Professional Documents
Culture Documents
Since 1993 I had intended to keep the "hobby" web site Valve Audio only but in 2002 I was handed yet another QUAD
transistor set–up consisting FM4 tuner ~ QUAD 34 pre–amplifier and QUAD 405 power amplifier ~ All were early "brown"
models with DIN plugs which normally attract little interest and would be difficult to sell for what they are really worth ~ I
decided to keep the set together and use it in my study office but first I wanted to experiment with a few modifications to the
405 to see if the rumours were true
Many QUAD 405 PCBs had the input op–amps mounted in sockets which of course encouraged the substitution of every
pin compatible OP–this and AD–that with the usual subjective analysis and discussions ~ Often with the most expensive or
lowest distortion devices "sounding" best ~ even if still very noisy ~ Due to the input topology simple op–amp substitution
will NOT fix the problem of high input stage noise as I shall try to explain below
After applying some of the web published mods or "upgrades" I decided to tackle the problem of the input stage noise which
further improved the amplifier both objectively and ~ surprisingly ~ subjectively ~ The QUAD 405 now sounded very
impressive and worthy of further work ~ It was treated to new larger value smoothing capacitors ~ Better output connectors
and gold phono inputs paralleled to the 4 pin DIN so that it could be readily connected into most systems allowing direct
comparison with other amplifiers { I know but we all do it }
There are a number of "modifications" to the QUAD 405 that can be found ~ Most are not modification or even upgrades
but simply replace the capacitors and resistors ~ The best information that I found on the web in 2002 was from Bernd
Ludwig and I suggest you read and digest his words along with mine ~ The information below is based around Bernd's work
except for the non inverting input stage design ~ the reposition of C8 and the input/output terminal changes ~ Other changes
like removing the current limiters and providing a better current source in place of Tr1 have been subsequently tried and are
mentioned on the 405–2 page
If you wish to repair or upgrade your QUAD 405 or just want to learn more about the 405 and 405–2 you can download the
QUAD 405 Service Data that I have recreated in pdf format ~ Also see my QUAD 405 evolution which is a set of pdf
schematics from the second (production) issue of QUADs M12333 405 up to some mods ~ Also you should read this article
by Peter Walker and Michael Albinson
NOTE: The following design changes were made to a QUAD 405 with M12565 iss.3 PCBs ~
other issues of the PCB will be similar but different ~ see examples
When the 405 was switched on it was clear that something was wrong ! ~ QUAD had even placed a bright red LED on the
front panel to indicate so and this was the first thing to be changed ~ It was replaced with a Yellow HP LED run at a lower
current of about 7mA by changing the series resistor to 6.8kΩ ~ The old Green and Red LEDs on the 34 pre–amp and FM4
tuner were also changed for Yellow and Red HP LEDs run at lower current
With extensive listening this QUAD405 sounded . . . reasonable ~ Just like a QUAD606 or QUAD306 at domestic listening
levels but with noticeably more noise at lower levels ~ There was also some hum on the left channel but other than that it
appeared to work fine and was given a further listening tests with friends and several pairs of speakers before being
transferred to the workbench
Testing with a Tektronix AA501 and HP3582A (I now use an HP35670A) showed the distortion at 30W to be 0.005% on the
right channel and 0.029% on the left while using the AA501 400Hz high pass filter to eliminate any hum from the
measurement ~ Without the 400Hz HP the left channel residual showed sidebands of 100Hz around the fundamental and
harmonics which explained the audible hum and contributed to the worse distortion measurement
When attempting to test the distortion below 1W noise and hum affected the readings with each channel measuring about
0.03% (400Hz–22kHz) ~ The hum on the left channel was cured by replacing C5 ~ A better solution would be to fit a better
current source in place of Tr1 but for now 100µF 16V tantalum capacitors that were to hand were fitted for C5 in both
channels ~ The hum was greatly reduced and the remaining noise was clearly from the input stage op–amps and ~ due to the
design topology ~ would not be improved by fitting "better" op–amps
C10 was changed for 100µF 100V electrolytics ~ NOT bipolar types ~ Some mods suggest bipolar capacitors all round but
this is crazy where the capacitor is clearly biased with a d.c. polarising voltage and at a.c. will never be subject to a reverse
voltage even on the lowest bass notes ~ Electrolytic bipolar capacitors by design have higher ESR and hysteresis than
equivalent rated standard electrolytics and when subjected to a permanent d.c. bias can be worse and may even suffer earlier
