When the KPA was new (I picked mine up very shortly after performance mode was released), there was a lag of around 250mSec between rigs and everyone immediately jumped on this as being impossibly long. 2 Seconds is an eternity.
FYI, the last time I measured the KPA switch time in performance mode, I believe it was around 40mSec which I feel is pretty darned close to "instant"
When the KPA was new (I picked mine up very shortly after performance mode was released), there was a lag of around 250mSec between rigs and everyone immediately jumped on this as being impossibly long. 2 Seconds is an eternity.
FYI, the last time I measured the KPA switch time in performance mode, I believe it was around 40mSec which I feel is pretty darned close to "instant"
2 seconds would be substantial, but I have not been able to find neural dsp giving out such info anywhere, and the poster isn't sure (it seems) about the 2 second figure anyway (to poster: feel free to post the video as can't find it). About to email them to ask about the gap time. But sure -- if this is about any perceptible gap, that's not something I didn't expect. No surprise there considering how cortex works.
Whether it is 2 seconds before the patch changes, or 2 seconds of silence is more important.
2 seconds would be substantial, but I have not been able to find neural dsp giving out such info anywhere, and the poster isn't sure (it seems) about the 2 second figure anyway (to poster: feel free to post the video as can't find it). About to email them to ask about the gap time. But sure -- if this is about any perceptible gap, that's not something I didn't expect. No surprise there considering how cortex works.
from their website: “Preset changing will incur in the inevitable, familiar gap. You can use Scene and Stomp modes to change the sounds while maintaining delay trails without any crossfade/gaps.“
It doesn’t say here how long that gap is. They talked about the gap in a very long livecast video during its development, that I had found on YouTube. Definitely not spillover. So it defeats the assumption that this thing has gobs of untapped DSP. It does in a sense, but reserves it for within presets by giving you four amps and cabs and ridiculous amount of effects to go with them. So even if the gap is super short there’s still no spill over and even though I don’t always need spill over in the middle of songs, I need it enough that in general I have to organize around that common need, hence trying to (in general) cram everything thing I need for a song within one preset.
from their website: “Preset changing will incur in the inevitable, familiar gap. You can use Scene and Stomp modes to change the sounds while maintaining delay trails without any crossfade/gaps.“
It doesn’t say here how long that gap is. They talked about the gap in a very long livecast video during its development, that I had found on YouTube. Definitely not spillover. So it defeats the assumption that this thing has gobs of untapped DSP. It does in a sense, but reserves it for within presets by giving you four amps and cabs and ridiculous amount of effects to go with them. So even if the gap is super short there’s still no spill over and even though I don’t always need spill over in the middle of songs, I need it enough that in general I have to organize around that common need, hence trying to (in general) cram everything thing I need for a song within one preset.
"Maintaing Trails" = "Spillover"
Having to build a single preset and try to incorporate scenes/snapshots was what finally killed the Helix for me, in the time I had one. I hate having to strategise my sound with regards to gigs; I want to just find sounds that work and not have to think about workarounds.
Don't know, of course I'll better wait for "real" reviews but I expect this to be impressive for Kemper users. The least.
Not that it would make me better guitar player, but I wonder why they didn't put optical connection...
Not that it would make me better guitar player, but I wonder why they didn't put optical connection...
I'd probably buy it if it had SPDIF.
If it captures really well then you could have a few of your kemper rigs in there too and in one preset.
There no denying, it does sound awesome.
'Short delay' has been announced, with no new date.
I have to say, communication from Neural has been utter crap. Certainly left a very sour taste in my mouth.
I have to say, communication from Neural has been utter crap. Certainly left a very sour taste in my mouth.
Yes, definitely. But I would add, as a new company they deserve a true chance to prove themselves. I wonder how Kemper handled everything 9 years ago with the introduction of the Toaster and if everything was perfect from A to Z. I wasn`t around till 2007 when everything was rolling perfectly.
Having to build a single preset and try to incorporate scenes/snapshots was what finally killed the Helix for me, in the time I had one. I hate having to strategise my sound with regards to gigs; I want to just find sounds that work and not have to think about workarounds.
That is the reason the kemper has become industry standard.No "strategies" and "oh shit"s because you spend crucial time with "workarounds" while in "real action" during live and recordings.
Btw..Two seconds?I don't believe that.Even half a second would be a disaster anyway..can't be..not in freakin' 2020..
