Fix price offer

Quad 405 Revision

complete revision with 3 years warranty*
*explanations see warranty disclaimer

german version

refurbishing base price ca. 650€, not yet exactly calculated
includes: 2 new power modules 230€ each, 2 new smoothing capacitors 25€ each, assembling about 1 hour.

first sample overview with the new boards

Quad 405 - a breakthrough innovation

This is how you could call its design at least in the mid seventies, what I did now to it is virgin soil as well. With the model discussed here at that time the conception was first applied to a serial production. Later on below I'll respond to its "current dumping"-design. Tens of thousands Quad 405 power  amplifiers in two variations were sold for more than a decade and were a reliable dray-horse for the manufacturer, even if it may not have brought the economical success the technical ingenuity promised. What also may have kept it on the "insider's tip" level is the slightly bullish British layout. Also some unnecessary technical deficiency would have better been removed. Let me list some of them:

  • the low mass, but also in British scales a bit thin/high impedance internal wiring using the case as central ground "star", connected beside the rectifier
  • the speaker terminals are of an acquired taste (same thinking)
  • for a long time the missing power switch and the 240°DIN input socket
  • the inverting and being all but "low noise" OP input stage of the power modules

All that may be acceptable. Harder to bear is the meanwhile no more negligible wear of almost all examples:

  • eroded, bursted open and leaking main power supply smoothing capacitors,
  • cracked, ruby coloured Röderstein miniature plastic caps and other small electrolytic capacitors with more or less capacity loss, followed by humming and lower open loop gain resp. a bad slew rate of the negative slope.
  • several heat worn or by former repairs damaged printed board regions.

Keith Snook

With an own Quad 405 a British Hifi enthusiast (obviously engineer) brought me to the idea to try implementing his proposals beyond the standard revision. You find the description of Mr. Snook's perceptions on his internet page. There amongst others he reveals the not so beneficial compromises made during construction and assembly of the 405 which especially in that model foil a bit the outstanding theory of the invention - which was much less the case with later models, thus he takes some of his circuitry change proposals from the successor models. But I think, that the 405 in its more conventional layout with some local electrolytic buffer capacitors - compared with successors and their totally (hyper-)activated capacitor replacement trick circuitry - also shows some advantages. Buffer electrolytics may have wear, but as long as they work, they also "calm down" and "warm up" the sound stage, that is even more true for the modified version of the 405, if e.g. you compare the character with an intact Quad 606. I tested the latest of Snook's modification variations and was very taken by the result. Still, unfortunately, the reversion to the proposed resp. intended stage of development was everything but easy, at least at the 405s first generation boards Quad had bended all component wires flat on the backside of the board prior soldering. It makes a hell of effort to remove a greater number of parts.

So years ago I already decided to transform Snook's latest proposals into an own board layout one day. This decision remained unrealized for a long time. In the End of 2015 resp. in early 2016 then two customers applied for a revision of their 405s. Even then the development still didn't proceed because of missing capacities until the first amplifier had passed the queue. Since the customer had no replacement then I took the case on my desk with some major priority.

Own replacement board

In this Quad 405 the power stage boards were in a really too miserable condition for a combination of repair and modification, so I even more concentrated my effords on the development of an own board. Further motivation: for donkeys years Google shows the most often used keyword leading to my homepage is "Quad 405". There were two options: transforming the positive tested Snook modification version one-to-one into a printed board layout, or the last proposal in his development list.

No option was to use one of the many offered cheap far east replacement boards (which are mainly pure clones of the original) or of expensive competitors (where you have no control of the details).
Further there are some personal preferences, e.g. relative big buffer electrolytic capacitors after the fuse in style and choice similar to Symphonic Line power amplifiers, all that in the driver section as well as in the power stage, if possible in a way that it makes no noticeable impedance change in the audio range. Such wishes can only be realized by doing it yourself, they cannot be bought or even be upgraded in an acceptable way.

