Harmonic Converger -- Warm up Amp Modelers and Kill Fizz

[lovekraft0]

Judge for yourselves:
http://www.bossgtcentral.com/hctestdrive.php
Give the clips a listen, and decide if you hear anything miraculous happening. It's interesting that the clips are recorded with the effect applied first and then with it bypassed.

[stm]

I listened to the samples, specially the Lead Patch, and after listening to the differences the first thing that came to my mind was the effect obtained while developing the Condor Cabsim.  I did a little testing with the Windows Media player equalizer, and in the second part I was able to make it sound very close (80% there) by fully cutting the 4, 8 and 16 kHz bands, then lowering the 500 Hz band by 12 dB, and raising the 100, 1k and 2k bands like 1 or 2 dB.

Based on this, my guess is that this device contains a passive implementation of a carefuly chosen cabinet simulator response, made with inductors, capacitors and resistors.  The 3-posiion switch changes some components to alter the high frequency cutoff--a key parameter when we defined and adjusted the response of the Condor.

Now to the mojo part of the "nonlinear" device in there, well, in order to have passive filters made of inductors and capacitors you need very large valued inductors, which can introduce some coloration due to saturation if the input signal goes beyond a certain threshold.

I'm pretty sure it is a matter of measuring the frequency response of the device at different input levels to be able to infer about the internal topology and component values. If anyone dares to lend me a device like this I would gladly measure it and return it unharmed.

This is my modest opinion.

[Nasse]

Sound clips with that gadget on does sound good. I thought too that it sounds like some speaker simulator recordings.

This one was found by search...


But maybe there is that passive diode compressor thing, never tried such zener arrangement.

Years ago I tried old Elektor circuit, a passive ge diode compressor. I think I got it working, but liked it not much. I have it in some junkbox, but it attenuates signal...

There is need for such boxes because people are reporting difficulties with modelling boxes

[stm]

Interesting circuit. I wouldn't say it is a cabsim per se, however it will certainly smooth out any digital multieffect or harsh distortion.

I simulated the circuit and can say the following:

1) Insertion loss is 21.5 dB (around 1/11th of the original signal goes out at maximum gain)
2) Cutoff frequency is 6 kHz--I bet the HC is less than that, maybe 4-5 kHz
3) The response is approximately a 1.5 dB Chebyshev 3rd order filter
4) It has no mid-frequency notch (300-700 Hz ?), as opposed to the HC (based upon listening tests)

I redrew the schematic to make it more readable, combined the 10k volume pot and 1k parallel resistor, and plotted the response as shown here:



Regards.

[stm]

Well, I've got tempted to take a guess and I got something interesting.  I think this circuit should be within the ballbark. Usually one tends to get overcomplicated by mojo and esoteric descriptions; I tried to keep it reasonable simple but effective.

If you want to place your compression diodes, I would do so just after R1 to ground, or maybe even between R1 and ground (in parallel with L2-C2), so any harmonics generated here may be tamed by the lowpass filter formed by L1/C1.  I think they are not needed, though. I would better wire my inductors in a saturable small donut core if I desire to introduce some harmonics, but this is like a second order derivative, to put it in a mathematical perspective.

As you will see, insertion loss is very low, except at the 500 Hz notch and the frequencies above 5 kHz. I managed to keep component values within the typical "standard" series.

If you change the value of C1 you can tune the high frequency response to your liking  (as in the real HC), so here you place your 3 position switch!

The notch can be tuned with R2, L2 and C2.  R2 controls the maximum depth, while changing L2 or C2 only will vary the central frequency. To broaden the notch without affecting the cetral frequency, increase C2 by a factor and decrease L2 by the same factor. Narrowing of the notch can be done in the opposite direction.

The input impedance is realtively low (3.3k), so a buffered output is required, such as that of a non true bypass pedal or multieffect.

This is the monster:



Please report if you come to build it.

[stm]

Somebody stop me!

Thinking more about the frequency response, I'm pretty sure that with a decent audio analyzing software (which I don't have) you could get the actual frequency response of the HC by subtracting the average spectra obtained from the "with" and "without" audio samples. In order for this to work the two spectra to be subtracted must be logarithmic, i.e. expressed in dB, which is quite typical anyways.

Spooky!

[Paul Perry (Frostwave)]

Plotting those response graphs, one has to make fairly heroic assumptions about the source and load impedances. One of the reasons theory and practice so rarely converge. It's certainly true the source had better be very low, or there isn't going to be much left..

[stm]

You are absolutley right, impedance matching with passive filters is a serious issue.

In the case of the suggested circuit, I think it would be wise to spend a dual OpAmp to buffer both the input and output of the filter, thus making it independent of whatever you connect around.

[GFR]

Somebody stop me!