failure
After reading Bernd Ludwig's notes about D13 which was added
from PCBs M12565 issue 6 to increase the standing bias on the
dumpers and with my PCBs at issue 3 it looked like a good place to
start some mods
A slight improvement in THD was measured especially below 1W but the crossover spikes previously seen in the residual
trace output from my AA501 analyser were only slightly reduced ~ The fundamental energy of the spikes is way above the
audio frequency range but this is no excuse to ignore them
After reading all I could find about the QUAD 405 on the Web (2002) I turned to my supply of Wireless World magazines
and re–read the original articles about current dumping and the QUAD 405 from P. J. Walker and M. P. Albinson WW
December 1975 I had in mind something about the "crossover" spikes being related to current dumping bridge balance and
wanted to see if they could be further reduced or even eliminated
A better clue was found in follow up letters and the article from J. Vanderkooy and S. P. Lipshitz (WW June and July 1978)
But note that their tests were done on a very early QUAD405 with C9 and R19 fitted as were many other critiques like the
reports from Gordon J. King and Hugh Ford ~ Bear this in mind when reading magazine articles about the QUAD 405
If you compare the "Bridge" L2 ~ R38 ~ C8 and R20//R21 as implemented in the QUAD 405 and 405–2 with the QUAD
306 and QUAD606 . . . onward ~ It is obvious that compromises were made with the QUAD 405 ~ Probably due to
instability problems when applying "feedforward" to a transistor amplifier via a bridge in its "feedback" loop ~ If this is
taken into account much of the analysis (especially using the first production units) of the design can be ignored ~ When
implemented correctly the current dumping Maths works and it works well
I say the Maths works but 120pF x 500R does not exactly equal 3uH / 47R as it should ~ The problem is ~ as with most
problems solved by formulae ~ the initial assumptions are wrong ~ Although it is reasonable to assume the value of L2 is
3µH because it is marked on the schematic it is more likely to be 2.8uH
If you follow the QUAD Change History for the 405 and compare the schematics with the QUAD 306 and QUAD 606 etc.
you will see that C8 is not placed optimally in the Bridge circuit ~ Ideally C8 needs to be returned to the emitter of Tr2 ~ The
75Ω resistor R44 placed across L2 in later versions of the QUAD405 was a change intended to redress the bridge "balance"
but after applying the C8 move below I could get no better results at several output levels and frequencies with R44 fitted so
left it unfitted
A variable capacitor for C8 in its original position could not be used to reduce the crossover spikes and the best results were
found to be with a 1% 120pF capacitor for C8 either side of R17 ~ Low level crossover spikes are an indication of good
Bridge balance but the spikes can also be reduced by increasing the bias on Tr9 and Tr10 ~ However the QUAD405 would
then be more a class B amplifier which could suffer thermal runaway and without other changes may also be unstable
Looking at the change history you can see QUAD Experimented with the position of "bias" diodes D5 and D6 and later
added D13 to reduce the spikes but long before that Walker and Albinson had chosen an inductor L2 and capacitor C8
instead of resistors in the bridge and these components also reduce the crossover spikes and allow negative feedback to be
used around the class A stage
Even after adding the extra Diode D13 and 10Ω resistor to my M12565 iss.3 PCBs as the spikes due to dumper switching
were still present but after moving C8 to Tr2 emitter they are greatly reduced
With C8 in its 'correct' position in the bridge the distortion on both channels was about –87dB (0.0044%) at 30W and the
nasty spikes were not seen even when varying the load ~ The noise from the front end op–amp still prevented accurate
measurements at lower than 1W and I was convinced that the distortion and noise could be reduced by a change of design
around the input op–amp
The QUAD405 input op-amp forms part of a d.c. feedback loop to ensure 0V output at the speaker terminals ~
At d.c. the input op~amp is operating at maximum gain and its output is d.c. coupled through the current dumping stage to
the output terminals ~ The output terminal voltage is fed back to the inverting input of the op–amp via R4+R5 = 26.7kΩ to
be compared with the ground reference at its non–inverting input ~ As there is no corresponding high resistance at the non–
inverting input any input bias current will give a d.c. offset at the speaker terminals
If you feel you must change the op–amp consider the input bias current specification which generally gets worse as gain
bandwidth product (GBWP) increases ~ FET input devices like TLE2071, OP134 and OPA604 are all good for audio and
have very low input bias current required for the low output voltage offset in this design
