The Audio Research Classic 30
The Classic 30 was the baby in a family of hybrid (valve/JFET) power amplifiers released by Audio Research between 1988 and 1990, which also included the Classic 60, and the monoblock Classic 120 and 150. The whole range uses pretty much the same circuit (the 120, for example, uses an almost identical power supply circuit and components, apart from transformer rating and reservoir capacitance, as the 60). Unusually for ARC, there were no Mark II revisions for the Classics; however, revised versions of almost the same circuits but with balanced inputs instead of the single ended inputs of the Classic series were released in 1992, but renamed V35, V70 and V140.
Unlike the all-valve VT series the Classics were a hybrid design, with JFETs at the inputs, appearing in the same era as the hybrid SP-9 and SP-14 preamps. All the same, despite the complex circuitry and the widespread incursion of solid-state devices in the signal and power supply stages, the Classics were positively reviewed, and are indeed still well regarded.
The Classic 30 is a kind of hybrid development of the D70, with the three input triodes of the latter replaced by JFETs (although the D70 uses its single pair of 6550 per channel in tetrode, rather than triode, mode, giving the higher output power). The output half of the D70 itself seems to be inspired by the classic Williamson amplifier. In the ARC circuits, though, the phase splitter is before the input stage, instead of its usual position before the drivers, and ARC's patented cross-coupling is added between the drivers to reduce distortion and the sensitivity of the circuit to component variation and valve ageing.
At the input stage of the Classics there is an inverter, consisting of the N-channel JFET Q1 and source-follower MOSFET Q2, which generates the inverted polarity signal; from this point on, the circuit is fully differential, The Classic 60, 120 and 150 have servos in the input stages to maintain DC balance, but in the CL30 these are absent. The input stage of each leg is a common-source n-channel JFET (Q3 and Q4), providing some of the voltage gain, and the two global negative feedback loops are attached the source of each of these JFETs. The input impedance is R1 plus the parallel combination of R2 and R4, or about 100K, and there is a low-pass filter in the input with a turnover frequency of around 500kHz. These are followed by a driver stage consisting of a pair of paralleled source-follower p-channel JFETs cascoded with a 6FQ7 triode, but with the JFET and triode fed with opposite polarity signals.
The functioning of the circuit is quite sensitive to the parameters of the JFETs in the input stages, which aren't identified in the parts lists (and presumably these components are specifically hand-selected by ARC for this application anyway). For instance, Q1 is operated at about 2mA, which means that its Vgs must be 0.33V at this current. With a different Vgs Q1 may not conduct - in this case the drain voltage will rise to over 50V, which is likely to be fatal for the JFET. Similarly, Q3 and Q4 need to be well matched in their parameters, and their Vgs will have to be selected carefully (or else the source resistors tweaked to get the right quiescent current). The AC gain of Q3 and Q4 is also dependent on their transconductance, but since the circuit has global loop feedback this is probably less critical, as long as it is high enough.
The 6550 output valves have their screen grids connected to their anodes to force them into triode operation. The 30 has two 6550's in the output stage (the 60 has four, and the 120 eight, with the latter two using fans to keep air circulating around the big output valves), and there is an op-amp based servo circuit to maintain the quiescent current on both of the power valves identical and so reduce distortion in the output transformer. Unusually, the cathodes of the output valves are connected across the secondary winding of the output transformer, and these are also the points where the feedback signals are derived. The cathode feedback of the Classics has been discussed in several online forums, and the consensus seems to be that this puts stringent demands on the output transformers - the secondary windings need to be strictly symmetrical, otherwise the cross-coupling increases the distortion, instead of decreasing it. A minor side-effect of the cathode feedback is that the standing current in the output valves passes through the secondary winding so that, as a consequence of Ohm's Law, there is a DC voltage present at the outputs. The exact value of this will depend on the DC resistance of the secondary and also on which impedance tap is being used, though it should be only a few millivolts.
There are three high-voltage supplies (+420V unregulated for the output stage, +330V regulated for the drivers and +320V regulated for the input stage) and a further adjustable regulated supply at -55V for the bias circuitry for the output valves. The three regulation stages are variations on the usual ARC op-amp/MOSFET circuit, with the op-amp U2 in the crucial -55V regulator having its own supply voltage shunt-regulated by the P-channel MOSFET Q11.
The valve filaments are all heated by 6.3VAC, which comes directly from the transformer secondaries, with separate supplies for the output valves and for the drivers. A final regulated 15V negative supply is used for the two LED indicators on the front panel: one of these lights when power is applied, and the other comes on once the 320V line has reached a threshold voltage, which takes a second or two - in effect, the latter indicator is a monitor for all of the series regulator circuits. There is a slow-start circuit that reduces the current drawn at switch-on, consisting of a thermistor on the primary of the mains transformer which is shorted after a second or so by a relay.
It's interesting that the Classic 120 was released a couple of years after the 30, but the circuits nevertheless are very similar, apart from the obvious doublings-up for higher power - this inspires confidence that the design was essentially finished in the first place. Some listeners have described the CL30 as the sweetest-sounding of the Classic range, and there are definite advantages of not doubling up valves when high power isn't required. The absence of the cooling fans installed in all of the rest of the Classics will also make the system audibly quieter.
