New Posts New Posts RSS Feed - 1970s Design Indulgence
  FAQ FAQ  Forum Search   Events   Register Register  Login Login

Welcome to the Graham Slee Audio Products Owners Forum

 

Open to all owners plus those contemplating the purchase of a Graham Slee HiFi System Components audio product and wishing to use this forum's loaner program: join here (Rules on posting can be found here)

This website along with trade marks Graham Slee and HiFi System Components are owned by Cadman Enterprises Ltd


1970s Design Indulgence

 Post Reply Post Reply Page  <1 8990919293 345>
Author
Graham Slee View Drop Down
Admin Group
Admin Group
Avatar
Retired

Joined: 11 Jan 2008
Location: South Yorkshire
Status: Offline
Points: 16298
Post Options Post Options   Thanks (0) Thanks(0)   Quote Graham Slee Quote  Post ReplyReply Direct Link To This Post Posted: 06 Dec 2019 at 10:18am
A rather big blunder...

Dinsdale did it; Reg Williamson did it; scores of others did it; and this time I did it!

"an inevitable consequence (of rail decouplers) is that rail-voltage variations cause current to flow into the ground connection" (Self. D, annotated by me)

See images:



Above image subject to copyright: Doug Self, Audio Power Amplifier Design Handbook, 5th edition
That none should be able to buy or sell without a smartphone and the knowledge in how to use apps
Back to Top
Graham Slee View Drop Down
Admin Group
Admin Group
Avatar
Retired

Joined: 11 Jan 2008
Location: South Yorkshire
Status: Offline
Points: 16298
Post Options Post Options   Thanks (0) Thanks(0)   Quote Graham Slee Quote  Post ReplyReply Direct Link To This Post Posted: 06 Dec 2019 at 11:57am
But is there still room for the stereo loop resistor? Does the ground distortion introduced by the decoupler, having been solved, remove stereo loop distortion?

Theoretically it must still exist simply because the smoothers have ESR (and ESL), and ground here is one of the HT rails (where it isn't in a DC coupled amp).

There is also the problem that ground being -HT it cannot be decoupled to itself. Therefore, it could be thought that we've only gone half way, and that is probably true.

If this is the case, then the ground will still want to shift slightly in sympathy with the current being drawn by the load, and taking the decoupler away from the star is simply kicking the can down the road - but audibly, it helps.

So, this current is bound to flow contra from one channel to the other (via the input ground connection).

This current has been reduced by removing the positive rail distortion (being unregulated) further from the star ground, but there is still the pull due to the load (the speaker) which is different to the other channel.

An early RCA design, and the Boosey & Hawk variant used -HT resistance mixed with output stage voltage at AC. If you think about it, all rails are the same at AC. But their's was probably a grounding blunder like above, as it didn't feature in later designs.

The stereo loop resistor can have unwanted consequences including a phasing effect which can induce tinitus. The larger the resistor, the worse this gets.

There is also the stability issue if not decoupled at very high frequencies. In this single rail design the decoupling is back to the star, where in the RCA it was to ground (dual rail supply remember).

10 ohms produced the worst phasing; 1 ohm did hardly anything, but sounded better than nothing; 6.8 ohms sounded least distorted but some phasing evident; 3.3 ohms was just about right.

With the rail distortion "half improved" perhaps 3.3 ohms will do the rest?

But it would be good to do this scientifically. If only we knew how.
That none should be able to buy or sell without a smartphone and the knowledge in how to use apps
Back to Top
Graham Slee View Drop Down
Admin Group
Admin Group
Avatar
Retired

Joined: 11 Jan 2008
Location: South Yorkshire
Status: Offline
Points: 16298
Post Options Post Options   Thanks (0) Thanks(0)   Quote Graham Slee Quote  Post ReplyReply Direct Link To This Post Posted: 07 Dec 2019 at 7:26pm
Commercially recorded vocals can contain occasional sibilance due to microphone proximity being too close. Essing as it is called can cause clipping of the microphone signal making the sibilance worse. In replay the amplifier will attempt to reproduce the signal which may be spread between left and right channels but might not be same in both channels, and might have been mixed over to one side using stereo panning.

