Quad 405-2 Upgrades

Quad 405-2 upgrades

• Introduction
• Hardware
• Power Supply
• Capacitor Replacement
• Input Stage
• Feedback Bridge
• Gain Reduction
• Results
• Parts List (Bill of Materials)

Introduction

This project stems from a piece of extraordinary luck (or misfortune) at a local boot fair in the summer of 2011.

For years I've designed, built & repaired amplifiers, both solid & hollow state, however, I've never owned a Quad. Maybe its just being fussy, but I've always been a contrarian - e.g. even though my career has been in cutting-edge technologies I've never owned anything by Apple. Not, you understand because I dislike or disapprove of them - it's just that when I see a bandwagon, I have this irresistible urge to set fire to it... Everyone else had Quads, ergo, I didn't.

So, I was strolling around the boot fair, looking at the piles of mostly complete tat, when something caught my eye...

"How much?" I ventured. "Ten quid mate". "I'll have it".

"It" was a stack consisting of a pristine Quad 405-2 power amp, Model 34 pre-amp & Model FM4 tuner. Bargain. 10 quid, by the way, is about USD 15. Total real value of the units? Well in excess of 700 quid/USD 1000...

Now the problems start - looking at the serial number, this amp was probably built in the early 1980s so, at the very least, any electrolytics will need replacing. And then there's the mods. Oh, yes. The mods.

As I'm an engineer, I believe in keeping things simple and to the point. If there's a mod that's cost-effective & actually makes an audible improvement, I'll do it. If its a esoteric maybe bit of audio-phoolery, such as solid silver or "unidirectional" OFC 99.99999% pure wire, forget it.

My objective is to bring the amp up to contemporary specifications with regard to noise & distortion etc., but under no circumstances to compromise the "Quad" sound to the extent that what we have is essentially a new amplifier in a Quad box.

There are many pages of stuff out there, but really it boils down to two people - Bernd Ludwig & Keith Snook. Both approaches have a great deal in common, and both acknowledge the others' input, so I decided to make a selection of those that Bernd & Keith propose, plus one or two of my own. Bernd doesn't seem to have a website, but he very kindly sent me his final (last ever, he says!) version (V14, July 2010) of his mods which I have made available here.

At this point I should mention that my particular 405-2 has a board type of M12565-6 and a serial number in the 71500s. The board revision is important - my boards are some of the latest that were made and thus already have some of the modifications that are recommended by the various authorities out there - Quad have already done them. However, there is a lot more to do...

Summary of intended upgrades

• Add decent phono input
• Add decent LS posts
• Split PSU so each channel has its own supply- new bridges & capacitors
• Lower input sensitivity to 1.5V (was 0.5V)
• Adjust feedback bridge to lower crossover distortion
• Replace internal wiring with something a bit more substantial
• Replace PSU caps with BHC ALS30s - smaller and much lower ESR & higher surge current
• Replace opamp with modern low-noise variant - OPA604 or OPA134
• Change opamp power to +/- 15V with regulated +ve rail
• Replace all electrolytics with direct equivalents or new values if functionality changing
• Add decoupling caps all over the place (see text)

Hardware

• Strip down entire chassis
• Remove all internal wiring except mains input
• Remove heatsinks and clean
• When refitting heatsinks, use Dow Corning 340 heat sink compound
• Fit insulated input RCA/phono sockets - red & black
• Remove old loudspeaker sockets & fit insulated new ones
• Remove old wire to board connectors and fit new silver plated ones
• Drill and tap holes for new bridges, capacitor clamps & earth post
• Cable upped to 0.75mm tri-rated (20 AWG)

Power Supply

Note that there is a maximum of about 90mm height available to fit the replacement main smoothing capacitors (C13,C13',C14,C14'), so do allow for any fixing etc., especially if using screw terminals. I wanted to use hex bolts (they do look good!), but they are too tall.

There are those who might say I've gone over the top on the PSU. My argument is that its the foundation on which the whole amplifier depends, a fundamental part, and regardless of PSSR etc. of different bits of the schematic, it actually costs relatively little to do it right. People may complain about the number of decoupling capacitors etc. - I'll argue that they are cheap and it's very worthwhile. If a touch obsessive.

• Supply split into 2 mono supplies rather than original single stereo supply
• PSU capacitors (C13,C14) changed to ALS30A103DE063 (10,000µF, 63V, 11A ripple, 11,000 hour life) with MAL804324031E3 clamps (35mm plastic, no flange) - Add C13' & C14' for second channel
• Add Cx8, Cx9, Cx10, Cx11 100nF @ 250V MKT across each of the main smoothing capacitors (C13,C14,C13',C14') (4 total)
• Add C15',C16' ECA2AAM221X (220µF, 100V) in parallel with C15 & C16 respectively
• Add Cx4, Cx5, Cx6, Cx7 470nF @ 250V MKT between each secondary and their respective centre tap (4 total)
• Change D7 bridge to CM2502 (25A, 200V) - Add D7' for second channel. I punched out the M5 inserts used for the old bridge and repositioned them for the new bridges (one per bridge) in the same vertical line.
• Add Cx12-Cx19 (100nF, 100V MKT) in parallel with each diode in D7 & D7', i.e. 4 per bridge
• Add 330nF @ 400V X2 capacitor between live & neutral on back of inlet IEC connector
• Add M5 Earth post. Note that this cannot use an M5 threaded insert as the heat sink is flush with the panel so there is no space for the protruding insert. i.e. the hole for the earth post must be drilled and tapped with no insert. Note also that the black anodizing must be removed from around this fixing so that good connection is made with the heat sink - about 1cm in diameter should be enough.

