Thoughts on this HiFi line amplifier?

[Synchrodyne]

The series vs. shunt feedback question is an interesting one.

As RW had said, H.P. Walker demonstrated (mathematically) quite decisively 50 years ago that for an MM/RIAA stage, series feedback was much better in noise terms than shunt feedback. The difference was 13.5 dB referred to a 2 mV input, assuming perfect active devices. The limiting noise floors were circuit limits, and were closely approachable with transistors of the time. The main bugbear was the high load impedance (typically 47k) that MM cartridges needed to see, which in the shunt feedback case, was in the input arm.

With the downward movement in MM cartridge specific output levels through the 1960s, the noise problem with the shunt feedback circuit increased in magnitude. Fortunately, it was to some extent solved with the transition to transistors, with the widespread adoption of Dinsdales’s two-transistor series feedback circuit and improvements thereto.

In the valve era, the MM RIAA stage was often a single pentode (e.g. EF86) with shunt feedback, with a sensitivity usually not better than 4-5 mV, as suited the cartridges of the time. But the same considerations applied. One maker’s noise specification was phrased thus: “-80 dB, or where applicable, the equivalent noise of the pickup load impedance at the input.” In most cases I think that the conditional part would have applied. And in those days 68k rather than 47k was a common MM load impedance. One could say that the basic circuit configuration was as much, perhaps more to blame for the noise apparent with low output MM cartridges than the pentode itself.

A small number of makers (e.g. Lowther as early as 1953) used a cascode double triode with shunt feedback. But that would run into the same noise floor limitation as the pentode, so in practice might not have made much difference at all. The realizable specification might have been something like: “-90 dB (maybe optimistic) or where applicable, the equivalent noise of the pickup load impedance at the input”, with the conditional part still applying in the majority of cases.

But where the cascode was used, it usually also doubled as a microphone and/or tape head preamplifier, in which roles it probably was better than a pentode. But that distinction was not always made in the sales literature. These days I suspect that in the boutique world at least, there is some thinking that as the cascode is inherently quieter than the pentode, it must be better, regardless of the circuit in which it is used.

Another valve era approach that would have lowered the actual noise floor was a triode pair with series feedback. The disadvantage was that DC heating was required on least on the input triode, so it was not used all that much.

The pentode/cascode “fallacy” in respect of the shunt feedback MM/RIAA stage was somewhat repeated in the transistor era, in that some makers used a quieter PNP transistor for the input stage of the series feedback pair, even though in situ it would not have made a material difference. In at least some of those cases though, the stage doubled as a microphone input, where the PNP input would have been beneficial.

A look at H.P. Walker’s 1971 mixer circuitry is instructive. For the MM/RIAA stage he used NPN transistors, and achieved very close to the theoretical noise floor. But for one of the microphone preamplifier options, of the series feedback type, he did use a PNP input, because in that case it was advantageous.

For “line-level” stages, including tone controls, the shunt vs series feedback choice seems to be less of an issue, perhaps because of the combination of higher signal levels and the possibility of much lower input impedances. The Baxandall tone control was built around an inverting stage with shunt feedback, and as far as I know Baxandall himself never departed from that. H.P. Walker also used shunt feedback tone control circuit in his 1971 mixer, using two transistors with the second bootstrapping the collector resistor of the first. That was essentially as Quad had done with the 33, and which Quilter later became famous for. (H.P. Walker got into a WW “stoush” over bootstrapping, although of smaller magnitude than his feedback stoush with Linsley Hood.)

Shunt feedback seems to have remained the modal choice for (feedback) tone controls. Series feedback circuits have been used here and there (one or two in this thread: https://vintage-radio.net/forum/showthread.php?t=156244), but the apparent disadvantage is the unity minimum gain, which means that the mean gain needs to be quite high (maybe 20 dB) to allow room for frequency selective cutting, with enough headroom (maybe another 20dB) to allow for frequency selective boosting).


Cheers,

[Radio Wrangler]

This is a line level amplifier, not a magnetic cartridge or tape head input, so the noise issues are less significant, but if you can do something better, why not?

Baxandall tone controls are implicitly shunt feedback because the pots share resistance between the input and the feedback paths. It's rather elegant, and if placed where signal levels are moderate, then noise isn't a problem. One thing with the Baxandall, is that it works best with a low driving impedance from the stage ahead of it. Not common in valve implementations, but neither difficult nor expensive with semiconductors. The Baxandall implements a virtual ground and depending on where the control sliders go, the input impedance to the stage can go low at some frequencies.
The Baxandall arrangement uses the shunt feedback gain of the form A/B to not vary midband gain, and give a balanced effect, while the series feedback arrangement has gain of the form (1+A/B) which distorts the symmetry. So shunt is the right choice here.

I pondered tone controls some years ago. I synthesised a set of passive networks to give the shapes in the Quad 44 brochure slope control graphs and liked the result. I hadn't seen the Quad 44 schematic at that time, so my arrangement is different but equivalent. I used some Takemasawa miniature (DIL) relays to switch 4 separate networks and a bypass. Overall, it doesn't do much, but I've found it does enough for me.

David

[joebog1]

Mu of 70 can also be 12AY7 specifically for audio. Or my favorite 5751. Both pin for pin replaceable.

Joe

[Synchrodyne]

Interesting background article on the 12AY7 attached.

[Radio Wrangler]

Indeed, a very interesting article.

One observation is that their example amplifier has an input transformer as a step-up because the valve is still better for signal to noise ratio in a high-impedance environment. This means that the voltage step-up of the signal makes the circuit more sensitive to current noise at the valve input, and less sensitive to voltage noise. As the valve has lower current noise than voltage noise, this is a net win.

Jumping to magnetic cartridges for record playing, it's well known that phono stages normally terminate the cartridge inputs with typically 47k resistive. Less well known is that the effective source impedance presented by the cartridge is appreciably lower. So there is a mis-match here, but it is intentional and serves to control the frequency response so that only RIAA correction is needed.

From the point of view of signal to noise ratio, the phono stage needs to be noise-optimised for the impedance the cartridge presents, not for that 47k resistor. This is much lower than the regime in which most valves do their best.

So the ratio of the transformer for their 600 Ohm input amplifier is a useful indication of what impedance regime the valve gives its best noise performance in. The transformer in their design obviously provides balance and ground isolation, but it also provides this extra clue.

David

[RichardGM]

Originally Posted by Radio Wrangler

I pondered tone controls some years ago. I synthesised a set of passive networks to give the shapes in the Quad 44 brochure slope control graphs and liked the result. I hadn't seen the Quad 44 schematic at that time, so my arrangement is different but equivalent. I used some Takemasawa miniature (DIL) relays to switch 4 separate networks and a bypass. Overall, it doesn't do much, but I've found it does enough for me.

David

That sounds a nifty bit of design. Any chance of your telling us all the details (possibly in a new thread)?

[Radio Wrangler]

I'm not sure where the schematic may be. I think I have two, both A0 sie. The ink on mylar original and a paper blueline copy. It'll have to be the next time I come across iy, probably in the attic. I remember a cardboard tube of big drawings....

Alternatively I could take the lid off the preamp and just do some photos. It's all built flying over a copper groundplane. Real breadboard.

David