An Unconventional Approach To Distortion
Circuit Analysis
The VALVER
Hands-on Experience
Reference

An Unconventional Approach To Distortion

VALVER: a work-a-like of The Valve Driver by Paul Williams (click on the pictures to load a larger image)  This pedal uses a very clever and unconventional method to achieve signal distortion by shaping the transients rather than clipping them. It allows for continuous panning between even and odd harmonics and can thus mimic a valve amp's pre-amp (top trace) or power amp distortion (bottom trace) or any mix thereof. The intensity of the effect can be varied with the "intensity" control, without affecting loudness. The modernized version uses a VCA rather than an OTA, which makes it virtually noise-free. As different to common distortion pedals, the VALVER is indiscriminately with regard to frequency. In other words: it is not overloaded by too much bass.

The quest to find the ultimate solid state device that emulates valve tone (a.k.a. tube-tone) never ends.

Having read in Nick Boscorelli’s book [BOSCORELLI-1] about the benign clipping characteristics of OTA’s (operational transconductance amplifier), I was curious to hear what this was about. Having built such a unit by extracting the distortion generator out of the schematic of a the LAB series of amplifiers [AIONfx] with only mediocre sonic results, I was looking for more information on the subject. I stumbled over The Valve Driver by Paul Williams, which appeared in an English magazine for music makers way back in 1983 [WILLIAMS], but I had filed the subject out of frustration.

This, however, does not achieve distortion the conventional way. Williams’ design is truly remarkable, because it is one of the very, very scarce approaches to achieve distortion not by inflating a signal until it hits a boundary, as most other designs do (and by generating lots of noise alongside), but rather by putting gain into the control voltage of a gain control element.

►	One of these approaches is known as the Baldringer Drive [BALDRINGER], although Baldringer uses a very different actual control element.

Seeing the presence of a potentially noisy OTA, I did not bother building the original for reasons that will soon become clear, and I set out for a contemporary version right away. But before, let us try to understand what he did.

Back To Index

---

Circuit Analysis

In order to be able to copy the original unit’s behavior, we first must understand it. May I suggest to have Williams' schematic handy [WILLIAMS] for reference. Designators refer to the latter.

►

One word ahead; the whole circuit obviously has been designed with a minimal parts count ideology in mind. This inevitably leads to compromises. We will stumble over this ideology several times through the course. What we essentially see here is a compressor circuit with a side chain and a gain control element, but it predominately differs from such by being instant in action, meaning there is no integration time constants like attack and release involved. A stock compressor rectifies and averages the signal and is phase insensitive, whereas the Valve Driver uses the instantaneous signal value and is phase sensitive.

---

Description and terminology, children of time

The magazine’s description of the circuit’s function is potentially misleading. It describes the distortion phenomenon from a perspective of the input signal versus the output (since that is inverting), and not, as one would probably expect, from the input of the side chain to the output. When you bear that in mind, the description becomes more comprehensible.

Consider that the input amp inverts and the OTA inverts too. To achieve symmetric squashing you need an in-phase control signal (the precision rectifier inverts the inverted input signal, RV1 set CCW, for an in-phase relationship).

Williams uses the rail names -V, GND and +V,  which together with the unconventional battery symbol (misleading to think it was two battery units in series, while it supposedly suggests several 1.5 V cells in series) lead many readers astray by making them believe that he used a symmetric supply. One commentator even feared that this could lead to problems when interconnecting with other stomp boxes.

This is a wrong assumption caused by Williams’ awkward nomenclature and use of symbols and the fact that we are currently associating such nomenclature with a true symmetric power system.

So for understanding, replace GND by Vref and V- by GND and all falls in place and becomes familiar again.
It is just a single battery stomp box with an artificial center voltage generated by two resistors. Granted, back in 1983 there probably was no such standards established yet, let alone the non-existence of a semi-knowledgeable community.

---

OTA as a linear VCA

Williams used an LM13600 transconductance amplifier, which I suppose was en vogue at this time (and probably the only discrete part that could be used as an VCA, and Williams’ OTA is just used as a such, at least the main one). Synth heads subsequently jumped onto the band wagon for all sorts of voltage controllable filters at the time.

Unfortunately, all  (cheap) OTAs were all ridden by a major flaw, namely by their inability to withstand large differential signal voltages (we speak of 30mV or so) on their inputs without significant distortion. As I said, Williams’ approach was not after blunt distortion, so that was to be avoided. In order to digest those signals, huge attenuation had to be applied beforehand, and accordingly, a complementary boost afterwards. But this trick does not come without a severe noise penalty. OTAs (particularly the cheap ones like CA3080 and LM13x00) have a bad reputation for that.