That is the reason the kemper has become industry standard.No "strategies" and "oh shit"s because you spend crucial time with "workarounds" while in "real action" during live and recordings.
Btw..Two seconds?I don't believe that.Even half a second would be a disaster anyway..can't be..not in freakin' 2020..
I misspoke, my apologies. I could have sworn he said two seconds, he didn't, unless he addressed the issue more specifically somewhere else. However, I just found the part of that video, like 55:55 into it, where the question was asked. What he indicated was they'd have the same inherent gap you find in other modelers, it's not instant, as everything on the previous preset will be overwritten. But he didn't indicate how much of a gap there is. By other "modelers," does he mean the Helix which is unacceptably long? Or similar to the KPA Rig change, which is excellent, but with no spillover? (*Edit: the KPA HAS spillover on Rig, the Quad doesn't on preset changes.)
For example, the gap between rigs in Kemper, as we know, is so short is isn't worth worrying about, and whatever gap there is, is covered by spillover or "maintaining trails." So we can go between rigs in the middle of songs without thinking twice. Of course, the spillover of delays and reverbs on the KPA is limited to the dedicated delay and reverb blocks, which is where you'd want to place these in the vast majority of situations anyway. I have a handful of ambient rigs where I have used an extra delay in the mod block, and even these sound natural to change rigs, even though technically the mod slot doesn't carryover. But once again, it is the fixed signal path of the KPA that makes it able to have two dedicated blocks allowing trails.
I think it is quite interesting to realize that we today still talk about gaps between changing presets on high priced guitar modelers..this is funny at least..
Regarding the possible patent infringement of the "neural capture" process available in the Quad Cortex:
There is a legal part of the issue, and a practical part.
The legal part is basically about whether there is actually a patent infringement by Neural DSP when selling/offering the Quad Cortex with the capture feature, or whether they are legally free to do it.
The practical part is about whether, under the assumption that Neural Capture actually does or at least might violate one or some of the patents owned by Kemper, it would make sense to sue, or whether the trial costs - or even a possible risk of losing the trial - might be to high (if one assumes that there is no legal infringement, going to court does not make any sense whatsoever).
The practical part has to do with the fact that patent infringement lawsuits typically cost a freaking truckload of money, and you might get it back from the infringer, or at least to some amount, but if the costs for the lawyers and the court fees etc. are much higher than what you could get out of it, then you have to see whether a product can be so dangerous to you that you want to keep it out of the market at almost all costs (or whether you want to try to settle on an agreement with the infringer in order to not risk your own cash flow etc.). And: Sometimes, infringements are not that easy to prove, in particular in software-based implementations - i.e., to be totally proven, not some "it must be like this" guessing, but sufficiently solid proof so that a judge will rule over tenthousands (or hundreds of thousands) of $$/€€ on basis of the proof.
The more technical part is the legal part. It seems that the european patent is valid in Germany, UK, France and Sweden, and the protection of the patent (what the patent owner truely "ownes" as invention) is actually for the imlpementation of the profiling process as some Wiener-Hammerstein model of a hyperbolic tangent sandwiched by two frequency response transfer functions (the ones to be found by the profiling in order to represent the amp in question). If Neural DSP found some workaround here, they would appear to be free. They claim they are doing some neural-network-like approach, however without specifying what exactly the networks are "learning"/training (but a priori, it sounds different).
However, the US patent and a further German patent both have a protected invention that seems insanely general, it looks like it might cover all possible profiling/capture/younameit processes as soon as the process uses a comparison/difference of two frequency response curves (one for the real amp, the other for its profile). It is difficult to imagine any possible implementation of a profiling/capture process without comparing these two frequency responses. One probabely would have to implement the process completely in time domain rather than with frequencies (to be absolutely sure about not touching the Kemper patent), and I find it difficult to believe that such an implementation is technically feasible at all (in particular, when hearing the white noises involved).
Given what is protected by the US patent and the feasibility issues with what's left to do then, I could imagine that the Kemper US patent actually might be infringed by the Neural Capture process. Whether this is easy to prove now (I mean, "court-solid" proof), could be a totally different story
Regarding the possible patent infringement of the "neural capture" process available in the Quad Cortex:
There is a legal part of the issue, and a practical part.
The legal part is basically about whether there is actually a patent infringement by Neural DSP when selling/offering the Quad Cortex with the capture feature, or whether they are legally free to do it.
