HELIX IR RATES

[thomasnorman112]

Whats up guys, i'm new to the Helix and IR's in general. I am curious what the difference between the 2048 Sample Rate, and the 1024 Sample Rate. 

 

I am also curious about the difference between 44.1, 48. and 96 when downloading IR's. 

 

Does a higher sample rate give you a higher definition sound? And is there a difference in sound from a 44.1 IR download as apposed to a 96?

 

ALSO. I'm getting a lot of buzz and feedback running a 4 cable method into my 5150. Any ideas? I use a Furman Power Conditioner so i would imagine its not ground noise. It's only when i'm running an IN/OUT FX loop into my EVH head. 

 

Thanks dudes!

[spikey]

Think of bit rates (16 or 24 bit) as made up of dots and spaces. Each increase makes the spaces smaller between the dots. The fidelity of your audio after running thru plugins and such will be better at higher bit rates. On Sample rates, Heres a good copied tidbit-> most just use CD quality (44.1). Why? Because half of the sample rate at 44.1Khz = maximum frequency of 22.05KHz. 48 = 24KHz. and 96 = 48KHz. All of which are above the range of human hearing. 

[duncann]

Whats up guys, i'm new to the Helix and IR's in general. I am curious what the difference between the 2048 Sample Rate, and the 1024 Sample Rate. 

 

I am also curious about the difference between 44.1, 48. and 96 when downloading IR's. 

 

Does a higher sample rate give you a higher definition sound? And is there a difference in sound from a 44.1 IR download as apposed to a 96?

 

ALSO. I'm getting a lot of buzz and feedback running a 4 cable method into my 5150. Any ideas? I use a Furman Power Conditioner so i would imagine its not ground noise. It's only when i'm running an IN/OUT FX loop into my EVH head. 

 

Thanks dudes!

 

From a perspective meaningful to Helix, the difference between 2048 and 1024 samples will be DSP used. Or one will use half the DSP of the other. Doubtful if you would be able to hear a difference between the two.

 

For sample rates, Helix converts everything you import to 16-bit, 48KHz. So it doesn't matter what you choose to start with.

 

Can't help with the last part, sorry.

[zooey]

Bottom line: Use 48kHz, 16-bit, mono, .WAV  files if they're available, either 1024 or 2048 samples. If not, the Helix app converts all IRs to that format, so no worries.

 

More blah blah, if you care...

 

IRs are kind of a digitized sample of the audio response of a cab or acoustic space.

 

16 and 24 bit are the number of bits used to encode each sample, the number of discrete vertical steps used to digitize the audio. In the real world, audio isn't stepped at all, it's a smooth curve. 24 bit IRs approximate that smooth curve more closely than 16 bit. Here's a pic of that stepped approximation. You can imagine that the finer the steps are, the closer the digital version is to the original analog curve.

 

44.1, 48. and 96 kHz are the sample rate, how many samples per second are taken. Higher sample rates are higher fidelity, particularly at higher frequencies, but Spikey is right that CDs, typically the highest fidelity we're exposed to these days, are 44.1kHz. Since the theoretical limit on high frequency response is half the sample rate, 44.1kHz audio allows for response up to roughly 22 kHz, beyond what (most people think) humans can hear.. You've no doubt run into the fact that us guitarists often limit high frequency response much lower than that anyway.

 

1024 and 2048 aren't sample rates, they're the number of samples in the IR, i.e., its length, the time it lasts. Cab IRs are very short, where ones intended for reverb are much longer. They have to be, to represent the decay characteristics of larger acoustic spaces like a concert hall. Helix doesn't support longer, reverb-style IRs, just short cab ones. Cab IRs with 2048 samples are in some sense higher fidelity than 1024-sample ones, but they require more DSP power. I haven't compared 1024 vs 2048 versions of the same thing myself, but some people say they sound some flavor of better. If you have access to some IRs that are the same except for their length, listen for yourself and see if you hear enough difference to spend the extra processing on it.

 

Helix uses 48kHz, 16-bit, mono, .WAV  files internally, so any greater resolution in any dimension doesn't get you anything. If that format is available from whatever provider you're looking at, just use it. The Helix app automatically converts all IRs to that format anyway.

[Thurston9]

Might want to start a new thread about your noise issue. Might get better reaponses.

[PeterHamm]

It's super easy to compare a 1024 and 2048 IR with a footswitch and decide if you can hear the difference.

When I did this test with an acoustic guitar IR (I don't use 3rd party IRs for cabinets, preferring what is built in to the unit), I discovered that in the room I play in, I can NOT hear any difference at all.

Yet I still use the 2048 sample ones since I have the CPU power available.