A signal rich in such sibilance to which is added regular high hat and cymbal strokes, will have the output stage fluctuating wildly the current being drawn from the power supply, which will be in sync and out of sync with the current being drawn by the other output stage.

These signals ride on the low, more regular rhythmic sound frequencies. The power supply is therefore not optimal for both channels together, and the HT can fall such that its voltage is in the dips of ripple. The current flowing in the return - the 0V power supply connection - will be different in one channel with respect to the other, and rapidly changing.

The unavoidable loop formed at the input ground (the common connection shared by both left and right signals) will have current being pulled for one channel through the opposite channel's input ground, and this is an asymmetrical modulation voltage. Not only does the crosstalk suffer, the signal is badly distorted.

Breaking the connection by 10 ohm grounding resistors simply moves the ground common to the source equipment, and often the connection is made at the source's input. If that is a phono cartridge input, the signal will suffer a greater degree of distortion than a line level signal.

Should the 10 ohm resistors go open circuit due to mains breaking through because of a burning-out transformer, there is no earth protection, and the source equipment not being earthed can become live.

10 ohm input grounding resistors, although often used, should never be used.

Because of the resulting direct common (using the safe method) the loop cannot absorb the "current jolts" because of the music being played. The input ground wiring being the weakest point, will distort the input signal.

Shifting the decoupler ground helps reduce this effect, but cannot resolve it completely, as the outputs to the loads depletes the HT unequally. The complete cure is to use monoblock power amplifiers, but in designing a stereo amplifier there has to be some clever solution to achieve the same.

Separate transformers for each channel help greatly because they isolate the 0V's at the power end of each amplifier. In actual fact they don't fully isolate the 0V's because this is AC, and zeros are only relative, so the grounds actually connect at the amplifier mains input. This being the reason for supply polarity affecting, more or less, hum in some phono equipped, double-insulated integrated amplifiers.

Where separate transformers cannot be accommodated in say a compact case design, we could use dual secondary’s, but in effect the secondary’s are joined in parallel at AC without any external connection. This is because of mutual inductance - the intimate contact between secondary’s and the primary - and is why secondary inductance does not obey the parallel inductors rule in a transformer.

The "invisible" connection between "isolated" secondary’s is evidenced when adjusting quiescent current and output voltage centre point to minimise distortion, as adjusting one channel effects the other channel to some degree.

Therefore, in a dual secondary supply there is still power-end 0V commoning, and therefore some stereo-loop distortion might make itself present.

In a dual rail supply amplifier some averaging might take place which assists the power end 0V remaining at zero, but not in a single rail supply.

At worst the stereo-loop distortion could be blamed on other things, and I'm sure it has, and I think the output capacitor has often taken the blame. If you think about it, if stereo-loop distortion is worsened by a single rail supply, and single rail amplifiers have an output capacitor, then it might be seen as prime suspect.

Bringing back the 0V end of the decoupler via a separate trace demonstrates where the "levers" for this distortion are located, and tends to cement the argument for the existence of stereo loop distortion.

If we then add Dinsdale's observations, we can be reasonably confident that we've found the source of the problem.  Inserting a resistor between voltage and current stage 0V's (the VAS 0V to the star) changes the asymmetrical ground current that would flow the wrong way around the loop, into a voltage, which is an error voltage which the global negative feedback corrects for.

The problem in doing this, the mixing of ground currents between left and right, means we have created an accidental stereo-width enhancer.  It is enhanced by the asymmetry, and if too much is used, we will arrive at left-right distortion again.

As Dinsdale commented; "Although this causes a slight increase in overall distortion on mono signals the improvement on stereo signals is impressive". I think this is an indication of the accidental stereo-width enhancer to which I just referred.

The hundreds of millivolts Dinsdale also refers to are in part due to the grounding regime, which was improved to nearer what Doug Self found. This was in a later issue I have not been able to find yet. Dinsdale considered it sufficient to dispense with the resistor. However, when trying this myself I have felt cross-channel distortion is still present.