What would I do differently?
Provided one has no basic objections to the presence of solid-state amplifying devices in the Classic 30, there is little that could specifically be changed, although it could be argued that all the ARC designs of this period are unnecessarily complex! However, the fact that the circuit beyond the JFET/MOSFET inverter at the input is entirely differential means that it is crying out for a balanced input, which is indeed what ARC subsequently did with the V35. In this case, unfortunately, they shot themselves in the foot by removing the single-ended RCA inputs, which restricted the V35 to systems where the preamplifier had a balanced output. I would instal a relay in the input circuits to allow switching between single-ended and balanced operation at little extra cost.
The other major change from the CL30 to the V35 was the switch-on delay of the HT supply, which allows the valve filaments to warm up before the high voltage is applied. This has the potential to significantly lengthen valve life, so would certainly be worth doing.
The amazingly dedicated Manfred Persson has a website dedicated to ARC products, which is stuffed with information and photos on the ARC amplifiers and CD players. Here are a few links relevant to the Classic 30:
ARC's original flyer for the Classic 30 and 60.
Photo of the amplifier from the top
Photo of the amplifier from the top, without mesh shield
Photo of the rear of the amplifier
Parts list page 1
Parts list page 2
NOTE: there is an error in the parts list: R8, 25, 26, 44, 45, 61 and 62 are listed as "100 1% MK8". These must be 100K, rather 100R, since firstly they have quite serious voltages across them (in particular R8, R25, and R26 are the drain and anode loads for Q1 and V1 and V2 respectively) and secondly are clearly marked brown, black, black, orange on the PCB. Using the values marked would put most of the 320V HT across the JFETs and certainly blow them up, and would probably damage the valves too.
R37 and R38, on the other hand, really are 100 ohms, since they are the stoppers on the screen grids of the 6550s which convert the tetrodes to triode mode.
Biasing the output valves of the the Classic 30
The four 6550 output valves are intended to be operated with a standing current of 65 mA in each. The output stage of each channel has a clever servo circuit, based around an op-amp, which ensures that each of the 0.5 ohm resistors in the anode circuit of each valve has the same current through it. The op-amp output is coupled to the bias circuit through a photocoupler (the Classic 120 circuit diagram on the ARCDB site identifies it as a VTL5C4 vactrol). The offset null of the op-amp for each channel is then adjusted through a preset potentiometer (VR1) to set the voltages across both resistors to measure the same. Note that the common-mode range of the op-amp needs to be within about 30mV of the positive rail for the servo to work, which restricts the choice of device here: the specified LF441 obviously qualifies, but is now obsolete. Among more recent FET-input op-amps, the TLE2081 allows a rail-to-rail input range, as well as having the same offset null connections as the LF441. The setting of the VR1 pots should be independent of the valves chosen, so this only need to be adjusted as part of the initial setup, and then as part of occasional maintenance. The bias voltage is supplied from a well-regulated -55V supply, common to both channels, and can be finely adjusted through a further preset pot (VR2).
The only trouble with this arrangement is that we only have one adjustment to set two quiescent currents! I guess that the DC balance in each output stage is more important than getting the exact standing current right, though the bias does set the operating point of each valve, and so would be expected to affect the distortion characteristics of the amplifier. Of course, valves supplied specifically for this amplifier by Audio Research will be matched in more parameters than their bias voltage at 65mA, so they will match pretty closely anyway and little adjustment will be required. I suspect that the best way to balance the output stage, given four randomly selected 6550s, would be to shuffle them around so that the currents in V5 and 6 (the non-adjustable ones) are as close as possible, and then the servo circuits will ensure that the standing currents in V3 and V4 are also 65mA.
If I were designing the CL30 from scratch (and if there was space on the PCB), I would have an independently adjustable bias voltage for each channel; it would then be much less critical to have matched valves.
A Tube Asylum thread on Biasing the Classics...
A page on biasing the (rather different) VT100
Reviews of the Classic line
Review of the Classic 30 on Audioreview
John Atkinson's Stereophile review of the Classic 60
A review of the V70
Links (Audio Research stuff)
Audio Research Corporation's home page.
Manfred Persson's wonderful ARC information pages
Manfred's page on the CL120,with a lot more specific information on the schematics than on the ones for the similar Classic 30.
Discussions on the CL30 and similar ARC amps
An interesting thread on diyaudio.com on cloning ARC amplifiers
A thread on the Tube Asylum on output valve choice for the ARC VT series power amps
Another thread on the Tube Asylum on choice of output valves for (earlier) ARC power amps.
Another thread on the Tube Asylum, on the Classic 60
Yet another thread on the Tube Asylum, this one on valve choice for the Classic 30 and SP-9...
... and another one
A thread on selection of 6FQ7 driver valves for the Classics.
A Tube DIY Asylum thread on servo control of bias, a little like ARC's approach in the Classics.
A discussion of the cathode feedback arrangement used in the Classics and other ARC amplifiers.
Back to DIY page.
UK valve and transformer suppliers
The offputtingly-named Banzai Effects, who nevertheless stock a great range of valves and good-quality parts at excellent prices.
An output transformer, described by Sowter as a "Williamson" style, which could be used in a Classic 30 clone, though it would need slightly different primary and secondary taps.