I decided to use the last, much simplified Snook version, in spite of his "partly VHF unstable" note. I presumed the scheme was mainly tested, provided suitable compensation points in the layout and expected it to be done with some fine tuning of the dimensioning.
Far wrong, it absolutely didn't work. The 30A dumper transistors simply died during the load test at reasonable high frequencies or steep down slopes. The behaviour was so much asymmetrical, that I pondered: developing forward or or back to the modified version? Hence one will loose in both cases according to Murphy's law, I thought the more innovative solution to be the more attractive one. After all these tests and perceptions with the first attempt I preferred going forward at even higher risk and exploring, how these actually perfect chosen high power dumpers are applied correct, how to proceed with the core of the Quad conception.

Consequence: a totally symmetrical version of the class A stage without a bootstrap capacitor current source (as Quad does) or a ring of two current source (Snook) as class A load, but two complementary driven branches with automatic idling current control.
After having ordered the first 10, now obsolete boards for about 120€ the next 10 samples now were of much more use, but the reduced and now absolutely symmetrical problem had not vanished, was not yet even explained. Obvious design errors were soon corrected, but only the correct coupling of the dumpers made everything run smooth and then was tested with speakers. All together at this board were again too many changes mounted to expect that to a customer.

So the third board went into production. This finally was in any aspect usable and rugged from the very beginning.
When the hearing tests had showed the same result, I dared to send the amplifier back to the customer before I went on holiday.

last version of the own Quad forward development, the complete symmetry of the very power stage made the use of some SMD parts the better choice for the lack of space.

The result uses the same base circuitry still with some major improvements. The amplifier was even tested with 20kHz full level into 8 Ohms, the crowbar safety circuit is integrated in the board as it is in the Quad 405 MK2. Hiss, on- and off-blopp are things this amp doesn't know. The input OP amp on its well buffered and low noise supply is the TI/BB type OPA134, which as dual version already convinced with a very good performance in a similar function in the Musical Fidelity A1.

Not before the third layout I had everything placed in a way, that allowed the high buffer capacitors to fit exactly between the the power supply smoothing capacitors. (Still) these buffer capacitors are are bit in the way of the cooling brackets' screws.

Additional perceptions:

I found that the damping factor of 175 is far too high at my 2 way Bluesline floor standing speakers. For the use in parallel mode (mono block) a second output pin connected via a 0.22 Ohms 5W resistor is already provided on the boards. That equals e.g. the output wiring of all Farlowe-Exposure-amplifiers. With this connector at that small speaker everything played more dynamic, cleaner and aerial. In that mode I finally sent out that first amplifier, with other speakers maybe the other connector is the better one , but this can be only found out by try and error.

at the other channel you look on the board from the other side

Current Dumping by Quad

Established methods to reduce distortion

The designers purpose is like with any analogue amplifier to have the input signal voltage just multiplied by the gain factor, but having no changes in the curve apart from the scale. A recurrent problem with this is, that in energy terms most saving solutions obtain their efficiency from circuitry parts switch off (in half waves or pulse width). Thus always edges of the characteristic curves of the involved power devices are involved. The traditional methods to avoid or correct the resulting distortion are either avoiding such curve edges (e.g. by applying more idling current up to class A) or by reducing gain resp. assigning most of the gain for reducing distortion. That is usually achieved by putting a high gain amplifier into a comparing mode, where permanently the input signal is checked for identity with a resistor divided part of the output signal. Only if the input stage "sees" no more difference between these, the output is no more driven further applying the whole open loop gain. Hence the resulting gain factor equals exactly the division factor of the resistors. Given an amplifier with an open loop gain of 100000 (= +100dB open loop gain) and applies a divider of 10:1 (= -20dB), 80dB of the gain are used as negative feedback for linearisation. By exactly this scale also the (linear) distortion is reduced, given the amp had a THD level of 15% without negative feedback, it is reduced (all side effects neglected) by the factor 10000 to 0,0015% - quite usual values.
In the same scale the linearity error is reduced the bandwidth and frequency response grows. At the same time the (virtual/differential) source impedance of the amplifier is reduced by the same amount. You could think that's ideal, but the sad truth is that you seldom get free presents in physics, for the really important physical property of the arrangement, the bandwidth-gain-product, stays invariant anyway.

