Quad 303 triple cascade


2007-12-19 8:53 pm
Dear all
in long gone days when I was a student I owned for a short time a Quad 303. I had to sell it when I went overseas to another university. Now that I am retired and still have the schematics
I want to do something entirely different from what I did during my active occupation, computer theory , mathematics and that like, so why not build an amp like the old Quad?
I do still remember it had a wonderful musical sound.
But apparently the special triple cascade of the Quad output
was a particular charcteristic of only this amp and was not utilized in other amps. Although it should provide excellent quiescent current stability 'online' so to speak as the power transistors junction temperature is the input variable of the feedback.
Does anyone know major disadvantages of the Quad concept?

Dieter F.
Interesting and coincidental.

I had a Quad 303 long time ago and enjoyed it very much as the sound
was very smooth and pleasant. For some reason, I remember, it had
a problem at the time ( I'm talking about 1980's) and I decide do scrap it, I didn't see no reason of keeping it around because of the output
cap technology.
I don't know too much about the 303 topology, what I know is the beautiful
sound it produced. After all these years after founding one of the
boards I scraped in one of my boxes I kind of thought "why don't I revive this amplifier that gave so many hours of pleasure?", this was
my motivation of doing what I did: I made another board from the existent one, copying the layout and constructing it from scratch also following the layout and schematic.
Fortunately I still had kept the original transistors (38495 and 38496) so it wasn't too difficult to reproduce the original. The rest
of the transistors was a matter of some criterious substitution.

Coincidentally, because I just finished the whole amplifier a couple of days ago, I improved the power supply with beefer capacitors
and made the output caps much bigger.

The result? I think I've got to make an even better 303 that it was before,
I can't tell you how much pleasure it gave me after I turned it on and started playing my favorit CD's.

I've never seen too many posts about the 303, I know there are
a lot of mods that can be done on this amp, but for now I'm going to enjoy the sound until I I think it's time to tinker with it.

That's why I thought there was a coincidence of finding a post about this amplifier.
I have built Quad-303-alikes with (epi-based ST) 2N3055, BD139/BD140's and although I was not able to compare an original the newer epi-base transistors allow higher frequency response (at least 2x from 30 kHz (I think?) to > 70 kHz) than the original, and with a some mods. it seems possible to reduce the susceptibility to transient distortion (slew rate if you prefer).

Maplin (still) sell TO-39 case transistors 2N3019, 2N4033 which were ST and MOtorola similar to RCA's 40361/40362 which others seem to have used at the time. Don't know how these compare (yet) with the BD139/BD140 but they may replace old Quad transistors. Rated at 80V.

Quad 303/33

I like these amps i have owned a couple of them, and repaired a few. mainly capacitors drying up.
features that are good or "odd" about them, depends on your veiwpoint. still not bad for 40 years old.

0.5V input level @100kR (nice with a 10k vol control "passive" pre)

Stabilised power supply

single supply (output capacitor) fit a good one

Inverting (i think ?)

Use MJ15003 for output

Use 2n3773 for regulator

links to info etc http://www.net-audio.co.uk/quad303caps.html

and http://www.richardbrice.net/quad33&303.htm

hope this is of interest


2007-12-19 8:53 pm
There are a couple of variables to consider why the old 303 produced such a musical sound.
The foremost important imo is the spectrum of distortions.
If this is like the spectrum of harmonics which a musical instrument
produces with acoustical output power as a variable then the amp's distortion were basically inaudible as distortion but where subjectively interpreted as a property of the musical instrument, either pleasant, or unpleasant. That would require crossover distortion must be a minimum.
Insofar the property of the Quad triple cascade , to keep the quiescent current constant , could play an important role and may even be improved when the driver stage transistors are kept at a constant junction temperature by means of a sort of thermostatic device.

I will post some design ideas soon

Dieter F.
Hi Andrew T

The Quad 303 was a quasi-complementary with triplets.
The first I made used ZTX304/40362/2N3055 for the upper and ZTX504/40361/2N3055 for the lower. At that time a 2N3055 was a real 2N3055 (1 MHz).

Another version used MPSa06/MPSa56 small signal first stages. I've also made some with BD139/BD140 in first and second stages recently, to check the performance with the newer 2N3055's.

The newer 2N3055 is an epi base and not dissimilar to MJ15003. (2.5 MHz min). Not only does this make an audible improvement it is clearly measurable as having double the bandwidth (when the capacitors are reduced as well).

Hi Bigpanda

I don't think that any changes are needed to use the BD139/Bd140 in old 303 circuits, except that the pin-leads are different from the old TO5/TO39 cans, and therefore some track hacking is needed, if you use these in an original board.