The practical part is about whether, under the assumption that Neural Capture actually does or at least might violate one or some of the patents owned by Kemper, it would make sense to sue, or whether the trial costs - or even a possible risk of losing the trial - might be to high (if one assumes that there is no legal infringement, going to court does not make any sense whatsoever).The practical part has to do with the fact that patent infringement lawsuits typically cost a freaking truckload of money, and you might get it back from the infringer, or at least to some amount, but if the costs for the lawyers and the court fees etc. are much higher than what you could get out of it, then you have to see whether a product can be so dangerous to you that you want to keep it out of the market at almost all costs (or whether you want to try to settle on an agreement with the infringer in order to not risk your own cash flow etc.). And: Sometimes, infringements are not that easy to prove, in particular in software-based implementations - i.e., to be totally proven, not some "it must be like this" guessing, but sufficiently solid proof so that a judge will rule over tenthousands (or hundreds of thousands) of $$/€€ on basis of the proof.
The more technical part is the legal part. It seems that the european patent is valid in Germany, UK, France and Sweden, and the protection of the patent (what the patent owner truely "ownes" as invention) is actually for the imlpementation of the profiling process as some Wiener-Hammerstein model of a hyperbolic tangent sandwiched by two frequency response transfer functions (the ones to be found by the profiling in order to represent the amp in question). If Neural DSP found some workaround here, they would appear to be free. They claim they are doing some neural-network-like approach, however without specifying what exactly the networks are "learning"/training (but a priori, it sounds different).
However, the US patent and a further German patent both have a protected invention that seems insanely general, it looks like it might cover all possible profiling/capture/younameit processes as soon as the process uses a comparison/difference of two frequency response curves (one for the real amp, the other for its profile). It is difficult to imagine any possible implementation of a profiling/capture process without comparing these two frequency responses. One probabely would have to implement the process completely in time domain rather than with frequencies (to be absolutely sure about not touching the Kemper patent), and I find it difficult to believe that such an implementation is technically feasible at all (in particular, when hearing the white noises involved).
Given what is protected by the US patent and the feasibility issues with what's left to do then, I could imagine that the Kemper US patent actually might be infringed by the Neural Capture process. Whether this is easy to prove now (I mean, "court-solid" proof), could be a totally different story
Writing a good patent is like a balancing act:
1. Making in general enough that it can cover a lot of things, but hasn't been previously done. But also ...
2. Making it specific enough so that it is patentable
For Kemper, I would say it is easy enough to patent a *very specific* profiling process. As you mention, some specific mathematical model and implementation. I can see this as more easily defensible. But the more specific it is, the easier it is to get around it.
The thing you mention on Kemper patenting a process whereby two EQ curves are compared -- that seems pretty general. If they were granted a patent on that, I could see that be defeated in court based on previous SW or HW that does this. Many SW or HW equipment today do this sort of thing (i.e. tonematching, Tonedexter, etc.), or any series of audio SW plugins or apps that compare EQs of incoming songs or signals and create EQs to match.
In any case, I am speculating here but I think Neural have some profiling method that does some some of matching to do capture comparisons with the reference, and does so using a NN model rather than something more specific in Kemper's case. In this sense, it isn't violating their patent on the specific profiling method. But would violate some general patent that compares frequency responses. But as I said, the latter seems so general that it would be invalidated if brought to court.
I have 5 patents and have gone through the legal process that many times.
First, you can patent anything. You just can't defend it so the patent isn't worth the paper its printed on if you don't perform due diligence.
Second, to get the patent, you have to expose how you did it. Anyone that can figure out from that how you did it can tweak the process enough to bypass the patent.
Third, you can't get blood out of a turnip. If the company doesn't have enough money to shake a stick at, why file suit?
I personally LIKE having my name on patents. Having patents makes your company value higher; however, unless you are prepared to take legal action, and have patents that you and your company understand well enough to defend in a court of law, it is better just to have trade secrets IMO.
CK could send them an "invitation to participate in the Kemper licensing program" from his lawyer. This is how most suits I have been involved in (2) started out.
Well, fractal will have their own version of capturing source tones (other than tone matching alone) in the future. Before, they said they didn't do so "not to get sued". Apparently now they have a distinct enough method; as I believe does neural dsp anyway with machine learning.
