Cambridge Audio P40

[Craig Sawyers]

Originally Posted by Beobloke

I think you’re missing the point. No one is saying that the Amstrad is the be all and end all of amplifier wonderfulness, merely that it is a sound basic design because its circuit comes virtually straight out of the Toshiba applications book! The fact that it appears to have been thrown together in batches on the Friday afternoon before a bank holiday weekend does not detract from the fact that its basic circuit is decent enough.

Fair comment. My reaction was to the assertion that the P40 was unreliable, badly designed rubbish that sounded awful, and that the Amstrad was a whole world better. Which, certainly as far as reliability is concerned, seems not to be the case.

As an aside, I’ve just noticed that an, admittedly very tidy, Amstrad TP12D sold today on a well known auction site for £194! The world truly has gone mad...

Good grief. I'm lost for words.

Craig

[Radio Wrangler]

I think we need a name for it.

How about "The Dansette Effect"?

I had a quick shufti to see what the going rate for a P40 was... winning bid £100.

As our Atlantically-challenged cousins say, go figure!

David

[Ted Kendall]

As John Culshaw said in another context - "what is trivial to one generation is sometimes very valuable to the next"...perhaps people think the thing is a proto-Rega or related to the Pioneer PL-12D, which also seems to go for mad money.

As I said earlier, regrdless of topology, Amstrad was built cheapo. The Cambridge wasn't.

[Trevor]

I sold a P40 a few months ago did not get much for it I had the thing in the garage for about 10 years and needed to reduce the size of my collection.
Visually a super design inside a travesty lots of hard work and bits of glue thermally a shocker but it did sound nice only used it with a CD player for testing
I used to be in the Hifi trade during the 60- 90's They has a good reputation and went through the doors by the shed load
It seems to me we are only happy when knocking our manufacturers
Trev

[Hartley118]

I did once enjoy an extended discussion with the Chief Engineer of Amstrad. He was very forthcoming about the market addressed by the company. Alan Sugar's target customer was the 35-year old lorry driver, who wanted a product that looked the part, but wasn't too concerned about technical performance. I guess that, on average, that objective was achieved.

Regarding Cambridge Audio, my personal experience has been mixed. Around a decade ago, I bought a couple of Cambridge Audio units - from Richer Sounds of course. The CD player has been excellent, and is still going strong today. By contrast, their DAB/FM tuner was a disaster. OK on DAB, but a dreadful on FM, even with my external 4-element aerial. Background noise level was awful - much worse than a worn 78 record. Thinking I had a bad example, I returned it and Richer Sounds readily exchanged it for another sample. That was just as bad. Again, Richer Sounds were very helpful and exchanged that for a Denon tuner which worked fine and has continued to do so ever since.

I guess that our judgements are based on personal anecdotal experiences. So mine on Cambridge Audio are: capable of excellence, but production quality unreliable: be very careful. On Denon: excellent in my experience.

Martin

PS I too knew and liked Gordon Edge - Impressive chap and a great achiever. I worked across the road from his PA Technology lab.

[Radio Wrangler]

Originally Posted by Hartley118

OK on DAB, but a dreadful on FM,

I bought a Roberts Colourstream bedside radio and those words describe it perfectly.

It's Philips predecessor was a mono AM/FM alarm radio and worked from in internal ferrite rod for the AM bands, and used the mains cable for VHF. It worked fine.

The Roberts has a large whip that needs to be well extended and slewed around to get DAB working properly. It seems to make no difference to VHF/FM reception which is simply unusable.

Is it just cheapening or are designers under instructions to make DAB look as good as possible? No, too complicated. Occam's razor says cheapening.

David

[Hartley118]

Originally Posted by Radio Wrangler

I bought a Roberts Colourstream bedside radio and those words describe it perfectly...........
Is it just cheapening or are designers under instructions to make DAB look as good as possible? No, too complicated. Occam's razor says cheapening.

David

I certainly drew the conclusion that the Cambridge Audio factory test spec didn't bother to include the FM band. Must have made the assumption that if DAB was OK, then so would FM be. Wrong!

Martin

[Craig Sawyers]

Originally Posted by Hartley118

PS I too knew and liked Gordon Edge - Impressive chap and a great achiever. I worked across the road from his PA Technology lab.

At Neve?

I worked at PA Technology from 1982 to 1986, and then left with Gordon and a group of other renegades to set up Scientific Generics in 1986 (Now Sagentia). First from our bedrooms, then from a new building in King's Hedges (now demolished). That was before I moved to Wharfedale in 1990 as CTO and then Oxford Instruments in 1993 in the same role.