Probably 2N5551/2N5401 can be used for the smaller signal stages, but I haven't tried these.

ST still sell (through Maplin) the 2N3019 and 2N4033 which might replace the old devices directly (80V rating, should replace 2N2102/40361 and 2N4036/40362), if anyone needs TO-39 replacements.

I'm working on the updated circuit diagram and hope to post later.

Tks John,

I think this is the first time I have more complete info on the 303 alternatives since I visited this forum. PCB is no problem. I had it made for to126 device but I just don't know what to use. The only thing that is missing is the BC109. But like you said, I can try using the 5551/5401. There seems to be no excuse for my 'waiting' now.

Tks again


2007-12-19 8:53 pm

The amp which features two design characteristics of the Quad 303 , the triple cascade topology, and inverting, is now designed.
A particular restriction is that the two monoblocks are supposed to sit right behind bookshelf speakers thus the mechanical dimensions are restricted, and thus the size of the heat sinks and their thermal resistance. The loudspeaker features a min. impedance of 4 Ohms at 200/sec and that is just the frequency where the maximal volume appears in most classical pieces of music . Hence the supply voltage had to be reduced and fortunately ebay auctioned 4 toroidal transformers with 25 V and 50 VA each. ebay also auctioned a bag full of Siemens 15000 µF/50 Volt and 6800µF/63 Volts electrolytics thus this case was solved.
The power transistors SC2922 SA 1216 are directly, without insulation, mounted on the heatsink and so are the 1 Ohms resistors. This spares about 0.3°C/Watts thermal resistance
and keeps the PCB off high current.
Transistors T2,T8, T11 are mounted on a 4 by 1 cm² copper
between two thermally isolating polystyrene layers and are kept at a constant temperature of 70°C by an additional power transistor and a thermostat circuit. Since the temperature of the junctions of these transistors determine the quiescent current
- it is independent of the junction temp. of the power transistors-
the conditions are stable.
The CMRR of a current feedback inverting amp is poor thus the
driver must be fed from regulated supplies which have to put out just 25 mA each.
The input impedance is 1 kOhms thus a preamp is required which can drive this low impedance. Such preamp is already designed.

Transistors: BC449/ BC 450 are Motorola with 100 V BVCE and fT 150 Mhz and hfe 80 min. BD389/ BD 390 are also Motorola which are a bit more rugged than the faster SA 1112 SC 2592 and feature a lower Vsat.

The objective, or rather, subjective is to have an amp which has a feeling for music. It is known that the absolute value of THD is not a measure for the audible quality.
It is rather so that the respective energy of harmonics must
decrease in a particular mathematical series and push pull triode amps for whatever reason produce just such a distribution of harmonics, where the THD is 10% but subjectively they sound better than an amp with THD 0.5%
Happy new year!
Dieter F.


2007-12-19 8:53 pm
I forgot to attach the schematics. Here it is.


  • quad303 new2.jpg
    quad303 new2.jpg
    52.5 KB · Views: 1,919
HI Bigpanda

The difficulties with 2N3440/2N5415 I think is that they have some frequency limitation at Vce < 10V. They also have a gain roll-off after about 100 mA (if I recall correctly) and this can affect drive. They are only about 30 MHz anyway, so I'd prefer the higher frequency BD139/140 drivers. These have a gain hold-up to at least 200 mA.

I've used these for the first of the triple stages, as well as the second. One reason is that if you use these the stabiliser can be a BD139 as well and they all fit onto a metal angle panel to keep them at the same temp.

I'll post a photo of my layout later.

Here's the diagram for my Mod Quad. Some points about this: the faster output transistors (epi 2N3055's) allow double the bandwidth of the original. Simulations show that the original has a BW of about 70 kHz. This one has a BW about 150 kHz.

Overall, the simulated distortion at 20 kHz is slightly higher: comparing simulations, the Quad original has about 0.05% H2, but about 0.07 H3. The Mod has about 0.1% H2 but .05% H3. And this is the point - crossover distortion components are monotonically reducing, so appears to be improved.

The 10 nF roll-off capacitor across the base of the second stage transistors has been removed. Since the first three transistors are inverting, the Miller capacitor can be placed on the input stage instead of the third stage. This prevents the first stage overloading even on a square-wave transient. The new Miller capacitor uses 220 pF. So, now the original 100 pF Miller capacitor can be unwired from the third stage transistor base, and instead is grounded to act as a "sprog stopper" for the output stage.