With the mods in place, the PSU schematic should now look like this (click to enlarge):

Everything fits quite nicely when installed. You can see the new bridge rectifiers on the heatsink at the top with the new earth post just to their left; the 4 new smoothing capacitors are bolted down and all the decoupling capacitors are in place:

Capacitor Replacement

See Power Supply section for PSU-related caps.

• Change all audio electrolytics to Panasonic ECA series
• C5 to ECA1VAM101X (100µF, 35V)
• C10 to ECA1JAM101X (100µF, 63V)
• C17 to Multicomp NP35V106M5X11 (10µF, 35V, non-polarised)
• C18, C19 to ECA1JAM470X (47µF, 63V)

Input Stage

• Remove IC1 and replace with gold-plated, turned-pin, socket
• Replace IC1 with OPA134PAG4 or OPA604APG4 (yes, one is "xxPAG4" and the other is "xxAPG4". Odd).
  Absolutely no point in using a faster or lower-noise opamp here - a faster opamp may instead cause HF instability and the noise floor is dominated by other factors. Some people use the AD797 (about 3 times the cost of an OPA134), an AD843 (about 4 times cost) and others the OPA627 (over TEN times the price! Madness!) - I highly doubt anyone on the planet could distinguish between them - I certainly can't measure the difference using several USD 1000s of test kit. Why spend extra for no effect, unless you want to venture into "audiophool" territory? Get a couple of plastic pyramids or some dodgy-looking crystals and leave them neatly arranged on the amp. Just as likely to have the same result!
• Replace D1 & D2 with 1N4744A (Zener, 15V, 1.3W)
• Add Tx1 - MPSA42 to regulate +ve supply - stops "thump" at power-off
• Add Cx3 - Decouple D1 (220µF, 35V)
• Replace R7, R8 with 2K7, 2W
• Add Cx1 & Cx2 - 2 x 100nF MKT on solder side of IC1 - between pins 7 & 3 and pins 4 & 3

With both the input & gain reduction mods in place, the final input stage should now look like this (click to enlarge):

Feedback Bridge

The Quad 405 is actually two amplifiers in one - a relatively crude but powerful "current dumping" amplifier that operates with no quiescent current (and thus with no complicated setup or configuration) and which is responsible for delivering the bulk of the output power, and a second low power very precise class A amplifier that is fed the error signal (the difference between the current dumping output and the input signal) and which thus corrects the output - the overall performance of the whole amplifier is thus determined completely by the accuracy with which the class A amplifier corrects the output signal, i.e. its quality and the quality of the error signal. The error signal is derived using what Quad call the "bridge" and which is described below.

In order to maintain stability, a compensation capacitor, C8, is included. Almost every decent amplifier has this in one form or another - its purpose is to reduce the open-loop gain of the driver stage at higher frequencies by -6dB/octave (-20dB/decade) and thus to prevent high frequency oscillation. This gives a nice stable amplifier, but at the cost of overall performance. As the two feedback paths, via C8 and via R20||R21, are not balanced in the frequency domain, it is nigh on impossible to attain decent distortion figures. Quad's solution to this was to introduce the "bridge" - C8, R20||R21, L2 & R38 - see the simplified schematic below:

Most notes on the Quad 405 just skip over this bit and don't mention what it actually is. The 405 uses an implementation of a Maxwell-Wien Bridge - normally the input signal would be between points FB and OP and balance occurs when the voltage and phase at points A and B are identical, i.e. when Z_L2 / Z_R20||R21 = Z_R38 / Z_C8. Note that Z_Cx = (2πfC)^-1 and Z_Lx = 2πfL.

Rearranging the equation and assuming that Z_R = R, we get balance when R38 * R20||R21 = Z_C8 * Z_L2 = (2πfC8)^-1 * 2πfL2 = L2 / C8. The "2πf" components cancel out leaving a frequency independent equation. This frequency-independence is a key characteristic of the Maxwell-Wien configuration and makes it very attractive for use in situations like this.

So, if the bridge is balanced, C8's effects can be compensated for. Substituting in the actual component values, we get 47 * 5e2 should equal 3e-6 / 120e-12. The LHS equals 2.35e4 and the RHS is 2.5e4, so even assuming perfect components, its never exactly in balance (depending how you look at it, 6% or 7% out) - if L2 were exactly 2.82µH and everything else was perfect, balance would be achieved. R20||R21 should be left alone as they control the overall gain and L2 is tricky to alter, but C2 and R38 could be adjusted by adding parallel components should you desire.

