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 Club

 

Open to all owners plus those contemplating the purchase of a Graham Slee audio product wishing to use our loaner program: join here
Subscribe to our newsletter here (Rules on posting can be found here)


1970s Design Indulgence

 Post Reply Post Reply Page  <1 110111112
Author
Graham Slee View Drop Down
Admin Group
Admin Group
Avatar
Telling it as it is

Joined: 11 Jan 2008
Location: South Yorkshire
Status: Online
Points: 9517
Post Options Post Options   Thanks (0) Thanks(0)   Quote Graham Slee Quote  Post ReplyReply Direct Link To This Post Posted: 14 Sep 2020 at 5:24am
Actual real-life measurement requires the removal of the output inductor and input filter. It would need to swing a good percentage of its 63 volts P-P into the 8-ohm dummy load to give the result's credit, and fuse heating would be rapid - probably too fast for reading at its rated four amps.

The SPICE program allows simulated measurement without resorting to destructive testing.

There is also the formula for maximum slew-rate (I ips/C dom) to compare SPICE results (24.8V/uS).

Additionally, there is the Ft calculation by Horowitz and Hill, which could be manipulated, but that is based on a differential input where Re is shared between two transistors. Here it would result in a value close to 40.3V/uS. As Re applies to the single IPS transistor, do we halve the result? If so, the suggestion is 20.15V/uS.

Another way of proving the slew rate is to use a full power sine wave and see at what frequency it goes triangular. Unfortunately, I cannot reach the 100kHz required. My AP signal generator stops at 80kHz. As it is still sinusoidal in shape up to that point, I can claim 16V/uS.
Not simple enough for Google-Bot to understand...
Back to Top
Graham Slee View Drop Down
Admin Group
Admin Group
Avatar
Telling it as it is

Joined: 11 Jan 2008
Location: South Yorkshire
Status: Online
Points: 9517
Post Options Post Options   Thanks (0) Thanks(0)   Quote Graham Slee Quote  Post ReplyReply Direct Link To This Post Posted: 16 Sep 2020 at 10:46am
Unless suitable precautions are taken, every electronic circuit performs at least one other unintended function.

I think, by now, I can safely say that an amplifier is a circuit which, if taken one way, becomes an oscillator, and if taken the other way, it also becomes an oscillator.

Physics has provided us with inductance, which we can use to make oscillatory circuits to communicate with, and it's not too difficult.

An amplifier could work particularly well, if not for inductance. When thinking of inductance, most technically orientated people will think of coils, but inductance is possible in straight pieces of wire.

Such inductances are small but combine with parasitic capacitance, to make impedances which might peak, or become negative, just where they are needed least.

The human mind models phenomena with handy items, which might not be a perfect fit, but works within domains.

The remaining problem - or the only one I can now see - is the amplifier's tendency to want to oscillate in the high MHz, around 350MHz, to be more precise. Although that frequency is shown to rise, it is still considerably below unity gain, and producing only a few millivolts, should be of no consequence. However, something is drawing current and adding to the quiescent, even if only a tiny amount.

As all current flows back through the power supply, and 350MHz would be entirely unwelcome, mingling with the smoothing capacitors' inductances, then I think its removal might result in an improvement to the sound.

Doug Self's 33pf instability cure (VAS collector to ground) acts on this range of frequencies, and therefore, I believe we could be talking about the same thing.

I have found that as soon as the VAS transistor lead inductance is modeled-in, the 33pf capacitor is entirely ineffective.

If a 15pf capacitor is used to "cancel" the collector-emitter lead inductances, then the 33pf capacitor becomes effective.

However, we must model in the capacitor inductances because being connected by a wire; they also have inductance. Therefore, it is a matter of playing off one thing with another and hoping some of it works in practice.
Not simple enough for Google-Bot to understand...
Back to Top
Graham Slee View Drop Down
Admin Group
Admin Group
Avatar
Telling it as it is

Joined: 11 Jan 2008
Location: South Yorkshire
Status: Online
Points: 9517
Post Options Post Options   Thanks (0) Thanks(0)   Quote Graham Slee Quote  Post ReplyReply Direct Link To This Post Posted: 21 Sep 2020 at 6:00am
As far as op-amps go, a non-inverting amplifier can never have gain less than unity, and mostly it is at least 2.

( potential divider NFB is of the function (1/1) + 1 = 2 )

A power amplifier and an op-amp are very similar - even singleton input types.

The only parts of a power amp which push the unwanted high-frequency signal (read: stimulus) below unity are parasitically inadequate devices.

We then apply the result as negative feedback to cure the illness with the inverse of the illness.

Unsurprisingly, it doesn't work.