 

As far as bit rate, Helix changes whatever you throw in to 16-bit 48K. There's no benefit to using higher bit rate/depth ones as they get converted, and it could be argued that it's best to just use 16-bit 48K so they don't get changed at all.

[jeremyn]

The main audible differences between 2048 and 1024 sample sizes is the effective frequency resolution (which is quite noticeable in the lower frequency band) and the detail you get by not cutting off a reverb tail with an IR that is too short for the chamber being modelled. But, depending on the IR and what it is modelling, the two lengths may sound nearly identical.

 

Think of it kind of like a 10 band EQ vs a 5 band EQ. If the 5 band is already EQing all the right frequencies in just the right way for your situation, then adding more bands isn't going to improve anything. However, if the 5 band is just missing a few spots that happen to be between bands, then the 10 band will give you just enough extra control to make things better.

 

When using acoustic guitar body IRs, the tail is noticeable up to about 8192 samples, with 4096 being about optimal when taking diminishing returns into consideration. Once you start dropping samples lower than that, the acoustic body IRs sound less realistic, where at 2048 and 1024 they're still ok, but less detailed. Same applies to cabinets. Maybe it won't even matter when used live due to other instruments / reverb masking the subtleties.

 

Some speaker cabinets are not very resonant and therefore there isn't much audible difference between long and short IRs. While other more resonant cabinets can really benefit from both the extra low frequency resolution of 2048 sample IRs and the extra detail in the longer tail.

[Ka5par]

Related question: would mixing 2 or more IR-s in a DAW and adding resulting mono file into Helix, give the same result as if I had multiple IR blocks in parallel?

That would save the resources in Helix if a tried and tested combination could be in just one IR.

[PeterHamm]

Related question: would mixing 2 or more IR-s in a DAW and adding resulting mono file into Helix, give the same result as if I had multiple IR blocks in parallel?

That would save the resources in Helix if a tried and tested combination could be in just one IR.

 

Never tried this, but I've been assured it can be done.

[Digital_Igloo]

Try this:

1. From a brand new empty preset, route Path 1 into Path 2.
2. After the Amp block on Path 1, place a 1,024 IR block.
3. On Path 2, place a 2,048 IR block with the exact same IR and settings as the one on Path 1.
4. Assign both IR blocks to the same footswitch. While one is selected, press BYPASS.

Now you can toggle between the 1,024 and 2,048 versions of the IR to see whether you can hear a difference (and if you can, whether or not that difference is worth the DSP hit).

[Guitarzman]

It's super easy to compare a 1024 and 2048 IR with a footswitch and decide if you can hear the difference.

 

When I did this test with an acoustic guitar IR (I don't use 3rd party IRs for cabinets, preferring what is built in to the unit), I discovered that in the room I play in, I can NOT hear any difference at all.

 

Yet I still use the 2048 sample ones since I have the CPU power available.

 

As far as bit rate, Helix changes whatever you throw in to 16-bit 48K. There's no benefit to using higher bit rate/depth ones as they get converted, and it could be argued that it's best to just use 16-bit 48K so they don't get changed at all.

Agreed I tried this test with my Redwirez IR s which come in samples of  48 khz / 16 bit and also  48 khz  / 24 bit. I Loaded both samples 16 & 24 and saved them to the exact same patch side by side. One loaded with 16 bit  and one loaded with 24 bit. To me the 16 bit sound a bit better and closer to the real speakers. Which by the way sound awesome!  The 24 bit sound slightly boomier. 

[PeterHamm]

Agreed I tried this test with my Redwirez IR s which come in samples of  48 khz / 16 bit and also  48 khz  / 24 bit. I Loaded both samples 16 & 24 and saved them to the exact same patch side by side. One loaded with 16 bit  and one loaded with 24 bit. To me the 16 bit sound a bit better and closer to the real speakers. Which by the way sound awesome!  The 24 bit sound slightly boomier. 

 

 

The 16-bit probably only sounded better because it wasn't being re-sampled or anything. Good test, though! Surprised you could hear a difference at all.

[Guitarzman]

The 16-bit probably only sounded better because it wasn't being re-sampled or anything. Good test, though! Surprised you could hear a difference at all.

Thanks Peter! Yes just a slight difference, my set-up is through studio monitors. I'm wondering if anyone has tried this directly through an amplifier.

[phil_m]

I know this is only tangentially related to the post, but the whole "stair steps" analogy for bitrate and frequency as it relates to digital audio isn't a very good one... Increasing the bitrate doesn't really give you more detail, it just lowers the noise floor, and gives you more dynamic range to work with.