[Synchrodyne]

[

Originally Posted by Radio Wrangler

The P40 was an interesting amplifier at the time, and certainly was an offering at the altar of overload margin, but that feedback gain control allowed it to segue from reasonable low noise performance with the knob set to high gain into great overload performance with the knob set to low gain.

As Synchrodyne said, the reviewers wanted both low noise AND high overload margin at once. This circuit very cleverly gave it to them. They loved it!

If you're only interested in playing records, the P40 is a great amplifier. If you also have a tape recorder, it becomes a pain in the rear.

A better design would have needed some better way of rationalising the gains of the recording and the listening channels.

Looking back with 2020 hindsight, the P40 gain control was very clever, but not really needed. The reviewers were being unrealistic and a phono stage which can be set to suit the output of a cartridge does the job nicely - not only for the listen path but also for the tape path.

The P40 was designed to get good reviews and therefore sales. Managing the gain control business to make it suit tape recording would have been very complex and very expensive, and killed sales. It was a child of its times.

It is a product of a style of reviewing, the fixation on one parameter to the exclusion of other matters. The car industry has long been dominated by 0-60 times. You can 'improve' a classic mini by fitting a 5-speed gearbox... but those gearboxes fit it all in by deleting reverse! The P40 gets marvellous overload margin on disc, but you can't change the listening volume if recording. If you only listen to discs, buy a P40. It is a very clever design, but it is focused on only one thing.

The thing in my lounge has an NE5534 phono stage published by Peter Baxendall. Series feedback with that added pole on its tail. Its gain is scaled to suit the cartridge I use. After switchery to feed the listening and recording paths comes a feedback gain control stage using a silver-stud fader, then my version of the Quad 'Tilt' control, and a driver for the cables to the power amps.

Sometime, when I get round to fixing that B261, it'll go in the lounge. It has individual input buffers, a phono stage with settable loads and sensitivities, and a feedback gain control (and another for the record path) IT was certainly designed for a recorder owner!

David

I suppose the cleverness was that the volume control served simultaneously as a both a conventional volume control and as an input sensitivity control. With a higher output cartridge, lower normal volume control settings would be used than with lower output cartridges, so the overload point was automatically adjusted according to cartridge output. But on the other hand, being forced to use just a fraction of the volume control swing is a pain. Having individually buffered inputs with selectable sensitivities is a much more elegant solution. One may then switch between sources without major volume changes, and be able to use a sensible volume control setting range.

On signal-to-noise ratio, for moving magnet cartridge RIAA-equalized input amplifiers, H.P. Walker calculated theoretical best signal-to-noise ratios (relative to 2 mV at 1 kHz) of 72 dB for the series feedback case and 58.5 dB for the shunt feedback case. The measured numbers were 70 dB for his 1971 series feedback design and 58 dB for Linsley Hood’s 1969 shunt feedback design. In 1983, in connection with his Precision Preamplifier design, Doug Self measured 84 dB, relative to 5 mV (which equates to 72 dB relative to 2 mV), for his series feedback design based on an NE5534. The latter was I think the first reasonably available opamp that could match or slightly better discrete circuitry and so approach the theoretical.

I have a vague recollection that Cambridge quoted a 60 dB signal-to-noise ratio for the P40 disc input. This might have been relative to its 3 mV nominal sensitivity or the then-customary 5 mV number. 60 dB relative to 3 and 5 mV respectively equate to 57.5 and 52 dB relative to 2 mV. So if the former applied, then the P40 shunt feedback circuit was close to the theoretical.

Back in the late 1960s, there was a trend to lower noise in recognition that the better cartridges were trending lower in specific outputs, generally towards 1 mV/cm/s, with I think the Shure V15 Type II Improved being pretty much at the bottom of the curve at 0.7 mV/cm/s. So one might have expected the aim point for an avant-garde amplifier of the time to have been towards 70 dB relative to 2 mV.

I am not sure where the Darlington pair emitter follower input buffer used in the Cambridge P50 would have fitted in terms of noise relative to more conventional inputs. Insofar as an emitter follower effectively has nominally 100% series feedback, perhaps it was closer to the series feedback case?


Cheers,

[Radio Wrangler]

Structurally, an emitter-follower is a series-feedback case. Although tied to slightly below unity voltage gain it can still give power gain if the source and load impedances presented to it permit.