►	Later OTAs were superior (like the CA3280), and they are still available second source, but compared to current VCAs, they all fall short by a mile regarding ease of use and performance for voltage controlled amplification at least.  So here we go.

---

Input stage

The accompanying text speaks of input amplification, but the gain is only one, with R3 making only a fractional difference.

"(…) SW1a allows the amplified input signal (…)"    - see [WILLIAMS]

This is ok, since we don´t need a large signal for the distortion mechanism, and we need to knock down the signal by R6-R7 anyways to make it palatable for the OTA.

The diode bias on pin 2 relaxes that situation only slightly. The size of R16 is directly related to output volume (set for unity), but sadly also to the noise level. The configuration with this resistor to Vref (GND as he calls it) is inferior because it adds distortion (again of a type which is unwanted in this case). Also, the darlington wired as buffer is not as good as as real buffer.

The input amp is configured as inverting. The serial impedances are pretty high, so high noise  is programmed. The input could have been configured differently, but not with the minimal parts policy in mind.

---

Precision rectifier

Ic1b receives a  fraction of the input signal. It is a variant of a full-wave rectifier [MICROCHIP], but the diode is inserted in a way that it favors the negative half-wave (although both could have been used with equal result).

The rectifier circuit needs a load resistor on its output. What is called RL in the app-note, is pot RV2 here. RV1 is a specialty, it is wired as variable resistor, which continuously lowers its resistance (CCW) in parallel to the diode to a point, where the diode is effectively shorted out eventually.

In this position the OPA works as a plain inverter. In the CW position, the parallel resistance is more or less out and we have a full wave rectifier with negative polarity favored.

This very clever little circuit allows Williams to fold the negative half-wave up, with a degree continuously variable from zero (full wave) to max (full wave rectified). That said, this circuit, as clever as it is, only works so-so due to the minimalist approach, particularly in the in-between positions. But maybe this is good enough.

---

Multiplier

Rv2 (intensity) takes a variable fraction of that voltage, which basically could already serve as a control signal for the variable gain cell directly. But although the OTA runs as a VCA, its control is linear, meaning the relation between control voltage (CV) (actually the current thereof) and gain is a linear one.

Since our loudness perception happens exponentially, this is not a good idea, and many times some conversion to a log law is being called for.

Williams cleverly chose to use the hitherto unused second part of the OTA to yield a CV law that caters for that. The OTA is configured as a one-quadrant multiplier [ANALOGDEVICES], where the signal is multiplied by itself – i.e. a squaring function. While this strictly is no log converter, it seems to serve the purpose.

---

Bypass

Bypass has been realized by just removing the control signal. This has the advantage of being virtually pop-free, but has the huge disadvantage  of permanently having the noisy OTA in line.

►	With our revised circuit, noise is not of concern. One might well consider having an active bypass over a true bypass.

---

Summary

As mentioned earlier, I did not build the circuit, but I made a computer simulation. The overall performance is as follows:

    • expected noise: pretty high
    • impulse shaping works as explained (reminds of DAW plug-ins that call themselves "transit shapers").
    • overall gain: not as constant as proposed (volume control needed)
    • no full bypass – the OTA is always active

Using a contemporary VCA, a better input stage and a slightly more elaborate side-chain we can improve most of that dramatically.

So this understanding finally peaks in:

Back To Index

---

The VALVER

VALVER is an acronym composed of the syllables VAL(ve) and (dri)VER.

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

---

Overall topology

The input is a variation of the trusted Boscorelli input stage [BOSCORELLI-2]. Although the choice of some components contradicts current customs on a casual look, every component is very carefully chosen. I suggest, before you object on anything, to read the original design notes.

If you feel the bass range needs tailoring, resist the temptation to change the input capacitor. Change the ground leg cap in the feedback path.

The VCA (THAT 4305), was a leftover of a previous project. It is set up according to the datasheet [THAT], with a quiescent gain of = +1. An output volume control lets us dial in any desired volume. The VCA operates into a buffered I-U converter that presents a constant voltage to the VCA’s current output, which minimizes distortion [BOSCORELLI-3] .

►	The THAT4305 comes in an SMD package. I bought SMD to QSOP-16 to DIL-16 adapters.

There is no (noteworthy) gain in the whole unit, except for a small overall working gain, which is counteracted on the output for some quasi companding for the sake of noise performance.

---

 Supply

The unit thus could theoretically be operated using guitar levels and thus easily off a 9V supply without further ado. However, it is much preferable to work with somewhat elevated voltages, so a small working gain (of 6) has been used with hitherto good results. Even using a high output hum-bucker did not produce any unwanted clipping.

All signals are DC-coupled and referenced to Vref (½ supply). An active split rail has been chosen to prevent corrosion of Vref.