 

http://productionadvice.co.uk/no-stair-steps-in-digital-audio/

 

[willjrock]

Agreed I tried this test with my Redwirez IR s which come in samples of  48 khz / 16 bit and also  48 khz  / 24 bit. I Loaded both samples 16 & 24 and saved them to the exact same patch side by side. One loaded with 16 bit  and one loaded with 24 bit. To me the 16 bit sound a bit better and closer to the real speakers. Which by the way sound awesome!  The 24 bit sound slightly boomier. 

 

The 16-bit probably only sounded better because it wasn't being re-sampled or anything. Good test, though! Surprised you could hear a difference at all.

I doubt if there was any difference at all because the second he imported that 48/24 file into the Helix, it was converted to 48/16, therefore when doing his analysis he would have been comparing two identical files, and any difference in sound was most likely caused by placebo ,the positioning of his head - even an inch, or something similar..

[spikey]

I see and get the noise floor, so thanks for that Phil. Wow on the Dither difference! But If you will notice the analog waveform changed from clean @ 2:22, to a bit fuzzy looking 5:58 in the video. Look close. I would wager if he hit MAG you would see some tiny stair stepping noise there... :) Yes the frequency did go up and it is still the same overall shape on the scope, but the converted analog signal there is not as clean @ the higher frequency after D to A conversion is my point. So for argument's sake, if the analogy is wrong, and if we are using the best analog scope to test this, and if there is no stair step increase at higher frequencies, then why is there a bit more fuzzy looking waveform at the higher freq's? What else has changed (besides the frequency itself) to cause this then? Is the D to A converter not smoothing out the analog as well as today's interfaces do perhaps? Great geekoid video however, I enjoyed it!

[FlyingsCool]

I see and get the noise floor, so thanks for that Phil. Wow on the Dither difference! But If you will notice the analog waveform changed from clean @ 2:22, to a bit fuzzy looking 5:58 in the video. Look close. I would wager if he hit MAG you would see some tiny stair stepping noise there... :) Yes the frequency did go up and it is still the same overall shape on the scope, but the converted analog signal there is not as clean @ the higher frequency after D to A conversion is my point. So for argument's sake, if the analogy is wrong, and if we are using the best analog scope to test this, and if there is no stair step increase at higher frequencies, then why is there a bit more fuzzy looking waveform at the higher freq's? What else has changed (besides the frequency itself) to cause this then? Is the D to A converter not smoothing out the analog as well as today's interfaces do perhaps? Great geekoid video however, I enjoyed it!

 

Nope, it's not possible for stair stepping noise.  It is possible to change the waveform depending on how you bandlimit the signal, as he showed, but stairstep noise is a total myth.  The only stairstepping you will see will be due to the monitor, not what's really in the signal.  It's an analog signal... period (pun intended)

 

You can tell that there is no audible noise outside of the noise floor from the spectrum display.  If there were any other signals (i.e. that stairstepping you thought you could see) you would see peaks at frequencies other than the peak of the original sine signal at volumes higher than the noise floor.  He demonstrated this when he discussed dithering and showed harmonic distortion.  But, at 16 bits, that noise is pretty much inaudible with decent A/D -> D/A

[spikey]

Can you explain the noise @ 5:58 on the Oscope waveform that wasn't there @ 2:22? That's a good Oscope he was using. If there was not noise there it would not have shown it. So given there was noise, where was it coming from?

[FlyingsCool]

Can you explain the noise @ 5:58 on the Oscope waveform that wasn't there @ 2:22? That's a good Oscope he was using. If there was not noise there it would not have shown it. So given there was noise, where was it coming from?

 

I'm confused.  What noise at 5:58?  You will not be able to see noise in the Oscope, not in the display mode he is using.  You'll want the spectrum analyzer output to really see any noise or harmonic content other than the wave you see on the Oscope.  Plus at 5:58, he clearly states there are no stair steps. 

 

But to answer your question, As he points out at at 2:26, there IS noise in the original analog signal, with a floor of about -90 dB to the original signal, and some harmonic distortion about -70 dB lower than the original signal.  I'm assuming it's just electrical noise leaking / errors in the signal generator.

[spikey]

I'm confused.  What noise at 5:58?

 

Look at the trace on the Oscope and compare it to the one @ 2:22. It's much dirtier/noisier.

 

You will not be able to see noise in the Oscope

 

And yes you can see noise on an O-scope. That's why they make O-scopes bud. It's a troubleshooting tool for Electronic techs as well as Engineers. That's what most techs have used for decades looking at signals and dirty power/ripple. At 2:22 there is hardly any noise compared with the Oscope trace at 5:58. Have you ever used an O-scope before Cool, or are you just going by this video? No slam intended, but I've used one for around 40 years off and on and can immediately tell by looking at the traces whether they are dirty/noisy with ripple or clean signals. The one at 5:58 has noise on it.

