SLO100 vs NAM vs Tonex Blind Test, Time to Go Back to 2015 :)

So, the order was NAM -> Tonex -> Amp.

Odd thing, the wave form for Tonex was almost identical to the real amp making it very easy to line them up.

NAM was wildly different.

Any thoughts on this?
It's happended to me a couple of times. I think NAM is not perfect and that some times tonex does better.

That said, I do think NAM is overall more consistent. Too much presence can throw off my tonex captures by a big margin, and it's pretty annoying when that happens.

But yes. I've seen cases where tonex is closest. (After adjusting for the lack of gain, which i do every time. Kemper and nam more accurate for gain levels imo).
 
So, the order was NAM -> Tonex -> Amp.

Odd thing, the wave form for Tonex was almost identical to the real amp making it very easy to line them up.

NAM was wildly different.

Any thoughts on this?

You can only really gauge so much from the waveform for something as harmonically rich and full range as a guitar amp. There are many reasons for why it may or may not look similar, too many to pin on one reason.

Results were as I and several others guessed - the amp has a certain feel in the low end that’s hard for any modeller to get quite right. ToneX has some giveaway clues that can make it stand out - less gain, and a different low end and top end. I’m not sure why ToneX seems to need a bit of extra gain once a model is made to get it to unity, it’s been that way from the start.

How many epochs did you train with NAM? Have you tried xSTD or Hyper Accuracy models? They’ll close the gap a good amount more often than not.
 
You can only really gauge so much from the waveform for something as harmonically rich and full range as a guitar amp. There are many reasons for why it may or may not look similar, too many to pin on one reason.

Results were as I and several others guessed - the amp has a certain feel in the low end that’s hard for any modeller to get quite right. ToneX has some giveaway clues that can make it stand out - less gain, and a different low end and top end. I’m not sure why ToneX seems to need a bit of extra gain once a model is made to get it to unity, it’s been that way from the start.

How many epochs did you train with NAM? Have you tried xSTD or Hyper Accuracy models? They’ll close the gap a good amount more often than not.
Yeah, I could have added some gain when in the capture process for Tonex.

For the NAM Captures I used Tone 3000 with the 1000 epoch setting. It stopped at around 770 epochs due to whatever measure it uses to predict the benefit of further processes.
 
I’m not sure why ToneX seems to need a bit of extra gain once a model is made to get it to unity, it’s been that way from the start.

And Kemper adds gain. Even with known hardware, it is hard to get the calibration perfect to a fraction of a db accuracy. The thing is, we can can easily adjust gain and always have. The volume pots on amps are even more inconsistent in their sweep so we should be used to giving things a little tweak to get where we want.
 
Even with known hardware, it is hard to get the calibration perfect to a fraction of a db accuracy.
Half of it is down to the user to get their levels right, but even when that’s correct ToneX seems to be consistently under. Quad Cortex does this too but I believe that’s because of its reamp levels not being aligned to the input. Harder to say for Kemper, in my experience it’s just more random in general.
Yeah, I could have added some gain when I the capture process for Tonex.
Yeah I’ve found more often than not ToneX needs about another 2dB more gain once the model is made for it to sound right. Not much that can be done for closing the gap on the other stuff.
For the NAM Captures I used Tone 3000 with the 1000 epoch setting. It stopped at around 770 epochs due to whatever measure it uses to predict the benefit of further processes.
Yeah that sounds about right. I think for most people, example A would be satisfactory. It’s possible to get more accuracy but it involves doing some tweaks within NAM and maybe playing with different reamp files. The off the shelf reamp file and training are optimised for speed of reamping+training, as well as CPU, and you still get pretty good accuracy. But if accuracy is the goal then you can spend more time on the reamping and training.
 
It's happended to me a couple of times. I think NAM is not perfect and that some times tonex does better.

I have had a few NAM captures that came out poorly, but I had similar outliers with Kemper and Tonex. I haven’t done enough with a QC to say one way or the other, but I am sure it happens there too. I have no idea what causes it.
 
I have had a few NAM captures that came out poorly, but I had similar outliers with Kemper and Tonex. I haven’t done enough with a QC to say one way or the other, but I am sure it happens there too. I have no idea what causes it.
Maybe the amp has intermittent stability issues?
 
Tonex really stood out in a negative way there, compared to the NAM (that got incredibly close) — though, of course, we're talking about subtle differences. Was that capture just from the amp (with an IR used later), or was it amp+cab? How Tonex compares to the Quad Cortex captures in your opinion?


Edit: Just to be clear, not bashing Tonex at all. I use Tonex and love it, but in this case it really lacked some gain and low end compared to the real amp.
 
Last edited:
Maybe the amp has intermittent stability issues?

It's not always the same amp, and I am pretty sure none of my amps have stability issues. Now, are there issues with the digital side, home power, etc? Possibly. Bad captures are rare, so it is not obvious what causes it, but a recapture has never failed to yield a good result so who knows.
 
Before my turn to play the harp in the sky gets here, I need to play a SLO100 in real life!

Keep in mind that even though it is a MV amp, it still sounds best when LOUD!!! Forget Warhol's 15 minutes of fame, everyone should get 15 minutes in front of a cranked SLO 100 running into a quality 4x12!!! Its a fun as hell amp!
 
Tonex really stood out in a negative way there, compared to the NAM (that got incredibly close) — though, of course, we're talking about subtle differences. Was that capture just from the amp (with an IR used later), or was it amp+cab? How Tonex compares to the Quad Cortex captures in your opinion?


Edit: Just to be clear, not bashing Tonex at all. I use Tonex and love it, but in this case it really lacked some gain and low end compared to the real amp.
This was amp and cab.

Turning up the gain does increase the low end. It is possible to adjust this when making the capture but in this case I chose not to.

I think the Quad Cortex is great. Each have theor weak points.
 
Maybe the amp has intermittent stability issues?
I've had profiles turn out weird but it was due to the amp acting up during reamping.

If the power's stable & the amp doesn't do crazy stuff, I've had great & consistent results with NAM.
 
I've had profiles turn out weird but it was due to the amp acting up during reamping.

If the power's stable & the amp doesn't do crazy stuff, I've had great & consistent results with NAM.
I apologise for the following attempted derailment but the (NAM) reamping and training process can actually be a pretty useful tool in provoking weirdness in an amp that wouldn't show up at normal signal types and levels. Sort of like an automated, unmedicated signal generator with a readout.
I wonder if it could be used to make an automatic diagnostic tool by teaching it to recognise typical faults.
 
This was amp and cab.

Turning up the gain does increase the low end. It is possible to adjust this when making the capture but in this case I chose not to.

I think the Quad Cortex is great. Each have theor weak points.

Got it, thanks for the info! Either way, Tonex got close enough — especially considering you can tweak the gain and EQ to dial it in.

As for the Quad Cortex, I see what you mean. What I’m still curious about is: if you had used a Neural Capture in this test, would the difference compared to the real amp be as noticeable as it was with Tonex, or would it be closer to the NAM — which, in this case, was scarily close?
 
Back
Top