more design work on the reverb system

Here's the latest iteration of the front panel layout:

(That big blank panel on the bottom row, is for the "McGrath" logo.  I think this will be done with masking and black paint, so that the background will be black and the letters will be bare aluminum showing through.  Of course, for users of the amp, or for those who copy my design and layouts, this logo area also represents six or seven knobs-worth of extra panel space, for added circuitry.)

You can see the new four-knob layout for the reverb section (source, tone, mix L & R).

I've been planning out the circuitry to go behind those knobs, in more detail.  Looks like the system will need three 12AX7 tubes, not two as I originally hoped.  Two triode sections (one tube) will be needed for the reverb driver, in order to support the active tone/contour control.  The driver circuit will drive a single transformer, connected to both reverb tank inputs; I believe I will provide four RCA jacks from the transformer output, two in parallel and two in series, so that more experimentation with different reverb tanks is facilitated.  The reverb returns (L & R) will be amplified by non-feedback inverting gain stages (two triodes), and these stages drive the "wet" sections of the dual-gang pan ("mix") pots; the "dry" sections are fed directly from the original input signal (i.e., the output of the eq2 bypass relay), with no buffering.  The wipers of the pot sections are summed together through resistors, into the virtual ground of negative-feedback triode stages (two more, total six -- i.e., three tubes).

Because of the six-position source selector switch, the bypass wiring gets a little complicated.  Instead of one DPDT bypass relay like most of the subsystems, the reverb system needs two DPDT relays.  Both poles of one relay are used for switching the two outputs, L and R.  One pole of the second relay switches the input of the reverb driver, between the output of the source selector switch ("in"), or ground ("bypass").  The effect-send jack also feeds from the selector switch, but is not affected by the bypass relay: i.e., it is always-active.  And the other pole of the second relay disconnects the "dry" signal from the mix pots when bypassed, to reduce loading on this non-buffered signal.  When bypassed, instead of feeding these mix pots, the dry signal will be feeding the final master-volume pot sections: just about the same kind of load.  (The source of this signal might be any one of the prior subsystem blocks, depending on which ones are bypassed or in-circuit.)

To summarize:
reverb driver input: subject to source-switch; subject to bypass.
effect send jack: subject to source-switch; not subject to bypass.
dry signal: not subject to source-switch (always from eq2); subject to bypass.

Thus, the effect send jack can be treated as a "monitor" output, which can sample the signal at various points in the chain.  In addition to the obvious uses, this jack could be used to connect a tuner, a VU meter, other test equipment such as oscilloscopes or spectrum analyzers, etc.; or it could be used as a buffered line-out to drive other amps or signal chains, with the source switch providing the option for this signal to be clean or distorted.

The L & R effect returns are simply circuit-interrupting jacks, inserted at the last point before the output bypass relays (i.e., effect returns are of course subject to bypass).  Thus, there is no buffering or padding, of either the effect send or the returns; the external circuitry must be able to handle and produce tube-compatible signal levels (nominally line-level, but with possible very high peak levels).  I thought about putting the returns earlier in the circuit, so that they could be affected by the "mix" pots (and also they'd be tube-buffered); however, this would have made these inputs inverting.  This would have also required me to separate functionality and add another pair of jacks: power amp in L & R.  Instead, I can retain the dual functionality of the effect return jacks: they are also the power amp inputs when the preamp is off.  (By unplugging the reverb tank outputs, one can get access to high-sensitivity inputs which are subject to the mix controls, although probably only RCA jacks will be available.)

Depending on how much extra complication I want to add, there might also be a "post-tone" effect send jack, which sends the same signal that drives the reverb transformer.  Thus, the active tone/contour circuit could be used as a pre-EQ for external effects, if desired (and also, possibly my real motivation, it will be easier to monitor and instrument this tone circuit).  This post-tone send jack will be subject to bypass, like the spring reverbs themselves; unlike some guitar amps, the reverbs will not be driven when the reverb subsystem is bypassed.

