Mullard Audio Circuit

[davew]

I was sorting through my bookcase recently and came across my Mullard “Transistor Audio and Radio circuits” which I received for Christmas 1972. This brought back some great memories.
These circuits were going to form the basis of my first great Stereo set up!
I first built the “Pre-amp for use with the 10W AudioAmplifier” which used 2 BC109s and a BC108
I was quite proud of this – I made my own printed circuit board and really took some care with the construction – This worked really well
For some reason I decided to shoehorn the whole amplifier into the Turntable plinth – The motor of the BSR MP60 must have hung about half an inch off the circuit board!
As an aside, I had already sent off for the speakers – One of the Ads you use to see everywhere in PW, Radio Constructor etc – The Keledyne 9000 (or something similar!) full range speaker 30-30000c/s etc etc Only 59/6
When these turned up, what a disappointment - they werer eally poor. Mini tower sized, around 3 ft X 6in X 6in containing what looked like to two ex TV 5 X 3 in ellipticals and a cheap tweeter in a very poorly made cabinet covered in fablon.
There really was some tat being sold out there then..
The final part was building the power amps and this is where I failed miserably
I used the “10W Audio Amplifer” circuit and to cut a very long story short I never got a peep out of either of the two I built – I think I eventually used Sinclair power modules available at the time and the whole thing worked quite well.
It set me wondering. I’m assuming that the Mullard diagram would be correct and this failure was due to youthful ignorance but I’d like an expert opinion and have attached the diagram that caused me such grief!
Anything weird or unusual here?

Thanks
Dave

[Radio Wrangler]

The unusual thing with that diagram is that the output transistors are just single devices without drivers in a darlington configuration, so you just have the current gain of the BD131 etc, which is around 80.

To try to compensate for this, the voltage amplifier stage TR2 and the VBE multiplier are run at a fair amount of current..... R9 and R10 are quite low values.

I think the whole circuit may be a bit tight on current gain.

The wiring of R8 in the Vbe multiplier is a no-no. if the slider skips while being turned, both power transistors get turned full on and hopefully a fuse should pop before anything else.

Normally, the Philips/Mullard circuits are nicely toleranced with sensible margins. This one is a bit too close to the bone just to save a couple of low power transistors and a couple of resistors.

I think that edition had the 25W amplifier in it using BDY20 output devices. built one of these with a schoolfriend and it was very successful. The preamp was a bit crude and switching feedback networks on a single stage was problematical. Locating the input switch hard by the rear panel inputs and extending its shaft helped.

David

[joebog1]

I seem to remember white data books with red writing. I built a few amps from the "Circuits" section. First the AC127/128 amp, and later the AD161/162 10 watt amp.
Both were excellent, but did have a few problems with thermal runaway. I couldnt get heatsinks, so the amps were just mounted on scrap aluminium sheet that I found.
I was about 10 or 12 at that time.

Joe

[Radio Wrangler]

I think these were small bound books with dark blue cloth bonded to the outer surface of the covers.

I've have the later one, with the 50W 2N3055 amplifier in it, but vI've just found the earlier one with the 25W BDY20 amp in it.

Ah yes, it does have that BD131/BD132 10 Watt amp in it!

David

[Argus25]

It is just a bog standard complimentary output stage, where the output devices act as push pull emitter followers, with some signal bootstrapping (optional) to help the drive at full output, the vbe multiplier for bias and the feedback sent back to the emitter of the driver transistor.

You will see this configuration everywhere in radios and radiograms of the time. Hacker used it in their radios with AC176 & AC128 output transistors models RP25,26 etc.

Small stereos used AD161 & AD162 outputs with the same/similar circuit and bigger amps simply had power silicon devices, 2N3055, MJ2955 etc or other less beefy parts.

The AD161 & 162's were a little unreliable in my view and easily damaged. Oddly though they were about the same physical size as American TO-66 devices, they have a different lead out spacing and the mica insulators are not compatible.

Since every transistor's DC conditions in a circuit like this is affected by every other transistor, it doesn't take much to completely foul up everything with an incorrectly fitted transistor etc. When these sorts of amps fail, usually more than one device takes a trip to la la land. You may have had a sequence of events that caused destruction to a few of the transistors.

[Radio Wrangler]

Originally Posted by Argus25

When these sorts of amps fail, usually more than one device takes a trip to la la land. You may have had a sequence of events that caused destruction to a few of the transistors.

Indeed!

You can get trapped in a never-ending circle replacing one device at a time where the other devices kill the new one and then it kills the next one you fit as you try to work your way through it in small steps. This has discouraged a lot of people.

The best way is to find and replace all bad devices and to not put power on until they've all been done. Just swapping the whole lot is justified if time is being paid for, or if you can't test semiconductors at rated voltages. Little hand held testers can be misleading.

David

[Argus25]

DaveW ,

I have attached the basic circuit with a description of it. The last sentence is interesting about the notion of eliminating transformers to save on costs & weight etc.