---

Side chain

►	As mentioned earlier, the whole circuit strongly reminds of a compressor. Consequently, we will use the terminology customary with compressors - such as the term side chain for the part of the circuitry that creates the control signal for the variable control element.

The original side chain circuitry could have been used, but the design is so minimalist to the point of being crude, that it was inflexible. The present, slightly more elaborate design, isolates all inter dependencies, which makes the parameters very easy to be tweaked. No compromises have been made, so expect to see buffers galore.

An attenuator divides the signal down by an empirically determined factor. For very hot guitars the jumper might be removed (more attenuation). This attenuator helps to keep the signal in a range where the intensity pot can maintain a useful travel, without hogging its functionality to one side. The final attenuator then can stay at a value save for the VCA.
A negative lobe extractor (akin to the original, although technically any lobe could have been favored…), as it says, isolates the negative going signal components. As in the original, a “harmonics” pot allows to dial in the full amount of negative going signals down to zero.

►	Consider the ground leg series resistor for the moment as zero. We will attend to that later… The chart assumes that resistor being zero for clarity. Note that the conversion of the screen shot to a jpeg obviously dropped some fine lines on the chart.

The following chart hopefully helps to clarify what is going on.  I suggest to download it first to have it handy for reference.

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

After buffering, the signal is inverted.
A recombination stage comprised of a  regular inverter sums up the signals.
Now this is where the clever part is. The side chain signal (an attenuated copy of the input signal), if it were used on its own to control the VCA, would boost one polarity and compress the other one (basically column C in the chart). By adding a flipped-up copy to the negative lobe with a magnitude continuously variable from x0 to x2, we can achieve either a mixed scenario (dual = compress & expand, column C), an asymmetrically compressed scenario (one half less compressed to uncompressed, column B) or a fully symmetrically compressed scenario (column A).
Of course all sorts of in-between shades are possible.

This is only possible courtesy of the 1:2 weighting of the signals at the recombination stage, which you may view as a subtractor. The intensity control is wired across the inverter and again allows for adjusting the effect depth. It is interesting, that sweeping this control does not change loudness noteworthy, although it compresses signals increasingly.
In an earlier design stage I had used lobe extractors for both lobes and recombined them in a similar manner, but this does not allow for boost. I sounds plausible that this path was chosen by Williams to achieve a higher signal magnitude for compressed signals in order to maintain a perceived overall loudness of unity, which only works properly for some settings.

The minimalist (however elegant) original extractor comes with a load of compromises, so maybe this worked so-so, but it does not on the given design. Apart from that, a unit like that (emulating tube amp tone) may well be the “low man on the totem pole”, meaning the final tone shaping effect, where loudness variation is undesired. So a volume control comes in handy.

A final resistive attenuator, whose components have been fine-tuned empirically (in unison with the other attenuators), makes the signal palatable for the VCA. Note that in this case there must NOT be any large CV filtering capacitor. A 1-1.5nF capacitor limits the bandwidth of the CV to frequencies that are out of interest. Like all of THAT’s VCAs, THAT4305 demands to be directly driven by a buffer.

Too much boost in dual mode compression is counterproductive and sounds unbalanced. To my ears, this creates a sonic imbalance and does not sound good. I therefore decided to add a “stop” resistor at the ground side of the harmonics pot, so that only a certain amount of expansion can be chosen. A resistor as large as the harmonics pot (10k) would defeat the expansion function, but an 8.2k resistor allows for a small boost in the sense of Paul Williams. If you encounter clipping in this position, make it 10k. I didn't even with very powerful pickups.

A bipolar control signal into a VCA is equivalent to compression resp. expansion. Since both can appear, this is a variation of a dual mode compressor. VCAs are exceptionally low noise for compression, but with heavy expansion (20dB) they become noisy. We are far away from that region, besides the fact that the compressor action is transitory.

At the end of this side chain processing, compression mode is linked to a negative polarity CV signal, while expansion is linked to positive polarity. This meaning is arbitrary and could have easily be reversed using an inverter, but the VCA used can work with both senses anyway. Either can be used with no difference.  Akin to inputting a signal to the + and – input of a regular OPA, the VCA’s + and – control inputs achieve an effect either with non-inverting sense or inverting.

The output of the VCA including the inverter is in phase with the input regardless.
So the + control input has been used, just in case the kind reader is wondering, because most compressor designs use the – input. Probably just out of tradition.

The VCA’s control input is logarithmic by definition, so no further measures are needed.
The unused RMS processor has been tied to mid-rail similar to the treatment you would apply to an unused OPA part. However, the input is current sensitive, so a resistor has been used. This has been suggested by one of THAT.co’s design engineers in an e-mail communication.