So... it's obvious how the overall functionality of this subsystem works, right?  One can have either spring reverb, *or* external stereo effects, but not both.  The bypass relay (i.e., footswitch or front-panel switch) controls whichever is active, spring reverb or external effects.  The spring reverbs are operative until the external effect outputs are plugged into the effect return jacks: then the external effects take priority over the spring reverbs.  If desired, one spring reverb (L or R) can stay active, by plugging into only one of the two return jacks.  The effect send stays active, regardless whether spring reverb or external effects are being heard at the output.  The source selector switch controls the input to the reverbs and also the effect send, but not the dry signal.  Thus, the dry signal can have more distortion or a different EQ than the "wet" reverb/effects signal.  The "mix" pots vary between 100% dry and 100% wet, independently for each speaker, so the spring reverb can be effectively "panned" all the way to one speaker if desired, or placed anywhere else in the stereo mix.  Unfortunately, as discussed, the external effects will not be subject to the wet/dry mix knobs, for pragmatic reasons; stereo panning and wet/dry mixing will have to be accomplished by the external effects unit, itself.

reverb tone control

More thoughts on the reverb subsystem.  I think it will need some kind of tone control.  A bright reverb will sound more responsive and immediately gratifying, but many real acoustic spaces have a high frequency rolloff characteristic; thus, a darker reverb, although it may not call as much attention to itself, may ultimately prove to be more realistic-sounding and more supportive to a good guitar tone.

So at the least, a control which can reduce the treble frequencies seems important.  However, since the reverb already requires tubes for the drive and returns, the possibility appears to be available, to make the tone control active: i.e., able to boost treble as well as cut, with "flat" reliably centered in the middle.  This amounts to half of a Baxandall tone circuit, just the treble knob, with bass effectively hard-wired to "flat".

But, as long as we are adding facilities to flexibly alter the tone of the reverb in useful ways, within the space of one knob, we can make more options available by using a dual-gang pot with a pull-switch.  When pushed in, the control acts as a simple Baxandall treble knob, as described above.  When pulled out, the two gangs of the pot become Baxandall treble and bass controls, both moving in parallel: thus, when rotated clockwise, the control boosts both the bass and the treble, creating a response curve akin to the "loudness" curve on a home stereo, i.e., a curve which is mildly mid-scooped.  Rotating the control the other way, counter-clockwise past the 12:00 "flat" position, the bass and treble are cut while the midrange stays the same, resulting effectively in a mid-boost curve.  This pull-switch functionality is called "pull contour".  I have not decided whether to label the overall control "treble", or "tone".

So, the reverb module now contains four knobs: the six-position source selector switch, the pull-contour tone control discussed here, and the two wet/dry mix knobs for left and right.

taking shape...

 

overall shape & size

I'll be referring to the front panel diagram in the previous post.  Panels are 1+1/2" wide (I'm pretty sure: there's also the possibility to use 1+3/4" "U" channel; I'll only go to that if space requirements turn out to demand it).  Panels are separated by 3/4"-wide strips of wood (Poplar or Tulipwood 1x2s).  The longer horizontal panels (which are the three gain stage modules -- but other things can go here) are 7+1/4" long.  From this, the other dimensions are implied: the overall perimeter of the collection of modules (not counting the surrounding 3/4" edge framing) comes to 8+1/4" high by 18+1/2" wide.  There will be a 3/4" outer frame, and then the case of the amp may add another 3/4" to each edge, depending on my exact case design, yet to be determined in detail.

Thus, approximate overall width of the amp will be around 22".  If the amp is also about the same height, i.e., a square front face, with the module panels in the upper half of the face, this leaves about the same area beneath the panels, where a side-by-side pair of 8" speakers would nicely fit.

As for depth, the plan is to use the narrowest standard lumber width which will work: possibly 5+1/2", but more likely 7" or even 10".

The plan is for the speaker cabinet to be closed-back, for the tonal and power-handling characteristics that this imparts (but both stereo speakers will be in a single enclosure, with no baffle or partition between them).  Power-handling is a concern in particular, because I plan to use speakers that are just barely rated to handle the power that the amps will generate: this, to obtain speaker breakup and distortion at high volume levels, when desired.

However, into the closed speaker enclosure, other items will also need to fit, to keep the overall "square" profile.  The two spring reverb tanks will be mounted in the bottom of the case.  And, unlike most amps, in this design the tubes are "topside" of the chassis, but the transformers will hang underneath; i.e., the transformers mount to the underside of the partition between the tube circuitry, and the speaker enclosure; the wires pass through holes drilled in the partition (plywood).  Note that there are four major transformers: the left and right stereo output transformers, and the two power transformers, for the preamp and for the power amps.  (As noted elsewhere, the two power transformers are an extravagence mandated by the large number of preamp tubes -- and thus, the large amount of heater current required, by this overall circuit.)

