1970s Design Indulgence

I remember reading that toroidal transformers have a wider frequency response than e and I transformers.

Edited by Dave Friday - 30 Nov 2018 at 2:29am

Feeling like I was missing a trick but rather worried about damaging the amplifier, I still felt this must be tried.

What's that you ask? It is the position of the dominant pole capacitor C7, 47pF back on page 23 - best to open it in a new tab/window for reference - and also do the same with this link going to page 7 fig.12 at https://www.americanradiohistory.com/Archive-Wireless-World/60s/Wireless-World-1965-01.pdf

You might also note C5 (starred) which gives you a clue.


Going back to page 16 of this topic (you may wish to open that in a new tab/window) and referring to the gains in the amplifier input stage comparison drawing, it can be seen that the modernistic differential IPS has below unity gain, whilst the pre-differential IPS has greater than unity gain.

A common emitter (T1 is sort of) with gain exhibits miller capacitance which is Ccb (the parasitic collector to base capacitance) multiplied by its gain. And this is all we need to know about T1 for the purpose of understanding this. If it has this capacitance, and gain, therefore it exhibits an important pole in the response somewhere.

That pole is not dealt with by C7. It can't be because C7 deals with T2's parasitic miller capacitance.

So we have a two pole circuit and that doesn't always guarantee unconditional stability.

To understand stability it could help to see this video which I watched as a refresher:



If we reach -180° we almost guarantee instability, and we need at least 10dB gain margin, or with zero gain margin we must have at least 45° phase margin. In my own personal opinion I say it should be at least 90° phase margin.

Simulating with C5 position replacing C7's position and using the same 47pF value, the phase plot shows it never exceeding about 66°, which is (180 - 66) a phase margin of 114°, but it does show the gain margin being very slim at much less the 10dB - nearer 1 or 2 dB.

However, stability criteria says that's OK.

The reason the gain margin will not increase - it will stay at just under unity - just under a gain of 1, is because of the type of negative feedback. It can only reduce gain to unity, and the current stage having gain of just under unity takes it just below.

So, will the amplifier survive given this dominant pole repositioning?

With trepidation I tried it, taking the precaution of using a power resistor in the HT fuse position to soak up any excessive current which would indicate a fault.

It didn't do anything untoward so I replaced the HT fuse and measured the result, which was disappointing as the distortion had risen. I was at a loss to understand why until I realised I was not taking it to the correct place.

Recalling the discussion about Re and RE - intrinsic and real emitter resistances - the lower position of C5 connects at the junction of Re and RE, in other words the physical emitter of T1, and this is what was wrong. Connecting it to the bottom of RE - the self same place where the global NFB connects - cured the problem.

1kHz THD now measured below 0.02% at 50W into 8 ohms (both channels driven), and all other measurements I could think of at the time, were just about the same as before.

At one point 1kHz THD got just below 0.03% at 75W into 4 ohms (both channels driven).

A quick listening test revealed a "less over-sweet" presentation, which if it continues like that would indicate that the "un-tamed pole" was the problem.

One thing to note however is distortion reduces with warm-up, and with no warm-up it isn't as low as I report here. It isn't affected much by output stage standing current, but by increasing it, the output stage will run warmer, and here I need to find a happy medium which warms the output stage up quickly so that it operates at the best THD, without it getting uncomfortably hot. Also bearing in mind that in the heat of summer its heatsinks must not get painfully hot. Plus we have to consider the effect of heat on the electrolytic capacitors inside the unit. We want the longest possible lifetime before having to re-cap.


Edited by Graham Slee - 02 Dec 2018 at 4:23am

The history of the progress - 100W amps on the cheap for everybody

Cheap op-amps, cheap transistors, cheap printed circuit boards, Hitachi, Quad's output triple (with the advent of matching performance PNPs) and cheap labour all coalesced around the same time to bring you today's pattern-part amplifiers at more than affordable prices (sometimes).

Without cheap op-amps DC coupled amps would not have happened. They're needed to keep the output voltage of a DC coupled amplifier at zero volts, because if it isn't, your speaker voice coils would become nice and warm courtesy the output DC offset. This hadn't happened with ac coupling but ac coupling cost money, and money had to be saved.

The op-amp serves the purpose of servo, so if the output drifts positive it steers it negative, and if it drifts negative it steers it positive.

The other thing that was required was the Hitachi protection chip and Hitachi (also mass producing amps at the time) obliged (nowadays it's a section of the board stuffed with surface mount discretes after Hitachi pulled out).

The stage was set.

The new DC coupled amps made hay using the centre-tap transformer rectifier relationship which make DC amps and ac amps reach parity.

The IPS became the long-tail-pair and in doing so you could swing almost the entire power supply volts into the load, rather than losing a bit.

100w into 8 requires 28.3v rms and requires a 30-0-30V transformer, but toroids can be forced to do more with fewer amps if you only increase the volts. So cheap 35-0-35V toroids became the norm.

One bridge rectifier and a couple of 10,000uF 63V electrolytics and you were away. As long as you could find high voltage transistors - the Japanese were happy to make them - the West following hard on their heels.

As to the smell of burning silicon of the VAS transistor this was touted as being "because it's new".

The transformer should be 495VA but at 35-0-35V a 300VA (or often much less) could be made to stand up for just long enough to do the factory set-up.

Mass produce - pile 'em high - sell them cheap - or... bling them up and sell at high audiophool prices.

No wonder valves made a comeback!

Makes this topic look sensible don't it?

(and that's without broaching the subject of MOSFETS)