Thinking more about the frequency response, I'm pretty sure that with a decent audio analyzing software (which I don't have) you could get the actual frequency response of the HC by subtracting the average spectra obtained from the "with" and "without" audio samples. In order for this to work the two spectra to be subtracted must be logarithmic, i.e. expressed in dB, which is quite typical anyways.

Spooky!

You mean like:



thanx image shack

This is comparing clip HC1stRiff.mp3.

software used: Audacity (free) and OpenOffice Calc (free).

Note that as the input signal already has a treble roll off, the estimated transfer (in green) for high frequncies is just garbage, because I'm comparing nothing to filtered nothing

The "proper" way of doing such an estimate would be

1) feed the device with a source signal that contains energy in all the frequencies of interest. White noise is a good choice as we shall see.
2) record the input and the output (filtered).
3) calculate the corr(x,y) and the corr(x,x) - the correlation of input with output and the autocorrelation of the input.
4) calculate FFT(corr(x,y)) and FFT(corr(x,x)). If your input is really good white noise then FFT(corr(x,x))=1, with phase zero, you can use that to simplify the method.
5) the FFT of the transfer FFT(h) = FFT(corr(x,y))/FFT(corr(x,x)).
6) Do this a couple of times and average the results.
7) Using a good window (like Hanning) and lots of points in the FFT helps a lot.

edit: If you prefer just the FFT instead of correlation then FFT(h) = (FFT(y)*conjugate(FFT(x)))/(FFT(x)*conjugate(FFT(x))) and
(FFT(x)*conjugate(FFT(x))) = 1 for really good white noise.

This gives an estimation of both magnitude and phase, and works really well.

If you have a full duplex sound card and a software that can do the FFT (octave or matlab or even Audacity/Cooledit if you don't care about the phase) that's easy to do. I once did this to the BOSS SE-50 speaker emulator, maybe it's somewhere in the archives. Be careful so you don't clip anywhere.

[Ge_Whiz]

Note that the ROG boys have always had the cunning to place a simple two-stage filter at the output of their amp emulator circuits... 

[lovekraft0]

Plotting those response graphs, one has to make fairly heroic assumptions about the source and load impedances...

Well, the HC (according to the article linked above) is designed to be driven by the headphone output of the modeller, and its output requires a high impedance input, so those assumptions may not be unreasonable in this case.

At any rate, Radley is being hailed on the modeller forums as the bodhisattva of passive electronics, and any legitimate questions about how it works are being shouted down by the faithful, so unless somebody buys one and does a spectral analysis on it, I suspect we know as much as we're going to about it. Then again, it may simply be beyond our ability to understand as mere mortals:

The level of arrogance displayed by those demanding to know the exact principles of operation of my device (or any other device for that matter) defies description - along with the underlying assumption that one undoubtedly has the electronic/physics smarts required to comprehend it if shared - truly breathtaking.

Anybody besides me having a little trouble with that statement?

[stm]

Holy Sh!t

Good job GFR. You are right with respect to using a source that contains all the frequencies, however in this case since we don't have the actual unit you have to resort to what you have, which are the actual audio samples.  In fact, in room equalizing it is common to use a pink noise generator for these purposes.

Would you be so kind to send me the spreadsheet to stepper(at)ing(dot)puc(dot)cl ?  (in EXCEL or CVS format if possible)
Looking at the green curve I did the following adjustments to the windows media player 10 equalizer:

1) Start will all bands flat
2) Raise the 2 kHz band to +3 dB
3) Fully cut the 4, 8 and 16 kHz band (-14 dB)
4) Make sure the SRS and WOW effects are off, as well as any volume leveling option

I listened to the first half of the Chord of the Rythm AB Comparison file with the Eq off, and then engaged the equalizer for the second half...

Wow!  With my previous settings I said I was like 80% there.  Now I'm not less than 95% there.


To lovekraft0:

Based on the above experiments, I think the hype and mystery is mostly unveiled.

I'm not sure how to feel about Radley's statements; on one hand the guy did find an equalization curve that did a good job and then did a commercial implementation of it.  Maybe a sound engineer or an audio enthusiats would have come with a similar equalization curve by experimenting with a third octave equalizer.  Anyway, I give credit to Radley for being able to go from a concept to a (lucrative?) product, and as such I think he is defending his work with passion. Let's not forget that it is quite easy to come up with something reasonably good sounding, but turning it into a product that can be sold for good bucks is a different story.  I don't want to be polemic on these issues.

Now that a practical and reasonably close DIY implementation is a reality, I just don't know if it is ethic to go further, since I think his device has been reverse engineered here (in an oh so clever way, though!). One important part of the fun for me is in the challenge posed by the problem itself, and in the possibility of increasing my knowledge and understanding.  Nevertheless, don't take my mixed feelings too seriously. It's just I'm going through another depressive-reflexive episode

A great synergy has been produced here in this thread.

[lovekraft0]

 

Now that a practical and reasonably close DIY implementation is a reality, I just don't know if it is ethic to go further...

I tend to agree. And since I don't use digital modelling at all, it certainly doesn't matter what I think, and I have no wish to interfere with Mr. Hockensmith's ability to make a living, since it is obvious that a large amount of his effort has gone into producing a marketable device. His product, although expensive, seems to fill a need for some people, and I can't criticize that - his attitude, however (at least IMHO), needs a serious adjustment. After all, one can describe the process of compression, even in terms a total neophyte can understand, without publishing a schematic.

stm, just in case nobody else mentions it, your insights are truly amazing, at least to a slow happy boy like me - thanks for pushing the envelope!

[puretube]

 

[GFR]

stm I'll post the data to you.

BTW this device does take away the "fizz", and some of the brightness etc with it (of course). Listening to the samples I'm not that sure that it sounds "better" with it than without it. One thing I noticed is that when "Administrator" created a patch from scratch for use with the HC you can hear he boosted the high end a little to compensate...

[David]

Radley's response was based upon "reviews" that his device was given by the H-C crowd.  Apparently their opinions were ruthless, toothless and useless.

[MartyMart]

This is a very interesting thread, I've been away for a few days, so excuse the "late"
comments ...
I've been recording/mixing a lot of guitar based tracks in the last few years and I use
some "modelling" products along with amp/JMP-1 also, both hardware and software
within Logic Pro.
The JMP has a cab sim output based on the Palmer speaker sim ( so I read somewhere )
and this has a certain "tone" and "fizz" to it, Pod/Korg modellers do have that problem
and when I read stm's "eq" description, I nearly fell over !
After recording parts with the Pod/Korg, here's my "Eq" curve which gets used 90%
of the time :
Low roll off, of everything under 60 cycles ( if there's any rumble )
Flat 'til 500Hz with a 1-2 db boost
1-2k slight boost by 2 db
5k upwards rolling off, 24db/octave with nothing above 12k at all
Instant "warm" and "unfizzy" result
With the JMP-1 it's more of a simple 8-10k (and above) roll off, which seems to work fine.
My main problem with "modelling" recordings is that one or two "parts"
recorded this way is fine, any more and you can't get enough separation between
them. ( which explains why most guitarists think they sound fine - playing one patch! )
They "mush" together in a mix and just dont have any depth to them, while "real"
parts can be "sat" in the mix quite happily, can be heard and dont become so mushy.
Interesting stuff .. !

Marty.

[GFR]

I have a personal rule: when recording, I always use less distortion than I would use live (no matter if using a modeler or micing an amp). I think it helps the tone to "blend" with the rest of the mix and also makes it clearer. And when recording, you don't need so much distortion to cover up your errors - you already have protools for that

[lovekraft0]

Radley's response was based upon "reviews" that his device was given by the H-C crowd.  Apparently their opinions were ruthless, toothless and useless.

If you're talking about the quote I posted earlier, it's from a Line 6 forum thread, in response to a user who was trying to find out what the HC actually did before he shelled out his hard-earned cash sight unseen, and precedes the passage quoted in the first post on this thread (from the same Line 6 forum thread). If you want to wade through the whole thing, by all means, read on:
http://line6.com/cgi-bin/ultimatebb.cgi?ubb=get_topic&f=3&t=015568&p=
It really starts to get stupid about page 8, but there's some info amidst all the noise, if it's worth sifting through. Just another reminder of why I enjoy this place - it's 99.4% drama-free!

[jimmygtr]

I the price point is what is killing everyone.  Some posts have been extremely bitter, some are doubters  and others are curious, some excited that the unit exists, but the challenge of $250 is tough even those who really want it (me being one).  I'm confident it works just as it states, but the price point is just so hard for some of us.

Wow,
I really appreciate the input and time you've spent thinking this out. It would seem the circuit is pretty simple. The inventor epoxies the circuit once assembled to prevent anyone from taking a peek.

And as far as the fizz part, we know that the digital fizz is registering at around 5K to 5.3K. You can actually EQ it out. However, owners of the HC device say fizz removal is just one part of the device, they say the device significantly warms up the tones to make the digitial device truly sound like a warm tube amp.

Snake oil? Perhaps. But if so, there are a lot of pretty tone-savvy top-notch guitar players that swear by the device. I would buy it a New York minute if it were $100, but at $250 I just can't justify it, but would love to breadboard something together.

I believe he has incorporated some kind of tone stack with a presence boost. Perhaps diodes or something to smooth out the dynamics. Digitial produces a stairstepped output--if some how you could round off the edges of the steps perhaps it would sound smoother, more like a real tube amp.

As far as the Neil Diamond tone, not sure the inventor uses the HC live. Oh, by the way, he is an accomplished session player so evidently he is no slouch guitar player. And evidently has been modding amps and effects for many years. Detractors wish he would come up with a web site with clear sound clips that show the device's performance. Chances are Behringer will copy it and sell it for $19.95.

Jack Loganbill