Extraordinary POQX Analogue Signal Processor


Introducing another little piece of genius – “the Precision Orthophonic Quasi eXtrola” AKA “Plaster the Art of Timing”. 😆 😆 😆 😳 It isn’t one of these, but the Mk II form of the analogue stereo pre-amp/crossover chassis described here.  And it reveals a lot more musical detail than certain commercial digital offerings whilst providing useful gain!  Unlike commercial systems however, it was built specifically for this active 3-way speaker system and is not adaptable to others without a soldering iron.  Therefore it isn’t worth ten bucks!  Well to me it’s worth a bit more.  😆

Here’s what it has:

  • Stereo pre-amplifier module;
  • Twin Linkwitz Transform circuits for the sealed stereo sub-woofers;
  • Active phase-coherent fixed frequency Linkwitz-Riley 24dB/octave crossovers (3-way);
  • 6dB baffle step compensation (tuned to measured in-room response);
  • Tweeter phase shift network for very flat mid/tweeter crossover response despite physically offset driver acoustic centres;
  • Remote analogue volume control;
  • Infinitely variable Robinson-Dadson loudness compensation approximation curves adjusted remotely and independently of the volume control while the music is playing;
  • Power On/Off output muting;
  • No annoying switching transients; and
  • It can take the analogue signals from an SACD or DVD-A player or high-end DAC processing even 384kHz/32bit PC sources (if that kind of silliness appeals) or DSD (more silliness) and do its thing without multiple D:A conversions or down-sampling.

Here is what it generates:

Crossover line out

Those are actual measurements and show the respective line level outputs.  The levels are not normalised since these are fed to different amplifiers and speakers which each have different overall gains/efficiencies.  The upward swoop at the left is the Linkwitz-Transform and the downward slope across the midrange is the baffle step compensation.

THD is exceptionally low.  Check this out:

Measured distortion

THD of a mere 0.003% at the middle of the midrange and this is fairly typical across the whole response.

If interested, the Mk 1 page should be read first.

Remote Volume Control

I decided to include a motorised volume control because the up-front D:A converter’s digital volume control was rendered incompetent by the manufacturer’s smarty pants firmware programmers and had to be disabled – else I’d lose my mind! 😆 The pot is a simple motorised dual gang Alps RK27 10K log taper pot that works without injecting any audible noise and sounds a lot better when turned down, can be manually overridden, and stays where you leave it when the system is turned off:

It is controlled by an ESP P110B PCB.  Its IR receiver (U1) aligns with a light pipe passing through the front panel:

It has a frosted dome to pick up a signal even from a wide angle:

The PIC chip works with Sony “Device 1” (TV) remote control codes for volume up/down and “mute”.  After much confusion, clarified by a kind person named Jeff, it was found that the correct buttons are 2 for up, 3 for down and 4 for mute.  Anyway, I re-assigned the up/down and one other key of a programmable remote to transmit the Sony TV 2, 3 and 4 codes and that works (so long as you cover the remote sensor of your Sony TV).

The PCB takes 21Vdc from the main power supply’s auxiliary output.  I used a switch-mode voltage regulator to reduce that unregulated voltage to 5V for the motor.

It’s tucked in behind the power supply transformer and its switching frequency is ultrasonic (350kHz) so even if despite its location it does leak into the audio ground it won’t matter.

Equal Loudness Compensation

When the playback volume is changed from the mastering reference level, the intended EQ is lost and that’s no doubt why my system sounded pretty ordinary or “thin” at low listening levels with most music.  In the olden days, many pre-amps came with a “loudness” button that could be pressed for low volume listening.  But these simply clicked in a compensation profile approximating a single one of the Fletch-Munson or Robinson-Dadson curves.  Here is a typical representation of the Robinson-Dadson equal loudness profiles:

The dips are all around the 4kHz mark which is where human hearing is most sensitive.  Where the curves rise to either side, this indicates that at both ends of the spectrum more power is needed for all frequencies to seem equally loud.

So a “loudness” button might have clicked-in say an “A-weighting” compensation to approximate the 40 Phon (40dB @ 1kHz) curve, but that one is not correct for any other listening level.  Most of those systems added gain by inserting an opamp or two into the signal path.  But change the volume and it’s all wrong and could easily be out by 25dB depending on the preferred listening level and it’s relationship to the reference level used during mastering.