If the improved grounding regime leads to a smaller distortion voltage of only a few millivolts, then a smaller resistor (than 10 ohms) should suffice, and perhaps the extent of stereo "enhancement" will also be less? In experiments using the 10 ohm resistor I was aware of sufficient mixing because it gave rise to a phasing effect which induced tinnitus.

With this "large" value, compensation by a bypass capacitor was also required to maintain gain margin. The value of capacitor gave a frequency far beyond the audio spectrum, so I don't think that contributed to the phasing effect. Smaller resistor values improved the subjective results, and I am now at 2.2 ohms, having gone as low as 1 ohm, where in my opinion, stereo-loop distortion rose again.

Returning now, to the transformer dual secondary's, and that paralleled windings indicate the same inductance as a single winding. If this was exact then it would indicate zero impedance between the two, thus making the power supply no different to one where both channels are supplied from the same secondary.

However, there are losses in all magnetic circuits, and the inductance isn't 100% the same for parallel versus single - there is a very slight difference in measurement. The inductance meter measures at a frequency which allows it to provide the most accurate result, this not being mains frequency, but around 1kHz.

There might be sufficient difference at other frequencies such that the impedance between secondary’s varies from zero. In that case the loop-resistor value can have very different effects at different frequencies. It was at this point I started to consider bridging the secondary’s which effectively makes it a single, but higher VA, single secondary winding.

So now it's time to try that.
That none should be able to buy or sell without a smartphone and the knowledge in how to use apps
Back to Top
Graham Slee View Drop Down
Admin Group
Admin Group
Avatar
Retired

Joined: 11 Jan 2008
Location: South Yorkshire
Status: Offline
Points: 16298
Post Options Post Options   Thanks (0) Thanks(0)   Quote Graham Slee Quote  Post ReplyReply Direct Link To This Post Posted: 08 Dec 2019 at 11:42am
The left-right differences need to be assessed further. My last conclusion seems a little silly.

The main difference between two monoblocks and a stereo power amplifier are now obvious, and that is the apparent lack of stereo-loop distortion in the monoblocks.

The stereo amplifier has to be made as near to two monoblocks in one case as possible.

There are practical limits however, one being the cost. Go the whole hog and it becomes an unreachable dream. The idea from my point of view, is to make the best available to all.

Currently we have two transformer secondary's and each amplifier has its own bridge rectifier and smoothers.

The unavoidable stereo loop joins at the transformer secondary's, and at the signal input grounds, and we need to see further into what's happening.

One thing I have not dealt with so far is the interaction between the bridge rectifiers: pull more current through a diode and its forward voltage (Vf) increases.

This might give some benign assistance to the stereo loop for an nth of the time; then at other nth of the time, it might be harmful.

Either way, it makes the case for separating the secondary's, and shows why the two transformer solution worked.

But let's say I want to keep it as a single transformer, what option do I have of unsticking the secondary's from each other?

It has to be non-invasive, and dual secondary's cannot really be separated, so I can only see the use of series resistance as being the way.

A single resistor on one end of each secondary can't really do the job, because, in this theory, the ends are just as stuck together as any other place in the windings.

How much voltage can we afford to waste? Will 1% do it? The price paid will be in the loss of maximum power (again). What would we rather have, quality or power?

All I have to hand is some 0.27 ohm 3W resistors. One either end of each secondary will lose 1.2 volts AC in total, and might translate to a 2W loss in output.

But, as we now know, this is part of the lever which works; it will have to be tried.
That none should be able to buy or sell without a smartphone and the knowledge in how to use apps
Back to Top
Graham Slee View Drop Down
Admin Group
Admin Group
Avatar
Retired

Joined: 11 Jan 2008
Location: South Yorkshire
Status: Offline
Points: 16298
Post Options Post Options   Thanks (0) Thanks(0)   Quote Graham Slee Quote  Post ReplyReply Direct Link To This Post Posted: 08 Dec 2019 at 5:18pm
After a lot of cross-referencing and thanks to http://www.douglas-self.com/ampins/library/ampartew.htm, I eventually found the Dinsdale ground correction presented by C Artus in Letters to the Editor WW Feb 1965.