And this procedure doesn't change the insufficient linearity and the harmonics spectrum of the base construction, but it just tempers them in the scale of the negative feedback and cannot even theoretically ever reduce them to zero. And with this method at the same time immediately other than the original linearity problems appear. For now factors, that were absolutely unimportant for an non feedback amplifier, like the signal delay come annoying into the game. At the input of a amplifier with negative feedback basically the actual incoming signal is compared with a signal delayed by one amplifier pass-trough time, an amplified and divided input signal from the past. As soon as such a delay is combined with high degrees of negative feedback the addiction to a certain life of its own grows, the delay equals a phase shift, the negative feedback becomes a positive one and in the worst case the amplifier becomes an oscillator. Thus you cannot speak of negative feedback as a universal remedy, here you pay for the lower THD with dynamic problems as tendency to oscillate, resulting mask effects or certain intermodulation distortion types. That may not annoy at all with uniform test signals, here THD as single parameter may still count a little. Anyway this value shouldn't be too bad, one can easily agree without any discussion on mandatory less than 0.1% (<-60dB). But whoever on the other hand side likes listening to pure sine waves in his/her spare time?

The linearity problem can at least partly be eradicated by the choice of components, the layout and applying high idling currents, than by a "much-helps-a-lot" negative feedback with all its side effects. Mostly this is also coming with more financial expanse for parts and/or energy.

Forward correction and combination of two amplifier types

Thus the "current-dumping" approach goes a partly different way, here is counted on a high quality, full voltage swing but low power class A amplifier, which is linearised by negative feedback on the one hand side, on the other hand side it drives the speaker like a push pull valve power stage with anode output, having no local negative feedback by the speaker current (similar Musical Fidelity resp. according to the Abacus Rieder concept). Up until currents of some few tens of milliamperes this precision amplifier still manages to serve the speaker alone via a coupling resistor, there we just have a low volume. If more current is needed the so called "dumper" comes into the game, a pure class B amplifier with powerful output devices, which is driven by the voltage produced by the load current at the coupling resistor. A class B amplifier may have a outstanding efficiency but also significant zero cross distortion, which you don't like to hear.

And now comes the trick: the low power class A precision amplifier is set on the even less powerful errors of the class B dumper, to compensate them as completely as possible, this is done with negative feedback on the one hand side and with a forward correction on the other. To achieve a mathematically and practically precise cancellation of all dumper errors and to get in sum a efficient high power amplifier with a pure class A characteristic and without any distortion by local current feedback, a bridge circuit with four components is applied. The simple variant would be using just actual resistors, but at Quad they had the additional purpose to allow the use of cheap, powerful (and as a result usually quite slow) dumpers. So in the bridge they used also apparent impedances to the best possible advantage: the dumper stage contains in its power bridge branch a (delaying) coil, the class A negative feedback bridge branch balances that with a (accelerating) capacitor. So higher frequencies will be preferential assigned to the precise A-amplifier, the dumper remains for the blunt. The whole construction is balanced in a way, that allows a almost random choice of high current power devices as dumpers.

With perfect bridge balance now the double track negative feedback provided class A amplifier will force its character on the complete circuitry. The class B dumper driven and forward corrected by it hast to take the high currents then, which the A amplifiers always "dumps" to it after a short period to be able to care unburdened for the exact further progress of the voltage curve. Both components have not the slightest tendency to thermal instability and the A amplifier develops just some 4.5 Watts of power dissipation per channel, the dumper doesn't heat up at all without signal. By the way it doesn't matter at all for the principle, if the dumper current has a direction according to the output voltage, for reverse going reactive currents we have exactly the same conditions, the distortion doesn't grow at all that way.

Tech Talk about the internal development

The differences between the tested Snook modification and his more advanced proposal consist of the following points:

  • the bootstrap circuit of the class A amplifier, comprising of a power voltage divider an a electrolytic capacitor was replaced by a transistor current source,
  • the dumper transistors became complementary 30A power Darlington types with simplified input coupling.