It is true of course and also transformers have many other unfavorable properties, limitations in frequency response, problematic phase shifts especially with overall negative feedback applied, less than idea electrical properties, leakage inductance, copper and iron losses etc.

The odd thing is though, that despite all that, people in the audiophile world find the transformer sound appealing. And the technician likes them, as they afford DC isolation between stages preventing an avalanche of transistor failures with a fault in one stage. So, they are not all bad, but the predominant market forces of $, ultimately turned many manufacturers away from transformers for audio work with semiconductor based amplifiers.

[davew]

Many thanks all - That does sound the most feasible explanation with a "never ending circle" of devices taking each other out.
I may try building it again one day to finally nail it!
Dave

[Craig Sawyers]

One of the problems is the quiescent current setting arrangement. Since there are no end stop resistors at each end of the pot R3. So when at the low end of the resistor, Q3 is cut off. That means that the base of the upper power transistor is connected to rail via 370 ohms. Even if we assume that the output is half the 30V rail (ie 15V) the base current will be 15/370 = 40mA. The spec of the power transistors is an hFE of 70 up to 1A, and 40 at the maximum continuous Ic of 3A.

So at one end of the quiescent resistor the output transistors will be pulling at least 1.6A and possibly more.

Even if it is 1.6A at 15V those power transistors will dissipate 24W each. Since the specified maximum collector dissipation is 15W (junction to case is 6C/W) they are guaranteed to blow up thermally even if on an infinite heatsink, if not just go into second breakdown.

Not a wonderful design by Mullard, and possibly never actually built and tested by them.

Craig

[Argus25]

Yes the pot on the 2Vbe multiplier is assumed to be close to the center of its range so that the voltage dropped across the C-E is about 1.4V. If the pot is cranked around to one end where the B-E is effectively shorted out, then you have to rely on the C-E voltage, set by the 100R pot resistance, and the current passing by that, in this case, to limit the max quiescent current of the output transistors.

I guess they were just trying to save on resistors. In most good designs it should not be possible to tweak a pot and induce failure. Mind you, you could also argue that software outputs should not be able to induce a hardware failure. However this has happened on a number of occasions with satellites and spacecraft.

[Craig Sawyers]

Indeed. Remember the failed Ariane 5 launch in 1996? They got a sign wrong in one term of a polynomial that is used for course management. Naturally there is fail safe software if one crashes for some reason. But the second software copy was identical to the first.

37 seconds after launch - the rocket turned 90 degrees one way and boom.

Back to the Mullard amp. I had not spotted that the Iq pot was only 100 ohms, I'm used to them being around 4.7k with end stop resistors. In fact all the resistors are low value in the Mullard design.

The power transistors are actually not half bad (within their ratings). They are an early example of sustained beta devices. The hFE is almost flat up to Ic of 1A, which reduces output stage distortions.

Craig

[Craig Sawyers]

I put the circuit into spice. The original devices were not in the parts library so I used BD139/BD140 as the power devices. These, operating in emitter follower mode are not sensitive to the precise value of beta. I used a load resistor of 8 ohms after the output capacitor.

Quiescent current by percentage setting of the pot

0% Ic = 740mA
10% Ic = 730mA
20% Ic = 730mA
30% Ic = 525mA
40% Ic = 247mA
50% Ic = 70mA
60% Ic = 3.4mA/1.7mA
70% Ic = 1.7mA/0.4mA

Anyway, this demonstrates the key problem in the design. A sensible current of maybe 30mA is almost impossible to set. Below 50% setting the current goes up very quickly - and at 0 to 30% setting of the pot, each power transistor dissipates nearly 12W

Setting the pot at 50%, the amp has a bias point of 70mA in the output devices so a standing dissipation of about 1W each, which is fine. The gain is 26.8dB agreeing with the gain setting ratio of 470 ohms in the feedback arm and 22 ohms to ground: (470 + 22)/22 giving 27dB.

The frequency response is 25Hz to 33kHz at -3dB, so that looks OK.

Clipping is asymmetric, and starts at 0.35V rms input, so 7V rms on the output. That corresponds to 6W into 8 ohms, so 10W was a little optimistic in Mullard's spec.

Anyway, an interesting exercise. And indicates that transistor thermal death is more or less guaranteed because of the quiescent pot characteristics.

Actually 10W is pretty much impossible. 10W into 8 ohms is 8.9Vrms, or 12.65V peak - and would need the amp to swing within 2.35V of the rails. And that is a tough ask even with today's amplifier technologies.

Craig

[Craig Sawyers]

OK - I've got 8.5W out of this little circuit. You need to halve R9 and R10 to 135 and 50 ohms (or the nearest practical values). That bumps up clipping to about 11.7V, which corresponds to 8.5W at the onset of clipping. Reducing them further brings no further benefit.

You need to put T2 on a heatsink. With its collector loads halved it takes 77mA, and dissipates 1W so needs to be heatsunk. Thermal resistance junction to ambient is 100C/W and that is sailing very close to the wind. From room temperature the junction will be at 120C!

