Quad 33 and FM3

[ppppenguin]

Craig, thanks for the info on modding the 33. i think I'll keep it original for now.

[Synchrodyne]

That improves headroom and distortion (see douglas self's site on RIAA). It also includes replacement electrolytics for all other boards, lower noise transistors and a number of resistors to adapt the circuits to 16V, and to optionally change the gain structure.

Re pickup input headroom, as I recall, the Quad 33 arrived on the scene before that became a “hot” topic in the hi-fi magazines of the era.  Then when it emerged, the 33 was sometimes criticized for its apparently modest headroom (40 mV) in the M1 disc input position.  Rarely mentioned was that the M2 position, which probably suited a wide range of cartridges, offered 120 mV.  The pundits appeared to want maximum sensitivity (2 mV or better) and maximum headroom in the same input as a measure of “goodness”, with no mention of adjusting input sensitivity to suit the actual input level and by so doing allowing sane volume control settings for normal listening.  I suppose though that also emerging at the time was the Japan Inc. approach of having relatively sensitive inputs that required very low volume control settings for normal listening, perhaps to convey the impression of a large power reserve to non-technical users.

The Quad 33 disc input stage looked, to a first approximation, to be a silicon version of the Dinsdale 1965 (1) two transistor design.  It retained the input bootstrapping that Dinsdale had used, but evidently in much “tighter” form, I’d guess providing a native input impedance of several megohms.  Thus, the disc input impedance was essentially defined by the input resistor, 68k for M1 and M2.  Quad later claimed that the input was purely resistive within ±5 degrees over the audio range.  One may ask why did not Quad adopt the three-transistor design offered by Bailey in 1966 (2), in which an emitter follower was added within the feedback loop to reduce RIAA feedback network loading on the second transistor gain stage.  One possible answer is that the Quad 33 was developed at a time when device cost was non-trivial, and so device count was a major consideration.  Thus, the extra transistor that might have been used in the disc input stage was rather required for the following multipurpose buffer stage, which provided a fixed and highish impedance load for the disc input amplifier, the same for the pair of radio inputs (bearing in mind that in 1967, it could be expected that the Quad 33 might be used with older radio tuners that did not have buffered outputs) and which also served as a tape output buffer.  That the tape output buffer was in the main signal path and not a separate side-chain was another manifestation of the need for device-count economy.

Something that the Quad 33 did have was a passive RC roll-off circuit following the disc input amplifier that compensated for the fact that a series feedback circuit has an inherent error in the HF end of the RIAA curve.  Baxandall (3) reported that this had a turnover point of 32 kHz.  I think that it could not have been precisely right for both the M1 (34 dB) and M2 (25 dB) gain settings, but presumably the error either way was negligible.  That refinement was not often found; for example, H.P. Walker did not use it in his 1973 design (4), the RIAA curve for which did show the error.  But Self (5) did use it in his 1976 version of the Bailey three-transistor circuit, to which he added a bootstrapped output load resistor for the second stage.

The ceramic cartridge input C1 was an early example of an approach later advocated by Burrows (6), and also used by H.P. Walker.  It assumed that such cartridges were self-equalized for the RIAA curve (true for all but about one by then) and so did not require such equalization in the disc input amplifier on this front.  But the 100k input impedance differentiated the cartridge output with a turnover within the audio range, and this was corrected by complementary integration done via the amplifier feedback loop.  This process was dependent upon cartridge self-capacitance, and the Quad 33 C1 input was said to be suitable for cartridges with self-capacitances in the range 450 to 900 pF.  I’d guess that it was exactly right for something like the geometric mean of that range, 636 pF.  (H.P. Walker assumed 600 pF for his 1971 design.)

That was a sea-change for Quad, in that the earlier QCII and 22 control units did differently.  With those it was assumed that higher quality ceramic (and crystal) cartridges were not self-equalized (probably reasonably true when the QCII was released).  Thus, they were subject to the same set of switched equalization curves as used for magnetic cartridges, but those curves were effectively rotated pi/2 to suit the amplitude responsive outputs of the piezo cartridges.