Noise gain is the technical term for the circuit's gain without a signal, where the function is simply 1/1 = 1. That means the gain never falls below unity. Still, we need it to fall below unity to prevent "monsters from the deep" arising, and making the circuit active - using power to produce out of band oscillations.

Those oscillations burn-out power supply storage, and therefore, bass cannot develop. We end up with a scratchy, squeaky, typically transistor sound.

Many class-B amplifiers can get away with it for a few hours but then rely on being switched off to recover. Some class-A amplifiers simply oscillate all the time.

Amplifiers are prone to oscillation in the MHz region, with modern devices going parasitic in the high MHz. Obviously, this is known, or we'd be using AM transmissions for VHF and UHF.

Early audio amplifiers might rely on poor power transistors, but now that's not an option.

The voltage stage having a noise gain of unity, allows the power stage to have noise gain of unity, where the parasitics spring to life.

Op-amp design engineers seem to understand this, but power amp designers seem not to ever think of it.

In this 100 plus page indulgence, I have looked into just about everything the world knows about amplifiers, and in doing so, I've learned a bit more than I knew, and I've learned that we're all still learning.

But one thing I did see recently was a half acknowledgment of noise gain, and one way to stop the "monsters from the deep" arising. The author, Doug Self, was unable to explain it, but the addition of a small capacitor from the VAS collector to ground, which stopped high-frequency instability - as he says, it works!

But isn't it merely reducing noise gain below unity at the VAS instead of it continuing to have a gain of unity? That is my way of looking at it. Of course, the gain should not continue because the output stage cannot do it, but the voltage amp can and is pushing it along.

One might ask, what do I mean by "monsters from the deep?" These are parasitic behaviours, or bad behaviour, due to things we amplifier designers are dumb about. The RF engineer would never use our semiconductors and our circuit drafting techniques, but we do it the way we do because it works for audio.

The "monster's" energy has to be curtailed, and the Doug Self "capacitor" does that trick. The particular part of the energy - the exact frequency - is unknown, but it seems to remove sufficient energy to prevent the onset of parasitic oscillation.
Not simple enough for Google-Bot to understand...
Back to Top
BAK View Drop Down
Senior Member
Senior Member
Avatar

Joined: 14 Mar 2010
Location: Kentucky, USA
Status: Offline
Points: 1128
Post Options Post Options   Thanks (0) Thanks(0)   Quote BAK Quote  Post ReplyReply Direct Link To This Post Posted: 22 Sep 2020 at 2:29am
"One might ask, what do I mean by "monsters from the deep?" These are parasitic behaviors, or bad behavior"... 

 Graham's "monsters from the deep" are noise micro-voltages hiding in the mess of static frequencies that are what is called the "noise floor". The noise floor is displayed as static on a frequency spectrum analyzer at the bottom of the screen.
 If any one of those noise micro-voltages are at the "wrong" frequency (or frequencies), they can excite some vulnerable component or un-damped part of a circuit to cause oscillations and cause added heat. 
 If the oscillation happens at a very high frequency, above our hearing, it can go un-noticed and the amplifier can run for years gradually cooking itself to death. 
 (Some HF oscillations happen at high enough frequencies that they aren't even seen on test equipment.)
 Yes, I have experienced this very problem.


Edited by BAK - 22 Sep 2020 at 2:30am
Bruce
AT-14SA, Pickering XV-15/625, Technics SL-1600MK2, Reflex M, Lautus, Technics SH-8066, Dynaco ST120a, Eminence Beta 8A in custom cabs;; Using Majestic DAC
Enjoy Life Your Way!
Back to Top
Graham Slee View Drop Down
Admin Group
Admin Group
Avatar
Telling it as it is

Joined: 11 Jan 2008
Location: South Yorkshire
Status: Online
Points: 9517
Post Options Post Options   Thanks (0) Thanks(0)   Quote Graham Slee Quote  Post ReplyReply Direct Link To This Post Posted: 23 Sep 2020 at 7:14am
To be able to hear on my M20 monitors a very similar "involving" experience as a ULDE driving my HD250's - at day 6 - cheered me up a bit.

The 15pf capacitor, in parallel with T2 - the VAS stage transistor - and its emitter circuitry, to the VAS star point, seems to have achieved what, to my mind, was probably the impossible.

As inferred in my last post, it seems to be taking out the energy required for the output stage transistors to self-oscillate. That the output transistors, which are no different to numerous other designs, eventually self-oscillate, and that this is not an immediate thing, should be a worry to any designer.

Whatever your feelings toward the designs of Doug Self, it has to be realised that he has provided more science about audio amplifier design than anybody else I can recall. To have reached the year 2008, and then having to find this solution, should ring alarm bells. This answer raises plenty of other questions.