That none should be able to buy or sell without a smartphone and the knowledge in how to use apps
Back to Top
Graham Slee View Drop Down
Admin Group
Admin Group
Avatar
Retired

Joined: 11 Jan 2008
Location: South Yorkshire
Status: Offline
Points: 16298
Post Options Post Options   Thanks (0) Thanks(0)   Quote Graham Slee Quote  Post ReplyReply Direct Link To This Post Posted: 08 Dec 2019 at 5:21pm
The problem with the above is there is no true a.c. earth because of smoother ESR.

(Edit: my assumption here is wrong and corrected near the end of my next post)


Edited by Graham Slee - 09 Dec 2019 at 7:49am
That none should be able to buy or sell without a smartphone and the knowledge in how to use apps
Back to Top
Graham Slee View Drop Down
Admin Group
Admin Group
Avatar
Retired

Joined: 11 Jan 2008
Location: South Yorkshire
Status: Offline
Points: 16298
Post Options Post Options   Thanks (0) Thanks(0)   Quote Graham Slee Quote  Post ReplyReply Direct Link To This Post Posted: 09 Dec 2019 at 5:55am
If components didn't require a burn-in period, this job could have been condensed into a few weeks rather than more than a year.

The secret of analogue audio design is being accustomed to the mechanisms which result in particular shades of distortion.

However, those "shades" take time to reveal themselves, so nobody can claim a particular performance level on a 9 'till 5 basis.

I truly wish it were not like this, and I can seriously understand how damaging to the bottom line the admission of the existence of "burn-in" would be. But it's simple; what sounds like an improvement might not sound like an improvement given sufficient time.

Like all designs, this one has been no different, except it's easily the "longest-distance" one I've done to date. They usually take time, but it’s easier to find the mechanisms, or levers, which move it back to convergence.

Self on decoupler distortion explains that you can do everything else that's possible and get nowhere until decoupler distortion is solved.

I will now state that you can do everything that's possible and get nowhere until "ground" fluctuations are solved. This might have to be qualified by saying "in the application of a single rail supply".

I owe it to Dinsdale for providing the clues. Others took him to task suggesting a degree of incompetence, but those others never explained the mechanisms responsible for grounding distortion in such an understandable way. Do it their (now) textbook way and spend the rest of your life wondering why it's not quite as expected.

A big clue was the way in which he later defended himself by proclaiming this wasn't such a big problem with valve amplifiers, that being due to the much lower currents involved, and that is true. I had already sussed that one, but was pleased to see it confirmed in print.

Transistor amps are by comparison high current - low impedance devices, so impedance matters, and the textbook grounding scheme might not yield the best results. Parasitics describes the infection of a device by parasites, and there are a lot of them. They are due to components which will never be perfect. Every passive is also two other types of passive - take your pick from L, C and R. L has C and R; R has C and L; and C has L and R.

(L = inductance; C = capacitance; R = resistance)

Every active device, such as a transistor, has all three. Wire is L and R; PCB copper traces likewise. L and C have time functions, and the current at the hard work end is in amps, not milli-amps.

AC; alternating current, and whatever its frequency at any particular time, cross connects where you might not think it can; because basic understanding treats connections as being real-world "wired". In AC the imaginary factor j is found (as in jw). What it is saying is use your imagination - test all things - hold fast what is good - then keep doing it!

This adventure has looked at every conceivable permutation possible using a six transistor single rail amplifier configuration of the type found in thousands of 1960s to 1970s amplifiers and receivers; but at the end of the day, the circuit common has been found to be the Achilles heel.

Assuming the negative end of the smoother is a fixed reference might not be the correct assumption. OK, it looks like a reference point, but is it?

Everything can be all over the place but at that star point's convergence, all that's connected to it is still, which it will be. The problem is that star point connects to at least one other.

This is where it should dawn on us that separate power supplies are the problem.
That none should be able to buy or sell without a smartphone and the knowledge in how to use apps
Back to Top
 Post Reply Post Reply Page  <1 8990919293 345>
  Share Topic   

Forum Jump Forum Permissions View Drop Down

Forum Software by Web Wiz Forums® version 12.01
Copyright ©2001-2018 Web Wiz Ltd.

This page was generated in 0.141 seconds.