Copyright 2014-2022 by H. Gragger. All Rights Reserved. All information provided herein is destined for educational and D.I.Y. purposes only. Commercial re-sale, distribution or usage of artwork without explicit written permission of the author is strictly prohibited. The original units  with their associated  trade-names are subject to the copyright of the individual copyright owner. The Author is by no means affiliated with any of those companies. References to trade names are made for educational purposes only. By reading the information provided here you agree to the Terms of Use.

As usual, no circuit I clone passes my hands without modifications, usually straightforward basic engineering improvements.

"As one might expect, there is very little new in electronics, at least in analog circuitry. (...) Sometimes the innovators try to make their transfer of known electronics methodology from one discipline to this one seem revolutionary. The revolution is not actually the technology, but rather its revelation to the rest of the world." - Kevin O´Connor: The Ultimate Tone Vol 4, Power Press Publishing, Thunder Bay, CN, 2006, p. 4-15f

In direct relation to the Fetzer Valve, I changed the fixed value source resistors to trim pots with a fixed portion (series resistance) to accommodate a wide range of j-fets (I did, however stick with the J201 because all others I measured would have needed substantially higher quiescent currents - not a good thing for a device that is operated by battery. Since the J201´s did not exhibit unbearable noise and performed well, those stayed...

The drain pots were changed to 50k trim pots because it turned out that that the whole adjustment happened in the very low range of the pots.

All j-fets were measured with a jig according to the one suggested in the Fetzer Valve document. The appropriate values for source resistors were calculated using the calculator on the bottom of this document, although it turned out that this was less than ideal sonically.

It turned out that setting the drains to half-supply would not yield the best tone. A closer look with an oscilloscope and a frequency generator revealed that the resulting waveforms measured at the drains were nowhere near idling at half supply.

Consequently, the adjustment range for the drain pots was narrower than expected, in fact for one transistor the optimal working point window was so narrow that it started sputtering as soon as the supply voltage changed even a small amount - unacceptable for a battery driven device.

(This explains by the way why I never got my version of the Peppermill running satisfactory. It uses a fixed source resistor which certainly has too high a value).

I fixed the issue by setting the source resistors to a lower value. Then the drain pots would adjust a balanced half-supply waveform. I set both of them so that  the bottom half of the sinusoidal waveforms becomes more deformed upon excitation than the upper half, just as valve amps are said to behave, and that the idle voltage approaches about half supply. This sounds great. The source resistance thus may deviate substantially from the calculated amount, but this serves well as a starting point.

Back To Index

---

Further Deviations / Improvements

• The "bright" cap over the volume pot was wired to a front panel switch, akin to the "bright" switch on the real amps. This is a kind of funny feature because it only works for low gain settings. At higher settings, it becomes increasingly bypassed by the pot wiper.
  
• I installed a spst switch that bypasses the third stage for less gain akin to the "jazz/rock" switch. The gain is substantially lower when this stage is bypassed, which requires high gain pot settings. The before mentioned "bright" switch does not do much on those settings so tone is noticeably less bright. Distortion tone is somewhat different, but the difference is so subte, that one may decide to omit this control entirely without much loss.
  
• I modified the output filter stage that emulates a speaker roll-off for more highs because the whole circuit seemed to lack treble compared to similar circuits down to 10k vs. the original 15k.
  
• I made the interstage coupling caps the same size as the xUmble schematic I found on the internet (i.e. bigger).
  
• Supply voltage and working point: The biggest improvement was a voltage doubler with regulation. Initially some of the stages appeared very critical to working point adjustments, as mentioned earlier. As soon as the working point was changed even slightly, the circuit would cross from "very transparent and touch sensitive" to "farting". This means that it could not reliably be used with a battery that was subject to aging despite its small current demand. I tried a low power switching microregulator from Linear Technology (LT1615) which worked apart from producing horrendous switching noise. Using ye olde MAX voltage doubler (MAX1044CPA = ICL7660SCPA) seemed counterintuitive at first glance because this circuit´s output voltage is very load and supply dependent. It was therefore unusable too despite the fact that the Umble would work better at elevated supply voltages.
  
  Madbean provides a voltage doubler with a subsequent voltage regulator (road rage) which would do the job basically. Unfortunately he employs a stock 78xx series regulator which wants to see some 3 Volts difference between input and output for regulation (a.k.a. dropout voltage). If I wanted to have some safety margin for an aging battery (say, 1 Volt battery drop) on top of its dropout voltage, I would have to throttle the output voltage back to a voltage that was barely higher than the one I started out with (11V), thus not really paying the extra effort.
  