It is not known why it works, but it works. It is an empirical solution whose mechanism is unknown simply because there is not any real proof. One can conjecture and possibly get it half right.

You may remember me using base stopper resistors and digging out the Clapp oscillator formulation, from which the value of base stopper resistor could be calculated. I say could, because, through lack of datasheet support, it could not be calculated!

We thus enter into intuition, which, although it is a marvelous thing, it alone cannot provide definite answers. Without information, intuition does not result in precision. It is precision that we need to be able to ensure the design is correct.

When we get to a point where even the real guru of amplifier design (a title Doug Self is worthy of) cannot fully answer, it becomes a serious worry. We need to be able to predict how long our circuits are going to deliver on their promise.

There is at least one patent that discovered and made use of resonance in the silicon die itself. Trying to find the page result in Google, when you need to, has proven impossible, so I cannot give it as reference.

I am pretty sure a 6mm silicon crystal sandwich, even when encapsulated in "solid plastic," will have a resonant frequency. With electrodes already stuck to it, some piezoelectrical effect must take place. Unfortunately, I am no RF engineer, so I have no authority to speak about such things. It would help, however, if the semiconductor manufacturers enlightened us.
Not simple enough for Google-Bot to understand...
Back to Top
Graham Slee View Drop Down
Admin Group
Admin Group
Avatar
Telling it as it is

Joined: 11 Jan 2008
Location: South Yorkshire
Status: Online
Points: 9517
Post Options Post Options   Thanks (0) Thanks(0)   Quote Graham Slee Quote  Post ReplyReply Direct Link To This Post Posted: 24 Sep 2020 at 8:47pm
Can we know if SPICE calculates class B output stages like it's done manually?

We can calculate the current gain of driver and output transistor for low frequencies, and plot the beta break frequency, considering the driving impedance of the voltage stage. Here, we can see that there ought to be a difference between the upper and lower halves of the class B stage.

Simulating in SPICE, we don't see this difference. It is essential to have sufficient phase and gain margins -- what if one half was considerably different from the other? For a start, the negative feedback would not only be commutating between positive and negative-going signals, but its error would be different on each half-cycle.

When it comes to stability, the 15pf VAS collector to ground capacitor will directly affect the PNP EF2, than the NPN EF2, because the NPN EF2 connection is via the bias spreader and its resistance. The bias spreader introduces a pole in the NPN that's not in the PNP, which is directly driven.

Using a diode chain, there is diode resistance due to diode thermal voltage, which acts like "little Re" of a BJT emitter. The three diodes might add 10 ohms, and the trimmer resistor 40 ohms. We could think of it as an equivalent real resistor of 50 ohms. The pole formed might be at 12MHz, but will tend to pull down the closure gradient from a stable 90 degrees to something steeper.

The PNP EF2 might cross at 4MHz with 100 degrees slope, but the NPN EF2 might cross at 3.5MHz with 110 degrees slope.

The 15pf collector to ground stability capacitor might only stabilise the PNP EF2, so the NPN EF2 might still go parasitic. The 15pf capacitor might make it sound better because it tames parasitic problems, but only half of them.

In fact, there are two ways of looking at this.

1. the PNP EF2 is stable, and the NPN EF2, being removed resistively, isn't.

2. the NPN EF2 is stable due to the bias spreader resistance acting as a base stopper, but the PNP EF2, not having a base stopper, is unstable.

If 1., then the resistance to the NPN EF2 needs to be bypassed by a suitably sized capacitor. If 2., then a base stopper of the approximate value of the bias spreader will be required in series with the PNP EF2 base.

How can we tell? As usual, due to the lack of information, I have to trust my ears yet again, for another week. Unhappy
Not simple enough for Google-Bot to understand...
Back to Top
Graham Slee View Drop Down
Admin Group
Admin Group
Avatar
Telling it as it is

Joined: 11 Jan 2008
Location: South Yorkshire
Status: Online
Points: 9517
Post Options Post Options   Thanks (0) Thanks(0)   Quote Graham Slee Quote  Post ReplyReply Direct Link To This Post Posted: 25 Sep 2020 at 7:13pm
Originally posted by Graham Slee Graham Slee wrote:

Can we know if SPICE calculates class B output stages like it's done manually?


I forgot to explain this.

SPICE AC modelling is small signal, and as such it models both sides of the power stage as if they're two amps, mixing at the output. So it isn't done like it is manually.

There is a way to separate the two halves, by using a couple of 10 terra Henry inductors. Such an item could never be made, by the way.
Not simple enough for Google-Bot to understand...
Back to Top
 Post Reply Post Reply Page  <1 110111112
  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 0.109 seconds.