Back To Index

---

Hands-on Experience

As usual, tube tone is a tall order and hard to fulfill. This partially stems from the fact, that there is no exact definition of what tube tone exactly is – except being a mystery. Note that this unit does, unlike others, not contain any further equalization, it is thus full range. An EQ before and / or afterwards can be necessary. Due to the enormous speed of the VCA, it treats all frequencies the same. Unlike the vast majority of other dirt boxes, it is not overloaded by bass frequencies and consequently does not distort bass more. There is no need to restrict the bass range for a more balanced overdrive tone. Signal peaks and large bass signals are just squashed more:

Handling of large signals: (click on the pictures to load a larger image)  With increasing signal amplitude, the VALVER starts to level out straight, and then fold back. Note that the wave form lacks fast signal components, which a conventional clipper would no doubt generate (straight horizontal line). All parts of the signal are still rounded. Another nod at the behavior of tubes. Eventually, the bottom lobe (even harmonics) is being affected too.

Little experience exists with purely even harmonics setting (harmonics pot full CCW), since this sounds unnatural anyway. A center setting ( with the dropper resistor in place) sounds best.

In addition to the described unit, there is particularly one (mostly overseen) measure I can recommend here, which is using a steep 4th order HF roll-off at about 4kHz afterwards (or thereabouts), such as a speaker simulator has it.

Incidentally, this is not the first time I happened to need such a device, so I dedicated a separate article to it [AQUATAUR]. This can be made part of the circuit, or being added as an "afterburner".

A setting of the harmonics pot towards slight asymmetry (about center with the stopper resistor in place) sounds very organic. On the ‘scope that looks like a typical non-symmetric preamp valve distortion (see oscilloscope trace at the top).

This unit is in my humble opinion destined for low overdrive. It works well for stacking, even in an early position, because it has no massive gain and thus is virtually noise-free.  Heavy peaks are squashed in a benign way and, as different to many other dirt-boxes, does not produce unpleasant artifacts. Its immunity to powerful bass frequencies may make it particularly useful for devices like organs or bass guitar.

By turning the "intensity" pot up, the signal is fattened without frequently crossing into heavy distortion on peaks, and without major change in loudness. For that reason, Williams' original design did not have a volume control, but I considered this useful.

Recently I found another superior use for it. I was playing some slide guitar. I found that its tone was a bit thin and reached out for an overdrive as usual, but despite the abundance of such devices in my repository, none of them augmented the slide guitar’s tone. Most were too harsh and were overwhelmed by the bass content, even so-called low gain overdrives. They also could not cope with the powerful transits the metal slide produces.

My little Roland Cube-60 which I use as practice amp, has an “overdrive” input with a j-fet input stage. This sounded the best yet, but wasn't very transit-safe, so I thought up ways to mimic its (alleged) preamp-tube-like non-symmetrical distortion and lo and behold! The valve driver does exactly that! It indiscriminately digests all frequencies, is transit-safe and sounds fantastic for slide guitar.
A friend expressed interested for his Hammond Organ...

Back To Index

---

Reference

[AIONfx]
Building information on the Lab Series® L5 guitar amplifier, https://aionfx.com/project/l5-preamp/

[BOSCORELLI-1]
Nicholas Boscorelli, The Stomp Box Cookbook, 2nd ed., 
https://pdfcoffee.com/qdownload/the-stompbox-cookbook-nicholas-boscorelli-1999pdf-pdf-free.html

[WILLIAMS]
Paul Williams, Electronics & Music Maker (E&MM) December 1983, p.92f,
https://www.muzines.co.uk/articles/the-valve-driver/6239

[BALDRINGER]
Patent WO 2007/138105 A1, Dirk Baldringer, Circuit Assembly For Distorting An Input Signal,
https://patentscope.wipo.int/search/en/detail.jsf?docId=WO2007138105

[MICROCHIP]
AN1353, Microchip. Inc., Op Amp Rectifiers, Peak Detectors and Clamps, p.6, Fig. 17,
https://ww1.microchip.com/downloads/en/appnotes/01353a.pdf

[ANALOGDEVICES]
MT-079, Analog Devices, Analog Multipliers,  https://www.analog.com/media/en/training-seminars/tutorials/MT-079.pdf

[BOSCORELLI-2]
as above, pp.112ff,

[BOSCORELLI-3]
as above, p.220, fig. A139,

[THAT]
THAT4305 datasheet; https://www.thatcorp.com/datashts/THAT_4305_Datasheet.pdf

[AQUATAUR]
H.G. (aquataur) 2026, The Secret Weapon Against Fizz, http://me.aquataur.guru/musicstuff/secret_weapon.html