In the late 1970s Quad offered a modified disc input board with the M1 position matched to the requirements of the Shure V15 Mk III cartridge.  Input sensitivity was 3.8 mV (instead of 2 mV) and input resistance was 47k (instead of 68k), and there was a non-trivial added parallel capacitance, whose value I don’t recall.  This was featured at one of the late 1970s London hi-fi shows.  In a nearby room Shure was showing its V15 III, and using a Quad 33/303 combination.  I asked whether Shure was using the modified disc input board; the answer, given in something of a Gallic shrug manner, was no, although they were aware of it!  Perhaps Quad had been overzealous in this matter.  But then circa 1970, B&W had offered a passive equalizer that plugged in under the SME arm support to exactly match the Shure V15 Type II (unimproved version, I think) to the Quad 33 M1 input (7).

The incorporation of a tape output buffer was new for Quad.  The QCII and Quad 22 had used unbuffered tape outputs (as found on most, but not all valve equipment) that needed to look into an impedance of 500k or higher.  In fact the initial version of the QCII did not have tape input and output facilities; they were an early amendment.  Impedance matching was going to be an issue early in the solid-state era when new equipment was interconnected with older valve units.  This was more so for the tape recorder makers.  With valves, the provision of high input impedances, say 1M or more, was a routine matter, but it required special circuitry in the early transistor days, if it was catered for at all.  For example, the Revox A77 had an auxiliary input with 1 M impedance, achieved by bootstrapping the input of the signal input amplifier.  Ferrograph used a fet source-follower for the 2M2 line input of its Series 7, and as far as I know Uher used an input attenuator (divider) with a 1 M upper leg.

Back to the 33, Quad used an unusual approach to matching its tape output to recorder DIN inputs.  Instead of the usual series resistor to approximate constant current characteristics, it offered attenuated, very low impedance outputs (nominally 3.7 and 20 mV) that provided signal levels appropriate for tape recorder DIN inputs.  To provide one example, the Revox A77 DIN input (labelled “radio”) had a sensitivity of 2.5 mV at 33k, so the Quad 33 3.7 mV output would have been a reasonable match to it.  (Although in practice I imagine that the Quad 33 100 mV tape output would more likely have been connected to the A77 35 mV, 1 M auxiliary input.)
 
Next in the Quad 33 circuit came the tape monitor switching, with tape input attenuator network, followed by the volume control and a buffer stage that comprised an emitter follower with bootstrapped input.  The tone control circuit was basically of the Baxandall type, but using two transistors instead of the customary one.  The second transistor was an emitter follower that bootstrapped the first transistor load, thus allowing some gain, about 14 dB.  It also served as the output buffer.  That meant that the passive filter control was beyond the output buffer, but again, device count economy requirements may have precluded the addition of a second buffer stage at the very end of the chain.  (Not at the time, but later it might have been seen as desirable practice to have included an inverting buffer at the end of the chain to ensure that the control unit was not inverting overall.  But then the Quad 303 was inverting.)
 
The Quad 33 tone control may have been a very early example of a two-transistor Baxandall circuit, and one that was largely overlooked in the “history”, as it were.  Quilter published a similar two-transistor circuit in 1971 (8), essentially presenting it as an improvement over the Bailey single-transistor circuit, but he made no mention of the Quad 33 case.  H.P Walker also used a similar two-transistor circuit, and stressed the benefits obtained from the bootstrapped load.  That got him into a debate on bootstrapping in the pages of Wireless World (9), although his similar debate (stoush?) with Linsley Hood on the series- vs. shunt RIAA equalization question (10) was probably better known.  In 1973, Ellis (11) proposed a three-transistor circuit, in which gain was provided by a cascode pair with bootstrapped output load from an emitter follower.  His objective was improvement over the single-transistor circuity; he acknowledged the Quad 33 as well as Quilter’s work.

When comparing the Quad 33 with the QCII and 22, one may note a “reversal” as it were.  The QCII and 22 had a shunt feedback disc input stage but an (apparently unusual) series-feedback tone control stage.  The 33 had a series-feedback disc input stage, but a shunt-feedback tone control stage.