The Snook circuitry

I implemented that proposal almost unchanged in a layout, still with a relatively unprecise position of the power devices at the heatsink. Power supply ground and input ground each got a dividing star (close to the input one track was lost again by a change) and to remove the VHF instability mentioned by Snook positions for possible compensation capacitors were provided in SMD technology. 

to zoom in: right click and "show picture"

The schematics obviously accords widely with Snook's proposal. The big buffer electrolytic capacitors in both parts of the schematic are almost the only difference. This version did work to the extend, that it could be put into operation, showed the proper class A idling current and amplified absolutely perfect without load. But what immediately struck with load: totally different behaviour for the correctly running positive and the far to flat and distorted negative slope. At full throttle either with clipping or if one applied more than 5kHz at maximum level (with steeper signals even more critical) or both with a hushed "plick" both power Darlingtons died. Without destruction I could not find out more, than that the dumper was just inadequate driven. Since the biggest measurable and visible part of the problem seemed to depend on the different driving of both half waves the following consideration resulted:
Obviously a idling current of 44mA as in the Quad original design was not enough for the Darlington equipped dumper stage to accelerate fast enough out of the idle position as soon as its support was demanded. The difference of the half waves in the original design and in Snook's proposal depends on the fact, that the asymmetrical layout of the A amplifier allows to drive it to arbitrary high currents in positive direction, but in the negative direction the maximum current is determined by the dimensioning of the current source, just the traditional 44mA class A average idling current. To drive the chosen Darlington dumper in a complementary way, a mirrored design also would be far more appropriate, which immediately brings up the problem how to determine the (average) idling current, which in the original and also in the Snook's design remains absolutely constant. As soon as the average is no more the negative maximum at the same time, something active must be introduced, which regulates itself reliable. 

the first layout based on Snook

Something that didn't work well at all with this layout was also the convergence of the brackets holes and the board. I also didn't consider the absolute single sided mounting on the back of the bracket, making isolation spacers necessary after some file operations. And the biggest mistake were the mirrored symbols of the power transistors, I had to cross-tinker the connectors, the library element I used was totally inappropriate, the same about the coil (far too small drilling). So together with the missing ground track close to the input that was all not too good yet.

"Next generation" - first attempt

The following plan first turned out to be at least in parts principally wrong. Not that any catastrophe happened during the switch on of the first ready mounted sample, no ampere-puffs, fuses o.k., offset close to the lower limit of my measuring capabilities, without load also some kind of amplification.

to zoom in: right click and "show picture"

the second layout: the first totally symmetrical Quad 405

But at the same time there was far too much distortion and not the slightest load capacity, the dumper absolutely didn't join in. The error in reasoning was the direct connection of the class A Darlington with the current source load of the power stage input transistor. The first structural change then had to be an additional transistor to receive an independent "ring-of-two" current source. With the according change the open loop gain rose massively and the dumper took up its work. One resulting necessity also was a different quiescent point regulation, see final circuit variant. Would have been too nice, if that had brought perfect function. Now as if by a miracle everything was symmetrical, but with load things looked dangerous from the beginning: after even the slightest clipping the slope was dented, a sign for a "memory effect". There was a moment, in which above the clipping limit the negative feedback tried so much to correct the curve progression, that it made everything getting out of gear.
First consideration: the collector current of the input transistor in the Quad original is limited by a 3.3 kOhm resistor in a way, that makes an overload of this abd the next stage in this branch impossible. the bypass capacitor of this resistor helps to avoid a resulting delay resp. a unwanted phase shift. Snook had deleted it and by this compromised the class A amplifier as well as the input transistor. So this RC-member was refitted. That was a partial improvement anyhow. The other point was that, as mentioned above, even the genuine conception could deliver virtually infinite currents to the dumper, the original just for the positive half wave, in my variant on both sides. Thus in the interim version both class A output transistors could be driven into saturation. So I determined a maximum current with another additional transistor. Now for both half waves the average current was at the traditional 44mA, the maximum current was twice as high. Each half now could be driven up and down by the same amount. A little distortion in the crossover region remained, obviously by not yet perfect driving of the dumpers. With the installed types MJ11015/MJ11016 I had problems from the very start to determine their true gain with my HFE meter. It showed just 50, probably due to a too low test voltage. A measurement with a higher voltage then showed more than 20000, but logically only from 1.2V base emitter voltage on.

