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 8485868788 92>
Author
BAK View Drop Down
Senior Member
Senior Member
Avatar

Joined: 14 Mar 2010
Location: Kentucky, USA
Status: Offline
Points: 938
Post Options Post Options   Thanks (0) Thanks(0)   Quote BAK Quote  Post ReplyReply Direct Link To This Post Posted: 26 Sep 2019 at 6:30pm
Originally posted by Graham Slee Graham Slee wrote:

My guess right now is that the power transistors are quite susceptible to parasitic oscillation, and the phase lead compensation might have addressed this by extending gain margin considerably, minimising the parasitic effect. My problem is not being able to see this on the test gear because I'm unable to filter out all the ambient garbage which seems to live at the frequencies of intererst. Filter it out and you filter out the parasitics. So we are in imagination land.


 A Faraday Cage may not help in GigaHertz Land.Wink
You can only imagine, conceptualize the problem, then try a possible solution empirically.
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: Offline
Points: 8905
Post Options Post Options   Thanks (0) Thanks(0)   Quote Graham Slee Quote  Post ReplyReply Direct Link To This Post Posted: 26 Sep 2019 at 9:52pm
Bode plot amp 260919

Using nearest available component values:

With loop gain of approx. 19dB, the open-loop cut-off frequency is approx. 16kHz. Preferably this should be 20kHz allowing 20dB loop-gain, but this is the best I could achieve.

cut-off frequency = F-3dB closed-loop/loop-gain

0dB crossing is at 4.1MHz; phase is -101°.

Unconditional stability is defined as </= -135° and </= -10dB. Here it is -135° at -26dB, so any parasitic oscillation near 60MHz has less chance of establishing itself.

Maximum phase displacement is -175° at 1GHz, so it never reaches -180°(the phase lag at the onset of oscillation). However, the gain has dropped by 68dB, which should overcome any parasitic oscillation if it were to occur at 1GHz.

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: Offline
Points: 8905
Post Options Post Options   Thanks (0) Thanks(0)   Quote Graham Slee Quote  Post ReplyReply Direct Link To This Post Posted: 29 Sep 2019 at 9:46am
Casting an eye at the IPS and VAS circuits with a "perfect" output stage (substituting a voltage controlled voltage source), the effect of different compensation schemes can be seen.

Lag/lead and lead/lag compensation schemes both achieve the same aims of high frequency stabilisation. Hereafter I will call it LL compensation.

Miller dominant pole compensation seems to be the lazy man's approach to obtaining low distortion, but by comparison it doesn't stack up.

Self and Cordell seem to think that it reduces high frequency distortion because it is local NFB, but Self's THD sweeps still show it rising. One might ask "compared to what?" because if Cdom wasn't there, neither would the amp: it would self-destruct. Therefore the local NFB argument can only be an assumption.

Otala rightly explains that Cdom charging is asymmetric: in one direction it is low impedance via the IPS transistor, and in the other, high impedance via the IPS collector resistor. Wilson current mirrors fix this, but at the cost of 5 IPS transistors.

Even so, slew-rate (inbound charging) is limited by IPS current: SR = I(ips)/Cdom. Example: Cdom 47pf; IPS current 0.5mA, gives 10V/uS.

Beyond the slew rate limit, the IPS drives the output stage via Cdom, with only a fraction of the required current - the VAS is paralysed.

It should be remembered that all amplifiers fall to unity gain eventually, and anything below unity is short lived. Some may argue that an inverting amplifier can, but for all amplifiers noise gain holds true. Noise gain isn't signal gain (NG = (Rf/Rg) +1, in either case).

That the signal (or "noise") dips below unity is only because of the short circuiting applied by Cdom, after which it rises back toward unity.

Regardless of transistor model, it is found that at roughly 40MHz and 400MHz transistors oscillate, and this happens as the "noise" settles back to unity.

Just before this, at 31MHz, a typical plot shows amplitude at -12dB with Cdom compensation, but due to its shorting behaviour, phase is -180°.

By comparison, using LL compensation, the same frequency (31MHz) has hardly dropped a decibel, and at 40MHz it tries to rise by a fraction of a dB, but the phase is only -12°, and has a long way to go to reach -180°.

Output then tries to fall because the lag compensation capacitor feeding back to T1's emitter loads the VAS (instead of the VAS loading the IPS), but at 400MHz oscillation tries to establish itself again. However, the phase here is slightly more positive at -10°, resisting it.

Finally, the voltage amplifier reaches unity noise gain in the high GHz.

In reality this will be affected by PCB capacitance and inductance, but so will Cdom compensation. By preserving phase it will be up to the real output stage to roll-off what is left of the output, due to its beta loss, which via LL compensation, passes through unity at 4MHz.

As this is the transition frequency of the output transistors, each driver then “pushes” the output, but at less than the input sensitivity and that voltage is falling, and is quite within the driver transistor's capability.


Edited by Graham Slee - 29 Sep 2019 at 10:03am
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: Offline
Points: 8905
Post Options Post Options   Thanks (0) Thanks(0)   Quote Graham Slee Quote  Post ReplyReply Direct Link To This Post Posted: 05 Oct 2019 at 6:41am
They say a poor workman blames his materials, but after yesterday's analytical testing I'm fuming!

New distortion reduction tool: generally abusing the left channel by using my right index finger to prod the big electrolytic capacitors really hard, whilst watching, on the AP, the left channel's distortion fall into line with the right channel, then stay there.

My thoughts:

Dry and weak joints don't mend with such abuse - they fail and things stop working - that's why in testing things get hit. This was the opposite.

Magic conducting fairy dust AKA solder trash? Because of solder flux it has to be scrubbed off - it won't just fall off.