On the device count issue, it may be observed that contemporary amplifiers, such as the Rogers Ravensbourne, showed similar economy.  Evidently the situation changed very rapidly in the few years following 1967.  By way of example, whereas the Quad 33 (of 1967) had 12 bipolar transistors, all in the signal paths, in 1971, the first iteration of the Quad FM3, in addition to two dual-gate mosfets and three ICs, had no fewer than 14 bipolar transistors.  One was the oscillator, one was an IF impedance-matching stage, and one was the mono channel output buffer.  Four were in the power supply regulation circuit and the other seven were used for the auxiliary functions of muting and tuning indicator drive.  So, it does not look as if there were any severe constraints on the number used.  Also in 1971, the Radford SC24 control unit used 50 transistors, but it was a “luxury” model that sold for nearly twice the price of the Quad 33
.
Judged by what was achieved with the Quad 33, one might say that the Quad 44 of 1979 indulged in quite a bit of “luxury”.  Each input and output was individually buffered, each was configurable by presets and/or changing modules, and there were two tape loops.  It came just before two-rail input switching arrived on the scene, so missed that useful facility.  An input module with isolated ground would also have been useful – that arrived with the Quad 66 a decade later.

Anyway, the Quad 33, whilst not above criticism, does come out as a carefully thought-out and in some ways innovative unit when analysed in the light of the norms and events of its time.  “Multum in parvo” might be an apt description.  Perhaps it’s a pity that its design and circuit features were nor covered in a detailed WW article at the time of its release. 

Cheers,

Steve

(1) Wireless World (WW) 1965 January p.03ff.
(2) WW 1966 December p.598ff.
(3) S.W. Amos, Ed; Radio, TV & Audio Technical Reference Book; Newnes; 1977; ISBN 0 408 00259 X; p.14-22.
(4) WW 1971 May p.221ff.
(5) WW 1976 November p.41ff
(6) WW 1970 February p.56ff; WW 1971 July p.321ff; WW 1971 August p.379ff.
(7) Earl, John (aka Gordon J. King); Pickups and Loudspeakers; Fountain Press; 1971; ISBN 0 852 42480 9; p.87.
(8) WW 1971 April p.199ff.
(9) WW 1972 May p.225; WW 1972 September p.423.
(10)  WW 1972 August p.389; WW 1972 November p.520; WW January 1973 pp.11,12; WW 1973 April pp.193,194.  
(11)  WW 1973 August p.378.

[Craig]

And excellent write-up Steve.

I'm actually quite a fan of the 33. Back when I was 16 (1972) I lusted after a 33/303 so badly it was almost painful, but could never afford them. Instead I built my own pre and power amps. Both are beautifully made.

However, Ross Walker, in an interview in Ken Kessler's book about Quad was very mealy mouthed about the appearance "When the 33 came out, people said "I'm not buying it with that bloody marigold thing on it". Well, Peter loved that, he thought it was great. And customers would come up and say "Well, I'm not going to buy that unless you take that marigold thing off. You'll have to change that". And he'd say "Well I'm not. Bugger off. Go and buy a Leak. Go on, bugger off"" He continues in the same vein. The interview is worth reading, if only for his reasons that the company bombed out under his watch.

Of course you can now buy these at affordable prices, so our kitchen system, fed by the LG television and an FM4, is a 33/303 with ceiling speakers. Both with new electrolytics throughout. And an auxillary headphone system is a 33/405II feeding Koss electrostatic headphones. And I don't mind the marigold at all ;-)

Craig

[Terry]

Thanks to Steve, I've learned a new word - stoush!

My spell checker doesn't recognise it and I don't think I'll have much use for it, though.

Wikipedia says:

Etymology

Possibly from stash. Australian from 1893; Boer War military slang. Also may be derived from stushie or stooshie, a Scottish term for a commotion, rumpus, or row.

Noun

stoush (plural stoushes)

(Australia, New Zealand, informal) A fight, an argument.