• However, a LDR (low dropout regulator) LT1121CN following the MAX1044 does the job. Those only need 0.5 V overhead voltage at a comparably neglectable quiescent current which leaves me at about 15 Volts regulated. The battery voltage would need to go below 8.5V to go out of regulation. For anything below, the output voltage would just start to drop at the rate the battery drops. At some point the umble would stop working, because the working points shift out of a usable region for some of the transistors at least. This depends on the j-fet types chosen and the specimen selected.
  
  The charge pump and regulator circuit themselves consume very little power. I shielded the whole power conditioning circuit in aluminium foil which shorts all radiated voltage to ground. Also, the local Umble power supply decoupling was chosen probably overkill (LRC filter with small coil and small series resistor together with an electrolytic and a small value  ceramic cap to ground). Welcome tube schematic ;-)
  

• The xUmble schematic found on the web (which the Umble circuit seemingly is crafted after) shows some interstage supply decoupling components (R/C) that may introduce some supply voltage shifts during sag, although sag is caused by the power stage. However, I initially installed some R/C time constants which were supposed to introduce some sag under load. This turned out a dead end (read more here: Intuitive Sag Circuit For The Umble) .
  

• On the oscilloscope I noticed spurious HF ringing on the later stages upon positive gate voltage so I installed 10pF ceramic caps between gate and source. This seemed to help against those effects without affecting sonic response. Those effects may have been not real, because I was measuring in an potentially electrically polluted environment so it is hard to tell.
  

Back To Index

---

More Tube Tone - Power Stage Sag

The Umble circuit does its job well, just like many similar overdrives surely do, but it only emulates the preamp portion of a tube amp. There are many other effects a tube (power) amp contributes to what is known as "tube tone", such as loose coupling of the driver compared to the stiff coupling of a transistorized amp. This cannot easily be simulated in the signal domain and I have never seen an attempt to do so yet.

Apart from graceful signal shaping, one of the further virtues  of a tube amp  is signal compression, in this context better known as sag. Linear power supplies consist of a transformer, some sort of rectifier and capacitance. Dependent on the implementation, this chain also contains a certain amount of resistance (deliberate or inherent), which, under load, leads to a drop in the power stage´s supply voltage.

A sudden  increase of the input signal therefore yields a current surge in the output stage which in turn drops the supply rail somewhat. Fast power peaks therefore get compressed to an extent. This works like a compressor with fixed attack and (mostly also) decay time. The effective resistance together with the filter capacitors determines its time constant. Since this seems an important part of tube tone (which we are after right from the beginning...), let´s look at what compression, or, in this case, sag, does to signals.

Back To Index

---

What Compression Does To Tone

Signal compressors in general, as they are used in recording studios, serve (amongst others) the following purposes:

Summary of compressor applications: accentuate the inner details of sound make sounds bigger, denser, fatter add warmth make louder change signal dynamics reshape dynamic envelopes (add punch, modify attack, modify transients, modify decay) and more                                                                                                                                                        - excerpts from Roey Izhaki: Mixing Audio, Concepts, Practises and Tools, Focal Press 2012, 2nd ed., p. 308

In short, they change the shape of the signal by potentially modifying its initial attack, its decay and its (perceived) overall volume. What is hidden within those actions is that by changing the signal´s waveshape you change its frequency spectrum.

This appears important, so let´s look at that more closely.

Signal envelope: "When we hit the string of a guitar, the string suddenly begins to vibrate at the fundamental frequency and at all of the harmonics of that note. Each frequency dies out in amplitude at its own rate, but because they all started at the same time, the initial energy release is very high. This is the transient of the envelope, or the leading edge. (...) the transient is about twice the amplitude of the sustained portion of the envelope. This relative height gives the note more or less attack. (...) Attack is perceived as the suddenness with which the note has appeared out of silence. A short or quick attack gives no warning of the note, (...). A long attack means the note rises slowly out of silence, building tension by anticipation rather than surprise." - Kevin O´Connor: The Ultimate Tone Vol. 4, Power Press Publishing, Thunder Bay, CN, 2006, p. 4-3

This shows us how important the shape of the first transient is and how much sag type (or any) compression may change the envelope.

How tone changes with a modified signal envelope: "Because the transient is taller than the rest of the note envelope, it will be the first portion of the signal to be clipped when the amp runs out of power. (...) As we nudge the signal level ´up´ we hear the attack change. (...) As we slowly dial the level upward, (...). The sound will be distinctly compressed and might seem to have lost some treble. (...) If the transient is missing, it must mean that those high frequencies are lost. Pushing the amp a little harder, the output stage clips the sustained portion of the note and the frequency response changes to that of a pure square-wave. (...) the harmonic balance is now skewed from the timbre of the clean instrument sound to that of a square wave.. This makes the sound a bit brighter than when it was compressed, but in a different way then when it was clean." - Kevin O´Connor: The Ultimate Tone Vol. 4, Power Press Publishing, Thunder Bay, CN, 2006, p. 4-3f

Compression by the way also makes the signal subjectively louder (since the decay portion of the signal appears more prominent), which, together with the graceful signal limiting of a tube stage, creates the impression that a tube amp can be played louder that a comparable solid-state amp (which may produce bad sounding distortion artifacts at the same power level...). From here stems the saying that tube amp watts are  better than transistor watts...

