Status
Not open for further replies.
Your null test seems quite loud so there are improvements to be had or the or it hasn't been volume matched etc.
You can try upping it to a 1000 epochs but that would take more time
It's not that loud compared to the recorded samples, and null tests made with other profilers are definitely louder. I don't know ho easy it would be for a capture or a model to completely null against a reference amp anyway.
 
I just did a capture of my 5150 block letter patch which has the Maxon 808 boost in front of it. It didn't come out so good. The over all distortion got dialed back somehow. When I did the Das Metall capture there was no boost in front and it came out well. I'm wondering if Tonex didn't know how to deal with the boost against the amp model or something.

Hopefully with regards to capturing a real boost, fuzz etc in front of a real amp it will fare better. Largely impulse buy for me for the price, but assuming it can do a decent job capturing the signal chain as a whole, I’m looking forward to being able to emulate the tone of some of my more unique and crazy fuzz pedals. Obviously wouldn’t have the cleanup etc, but if it can come close to the sound of my MK1 TB, companion fuzz etc that would be pretty cool as I’ve yet to use a modeler that really comes all that close.
 
Anyone tried or knows if it is possible to do a capture of the Axe Fx using it as an audio interface?
I mean choosing the right input/output on the app without patching cables?
 
Anyone tried or knows if it is possible to do a capture of the Axe Fx using it as an audio interface?
I mean choosing the right input/output on the app without patching cables?
Yes it is. You just have to route properly to capture the test signal TONEX sends to capture.
 
It's not that loud compared to the recorded samples, and null tests made with other profilers are definitely louder. I don't know ho easy it would be for a capture or a model to completely null against a reference amp anyway.
null test is also extremely sensitive to time alignment. I've seen some of Steve's that null to almost nothing at all - it can do pretty incredible. Obviously will not always be the case and requires probably some of his experience and skill. I'm able to get ESR in 0.006-0.008 at 1000 epochs locally (real JCM800) - I'm planning a comparison / null test vs tonex but haven't done it yet, though training is all done
 
FWIW tonex + HX isn't any smaller than the FM3 - the main reason to do it is to use models that aren't in fractal / captures of your own stuff, imo

I generally use only the Hx Stomp for live and band practice as I don't want and don't need to bring out the Axe Fx III.
The "problem" is that the 3 switches sometimes are not quite enough for me to go through an entire song and that requires me to do some hoops.

ToneX+HX would enable me to use the former for amp sounds and the latter mainly for fx, kinda "doubling" my switching possibilities.
 
I generally use only the Hx Stomp for live and band practice as I don't want and don't need to bring out the Axe Fx III.
The "problem" is that the 3 switches sometimes are not quite enough for me to go through an entire song and that requires me to do some hoops.

ToneX+HX would enable me to use the former for amp sounds and the latter mainly for fx, kinda "doubling" my switching possibilities.
Gotcha and yeah, the hx is highly limited in fs function. You can't use a tuner and snapshots at the same time 😥

Getting a small 3 button midi baby for mine
 
null test is also extremely sensitive to time alignment. I've seen some of Steve's that null to almost nothing at all - it can do pretty incredible. Obviously will not always be the case and requires probably some of his experience and skill. I'm able to get ESR in 0.006-0.008 at 1000 epochs locally (real JCM800) - I'm planning a comparison / null test vs tonex but haven't done it yet, though training is all done
Yeah, I aligned them as best as I could but that's not easy either as the waveforms are not 100% identical, so some zero points or peaks align perfectly while others don't.
I aimed at 0.01 for the ESR cuz that's what he suggests as a good enough threshold and cuz I just wanted to try it out quickly. As I have a bit of time I'll try to do a 1000 epochs training and see where it gets.

PS: I also made another profile of the VH4 ch.3 model and noticed that it got below 0.01 ESR quicker than the bassman one, after 100 epochs it was already there and ended at around 0.009. I wonder what influences the ESR other than the number of epochs.
 
Yeah, I aligned them as best as I could but that's not easy either as the waveforms are not 100% identical, so some zero points or peaks align perfectly while others don't.
I aimed at 0.01 for the ESR cuz that's what he suggests as a good enough threshold and cuz I just wanted to try it out quickly. As I have a bit of time I'll try to do a 1000 epochs training and see where it gets.

PS: I also made another profile of the VH4 ch.3 model and noticed that it got below 0.01 ESR quicker than the bassman one, after 100 epochs it was already there and ended at around 0.009. I wonder what influences the ESR other than the number of epochs.
The character of the amp will make it harder to reach lower ESR (better accuracy). In my tests, if you do clipping diodes, brighter/more aggressive tones, NAM nails them pretty good but you need more epochs to train vs a lower gain model which could reach an ESR of 0.01 in 200 epochs tops. Loudness seems to also affect the model's accuracy (how loud the output.wav is) AND training signal length - I've tried a longer training signal (spliced audio inside the v1_1_1.wav file Steve provided) and that got me better accuracy (lower ESR).
 
The character of the amp will make it harder to reach lower ESR (better accuracy). In my tests, if you do clipping diodes, brighter/more aggressive tones, NAM nails them pretty good but you need more epochs to train vs a lower gain model which could reach an ESR of 0.01 in 200 epochs tops. Loudness seems to also affect the model's accuracy (how loud the output.wav is) AND training signal length - I've tried a longer training signal (spliced audio inside the v1_1_1.wav file Steve provided) and that got me better accuracy (lower ESR).
In the two captures I've made till now I've found the opposite actually, the low gain bassman ended up having a higher ESR than the high gain VH4, both with 150 epochs. But as you suggest it might have been due to a different volume in the two files, the first I did was the VH4 and the profile came out pretty loud, so for the second one I tried to lower the level to see if it affected the profile final volume somehow, but it didn't.
I'll try with a longer training signal, I've already recorded the test signal of ToneX and I'll add that to the v1_1_1.wav
 
Status
Not open for further replies.
Back
Top Bottom