Or perhaps there is a faulty big electrolytic, or more than one? One or more out of eight? Doesn't bode well for manufacturing confidence does it?

It's not that they're cheap. These are Epcos branded capacitors supplied by RS. I wonder who they sub-out their manufacturing to?

Whilst we're on that subject, remember me telling you about a leading semiconductor manufacturer shutting down its production lines? Their own evidence showed this happened in 2014. Yet these transistors are still in production and being sold by the same manufacturer to distributors. Magic? No, sub-contracted. To who? Good question! ISO 9000 makes everything traceable, but my ISO 9000 supplier didn't have a clue where one batch came from. The manufacturer insisted they were made in Singapore, but the packaging clearly indicated China, and my guess is Shenzen!
Not simple enough for Google-Bot to understand...
Back to Top
DogBox View Drop Down
Regular
Regular


Joined: 10 Feb 2019
Status: Offline
Points: 58
Post Options Post Options   Thanks (0) Thanks(0)   Quote DogBox Quote  Post ReplyReply Direct Link To This Post Posted: 05 Oct 2019 at 12:05pm
You never considered: "Just one of those things..?" I wonder how long it will keep on going like that..? Probably until you go poking around again! LOL 
DogBox 
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: Offline
Points: 8905
Post Options Post Options   Thanks (0) Thanks(0)   Quote Graham Slee Quote  Post ReplyReply Direct Link To This Post Posted: 13 Oct 2019 at 7:30am
Brominated biphenyls!

Like I said between 2003 and 2006: "something's changed and it isn't just Lead".

I warned of this in two magazine articles (Janine Elliot and Geoff Husband might remember).

Hi-fi had become dependent on brominated biphenyls, and in-fact, so had the entire electronics industry. But whereas some parts of the industry could survive without brominated biphenyls, others could not.

Brominated biphenyls were the chemical additive to electrolyte which prevented "oil capacitors" doing what they ought to (get hot), and brominated biphenyls also appeared in virtually all dielectrics (plastics).

A little oscillation here, and a little oscillation there, did not result in overheating, enabling capacitors to keep their "composure".

The Elna Rod Starget - the superstar of audio capacitors - stopped production in 2006 (although it took quite a few years for RS to run out).

It's replacement, the Tonerex, was aptly named, as tone became ex.

The magic had gone and the search for replacements was on - at least it was here - actually I never heard any other audio company complain.

I don't mind admitting that the big secret of the original Era Gold was the Elna Rod Starget.

Now, bear in mind something. When all the amplifier books were written - when everything Wireless World did - brominated biphenyls were in virtually everything! Yes, even aspirin and Coca Cola.

Brominated biphenyls sound nice. Amplifiers could be just about thrown together and sound nice no matter how long they were left on. That's the beauty of brominated biphenyls.

Just about every record ever made up to 2003 would have featured brominated biphenyls in its production electronics, and during its sound balance.

Today we have considerable high frequency pollution of our mains supply and all signals entering and leaving our high fidelity equipment. We don't have brominated biphenyls protecting against oscillatory sound predators. We also have to work with transistors subbed-out to counterfeiters!

Brominated biphenyls are now rightly banned. They came from benzene - a coal tar by product of coking coal plants - Stavely, Manvers and Royston coking coal plants (quite local to me) are now gone.

However, when I read that brominated biphenyls are degraded by UV light, it reminded me of diazonaphthoquinone, without which there would be no technology at all! (Harmful when inhaled or swallowed; dangerous to environment, esp. aquatic organisms).

Legislators have to be careful what they legislate against. Wink


The tree of good and evil...


Edited by Graham Slee - 13 Oct 2019 at 8:44am
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: Offline
Points: 8905
Post Options Post Options   Thanks (0) Thanks(0)   Quote Graham Slee Quote  Post ReplyReply Direct Link To This Post Posted: 14 Oct 2019 at 6:41pm
As brightness can be/is emphasised by harmonic distortion, perhaps getting it down to -80dB (1/10,000th) might make it inaudible?

In such a low gain - low feedback design this can only happen by turning up quiescent current, and in doing so the output stage gets hot.

It also biases it into class-A, but before we get too excited we need to understand the "amount of class-A".

If this amplifier was to be true class-A at 38 watts, the heatsink would need to be the size of the average house door.

Pure class-A uses the word "pure" in a misleading way (IMO): it should be qualified in the specification as to how many watts it operates at class-A in, but seldom is.

Often pure class-A refers to the first watt, or perhaps two watts, but thereafter I would be wary of any claims.

The equal of 1 watt into 8 ohms is a quiescent current of 0.3535 amps (353.5mA). The math is easy: take the square root of 8 and divide it by 8.

As we mostly listen at around 1 watt, most of what we hear will then be purely class-A.

Its effect on THD is to lower it at 1 watt, dropping the 3rd harmonic to -80dB in this amplifier design, with the 2nd harmonic at around -75dB.

The only problem we have is the heat which needs to be dissipated. With 75 volts across the output stage and 353.5mA flowing in it, there's 26.5 watts of heat to get rid of. With a 1°C per watt heatsink the temperature rise is 26.5°C, and if ambient is 25°C, the case temperature will be 51.5°C. This will feel like its burning your skin, and so it is a non-starter.

But how about trying for 0.5W class-A? This brings the current down to 250mA, wattage to 18.75, and temperature rise to 18.75°C. Now, at 25°C ambient we have 43.75°C. Still quite hot, but it shouldn't feel like it’s burning.

THD won't be quite as low: 0.03% at 1 kHz and 1 watt.
Not simple enough for Google-Bot to understand...
Back to Top
 Post Reply Post Reply Page  <1 8485868788 92>
  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 0.094 seconds.