Some manufacturers (e.g. McIntosh and Yamaha) had a second dial being a variable loudness control, but to the best of my knowledge these acted solely on the bass levels.  Some used centre-tapped volume pots to add loudness compensation based on the rotational position of the volume knob.  Again a compromise, because different speaker efficiencies/amplifier gains/room sizes override any “standard” (or best guess) centre-tap position.  And there was no loudness compensation for half of the rotation anyway.

One of the most interesting loudness compensation ideas that I know of is the subject of US Patent 8,054,993 to Harman International (I guess it may be the DBX “Auto Warmth” feature).  Of course there is also Audyssey Dynamic EQ and THX Loudness Plus etc. which are adapted for multi-channel home theatre, but they sound dreadful for stereo music and you need a connected TV to set-up the menus etc. – a right pain.  The Harman system acts only on the bass, but takes the input signal and uses an RMS voltage detector to DC-illuminate the LED of a Vactrol on the basis of the overall signal energy.  The LDR of the Vactrol is part of a capacitance multiplier circuit that boosts the bass by shifting the corner frequency – i.e. a continuously moving loudness compensator that adapts to the dynamic level of the source.  Much too difficult for a DIY project like this.

Anyway, the best thing I could achieve as a mere amateur was a passive analogue profile adjustment.  A helpful person at the ESP forum came up with a couple of circuits to do it.  I wouldn’t mention his name without his approval, but anyone interested can go over there and search for the thread (and see that I was perhaps a little OTT with gratitude 😆 ).  One of the circuits generated arguably better curves but could not be “turned flat”.  The one that I liked required no switch and can do this with just 2 resistors and 2 capacitors around a standard 3-terminal pot (click for high res.):

It’s important to note that the curves are not intended to follow the exact contours of the equal loudness curves!  They are intended to track the difference between those curves, and they probably do an extremely good job of that.

Any one of the curves can be dialled in and the volume control used independently to raise and lower the curves without “bending” them.  It’s fantastic because not only can you correct for changes in your own playback volume, but also for incorrect mastering. 🙂 The “normal position” flat line is achieved at fully clockwise rotation of the loudness pot where the 4 passive components and the pot itself are bypassed automatically.

And the very astute observer will notice that the first half of the rotation (indicated by “Half turn” in the above graph) provides very useful baffle step compensation!

The suggested circuit had a 5K linear pot.  I could not find a motorised dual gang 5K linear pot, but a 10K version was available from Alps.  Rather than shunting a 10K resistor across it, I changed the RC values to suit.  I got hold of a circuit simulation package and modelled it myself.  Here is the circuit (and it simulates almost exactly the same):

(unwanted hotlinking (bandwidth theft) to this image forced me to apply the necessary blocks.  If you want it please just use the snipping tool or similar)

The input (Vs1) is taken from the preamp output and that 100K resistor simply represents the input impedance of the LT circuit that follows.  The 442Ω resistor is an odd but available value (which I had), but 430Ω would be fine – giving slightly deeper curves.

It is a completely passive filter – no opamps – and therefore attenuates the middle frequencies rather than boosting at either side.  For my case this is perfect because there is already ample “boost” immediately downstream in the LT circuit and enough is enough.

Here is how the amplitude and phase bodes appear for the actual circuit (click for high res.):

This of course is just what happens before the LT circuit and the 3-way crossovers where additional phase changes superimpose over the blue lines, but being “up front” it does not affect the phase alignment (i.e. no sudden step between drivers) at the crossover points and being quite shallow, it does not affect the acoustically summed responses of the drivers at either side due to the very steep crossover slopes.

Here is the loudness circuit duplicated for stereo on a tiny PCB designed for soldering directly to centre-tapped (4-pins per gang) Alps pots.  I had to hack the ground planes to separate them for left and right channels and of course there are no centre taps here.

The PCB has provision for 3-terminal screw blocks, but I installed 4-terminal blocks with each 4th leg snipped-off to provide a convenient place to connect the 442Ω resistors to audio ground (via the coax shields to the input of the next stage):

As an aside, the following is the kind of two-dial bass/treble compensation offered by many older preamps:

This is typical with a centre frequency around 650Hz.  A limited range of semi-correct profiles (like the one shown in red) could be achieved and compensation for bad/bloated recordings or a bright room could be achieved with some of the cut profiles, but two pots were needed and none of the profiles really follow normalised loudness curves such as these:

(some numerical errors there, but the idea is clear enough)

Of course the little circuit doesn’t follow any of the curves perfectly either, but it’s pretty good overall.