With 47 Ohms Snook had probably chosen the base resistors of the dumpers more intuitively for these not so much tested transistor types and he didn't find a very good value. My first assumption was, that this was a too high resistance, for there was a too slow reaction. I connected the dumper bases just with chokes. One set of defective power devices later I was aware, that this rather was a overdrive with memory effect and negative feedback reaction. Next try: 1 kOhm, no defective dumpers any more, almost no linearity errors. With a capacitor in parallel to the base resistor that was also history.

Now the output stage drove a load of 8 Ohms at full level and 20 kHz without visible differences to the free run on the scope. The level difference with and without load was extremely little, the resulting calculated damping factor was at handsome 175.

Third layout - all changes integrated

See the resulting schematics:

to zoom in: right click and "show picture"

last layout for present, designed very direct and for high currents

One simplification is the connection of the bias control to the reference voltage of the current source in the input stages collector branch. Due to the lack of space many low power components of the extension moved to the backside and tipped as SMD parts.

Many less stressed helper transistors and resistors are found miniaturized at the back.

The positions of the power transistors fit exactly, for the diagonally mounted TO3 transistors I had created exactly measured, single sided new library elements already for the second version, the coil now has got an even bigger drill diameter for its 1,3mm wire. The buffer electrolytic capacitors of the output stage for the left channel now fit exactly into the gap of the power supply smoothing capacitors. In all my board versions the fuses are positioned right and left on the board. In turn the connection points remained all in their original positions.

Mono block mode

For the parallel use of two modules according to the Quad mono block instruction I provided a trimmer, which of course is mounted only once per case and can be used for the exact gain adjustment. For the use as a mono block there is also an additional connector available, leading through a 0.22Ohms wire wound resistor. Theoretical after an exact gain balance adjustment one could connect two modules at the input and this output for a mono block that provides about 200W into 4 Ohms.

See the 0.22 Ohm output resistor behind the connectors, the isolates fuse holder and how the buffer capacitors fit between the smoothing capacitors.

Still one should "synchronize" some points in that case, especially the buffer capacitors in the input section and the idling current control could lead to differences during switch on or off due to unequal time constants and produce huge cross currents. Since there is no current limiting this may some time kill the dumpers. So this variant isn't yet tested and probably needs the connection of the right and left OP supply rails, but I already provided a better solution for a mono block:

Namely I gave the dumpers breaks in the emitter tracks in form of a big SMD pad, the power supply ground and the dumper output got extra soldering vias and two wholes in the board are provided for the passing of connection braids for to mount an additional dumper board instead of a second channel. Also two SMD pads for the connection of the remote dumper bases are available, connected by 1 kOhm in parallel with 10nF like the local bases. With 0.1Ohm at each dumper emitter now a single channel can be equipped with four 30A Darlingtons at both cooling brackets, unlike the original Quad mono block solution the damping factor is like in the stereo version to your free choice up to 175 by the serial resistor. Adngerous difference between different board outputs is not possible by default. The maximum output power again is about 200W@4Ohm. to say: all the transformer gives...

Preview: next board version

...there are just some minor changes. I'd like to place the buffer capacitors in a way, that allows the easy use of a long cross recess screw driver for mounting the modules again through the backside holes, but this is as much secondary as the more nice placement of the one resistor beside the IC socket and of both ground base point capacitors of the OP stage. The most important is moving one track from the fuse to the buffer bypass capacitor away from the aluminium bracket, for here the distance is too small, at the moment I use a second isolation above the solder stop on top of the board. With these three or four more cosmetic wishes then a status of sustainable usability is reached, only new perceptions may some day reopen the barrel.

Another improvement may be announced soon concerning the rectifier. I got a tip for the use of matched, fast Schottky diodes in TO220 cases from a chummy customer, who himself dose some screwing and research on an academical level. The original is mounted with a bracket to the heatsink, so a replacement could be made. In addition a smart suppression of rectifier generated or here inserted high- and highest frequency interference could possibly bring some sonic benefits. I will report if the according tests result in reproducible improvements.

Mono block experience

of course I was curious, if my new boards would behave as planned in mono block mode. Three used and unchecked 405s were still in my rack, so we decided to make two monos and one stereo from them. Here the mono story.

two of my used 405s became mono blocks

As they should be of the "one driver two dumper" kind two full equipped boards without level trimmpot were necessary - still the mounting of the more than 100 components per board takes the major part of effort and time. Instead of the complete second channel power module here the second cooling bracket is only needed to improve the current capacity and spread the produced heat better. The dumpers are intended to do the powerful but slower part of the amplifiers  work, so putting the second dumper pair a bit apart from the other should not be a speed problem. Until now I made no extra stump board, which in a serial production would make sense, but we had the left over second version boards, that perfectly fit to that purpose. Question was: with or without buffer capacitors - hence the board power ground star goes to the case, were also the power supply star and speaker minus leads meet we mounted them.

a version two board is used for the additional dumper

The contol board was mounted as usual, but got additional 0.1Ohms resistors for each dumper power transistor emitter connector to balance the dumper current between the two dumper pairs. The connection was made in the bottom between transformer and front. Of course both blocks got new BHC smoothing capacitors, about the same size as the built out blue ITTs but twice the capacity. The DIN connector was removed and a RCA socket mounted in the same hole on some left over second version board parts, of course isolated.

The control board side looks almost as the stereo version ones, beside some additional cables and parts

The first tests were made without load, no problem. But with a 8 Ohm resistor connected immediately a dissapointing behaviour appeared:
Some kind of humming seemed to spoil the performance, but no real explanation was obvious. As soon as the second dumper was removed by disconnecting its bases, all seemed fine again. A shielded base connection didn't remove the phenomena, a removal of the buffer capacitors on the dumper board also brought no improvement. A longer load test showed that the extra dumper became much warmer than the one on the control board.

Then I remembered Snook's VHF problem and studied the Quad original dumper circuitry a bit. There was a vast use of inductors in the dumper branch, obviously to slow down local current changes. Thus I replaced all four MPC low inductive resistors I had as  Emitter shunts until then by wire wound types and also added a ferrite bead at each resistor end. The shielded base wires were removed, the buffer capacitors mounted again.

The problem was gone. Not even gone, the blocks behaved perfect, on the scope as well as on speakers.
Obviously the perturbance was an internal very high frequency oscillation of the dumper stage, that showed just nasty demodulation products at the output and heated up the Power transistor dies.

above: dumper board with HF damped emitter resistors
below: the driver board gets the same measures

Speaker connection

There were four speaker connector holes, we used the all for 4mm sockets, but added a very interesting trick instead of a pure biwiring parallel switching:

one connector pair is directly wired to the power supply ground star and the amplifier output, the second red socket goes to the 0.22 Ohms decoupling output resistor on the power module, the second black socket is connected to the first by a 0.1 Ohms resistor, both resistors are 5W wire wound.

That gives the chance to chose from 4 different damping factors, depending how you plug in the speaker cable,
both right/ both left/right to left/left to right.

the speaker sockets can be used with four different damping factors

The resulting amplifiers

The first listen tests with these mono blocks - as ugly as their used cases may look were really convincing, British style understated, clean and fast but also with an emphasis I didn't expect. On the other hand side they presented a soundstage of a depth, I was not used to with the stereo version.

mono block inside overview

By the way an interesting detail of my mono variant is also, that there is just one class A stage and so the amplifier in idle mode heats up only on the side of the control module. You can feel the single side 4.5W idling dissipation after a while above the operation indicator LED. Of course with load all additional dissipation is equally distributed to both sides of the heatsink.

playing with dumping factor of about 65 into 8Ohm speakers

That superior behaviour immediately brings me to the idea to change the stereo version dumpers in a similar way, obviously there is some potential in cleaning up the current path, with that absolutely minimum trick the monos really began to "breathe" - maybe the stereo upgrade version will not have that grade of sheer power and control, but it will profit a lot anyway.
Let's try - my next step stereo version is to come...