[phil_m]

Look at the trace on the Oscope and compare it to the one @ 2:22. It's much dirtier/noisier.

 

And yes you can see noise on an O-scope. That's why they make O-scopes bud. It's a troubleshooting tool for Electronic techs as well as Engineers. That's what most techs have used for decades looking at signals and dirty power/ripple. At 2:22 there is hardly any noise compared with the Oscope trace at 5:58. Have you ever used an O-scope before Cool, or are you just going by this video? No slam intended, but I've used one for around 40 years off and on and can immediately tell by looking at the traces whether they are dirty/noisy with ripple or clean signals. The one at 5:58 has noise on it.

 

Well, that's the point isn't it? Decreasing the bit depth raises the noise floor... He even says there's noise at 9:17... It's not that bit depth doesn't matter. It's just that people misunderstand what it means. The most interesting thing to me in this video is when talks about the noise floor of magnetic tape in terms of bit depth and how you were lucky if a cassette tape gave you a noise floor equal to 6 bits...

[Guitarzman]

I doubt if there was any difference at all because the second he imported that 48/24 file into the Helix, it was converted to 48/16, therefore when doing his analysis he would have been comparing two identical files, and any difference in sound was most likely caused by placebo ,the positioning of his head - even an inch, or something similar..

Thanks for the reply willjrock your probably correct to many variables to dispute over!  

[PeterHamm]

I doubt if there was any difference at all because the second he imported that 48/24 file into the Helix, it was converted to 48/16, therefore when doing his analysis he would have been comparing two identical files,...

 

 

Not comparing identical files if the way that the extra information is truncated is different in Helix than in whatever computer program was used... Could be VERY different, and noticeable.

[spikey]

Well, that's the point isn't it? Decreasing the bit depth raises the noise floor... He even says there's noise at 9:17... It's not that bit depth doesn't matter. It's just that people misunderstand what it means. 

 

I think that there is still residual stepping there in 5:58 on the O-scope thus the noise seen, covered up and smoothed by some kind of internal dithering. Yea I know I'm stubborn (but not stupid) and so its clear what the guy shows. But, that's the only explanation for the noise I see on the O-scope @ 5:58 (versus 2:22) in the video. So, once again if that's not it, what is it then? 

[redmonda]

I think that there is still residual stepping there in 5:58 on the O-scope thus the noise seen, covered up and smoothed by some kind of internal dithering. Yea I know I'm stubborn (but not stupid) and so its clear what the guy shows. But, that's the only explanation for the noise I see on the O-scope @ 5:58 (versus 2:22) in the video. So, once again if that's not it, what is it then? 

 

I'm looking at the video... and I just don't see the noise that you do.  I see some fuzziness in the scope, but I believe this can be accounted by the analog screen having limited resolution. Put this on a high-res digital scope, and you'd get a cleaner display.

 

Edit: Forgot to mention, EE here with plenty of experience with modern and vintage scopes.  We have a sweet all-tube scope at my work from the 1940's! 

[FlyingsCool]

Look at the trace on the Oscope and compare it to the one @ 2:22. It's much dirtier/noisier.

 

And yes you can see noise on an O-scope. That's why they make O-scopes bud. It's a troubleshooting tool for Electronic techs as well as Engineers. That's what most techs have used for decades looking at signals and dirty power/ripple. At 2:22 there is hardly any noise compared with the Oscope trace at 5:58. Have you ever used an O-scope before Cool, or are you just going by this video? No slam intended, but I've used one for around 40 years off and on and can immediately tell by looking at the traces whether they are dirty/noisy with ripple or clean signals. The one at 5:58 has noise on it.

 

Yes, I've used all those pieces of equipment before (ME with some vibration testing experience).  Granted, I don't have the experience you do, but I'm with @Redmonda.  I just don't see a difference between the two signals other than video reproduction blockiness and light bloom perhaps.

 

And, mathematically, never think of noise as being due to any stairsteppiness.  There can be the quantization noise he explains due to round-off error and jitter and million other things, but no stairstep exists.  It's a figment of our imagination.  The D->A conversion creates a smooth analog signal.  There can be rippling due to the bandwidth limiting, again as he explains, but that's not in that signal as far as I know.  I don't think the sine wave input has any information outside the bandwidth that the signal is limited to.... But now I'm going beyond my knowledge...

[spikey]

Hey, it's all good.  I just wanted to point it out and as always YMMV.  :D