Motor Control to the Loudness Pot

The “mute” feature of P110B could be very useful.  Of course it’s intended use it to mute the audio output, but with a pause button on the CD player I didn’t want that.  So, instead of using the mute feature as suggested to control a signal-muting relay, it could be reassigned to “toggle” the motor power between the volume and loudness pots.  The motor wires from the P110B PCB connect to the Com terminals.  The NC terminals connect to the volume pot’s motor.  The NO terminals connect to the loudness pot’s motor.

Here is the miniature relay with the 5V coil attached via a pair of 75Ω series voltage-dropping resistors to a front panel LED alongside the loudness dial.  When the remote control’s “toggle” button is pressed, and the LED illuminates, that means “now the Up/Down buttons do me”.  With the PoM (Power-on-Mute) jumper set to On, the LED shines for 10 seconds after powering up also (looks like there’s something important going on inside).  😀

As the motorised pots were an after-thought to the original chassis build, there was little room for them.  I was going to shoe-horn them above the preamp PCB, but there was insufficient internal height.  After some careful measurement, I figured that with the preamp PCB raised, the motors could be inverted and squeezed in underneath, but the preamp board would have to be shifted back a bit too.  Problem was I didn’t want to drill new holes in the chassis floor only to have the legs clash with the crossover PCBs directly behind, so I made these little off-set plates:

Putting it all together

Well I decided to do it in a systematic way so that it could be pulled apart easily after the smoke. 😆 So I removed the front panel and one side and used that “L” as a module to hold the new parts:

Motor damping/circuit-quietening bipolar capacitors across the 5V motor terminals.

Motors then fit in under the preamp:

… so they ended up here:

… with the relay crammed in here:

Then all that was needed was to connect all the wires up.

What About the Baffle Step Correction and Phase Shift Networks?

They’re on a separate page which can be found here.


Attached to the scope and powered up.  No smoke.  No 350kHz noise from the switching regulator and the motors produced no visible noise when operated either.  The left/right traces at all outputs tracked perfectly at all volume and loudness settings so it all looked good.  These large RK27 Alps pots always seem to track perfectly and never get noisy (I’ve been twisting a manual one in a C-J preamp for 20 years anyway).

The high-contrast knobs were chosen for clear visibility from back in the lounge chair where I sit like a slug doing nothing for most of the day.  😀

A short video of it in operation:

Listening impressions

Well what can I say?  Despite the silly title of this page, the thing is seriously fantastic.  It is easy to use because you don’t have to switch the loudness circuit in and out.  At the fully clockwise position the circuit is “off”.  At low volumes “fullness” is achieved with an anticlockwise turn.  If you turn the volume up from there it can become bloated of course, but then just turn the loudness pot clockwise and the “fidelity” returns to where it was before (before this modification).  I find myself fine tuning on almost everything that I play.  Even Deutsch Grammophon CDs sound good now!  I could never stand them before, but now a correction can simply be dialled in.  There’s nothing I can do to fix Telarc recordings mind you, but very bright recordings that couldn’t previously be played loudly can now really fill the room without being shrill.  And it can be (ab)used to thrilling effect on certain discs to produce enormous bass.  Dorian CD DOR-90182 can be played with gross over-compensation of the loudness control and the volume set at a level that would deafen you otherwise and when that sub-contra C pedal comes in the floor flutters up and down under your feet but you can’t actually hear what’s causing it!  Absolutely extraordinary! 😯

Of course it can’t track dynamic changes in the music to adjust the compensation on the fly (my next project?)



2 thoughts on “Extraordinary POQX Analogue Signal Processor

  1. Hi Ian,
    Gosh, I’ve been looking for an “easy” loudness control for a long time.

    Please, may I have permission to use the one shown above? And save the ‘Loudness Control’ portion of this article to my hard drive?

    If you approve, is it possible to copy and paste the schematic, rather than having to redraw it?

    Ian, thank you very much for your time and the good work you have posted on this blog.
    Paul Croft

  2. Sure. It isn’t my design and it’s public domain now anyway. It was done for me by a scientist on the ESP forum who claims nothing too special in it. I still use it often. One thing to be aware of is that it lowers the overall sound pressure when the pot is turned anticlockwise, so the volume control has to be twiddled in conjunction with it. It’s simple once you do it a couple of times.

    Oh the schematic. I made that image “unclickable” because some jerk on diy audio not only criticized the circuit as being “hit and miss” (which it is not) but took it upon himself to hotlink to it thus exploiting my bandwidth. You can take a screen shot and do what you like with it.

Leave a Reply

Your email address will not be published. Required fields are marked *