Verb

stoush (third-person singular simple present stoushes, present participle stoushing, simple past and past participle stoushed)
(Australia, informal) To fight; to argue.

[alanworland]

Steve, most interesting but way above my head!
I always lusted as well and eventually bought mine from an outfit in Tottenham Court Road. Mine has always had a problem with the switch banks - has been back to Quad twice to be returned working well only for the intermittent fault to return. Servisol helps for a while but will always come back! I have another switch board to fit but haven't plucked up the courage!
My FM3 was always down on volume (I thought) so I installed an op amp (741 I think) inside to bring the level up!

Alan

[Synchrodyne]

That’s an interesting comment about the FM3 output level, nominally 100 mV at 30% modulation (316 mV at 100% modulation).  For its time, it was probably at the low end of the range, although it was consistent with Quad established practice.  A look through H-Fi Year Book 1972 shows a few, such as the Sugden R21/R51, at 0.5 V (presumably for 100% modulation), with 1 V maximum for variable outputs quite common, some in between and one or two even above that level.  There was nothing new about that range, though.  The Rogers RD Junior AM tuner of 1952 had 1 V available, and the original Leak Troughline of 1955 had 1 V variable.

Quad adopted the 100 mV at 30% modulation convention with its Acoustical AM model of 1953, designed to work with the QCII control unit.  Previously, its Quad/R tuner had a relatively low output level, around gramophone pickup level, in part required to fit the interconnection scheme for the original Quad amplifier.  However, an additional advantage of doing it that way was that the ratio of AC-to-DC loading was very little affected by the additional AC loading imposed by the amplifier input.  Keeping this ratio as close as possible to unity was an important element in minimizing diode demodulator distortion.  For a tuner with an unbuffered output that might be used with various amplifiers, the load presented an amplifier input was an unknown variable, so mitigation by tapping down the output was desirable.  In radio receivers, the AC loading was a known and fixed quantity, so it was not such an issue.  (Although it did become one when interconnection with tape recorders arrived in the mid-1950s – hence the German approach to diode outlets that eventually morphed into the DIN “constant current” standard.)

The 100 mV level still allowed quite a bit of room for the tapping down of diode outputs.  In the Acoustical AM case, the diode load was 150k over 4k7, so the additional AC loading added by the amplifier input (100k in parallel with the 4k7) was quite small.  And evidently 100 mV was quite workable within the gain structure of the QCII.  Quad then retained 100 mV at 30% modulation for all of its subsequent tuners through to the FM66.  This tuner output level was mostly associated with control unit radio inputs of 100 k impedance and 100 mV sensitivity.  That meant that the tuner would fully load the power amplifier with the volume control at -10 dB, which probably represented a higher volume control setting than for many or even most tuner and amplifier combinations.

The 100 mV convention was used by other makers, both for tuner outputs and amplifier inputs, an example being the Pye Mozart HFT108 FM tuner and Mozart HF10 amplifier.  An interesting case was that of the Armstrong 600 series tuners from 1973.  These had a “low” output of 100 mV at 30% modulation, which acknowledged an established convention.  But they also had a high output of 775 mV at 100% modulation.  (The previous 500 series had variable outputs, 0 to 1 V.)  By then the 775 mV number was being in seen in domestic audio systems.  Of course, it came from professional audio practice, and previous to that from telephone practice, where 775 mV represented the 0 dB level, corresponding to 1 mW (0 dBm) in a nominally 600R balanced pair line.  Armstrong did refer to the “high” output as being a “0 dB” output.  But as it was not associated with 600R, it was not 0 dBm.  (I guess that it could be described as 0 dBu, but that convention seems not to have arrived until the 1990s.)  I can’t think of any good reason for the adoption of 775 mV for domestic unbalanced and relatively high impedance interconnections, but it was established number and perhaps its use (along with 0 dB) conveyed an air of professionality.  Also around this time, the term “line”, which hitherto pretty much always meant the 600R balanced type, was coming into use for domestic unbalanced interconnects.  At one time “high level inputs” was sometimes used as the generic term for amplifier radio/tape/auxiliary inputs, but this seemed to fade out.  Maybe “line” also was thought to convey professionality.  It may have started with tape recorders before it was applied to amplifiers.  For example, the Revox A77 had an (unbalanced) line output, and the Ferrograph Series 7 had an unbalanced 2M2 line input.
 
When CDs arrived on the scene in the early 1980s, CD players had a standardized output of 2 V full-scale.  There may have been some uncertainty as to what the average programme output would be.  Quad I think tried CD input sensitivities of 500 and 400 mV before landing upon 300 mV.  That meant that a standard CD player, fed into a 300 mV sensitivity input, would produce about the same volume (at any given volume control setting) a as one of its FM tuners fed into a 100 mV input.  Extrapolating from that, a tuner with an output level of 1 V at 100% modulation, fed into a 300 mV input, would match a standard CD player volume-wise when that was fed into a 300 mV input.

Thus whereas the Quad 33 and 44 control units had been based upon a “line level” of 100 mV nominal, at which level the input switching was done, the Quad 66 was designed around 300 mV.   The radio input, still of 100 mV sensitivity, thus required 10 dB gain before the switching matrix.  Quad could have put the 10 dB of gain into the FM66 tuner, but probably did not in order to maintain backward compatibility between the 66 system and earlier tuners.

Generally, though the industry view seemed to be that tuner outputs in the range 775 mV to 1 V (at 100% modulation) would match CD player outputs when fed into typical “line” inputs.  The Meridian 200 series provided an example.  The 201 control unit line inputs had a nominal sensitivity of 150 mV (although variable).  The 204 tuner output was 775 mV, and the 207 CD player fixed output was the standard 2 V.  Meridian evidently preferred something like a 6 dB greater difference between nominal input level and input sensitivity than did Quad.

Returning to the case of the Quad FM3, there is a good chance that when used with a non-Quad amplifier, it would seem to be short of output level as compared with non-Quad units plugged into the “line” inputs of the same amplifier.  When used with the Quad 33, I guess it would depend what else was plugged into it.  I think that it would more-or-less match a standard CD player plugged into the tape input set at 400 mV, or to a radio input via a 10 dB attenuator.  You can get some sense as to how Quad approached level matching by noting that the special version of the disc input board with M1 configured for the Shure V15 Type III cartridge had a sensitivity of 3.8 mV.  This cartridge was specified to produce 0.7 mV per cm/s of groove modulation.  That equates to 3.5 mv at 5 cm/s, typically taken as the average programme level.  So, one might expect that the Shure V15 III used with the special M1 input would produce about the same average programme volume as an FM3 plugged into the radio input.  But with the V15 III plugged into the standard M1 input of 2 mv sensitivity, it would be 5.5 dB louder for the same volume control setting.  Against that, the FM3 might be perceived as being a bit quieter.

In terms of backward compatibility, the Quad 33/303 represented a notable break with the past in respect of the control unit-power amplifier interconnection.  This had been 1.4 V for the Quad II series, but came down to 500 mV for the Quad 33/303.  My guess is that Quad did not want to complicate the design of the 33 by specifying an output level that would have required a higher supply rail voltage, and perhaps an extra gain stage, whereas there was no real problem with including the extra gain in the 303.

Cheers,

Steve                                         

[alanworland]

It's interesting that you mention the pickup input sensitivity as I have only ever used the 33 with a V15 III which I guess would account for the level differences that I encountered leading to the Op amp installation.
I also think I can remember some mods to the pickup board to change its input specification.
Still got the intermittent switch board though! Might tackle it during the winter months with some replacement caps.

Alan

[ppppenguin]

Going back to my original question, does anyone think the 0.68uF caps are critical or will 1uF do fine?

[Keithstv]

I can't see any problem with replacing to 0.68uF with 1uF.

Keith

[ppppenguin]

Neither can I but Quad chose an unusual value of tantalum cap for these positions when 1u would have been more obvious. That's why I'm concerned about it.