So hopefully, the transformers will manage to stay cool enough, even though enclosed in an unventilated space.  If this is not the case, i.e., if cooling becomes an issue for some or all of the transformers, then some may have to be mounted "topside", with corresponding re-arrangement of all the other components; probably, the result would be a larger total size for the combo amplifier.

For the separate head/speaker configuration, I assume that the general layout will try to stay the same as the combo wherever possible, so this would point to a rather tall-looking "head", with the 20"x10" (approx) control panel, above a blank rectangular area at least 4-5" high, enclosing the transformers (and reverbs).  One nice thing is, the cabinets will be rather tall-looking, which may not be the most graceful proportion, but weight-wise, the center of gravity will be low.

design of active tube-based EQ

I've been working out the circuitry of the 3-band v-mid EQ, of which there will be two copies in this amp (the pre and post EQs).  I have been helped by several Internet "gurus" of tube-amp design, who I will credit in an upcoming post: there are a few excellent websites out there with tons of useful information for anyone working with tube audio.  Also, many of the famous-brand guitar amps such as Fender, Marshall, Orange, Vox, etc., have complete documentation including schematics, publically available (either from the manufacturers or from contributed reverse-engineering efforts by enthusiasts).  All of this has given me the confidence to design this amp, assembling the design from some directly borrowed circuits, and some created or heavily adapted by me.  In the latter category is the active EQ.

The issue is that the typical guitar amp treble-middle-bass "tone stacks", as found in the "F+" and "M" gain modules in this amp, are not able to really produce a midrange boost.  The passive tone stack circuit naturally produces a "mid-scooped" frequency response; twiddling the knobs changes the relative heights of the bass and treble peaks and the depth of the mid valley, but a basic scooped shape remains.  These response curves are an essential part of the "Fender sound" and the "Marshall sound", which is why I preserve the T-M-B tone stacks in the gain modules.  However, even amps with the scooped response in their electronics, will still ultimately tend to produce one or more peaks in the midrange.  These peaks, known as "formants", are crucial to the voicing of the guitar tone.  Various components in the signal chain, especially the speakers and cabinets, contribute to formant production, but in many cases these elements are hard to change; thus, a particular guitar and amp combination may have a distinct tonality, which persists despite any changes in control settings.  If the tonality is a good one, then this is not necessarily a bad thing; but some of us wish for more tonal flexibility, for a single system which can produce a great number of varied tone colours, perhaps including both the "familiar favourites" and also electric guitar tones that have rarely been produced or heard in the past.  To follow this latter path, for guitar, one quickly comes to desire variable midrange boost; and for this, active EQ is needed.

Studying the topology of various active EQ circuits on the Internet, both tube and opamp circuits, I've picked up quite a bit of useful information, though my own designs must be considered experimental and "naive" until I've done quite a bit more building and testing.  So beware...

One major advantage of active EQs, beyond the ability to more easily produce a mid boost, is that the level controls naturally end up wanting to be linear pots, as opposed to audio taper, and the midpoint of the rotation ends up being the "flat" position.  The basic characteristic of all the active EQ circuits, whether tube or solid-state/opamp, is that the level controls for each band select or "mix" between the input signal, and a negative feedback signal which is an out-of-phase image of the output signal.  By contrast, the corresponding passive EQ circuits will have level pots which mix between the input signal, and ground.  By considering the "virtual ground" principle, it can be seen that in the active EQ circuits, the midpoint of the level pot is where the circuit is in perfect balance, with the signal from the negative feedback taking whatever form it needs to, to completely cancel the input signal as seen at the wiper of the pot.  If there were no RC frequency dependence, it would be a simple cancellation with the same amplitude of signal at opposite phase; but with the RC action in the loop, e.g. the output will contain more high frequency signal if the RC network removes some from the input signal.

So with my 3-band design, the three level controls (bass, mid, treble) will each have a center-flat position.  The fourth knob will be mid frequency; now I am considering about a 20:1 range, of 200Hz - 4kHz.  It will probably take some field testing to determine exactly what range is needed.  The frequency knob will also have a pull-switch, "pull wide", which will widen the mid peak by changing one of the capacitors.  I.e., in effect, a choice of two "Q" values, though both are pretty low.

I believe I have worked out a good way to implement this EQ using two 12AX7 tubes.  There is nothing exactly like this on the Internet, but I've found enough confirmation for the various parts, that I think it will all work together as I am intending.  The design starts from what's known as the "active Baxandall" circuit.  This gives the bass and treble.  Then, I found an elabouration which provided 3 fixed bands, bass-mid-treble.  My contribution is to remove the capacitors from the midrange portion of this circuit, retaining the mid level pot and the fixed resistors at each end; thus, the overall resistance of the Baxandall circuit stays the same, which I hope will keep the bass and treble operating as they are supposed to.  The mid level control becomes a plain pot, selecting between the in-phase and the out-of-phase sides of the Baxandall network.  The output of the pot is buffered by another 12AX7 section, running as a cathode follower; this buffered signal drives a variable-frequency RC bandpass filter (Wien bridge circuit).  The dual-ganged frequency pot changes the "R" value in two points of the circuit; the "pull wide" switch changes one of the two "C" values.

The output of the Wien bridge BPF passes through a summing resistor, and recombines with the output of the Baxandall circuit; this summing point is the grid of the output amplifier, which operates as an inverting amplifier, and its output becomes the drive for the out-of-phase side of the Baxandall circuit.  Thus, as with opamp circuits, the summing point becomes a virtual ground, so the summing of the signals becomes "mathematically pure", and interaction between the signals is suppressed.

As I originally "borrowed" it, the active Baxandall and active fixed-freq 3-band EQ circuits were inverting, because they used a cathode follower at the input to drive the Bax circuit, and then the inverting-with-feedback stage on the output.  I have already added another stage, the mid-driver cathode follower above.  My additional changes are to add the available fourth stage as a cathode follower after the inverting stage at the output (i.e., a two-stage inverting amplifier, with better impedance characteristics).  And finally, the input stage changes from cathode follower (gain of roughly +1), to an inverting stage with local negative feedback, producing a gain of -1.  Thus, the overall circuit becomes non-inverting.

In my original design, I had planned to use a passive Baxandall circuit, and a buffered-but-passive midrange circuit, so overall there would be significant signal loss; thus, I planned to fit a gain-adjust trim pot at the input stage, to allow flat-gain to be trimmed to 1.  However, with the fully-active (feedback-based) circuit described above, I believe the flat-gain will inherently equal 1, so no trimming should be needed.

initial front panel layout

At an early stage, I started laying out the front panel of this amp, in order to decide which controls I want and how I want to lay them out.  The panel layouts are written by me, directly in Postscript: this is my preferred "graphics program" for things like this.  It is easy to change elements of the layout, such as the placement of controls or the choice of text for labels.  The original Postscript files (as well as PDF versions of the same) will be made freely available when this design is finished, as it is my intent to give away every aspect of this design as fully as possible.  I include a JPEG rendering of the layout here, to give a general sense of where the design stands, but don't depend on this image too much, many things are likely to change before anything actually gets built.

Each module is fashioned from 1+1/2" extruded aluminum "U" channel.  For convenience, I've made sure that no individual panel is longer than can fit on a regular sheet of 8+1/2"x11" paper.  Thus, DIY-ers can simply print the panel layouts and affix them to the extruded aluminum panels.  There's a whole procedure for doing this, in a way that results in a nice-looking and long-lasting panel; I'll detail the procedure somewhere down the road, when I start actually doing it for this project.  But the executive summary is: cut the paper to fit; spray a coat of white spraypaint on the panel; while the paint is still wet, carefully place the paper and let the paint adhere it to the panel (there's no chance for a re-do, so make zero mistakes if possible); press it down flat, taking care to work out any air bubbles; after a thorough drying period of at least 24 hours, cover with plenty of thin coats of clear-coat (clear spray enamel or polyurethane), allowing proper drying in between.  Probably the best time to cut the holes through the paper to match the holes which should already be drilled in the aluminum, would be after the white paint has dried but before the clear-coating begins.  If you're patient enough, this procedure can produce quite good, "near storebought quality", results (remember "NLQ" printers?).  (Thanks to the great Philip Williams for inventing this procedure.)