For any particular transistor and bias current, there is an optimum source impedance at which it achieves its best noise figure. For RF devices this is routinely expressed as a complex impedance and contours of constant noise figure can be plotted on a Smith chart. There is an interesting compromise as the impedance giving the best noise is rarely close to that for the best gain. Most ultra-low noise things are poor matches on their inputs. At audio, we just forget the imaginary term.

When Shure etc say "Please load this cartridge with XX kilo Ohms" They mean that literally. It doesn't mean that the output Z of the cartridge is anything like the specified load. The load is the Q-damper for the cartridge's resonances. The cartridge's actual output impedance is what the transistor sees, modified by the circuit it is embedded in.

The emitter follower circuit means that the noise-optimum source Z goes quite high. So I wouldn't expect an emitter follower buffer input for an RIAA stage to be at all good on the noise front.

Transistors can be made for very low source Z's (like MC cartridges) either by being very wide junction types or by paralleling a lot of them. There was one type made by Rohm for stepper motor coil driving which was very wide and had managed current sharing with a very low base spreading resistance parameter. It featured in lots of studio mixing desk mike inputs. That industry was very upset when they stopped making it. Can't remember the number 2SC7-something something.

David

[Craig Sawyers]

Originally Posted by Radio Wrangler

Transistors can be made for very low source Z's (like MC cartridges) either by being very wide junction types or by paralleling a lot of them. There was one type made by Rohm for stepper motor coil driving which was very wide and had managed current sharing with a very low base spreading resistance parameter. It featured in lots of studio mixing desk mike inputs. That industry was very upset when they stopped making it. Can't remember the number 2SC7-something something.
David

2SB737. Probably the lowest base spreading resistance transistor made, and the quietest as a single device by quite a margin, and superb 1/f noise. At 10mA Ic, 0.55nV/rootHz at 10Hz and 0.4nV/rootHz at 1kHz and a typical rbb' of 2 ohms (4 ohms max).

Lots of professionals and amateurs alike lamented its passing.

Like the SSM2220, monolithic pnp matched pair with similar performance to the 2SB737 - discontinued last year. So went the way of the dodo with the MAT02/MAT03.

Of course the problem with these low noise transistors is that the come with massive input capacitance, and are best run in cascode configuration

There are alternatives with similar performance, but since the electronics industry seems hell bent on discontinuing low noise pnp processes it is surely a matter of time before it all goes the way of all dust. If it doesn't go in a mobile phone or similar global high volume market they seem to lose interest.

Craig

[Radio Wrangler]

Thanks, Craig.

Hmm

"Is it used in a callphone?"
"Nope"
"Tablet computer?"
"No"
"Can we say it's useful for the Internet of Things?"
"Not really, but we tried that."
"Electric car?"
"'fraid not"
"Better obsolete it then..."

David

[Craig Sawyers]

It all comes down to industry consolidation, and focusing on specific markets. As an example, Vishay went through a massive acquisition phase, and each company they bought had many lines discontinued, usually the "useful" ones - low noise devices, high voltage pnp, dual transistors and dual FET's - the list goes on. I count 27 companies that Vishay now own - many of them well known names from the 70's and 80's when they were independent https://www.vishay.com/company/brands/ . And that says it all - "brands", and a company structure that appeals to investors but misses the point that they sell to customers who want product, not brands.

<soap box mode> = off

Craig

[Synchrodyne]

Originally Posted by Radio Wrangler

The emitter follower circuit means that the noise-optimum source Z goes quite high. So I wouldn't expect an emitter follower buffer input for an RIAA stage to be at all good on the noise front.

So it would seem unlikely that the Cambridge P50, with its Darlington emitter follower disc input stage, would have improved upon the P40 when it came to signal-to-noise ratio.

Cheers,

[Craig Sawyers]

I've put in a spice model for the P40 input buffer. I've included a moving magnet cartridge impedance of 500mH in series with 600 ohms. The 2N4288 spice model is complete and includes all internal noise sources.

Referred to 3mV at 1kHz, the signal to noise ratio is 58dB unweighted, so with around 2.5dB gain with A weighting an overall S/N ratio of 61.5dB meets the original published specification. Not surprisingly that is dominated by the Johnson noise in the input resistor. This of course will be modified by the RIAA stage spectrum.

The trackmeisters Shure measured on the most extreme vinyl a peak velocity of 70cm/s, which corresponds to an overload margin of 24dB (17x). With the volume control half way and 3mV nominal input 0.003%, and with 24dB overload (so 41mV input) 0.04%.

