali Rectifire (Light yellow), Cali IV Lead (Orange), ANGL Meteor (Lt Blue), PV Panama (dark blue)
Signal path for Helix tests: Toshiba laptop>Audient ID4>Helix>Id4>Laptop
Signal path for amp tests: Toshiba laptop>Audient ID4>Fuchs ODS50 hrm>Torpedo captor DI out>Id4>Laptop
Lastly, a few notes about all of this. This test is brutal, and probably somewhat unrealistic. First, using the peak hold function creates a graph that is a bit misleading, as it only maps the worst performance at each frequency interval, and not the average. Which is important because not all frequencies are stimulated at the same time. Second Amps do produce lots of harmonics, but this is really loud at 9-11k, frequencies that can never be the fundamental because they are beyond the guitar’s range. Just for an example the amp that Cliff used produces harmonics that are 12db lower at 9khz and 18db lower at 11khz. Point is the fundamental frequency will always be significantly louder than any of this.
At this point I really need to get my hands on an Axe Fx unit to compare to. If you are interested in seeing that test happen you can contribute here.
That’s all for tonight,