The QUAD 405 op~amp output operates in Class A ~ There is little or no benefit fitting some modern "distortion
cancelling" op~amps that "do voodoo" around the zero crossing point ~ IC1 output never crosses zero and the rest of the
circuit is effectively Class A including the Dumpers as they are in the Bridge with the Class A stage
The "new kid on the block" the National LME49710 may be a replacement option with only 7nA bias current but it has a
high GBWP which could make it unstable and you will not get the published specification in the 405 circuit ~ Many
application circuits for these "new" op–amps show current sinks connected from output to ground so the device output does
not cross zero and measures better as would almost any op–amp
Beware of bipolar input devices like the NE5534 which is often substituted in the QUAD 405 ~ The NE5534 and the dual
NE5532 etc. are very good op–amps and were probably used to produce much of the music you listen too but they may give
an output offset of 150mV when used in the 405
But that's not how it is so let's put these components back one by one ~ C1 (680nF) and R3 (22kΩ) form a high pass filter at
about 10.6Hz (a time constant of 15ms) which prevents d.c. on the input upsetting the amplifier ~ R6 ~ R3 and R4 set the
stage gain at 15x but when C4 (47nF) is back in series with R6 (330kΩ) the frequency response of the input stage becomes
in theory flat down to a few mHz because the C4+R6 time constant is also 15ms and negates the effect of C1+R3 for as low
as the amplifier gain holds out
Putting C2 and R5 back into position will cause frequencies below their turnover point to be subject to feedback from the
output stage ~ the closed loop gain from input to output is 56.67 and the turnover point of C2+R5 is 0.3386Hz which
multiplied by the closed loop gain gives 19.19Hz ~ it appears that the aim was to provide a high pass filter response with a –
3dB point at about 20Hz which fits with general audio power amp design philosophy but due to the interaction with the other
time constants the effect is a high pass filter with a turnover at 13.95Hz and ~12dB/octave roll off
Note a 75Ω source impedance is assumed for the calculation and a gain of 15 for the op–amp stage allowing for the effect of
R3 and R4 ~ The effect of input impedance is minimal but if the gain of the op–amp stage is changed by altering R3, R4 or
R6 then C2 will have to be scaled proportionally to restore the original response ~ You will also have to adjust C1 when
changing R3 and C4 for R6
If you look at the Bernd Ludwig modification that started me off on this pilgrimage you will see that he reduces the gain by
making R6=100kΩ and restores the 15ms time constant by making C4=150nF, R3 and R4 are unchanged and so the gain is
reduced by 3.333 and C2 is adjusted to 33µF in proportion to the gain reduction restoring the overall response of the original
design
The QUAD 306 to 606 including the 500 series of "professional" amplifiers use an op–amp "outside" of the signal path for
d.c. offset control (The QUAD 909 is just a 606 that IAG make with the case on upside down and with a buss lane)
The QUAD405 input op-amp is a buffer for the passive low pass filter ~
If any amplifier is subjected to signals that it cannot handle it will produce Transient Intermodulation Distortion or TID ~
Many valve amplifier designs attempt to push more bass through the output transformer than it can handle with horrible
results ~ While bass is generally not a problem with d.c. coupled transistor designs exceeding the high frequency capability
can be
The current dumping technique requires that the unbiased transistors that do the dumping switch in and out to aid the low
power class A stage, the effect of this switching can be seen in the oscillograms above where fast spikes are produced by a
1kHz sine wave ~ As Peter Walker wrote in his Wireless World article of December 1975:
"We have said that the dumpers have to be sufficiently fast to come to the rescue of the class A amplifier to prevent its
overloading. Clearly they must be sufficiently fast to achieve this over the audio spectrum of the programme. There is,
however, nothing whatever to say that they must do so at frequencies outside the audio range provided that steps are taken in
the design of the whole amplifier to ensure that any such frequencies that may be present do not embarrass the amplifier
performance within the audio range. If the system is properly designed it is possible to use relatively slow devices ~
inherently more rugged than fast devices ~ and to show in theory and practice that they will never fail to come to the rescue
of the low powered amplifier to any programme. If, however, the criteria are thought to be response to step functions, square
waves and other factors not relevant to programme, then of course faster dumpers must be used commensurate with the rise
times involved."