So with the proviso that the S/N ratio being -60dB ref 3mV, distortion and overload margin are very good indeed.

Craig

[Craig Sawyers]

I've now put in the buffer stage and RIAA stage. Since each of the three stages are based on emitter followers, distortion remains low even at maximum gain.

The RIAA accuracy is, in terms of today, very average at around +/-1dB, but typical of the day and limited by E12 resistor and capacitor values. The Quad33 is similar in accuracy to the P40. The Amstrad RIAA EQ is, candidly pretty dire, with critical design time constants in the range +38% to -25%.

All these designs were done well before Lipschitz's seminal paper "On RIAA Equalization Networks" JAES, June 1979, V27(6), 458-481.

Craig

[Craig Sawyers]

OK - I'm probably just having a monologue here - but does that stop me?

Fascinated by what the fuss was about the Amstrad being a great design and the P40 being rubbish, I tracked down the (Japanese language) Hitachi manual regarding the application of the TH9014P, in June 1968. Each device is a hybrid, with two identical amplifier stages (for L and R). These are each classic two-transistor gain stages, with the unusual feature of an NPN bipolar input stage (2SC732), and an N-channel JFET (2SK30) second stage. I went so far as to buy one (about a tenner), and by careful measurement it is possible to work out the resistor values.

Both calculation and Spice modelling confirms an open circuit gain of 76dB (6,300 times). Open loop distortion is low at about 0.01%. Interestingly, closed loop distortion (via the RIAA network) is about the same.

Input referred noise is about 4nV/rootHz, or an equivalent noise resistance of 1k - or about the same as the cartridge. So all good there.

Overload margin depends rather critically on the bias point, and therefore on the hfe of the transistor and the Vp and Idss of the JFET. Since none of those parameters is well controlled, the devices must have been selected to quite close tolerances. Although it has DC feedback, that can only correct so far. Since they were Hitachi transistors, selection was probably not a big problem.

The particular choice of component values means that to prevent loading effects causing increased errors (in an already poor RIAA response) means that in Hitachi's suggested design the total stage loading is ~82k. That is why in Douglas Self's improved 2-transistor RIAA he adds an emitter follower to drive the RIAA network and present a low drive impedance to following stages.

Anyhow, so much for theory and modelling. Next is to apply 30V and see what the thing does. I suspect that adding an emitter follower inside the loop (external to the hybrid) and putting an accurate RIAA network in place and it might have halfway decent performance.

Incidentally, the 1968 price for these was 1000Yen, or about £1.15 (at the 1968 exchange rate), which is £20 in today's money. The corresponding power amp, the TH9013P, was 3000Yen, so £60 each today (and you need two).

So the total for two (for Hitachi's suggested design) TH9014P and two TH9013P corresponds to £160, before you pay for the case, transformer and the rest of the hardware.

Craig

[Craig Sawyers]

Apologies - the three transistor RIAA incorporating an emitter follower was due to A R Bailey, WW, December 1966, and not Self. Self improved it with bootstrapping.

[Craig Sawyers]

Continuing my monologue. I plotted out the RIAA deviation for the Hitachi, Amstrad and Cambridge audio P40. This is Spice modelling using a precise pre-emphasis circuit using a perfect op-amp, using component values accurate to three places of decimals.

The Hitachi and Amstrad RIAA EQ have the usual R//C + R//C network and the P40 has R//C + C and an overall parallel R.

The values for the Amstrad and Hitachi designs using the TH9014P are 330k//6.8nF + 33k//2.2nF for the Hitachi circuit and 330k//10nF +33k//2nF for the Amstrad. The Cambridge Audio has 10k//6.8nF+ 22nF all paralleled by a 120k

Over 20Hz to 20kHz:

Amstrad +/- 1.13dB
Hitachi +/- 1.23dB
P40 +/- 1.38dB

The shape of the curve is of course different for each amp, so it emphasises different bits of the audio spectrum.

Of course these are pretty dreadful by today's standards. But the circuits for the Amstrad and Hitachi using the TH9014P are easily fixed.

330k//10n + 27k//2.7n give an RIAA error of +/-0.17dB which is an astonishing improvement by a fairly minor change in values - and keeps with easily available E12 series parts.

A similar improvement can be got from the P40, but I have failed thus far in keeping with simple values. One of them always ends up with an awkward value.

Craig