Actually, even with the original resistor values, T2 dissipates 0.56W, so the junction is 56C above ambient (76C if ambient is 20C). So needs a small dedicated heatsink for sure.

[Argus25]

Originally Posted by Craig Sawyers

Actually 10W is pretty much impossible. 10W into 8 ohms is 8.9Vrms, or 12.65V peak - and would need the amp to swing within 2.35V of the rails. And that is a tough ask even with today's amplifier technologies.

Craig

Yes with an emitter follower output it is difficult. Its not too difficult though when the output stage topology is such that the outputs come from the transistor's collectors.

[Radio Wrangler]

Originally Posted by Argus25

Yes with an emitter follower output it is difficult. Its not too difficult though when the output stage topology is such that the outputs come from the transistor's collectors.

Those circuits, while able to go closer to the rail, suffer from the output device voltage gain being dependent on the impedance presented by the load. This can cause some nasty stability problems.

When I was playing with audio power amp design back in the late 70's I was intrigued by Peter Blomley's design in WW and built one channel to try it out. I couldn't get it stable into all loads.

David

[Craig Sawyers]

Harmonic distortion is around 0.4% independent of frequency and at lower power levels, rising to 1.5% at 8.5W.

All this is from modelling by the way - I have not built the amp. If I did I'd think of a better arrangement for quiescent current setting.

[Argus25]

Originally Posted by Radio Wrangler

Those circuits, while able to go closer to the rail, suffer from the output device voltage gain being dependent on the impedance presented by the load. This can cause some nasty stability problems.

When I was playing with audio power amp design back in the late 70's I was intrigued by Peter Blomley's design in WW and built one channel to try it out. I couldn't get it stable into all loads.

David

I agree.

However, I know there were not only complaints about stability (HF oscillations etc) but one chief complaint of the collector output configuration was poor thermal stability too and the intrinsically higher output impedance from the collector being less favorable than from the emitter. And intrinsically higher distortion. Yet, if it was well executed, the results could be quite good. There is a way to stabilize the output stage gain and make it dependent only on the ratio of two resistors.

I'm sure I have a circuit example of this that was actually very good (despite the theoretical shortcomings) and it could swing near rail to rail and even in the hands of novices was stable. (I will look for it tomorrow & post the schematic/project).

Curiously you will find that the output configuration of some OP amps ( I think will have to check) use the collector out configuration, so the designers were able to "tame the Beast" so to speak in these cases too, it just might be more difficult requiring more thinking & more work.

[Argus25]

Here is the amp I was talking about.

It did require that the 2vbe transistor was thermally bonded to the heat sink and it could be either 50W or 100W depending if you added the two extra MJ2955's, it was known as the ETI-480.

The distortion was 0.05 to 0.1% up to at least 30W when it started to shoot up but at full power about 5%.

The output stage gain is simply set (as shown in the red box) by the ratio of the 110R resistor (two 220 R in parallel) and the 33R, so the gain at the output was Vin/33 = Vo/(33+110) , making Vo/Vin = close to 4.33.

(As an aside this topology is very handy in its single ended form to make a video buffer amplifier where you require a gain of 2 because you have a series 75R resistor at the output and its terminated into a 75R at the far end of the coax. Some video cameras used this, in a much lower power form of course, to drive their video out connectors).

Anyway, despite the bad reputation of this topology, I once experimented with the ETI 480, with no stability issues, I modified one to drive a reactive load (piezo electric crystals to 30W power) I had little difficulty with it. I quite like this design, though it appears frowned upon by others, for the reasons cited in my last post.

[mhennessy]

Usually for this sort of output stage, the Vbe multiplier has to be thermally bonded to the drivers, not the output devices.

Extracting gain from a CPF output stage has been widely done over the years - there are many examples of successful amps out there (e.g., see Bryston for an interesting variation). The notion of running the small-signal end of the amp from lower supply voltages is very attractive, especially as it's getting harder to source the decent small-signal, high voltage transistors that we used to use all the time. Of course, op-amps are an option too.

But there's no getting away from the fact that you have to put more effort than usual to ensure they are stable.

If going this route solely to chase voltage efficiency (i.e. how close you can get to the rails), then I'd say that's an entirely pointless endeavour for an audio power amplifier.

The ultimate expression of this is what ATC call "grounded source", or Hafler calls "Transnova". Widely used in PA amplifiers. https://www.stereophile.com/content/...oat-downstream

This is also "fun" to stabilise, but can make for a delightfully simple final design. Especially as you can contrive to have all the collectors connected to ground, thus eliminating the thermal resistance and hassle of insulation washers, which makes a big difference in a high power design.

[Craig Sawyers]

The Douglas Self blameless uses a CFP arrangement. And that manages <0.0006% at 1kHz 25W into 8 ohms and <0.003% at 10kHz at the same load and power. Similar topology to ETI-480 above, but with attention to detail and modern sustained beta power devices.