The exact mechanism of tube power amp sag and its time constants are well known and simple, but not easily to put into numbers. Also, this effect is distinctly power amp related resp. power stage supply current related, and can thus not easily be transferred to a preamp stage, be it tubes or solid state equivalents. However, this is not important if we want to emulate its effect.

Emulating SAG in the signal domain (as opposed to the power domain): "It is important and very useful to us that supply sag has the sonic attributes of compression.(...) Later, we will look at universally applicable methods that sound exactly the same [as power supply sag], but work in the signal domain exclusively." - Kevin O´Connor: The Ultimate Tone Vol. 4, Power Press Publishing, Thunder Bay, CN, 2006, p. 4-4f

Back To Index

---

Intuitive Sag Circuit For The Umble

The preamp stages are usually daisy-chained to the power amp supply and will drop too, but since we are talking of power amp compression and we assume the preamp runs at fairly low gain, this has not much effect on the signals passing the preamp stages.

If, in contrary, a tube amp uses the input stages as signal shaping devices, as is done on master-volume equipped amps, the power stage may run low power  and not exhibit any noteworthy sag.

So, although a tube preamp distortion mechanism does depend on supply voltage (like all signal bounding circuits work against a boundary),  one may not rely on the sag of the preamp for the effect.

So intuitively, one may be tempted to use an R/C filter chain, such as in the original tube amp schematic which will cause a voltage variation with varying signal amplitude.

However, in real life it is not the current drawn by the preamp tubes that cause a variation in their anode voltage (the j-fet drains in our case), but rather the the current drawn by the power stage. Preamp stages are biased about mid-supply, kind of class-a.

What happens if a healthy signal comes along? The voltage on the drains rises!. How come? A class-A circuit always uses most power when it is idle. When the signal swings rail-to-rail (particularly if it clips) the device is off half of the time -> the current drops and the drain voltage rises.

This circuit therefore is a clear failure. And I ran into it ;-)

Another reason for leaving the supply alone is the fact that our may j-fets shift out of bias. We don´t have the luxury of hundreds of volts of supply. So even more elaborate circuits like  the punisher will fail.
But the caps stay there with small resistors (56Ohms) - makes for a well filtered supply.

Back To Index

---

A Sonic Analogy To Power Stage Sag


So fiddling with the preamp´s supply is not working. But as mentioned above, the net sonic impact of power stage sag is compression, which can well be simulated in a preamp stage.

As mentioned above,  let´s apply known methodology to our problem ;-)

Although O´Connor does not go much into detail on preamp stages, he does apply it to a single ended power stage, which has much in common with a preamp stage for our purposes.

SE sag effects: "We mentioned that stock single-ended (SE) amps draw fairly constant current, and thus present little if any load variat6ion to the supply. (...)The same is true for high bias amps like the AC-30,(...) Although a fullwave -rectified signal guarantees symmetric control of the audios signal, it is not absolutely necessary for good results." - Kevin O´Connor: The Ultimate Tone Vol 4, Power Press Publishing, Thunder Bay, CN, 2006, p. 4-21f

Despite the fact that O´Connor strictly speaks of tube (preamp) stages in this context, we can safely apply this knowledge to our j-fet stages which are there to emulate tube preamp stages in the first place.

This circuit is an implemen- tation of a tube compressor circuit outlined by Kevin O´Connor: The Ultimate Tone Vol 4, Power Press Publishing, Thunder Bay, CN, 2006, p. 4-22 Fig. 1: Adding a compressor to the Fetzer Valve stage. (click to enlarge) The added complexity is astonishingly minimal despite its effectiveness. A control signal is derived from the output of the respective j-fet stage. The capacitor removes any DC content. A limiting resistor (which provides a maximum effect level...) feeds a pot that sets the amount of control voltage. A diode extracts the negative portion thereof (referenced to ground). In order to not waste any voltage in this already limited environment, a shottky diode is used. A subsequent R/C filter yields a voltage that has most of its high frequent content filtered out. Time constant will be about 100ms. (Note the original circuit outline in the book contains an error) Any voltage present at this point will pull the j-fet gate more negative. Upon zero control voltage, the gate will have ground level just as it would have for the unaltered Fetzer Valve stage. In order to not skew the balance of interstage signal attenuation in the overall circuit the upper half of the gate bleeder resistor is kept the same value as for the unaltered stage (in this case 100k) and the lower half the same value. Although the steady state impedance is doubled by this measure, the dynamic impedance remains unaltered due to the capacitor. The compressor´s function is readily observable on an oscilloscope.

