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 3940414243 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: 09 Jan 2019 at 9:23am
When it isn't fun

I think it's called a labour of love? It has been getting quite tedious lately, and I might have been chasing shadows because it looks like my new computer's USB interface hub is faulty, and the sound problems I've been having were not the amplifier... great!

Even worse is the fact that both channels are now hanging in tatters due to the numerous tweaking operations, and have refused to perform since the last modification, even after being reverted. Yes, it's a tough circuit that I've now managed to break.

I will now have to build two new prototypes, but before I do, I am going to look at a few changes. As you might be aware I have been working on ways to make my (unique?) Proprius configuration more powerful. However it might be more fruitful to look into the subtly different Tobey & Dinsdale AKA Leak Stereo 30 configuration which uses a different output voltage setting method, and in doing so the negative feedback loop is DC coupled. It relies on that for its DC stability rather than on the Proprius DC adjust bias take off point at T2's emitter.

Looking at it from a Doug Self standpoint, the voltage amplifier stage, using two transistors, should be capable of considerable voltage gain, and use of that is made in NFB to (quote) eliminate LF distortion, on the premise that lots of NFB reduces distortion to very tiny amounts.

Actually, we had/have that in the Proprius design, but I want to explore Mr Self's manipulation of the mathematics, as he sees it as T1 being a current source for T2 such that voltage gain is

Av = gm*beta*Rc

where gm is the transconductance: 0.038S; beta is T2 hfe; Rc is T2 collector resistor.

This being a faster way of deciding the voltage gain than my old fashioned methods. Without T1 emitter degeneration we would have come to the same answer, but in my opinion we need emitter degeneration for linearity, or do we?

Assuming NFB works instantly, that there is no time delay from input to output of the amplifier, then the input signal would see no variation in load, but there is time delay - it is called propagation delay in logic which is due to rise and fall times - and slew rate describes rise and fall time, so a large sharp input transient exceeding the 60mV peak to peak input limits, would not be captured by the NFB in time and be badly distorted.

The above describes slew induced distortion. So we need emitter degeneration, a form of elasticity or shock absorber to ensure we get painless transient performance.

So, I think no matter who's maths we use we will simply end up at the same place, but I'll give it a go anyway.
That none should be able to buy or sell without a smartphone and the knowledge in how to use apps
Back to Top
BackinBlack View Drop Down
Senior Member
Senior Member
Avatar

Joined: 05 Feb 2012
Location: Hinton, N'hants
Status: Offline
Points: 2020
Post Options Post Options   Thanks (0) Thanks(0)   Quote BackinBlack Quote  Post ReplyReply Direct Link To This Post Posted: 09 Jan 2019 at 10:31am
Graham, I truly admire your tenacity.

Ian
Just listen, if it sounds good to you, enjoy it.
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 Jan 2019 at 7:31pm
Thanks Ian.

The Leak Stereo 30 can be viewed here: http://www.angelfire.com/sd/paulkemble/stereo30.gif with further info here: http://www.angelfire.com/sd/paulkemble/sound7h.html, but here's the image if you don't want to venture out:


Thanks must go to Paul Kemble for keeping the interest alive.

The power amp is to the right of the volume control, but with mine I am using output voltage biasing similar to how the preamp (left of diagram) is done, and I'm not 100% sure it is the best idea.

It can be seen that the Stereo 30 isn't much different to what Dinsdale proposed in Wireless World Nov and Dec 1961. (https://www.americanradiohistory.com/Archive-Wireless-World/60s/Wireless-World-1961-11.pdf and https://www.americanradiohistory.com/Archive-Wireless-World/60s/Wireless-World-1961-12.pdf as referred to before).




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: 10 Jan 2019 at 5:00am
I said I was not 100% sure of the best output bias scheme, but after a bit of study I'm having to conclude I already had it. In the Dinsdale and Leak circuits the global NFB has a large ratio making for a voltage gain of 100 to 120 (40 - 42 dB), whereas for low distortion much more NFB is required resulting in the more common 25 to 30 (28 - 30 dB) seen since the mid-70s to date.

And trying to reduce gain by applying more NFB we run into a problem where T1's emitter gets lifted in voltage by the current produced. Squaring that circle is just about impossible, requiring much iteration. Better to ac couple the NFB, and in doing so DC stability is required elsewhere, and that elsewhere is by taking the bias voltage from T2's emitter circuit, which is what I'd already got. This appearing to be the definitive all NPN (with the exception of T5) power amp circuit.

Mid-70s just about every amplifier on the market became the same differential IPS configuration (discrete versions of simplified op-amps), and most featured exactly the same components but with one brand sticking with the quasi-complimentary output stage. But this is not the 70s I'm looking for. These mid 70s designs are the amps of the 80s, 90s, 00s and 10s. All seemingly clones of each other.

The "all NPN" amplifier is less efficient because some of the power supply is reserved for biasing the "mid-point", and it could be that me pushing for 50 watts, I am stealing too much of the bias voltages which must appear "below" T1 and T2.

One thing is for certain, so far this design has lacked the "magic" of the Proprius, and I think because of the greater VAS current, causing the emitter circuits to work with less voltage proportionally, some control is being lost. We can either throw more power supply volts at it, or reduce the available output voltage, which I'm afraid reduces the power output. With more power supply volts we can run into capacitor voltage limits (this being a single rail supply). As Ian pointed out earlier, better to lose a bit of power than sacrifice quality.

So a bit of number crunching using Proprius values looks to be the best course of action (hopefully).

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: 11 Jan 2019 at 6:30am
36dB NFB at 20kHz?

Right! Get yourself an A4 sheet of 6 log cycle by 7 linear division semi-log graph paper.