Quad take advantage of the above equations to drive the bridge slightly differently - by driving the A, B, and OP nodes, the FB node will reflect the OP signal minus any error (A-B). i.e. a composite negative feedback signal that is added to the class A drive and allows it to correct any errors in the output signal (caused by the current dumpers or any other aspect of the system).

Some analysis has been done on this it seems that leaving one end of C8 connected to R38 and moving the other to the emitter of Tr2 (as is done in the Quad 606) removes Tr2 from the bridge, moves closer to the current dumping ideal and improves crossover distortion. Ludwig states that if Tr2 is fast enough, it shouldn't effect the feedback path and may improve stability; Snook does the C8 move as part of a larger change involving sorting out the class A amplifier and Quad changed this as noted for the 606 - It's a simple mod, so that's what we'll do as well.

It might be worth pointing out that by default, C8 is a disc ceramic (unknown tolerance), R20 & R21 are 2%, R38 is 5% and I have no idea of the tolerance of L2. Certainly, replacing C8 by maybe a 1% polystyrene, R38 by a 50R 1% and checking the value of L2 might be idea worth chasing...

Gain Reduction

The ratio of R6 to R3 defines the input stage gain - Originally this was 330K/22K = x15 (+23.5dB) - now changed to 100K/22K = x4.6 (+13.2dB)

The class-A stage has a gain of 3.78 - this is determined by the potential divider formed by (R20||R21) & R16, i.e. (180+500)/180 = 3.78 so the overall gain was originally 15x3.78 = 56.7 (+35dB) and after reduction is 4.6x3.78 = 17.39 (+24.8dB).

Reducing gain by a factor of approximately 3 (15/4.6) reduces input sensitivity for full output from 0.5V RMS to 1.5V RMS which is in keeping with most modern equipment

Note that by reducing the front-end gain to 4.6 rather than 5 reduces the power output (into 8 ohms) from 100WRMS to a theoretical 85WRMS. It might seem a bigger drop than you would expect, but remember that power depends on the square of the voltage gain - to get a full 100WRMS you will actually need 1.63VRMS of input.

Leave inverting for now - Snook has a mod that changes IC1's configuration to non-inverting which reduces the noise levels still further - this is a mod I will do later.

C1 is used to block any DC on the input affecting the DC balance of the rest of the amplifier. Together with R3, C1 creates a high pass filter with a 3dB point of 10.6Hz. By making the feedback loop of IC1 have the same time constant (15mS), we use C4 & R6 to exactly compensate for the LF rolloff of C1 & R3 resulting in a flat response down below 1Hz.

• So, if R6 is changed to 100K as calculated above we must change C4 to keep the original time constant (15mS).
• C4 to 150nF (needed to maintain R6,C4 time constant at 15mS - same as R3,C1)
• C2 to ECA1AAD330X (33µF, 10V) - scaled by gain change - 4.6/15 x 100µF = approx 33µF
• Add C2' (100nF polypropylene) across C2 - compensate for RF impedance of C2

Results

Theory is one thing. Practical realisation is quite another! I used a fairly sophisticated test bed to quantify results from these mods - a Tektronix 2465B to monitor output and two audio analyzers - a GPIB-controlled HP 8903B driven by modified versions of Pete Millet's VEE software and a Tektronix AA501A mod WQ driven by an SG505 mod WQ & an SG505 mod WR (both ultra-low distortion sine wave oscillators). A non-inductive 100W dumy load was used, together with a custom test harness to ease swapping things around.

Completed amplifier:

After setup, gain was slightly lower than predicted at around x16.5 (+24.35dB vs expected +24.8dB), but distortion was at the floor of my analysers, i.e. around the 0.0025% level from 1W to 95W and from 10Hz to 20kHz. Error in the balance between the two channels was a very pleasing 0.8% and both channels exhibited a DC offset that was < 1mV under all load conditions.

So, how does it sound? Using my standard setup (KEF QX Fives), for a start, it's now very quiet with no audible hiss or hum. Bass is punchy, the sound stage is wide and well defined - in short, its a whole different amplifier, but in a very "Quad" way. More listening tests will come as I get time, but I'm not very good at subjective analysis - I just know what I like, and whereas I used to be indifferent about this amp, now I really like it. It may even replace my trusty Perreaux 2150B as the amp of choice...

In the photo below, the white box is the test harness, the HP 8903B is at the top left (looking rather boring as when under GPIB control the front panel does little), the dummy load is the largish black lump and there is a temperature probe and fan on the heat sink.

Test rig using HP 8903B:

Test rig using Tektronix AA501A & SG505 & Advantest spectrum analyser:

Underside & topside of modified and cleaned boards:

Bill of Materials

The following parts list has Farnell and RS part numbers as I live in the UK, but most parts should be available from DigiKey or equivalent suppliers.

Note that the DigiKey part numbers are not always the same component, but they represent a good compromise that is available from that supplier.