Installing the compressor in an early stage does not make sense, because we are not dealing with dozens of volts here like in a real tube stage. After the input stage the signal is small, not reaching clipping. After the tone stack it is even smaller and has to be recovered by the following stage. In order to have a useful control voltage range, the compression has to happen in one of the later stages.

I chose to use stage 3. The output voltage (at least with 15.5V supply) is  big enough to achieve a useful compression effect.
Compression can be dialed in gradually at taste. At full effect there is a noticeable drop in the attack and overall volume, since all one hears is the sustained part of the signal. This behaves like all compressors do and a little make-up gain fixes this. So overall, the system sounds louder, the sustain appears increased because the signal does subjectively take longer to die out. Also, since the shape of the signal has changed (some fast components have been removed from the spectrum) the signal appear less bright.

Since those portions of the signal have been reduced, that would otherwise been clipped, the signal also appears cleaner despite longer sustain. This virtue separates this distortion mechanism from pure signal clippers.

This is also the reason for placing the compressor in stage 3, because if it was installed into stage 4, we would probably have already much more clipped signal peaks and a different tonal spectrum.

Back To Index

---

How Does This Sound?

The umble works very well as described. I don´t know it it resembles a real xUmble amp, but I have heard people play over such a device and I noticed that it was very dynamic. However, the biggest tonal difference obviously stemmed from their choice of pickups during their playing and another setup would probably have yielded similar results.

There were some indications that the J201 transistors were noisy and indeed the circuit is not the quietest I have heard, but on a setting near unity overall gain the circuit is acceptably quiet. Maybe I try a low noise transistor in the first two stages.

I also compared it to my King Of Tone clone and in the basic overdrive setting it sounded so much similar, that I often had too look which one was on. The differences in voicing are subtle. If I were to recommend somebody one of these two stompboxes (for D.I.Y), I would probably recommend the King of Tone just for its sheer versatility. One half of the circuit (I think Madbean made a board for that) would be sufficient. Yet if the compression part proves any useful, this may shed a different light on the Umble again.

That said, and having my box sitting on the floor in a pretty case reminiscent of a valve head, I may just as well leave it on permanently as a very transparent overdrive. Don´t expect shredding metal tone, but you don´t expect that from a real xUmble head, do yo?

The funniest thing is, that when you turn the device on for the first time, it takes a few seconds until all j-fet´s bias correctly obviously, a behavior that is definitely tube-amp like.

When dialing the compression in, output treble content noticeably decreases, but the signal takes on a more "fluid" and softer touch.

Initially one may be mesmerized by this tone, but later on the wish appears to dial in a somewhat more conservative setting.
I like the Umble on a light overdrive setting with sag dialled in for playing slide, because it takes away some of the sharpness the brass slide introduces. Maybe this is the reason some big shots in slide playing use a xUmble.

Although the sound samples expose this effect less than expected, it is there. Leaving this modification out would be a shame because it is so easy to incorporate and it fits the concept of emulating a tube circuit with j-fets seamlessly.

In the following sound samples an attempt has been made to set clean and distorted sounds to equal perceived loudness. However, because of the spectral shifts (creation of artificial high frequencies) and the compression this is impossible. What appears equally loud on, say, chord work, appears different on solo work and even more different on the meter. When you adjust the output level so as to make passages with chunky rhythm chords appear equally loud, lead passages may appear too loud and vice versa. On the  final recordings the difference between sag and not sag may not appear dramatic, it is however present and even visible on the recorded waveform. Unfortunately canned recordings can never replace the real thing but it is hoped that they may prove useful.

The stomp boxes are set to a subjectively comparable distortion texture, loudness and voicing. Once set, none of the controls have been changed except for the chord work, where the guitar´s volume has been reduced slightly to prevent overpowering the distortion boxes. Note the subjective loudness differences in the sound samples despite this measure.

It is known that a certain distortion box or amplifier fosters a certain playing style since the player and the rig´s tone tend to become a unit. To eliminate the possibility of playing the riffs differently each time the samples have been recorded into a hardware looper and are played back from there instead of the guitar.

The subsequent recordings have been done using the following setup (in this order):

• Strat modified according to specs, neck pickup and Les Paul copy w/ single coil neck pickup for slide work
  
• J-FET buffer directly after the guitar
• Boss RC-50 looper for creating clean phrases. All recordings are a playback from the looper and thus identical in attack and touch, regardless of the subsequent dirt box.
  
• homebrew dirt boxes (Umble, Tone Queen) as indicated
  
• Lexicon Room reverb (Digitech RP-500)
• MOSFET amplifier through 10" detuned cabinet w/ built-in omni microphone (phantom powered)
  
• Recording device: Mixcraft 6 with Focusrite Saffire PRO 24 DSP front-end