To prevent high frequency signals affecting the dumper stage the QUAD 405 has a simple low pass filter between the input
stage and the current dumping stage, it consists of R12=3.3kΩ and C6=1nF the turnover frequency or point where the
response is –3dB is about 48kHz but at 20kHz it is about –0.7dB, I have changed R12=2.7kΩ which gives a –3dB response
to 58kHz and is –0.5dB at 20kHz ~ It hardly seams worth it and I was able to check the TID before and after the change, if
you are not sure then leave it as it is
The QUAD405 input op-amp provides a controlled drive for the power clipper ~
The "power clipper" provided to protect ESL57s from overload in this version of PCB consisted of two zener diodes put into
circuit with a link to ground ~ To function correctly the voltage limiter requires a point with sufficient voltage swing from a
known impedance R10 ~ For my input stage modification the limiter was not going to be used so the link pins were removed
and a hole left in the PCB was used to connect R3 to ground for the input modification as can be seen in the picture above
and as described next
Lowering the values for R4/R6 and scaling C4 will reduce the stage noise slightly more and also reduce any distortion
produced by IC1 input C changing with common mode voltage ~ providing the source impedance is low ~ If R4/R6 is
lowered too much the loading on the op–amp and the interaction with C2/R5 will have to be considered and the "magic" of
C1=C4 and R3=R6 would also be lost unless these are also changed
With IC1 configured as a non–inverting stage the signal to noise ratio referred to 30W in 22kHz BW is now better than
100dB (112dB ref. 100W) and 1kHz distortion is better than –90dB (0.0032%) indicating that the (max) 1Vrms. common
mode signal is not causing a problem with the non–inverting topology due to the op–amp ± inputs capacitance changing with
voltage (varactor distortion effect as described by Walt Jung)
Measurements at 1W were now more stable and meaningful but still dominated by 50Hz and 100Hz from the mains supply
~ The most likely source of this would be the supply capacitors or the Tr1 current source which relies on C5 for its power
supply rejection performance ~ Download the QUAD 405 input mod. schematic
As with all tweaks you have to consider at what level to stop ~ The final outcome will depend on the initial condition of the
equipment and the design changes and components used and of course your skill and patients
Try not to forget that the QUAD 405 is basically an excellent design and that some of the changes shown here and published
elsewhere by others could not have been reasonably be applied by a manufacturer at the time they were made ~ At least not
if they wanted to make a consistent QUALITY UNIT AMPLIFIER DOMESTIC at a profit
If the DIN socket had been removed the grounds of the phonos should still be connected directly to chassis to keep the
output voltage offset to minimum ~ See QUAD 405–2 modifications ~ With PCBs before M12368 iss.10 the output Zobel
network C12 and R39 was connected to the input ground side of R2 and connecting the input ground to chassis is required to
prevent speaker and transistor damaging oscillations
Some components that I didn't already have to hand were obtained from Farnell as some people have asked for them here
are the part numbers
Since 2002 experimenting with the QUAD 405 and 405–2 has continued along side my other audio experiments and
developments ~ In addition to the changes outlined above ~ lowering the open loop gain by bypassing Tr3 and fitting a "ring
of two" current source for Tr1 gives the "DCD–Mod3" which is a good level of modification for the QUAD405 or 405-2 ~
You can then fit higher rated output transistors and remove the current limiters
In 2004 I modified some M12368 PCBs replacing the Class–A load resistors R30/R31 with a 50mA current sink and the
output transistors for complimentary Darlington devices to make the DCD–Mod4 ~ similar to the NET audio Mk3 which
followed after its publication ~ see QUAD 405–2 modifications and QUAD 405 Evolution and Example Mods for additional
information about the 405 and 405–2
"Soon you'll throw down your disguise ~ We'll see behind those bright eyes"