I've kept diodes as the protection scheme, but other ideas may be better. The upper output stage triplet uses a compensating diode, to make this more complementary-like. This also reduces the saturation voltage of the upper output stage as well, as the original 10 ohms has quite a high voltage drop at high currents. LIkewise, I've removed the lower 10 ohms for the same reason.

I've added a small resistor and rather bigger capacitor in the power supply line mainly to filter the line from R11 which directly introduces supply noise into the amp. It is possible to run the amp. from an unreg. supply (if wanted) but if so, I recommend large capacitors (10mF) and a fairly heavy currenttransformer (I tested my Mod on a 50V 6A unit giving 72V quiescent.) so that there is not much voltage droop between quiescent and full load. I'm sure a stab. PSU will be better if anyone prefers to add one.

Anyway, this circuit is a suggestion - anyone is welcome to comment!



  • quadmod.gif
    95.5 KB · Views: 1,767


2007-12-19 8:53 pm
as a very simple trick if you have a sine wave generator ( or a program for a sound card ) and a two channel scope measure the difference of the input - the wave generator - and the signal at the base of Tr5 or Tr6 to determine the correct quiescent current
and to see the influence of the 10 Ohms resistor.

Dieter F.
Hi Dieter

One trick I have used to use to set the quiescent current of the (original) Q303 was to put the oscilloscope probe onto the VAS transistor. This is inside the feedback loop, and has much higher distortion than the output.

At low settings, and expanded origin showed distinct slope changes, for a sinewave input, around the crossover region. At low bias, the slope increased - and at higher bias, the slope changes to being shallower than the original sinewave. This illustrates the difference between crossover distortion and gm doubling.

(Incidentally, this is one reason to use a current source instead of bootstrap. The bootstrap resistor attached to the base of the upper stage always forces a resistive loading on the two bases. No setting can completely equate the two (exponential) inputs, but a constant current source cannot increase gm. Except, of course, the local Miller feedback or other capacitors reduce the impedance at high frequencies, but by then one hopes that the distortion components are inaudible.)

I found (again on the original) that you could set the bias like this at 1 kHz - but then it was slightly different at 20 kHz. I set the Iq to give lowest visible distortion at 20 kHz. I presume this was due to using "slow" 2N2055's.

I have not tried this on the Mod, I have to say, but I doubt that it will have the same crossover problems as the original. I found that symmetry could be improved either by using a diode (as shown) or by simply re-allocating the lower 1kohm resistor between the base and emitter of the lower NPN driver, instead of to ground.



2007-12-19 8:53 pm
Hi John
yes I had some good reasons to change the driver completely.
First I have lead lag compensation almost independent of each other and next no miller capacitors at all. However, with a software for a soundcard I am not able to measure distortion - such that the measure is repeatable - except varying noise floor.
This may be due to the very fast Sanken power transistors or whatever I am not an expert in taking exact measures.
Although in theory to put an emitter follower between the base grounded Vas and the complimentary triple cascades in practice
it yielded only problems as the amplification of the Vas became too high and I had to decrease the bandwidth significantly. This may be an effect of improper PCB layout. Nevertheless so far I am
very satisfied with the sound after all subjectively its better than
that of a high grade Mosfet amp. Emotionally better all else is
insignificant because I cannot hear numbers as accurate they may appear.


2007-12-19 8:53 pm
Heat source

Apparently a heat source is the opposite of a heat sink :)
As the amp has to have a standby power supply I put transistors T2,T11,T8,T4,T7,T5, and T6 in a kind of oven heated by a small power transistor to 75°C . The accuracy is roughly 2°C . As the devices are always kept constant at this temperature as the heater is fed by the standby power supply it is not a green design but it is very stable.
The "oven" is simply a copper tin strip thermally insulated on either side with 10 mm polysterene layers.
As the sum of power dissipation of all transistors is around 1 watt
the heating must be 3 watts. A next step could be a Peltier element to compensate the temperature increase by power transistor heatsinks .

Dieter F.
Only truly experienced (and unfortunately very few) developers (like Quad) achieve a design of such triple stages without any oscillation under any load condition.

I recall, that I have for checking a diy amp device, that uses also such - slightly different topology:
The name was "Equa" from German magazine "Elektor" December 1972, page 1216
This amp was an unwanted oscillator (even by p-Spice simulation and in real life only to test by current limited power supply without destroy the output devices).
Only after remove one stage in the output buffers I get satisfy operating.
About this address perhaps there is a complete Equa schematic available:
Audioclub Oost Brabant, clubavonden

I want to have basic articles about such three stage super ß quasi complementary stages. Who knows more about this?

I know too few English keywords therefore. I recall, that I have read some years ago a JAES article. But I don't know if it was the main topic.
Last edited: