A Surprising Problem

Recent work on tube sound emulation using solid state circuitry has brought an unexpected and surprising insight.
I have worked on a circuit emulating cross-over distortion (hereafter: CD) as it typically appears on push-pull, class-AB power amplifiers. Although this phenomenon is not limited to tubes only, the paper's claim was this this an integral constituent of "tube-tone". The paper itself is a (now expired) patent script. I have plans to describe the outcome in a later article.

I got this device working, and low and behold, it does what it should, and there is cross-over distortion galore.
I had the feeling that the effect was somehow overly bright sounding, although the distortion was low. A conventional treble filter did not help much. The feeling crept up that a higher order filter was, em..., in order.
How about a speaker emulator? But I did not care for the low end mangling and all that those things do.

I scanned the web for information. Bajaman's active version [Bajaman2008] of a unit called Harmonic Converger (hereafter: HC) is sort of one half of a full-blown speaker emulator - it only addresses the top end. The circuit is destined to kill fizz coming from a digital amplifier simulator. That sounded perfect for my purpose.

The small sub-circuit was quickly assembled and attached to the main board as an afterburner. I verified the function thereof with a signal generator and an oscilloscope, as usual. But then there came a big surprise.
I imposed a 1 kHz signal onto my overdrive plus filter - and there was no CD! I checked my circuit again, could see no problem.  Before the filter - effect is there, after the filter - gone. How can this be, so far below the 4 kHz cut-off?

Back To Index

---

A Surprising Answer

I was confused and poked around the web on all subjects regarding crossover-distortion.

I found Randall Aitken's article on CD [Aitken1999] and then it dawned on me: the filter removes all harmonics.

So at a test frequency of, say, 500 Hz, the effect would be visible, but from ca. 1500 Hz (given a 4kHz filter roll-off), the effect would diminish, leaving only the fundamental frequency back.

In retrospect, this is eminently logical, since harmonics are multiples of the fundamental, but it came unexpected.

I concluded that all of those observations and pictures of oscilloscope-traces on crossover-distortion (and other distortion shapes, as we will see...) are real and not real at the same time, because people oversee (like I did) the presence of a heavy forth-order filter - the loudspeaker. All those measurements are made electrically, such as by attaching an oscilloscope probe onto the output transformer, whereas the speaker is an electro-acoustic filter. Nobody makes such visual representations of oscilloscope traces using a microphone or a filter in line.

While an output transformer will have a restricted bandwidth, its impact is not as drastic as the speaker's.

Also, what all tube-sound emulations neglect, is the impact of negative feedback - which eliminates most of the effect again, probably dynamically variable too.

I wrote to Randall Aitken over the matter, and he confirmed my view. He also expressed unmistakably his aversion against CD.

So some steep low-pass filter will remove most of the unwanted frequencies.

A later project turned out to exhibit similar effects, although the distortion method was entirely different. It appears that both approaches are rather called dynamic impulse shaping than bland clipping. Both of them are meant to be inserted into the signal chain at a strategically carefully chosen place where there is some EQ before and / or after.
Both are Hi-Fi or full-range devices, that may on their own, without further treatment, sound unpleasant or at least unexpected.

►

A good example for the commercial application of such a device is the overdrive part of the Aion L5 preamp [AionL5], which in isolation sounds absolutely mediocre, but with enough EQ in front, allegedly sounds great. Incidentally, virtually all commercially sold distortion boxes are equipped with a minimal EQ at least.

The perceived brightness of such circuitry stems from the fact that they are hitting the front-end of an amplifier with an abundance of harmonics that would, were they generated inside the power stage itself (such as CD), never reach the ear because of being filtered out by the speaker. While the speaker swallows such harmonics at the place of their origin, an external device will dump all the harmonics into the input of an amplifier, where they get greatly magnified.

►	This fact surely contributes to the well-known observation that pre-amp distortion sounds  different to power-amp distortion.

A bare bone transit-shaping device therefore has to be judged (and eventually be EQ'd) always in a specific sonic context; a guitar setup may ask for a different treatment than one for an organ.
 
Back To Index

---

The Secret Weapon

Something like Bajaman's approach should help tremendously. I did not care for the presence part, which just bloats the circuit. Apart from that, distortion circuits are normally too bright anyway. Note that no amount of standard EQ-ing would work here, particularly not passive, cut-only filters. Those are all first order filters.

Without belittling Bajaman's work, it is only two cascaded second-order Sallen&Key low-pass filters, whose parameters can easily be changed by using an online calculator.

I had to do so, because the HC's corner frequency was placed a few hundred Hz too high for my taste, and also I aimed for slight peaking to achieve a more abrupt roll-off.

 

To download the schematic, click on the icon By doing so, you agree to the Terms of Use

Important Design Notes:

• This circuit snippet is not meant to be a stand-alone stomp box design. It lacks a suitable front-end at least.
  As is, it is rather meant to insert into the back of an existing circuit, ideally replacing the output capacitor in an existing build, and not e.g. behind a volume control.
  
• Select the proper powering scheme according to the board you want to augment.
  

The filter removes a lot of the brightness, while not noticeably (and objectionably) limiting the high range, because it targets the harmonics, while leaving the fundamentals intact.

Most of the current speaker-emulator designs and indeed Bajaman's obviously adhere to modern loudspeaker's curves, which exhibit a higher corner-frequency than was customary several decades ago. So I have been told by a veteran tinkerer.

►

Notice how much the corner frequencies vary on the different models I have listed, old style vs. modern style. Marshall amps are known for their trend from mellow to ever more ear piercing and brittle tones over the past decades, which reflects in the according speaker emulators resp. direct recording interfaces.

One needs to know that for designing a speaker chassis, parameters like

• sound pressure
• evenness in the pass band
• high roll-of frequency
  
• low roll-off frequency
  
• peaking at the roll-off
  

can basically be tweaked by the designer, but only few of them can be favored in one design. Favoring one parameter means making concessions somewhere else. It is obvious that loudness and HF range have been fostered recently to reflect the current taste.
The secret weapon's corner frequency that has been found to work best for bright distortion designs resides at 3.6 kHz.

Note that such a filter alone (much like the speaker itself) will not automatically make a distortion device sound good. In fact you may gain the impression that the filter does nothing much if you play quietly. The device it augments may still need some conventional (read: simple) EQ, but with the secret weapon in place, the former can bear fruits.

Back To Index

---

Reference