In landscape, bottom left, x-axis, start with the label 100Hz, next 1kHz, then 10kHz, 100kHz, 1MHz, 10MHz, and lastly 100MHz.

Up the y-axis mark the bottom left -20dB, the next one up 0dB, then 20dB, 40dB, 60dB, 80dB, 100dB and lastly 120dB.

The first major subdivision on each x-axis, after say 10kHz, is 20kHz.

Such amplifiers we are told have 28dB voltage gain, so go up the y-axis and put a dot there (hint, it is one minor subdivision below the halfway point between 20dB and 40dB).

Go across to where it intersects 20kHz on the x-axis and mark a dot there.

Now go up 36dB, that's 28dB plus 36dB which is 64dB, and mark a dot there.

This is where it intersects the Gain Bandwidth Product which is a slope falling 20dB per decade, so go to x = 10kHz and y = 70dB and mark a dot there. Then x = 100kHz and y = 50dB and mark a dot there. Then x = 1MHz and y = 30dB and mark a dot there.

Get a straight edge (ruler) and use it to join these three last dots with your pencil, but continue it down to the right and up to the left to the extremes of the graph paper.

Now, go up to 90dB on the y-axis and draw a horizontal line to where it intersects the sloping line you just drew.

It should intersect at 1kHz on the x-axis, and intersect 0dB at the second major subdivision which is 30MHz.

The area under the curve you've just drawn is the amplifier's open-loop gain.

At 30MHz with most high performance transistors their beta is unity (x1). But it doesn't matter because we're at zero gain.

No we're not at zero! 0dB is not zero. It is a voltage gain of 1. It is the amplifier's input signal level, which can be as high as 3/4 of a volt rms, or 2V p-p.

The driver transistors are therefore operating with 2V p-p into the loudspeaker load, not the output transistors, which are simply passing that current as they have no beta (current gain) left.

This is in actuality the point of instability when the amplifier is inside a negative feedback loop. And a simulator will not show the curve as a neat straight line like on your graph paper.

Now here we are being conservative because the knee where the curve goes from horizontal to falling away is at 1kHz. We are told (in an "expert book") it can be as high as 2kHz. This would pass through 0dB at 60MHz.

I searched online for a 60MHz power transistor, and found the miraculous Sanken 2SC3264. You'd have to use this or a number of "smaller" paralleled power transistors to achieve the same, and in-fact, as the Sanken 2SC3264 is 'not recommended for new designs', meaning it is EOL (selling-out), you would have to use the paralleled devices.

But what we have here in the MJL3281a is a 30MHz transistor. So its knee (or break point) is a maximum of 1kHz.

But we have still to draw the closed loop gain on our graph paper, so go up the y-axis to 28dB and draw a horizontal line to the right, and you'll see it intersects the falling line at just above 1MHz.

Now, remember this: the falling line at 20dB per decade represents a phase shift of 90 degrees (the horizontal line 0 degrees).

Now, the reason we have instability even though the drivers can handle the 3/4 volt signal into the speaker load (referred to above) is because the beta has dropped due to input capacitance "standing in its way" and its reactance has made phase rotate 90 degrees, and so if we add the slope's 90 degrees we have 180 degrees.

And if we send that 180 degrees round the NFB loop we get 360 degrees, so we have an oscillator at 30MHz.

So 36dB NFB at 20kHz is not going to happen. And the disappearingly low 20kHz distortion is a mirage that the textbook imagines.

But can't we stop the oscillations using a zobel network? Yes we can, but we will still have 'ringing' which gives rise to the artificial brightness in classical works, and dry (boring) presentation of rock music.

The solution is to set the knee a decade lower (at 100Hz) which means the curve crosses 0dB at 3MHz, a decade lower, such that 180 degrees appears at -20dB. This gives a 90 degree phase margin, a 20dB gain margin, but now we only have 16dB NFB at 20kHz, so 20kHz distortion will be a lot worse than what the textbook tempts us with.

The textbook is wrong! (IMO) And can you really hear the harmonics of 20kHz?

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: 12 Jan 2019 at 9:38pm

If you followed my instructions in my last post on this topic you will have ended up with the solid lines as my graph above.

On mine I have added the 40dB/decade slopes which indicate 180 degrees phase ensuring oscillation, and I've drawn them where experience of simulation tells me where they will happen.

A 30MHz power transistor, as said before, runs out of current gain at 30MHz, and here it depends on the drivers to "push" the "short circuit" of the output transistors, but the phase shift is the problem.

By utilising a "heavy" zobel network the phase can be brought down steeper than 20dB/decade such that the output crosses over the unity gain line before hitting 40dB/decade, but that is false use of a zobel network. It is intended to stabilise the Colpitt's oscillator tendencies of emitter followers.

It will however result in ringing.

This begs the question: do we want distortion or ringing? Because that's the compromise with power amplifiers. By compensating for stability as in the dashed curve we have less NFB to combat 20kHz distortion (16dB instead of 36dB), but ringing is substantially reduced.

We can design an amplifier with little chance of emitter follower oscillation which can improve the situation, but that would not be in the spirit of this 70s design topic. And that is to use complimentary feedback pairs in both top and bottom halves of the output stage.

But back to the question: do we want distortion or ringing? Ringing has a bad habit of "wrecking" electrolytic capacitors, and for this design to be authentic it is AC-coupled meaning it uses electrolytic capacitors in the signal path. Therefore distortion is the correct answer!

That none should be able to buy or sell without a smartphone and the knowledge in how to use apps
Back to Top
BAK View Drop Down
Senior Member
Senior Member
Avatar

Joined: 14 Mar 2010
Location: Kentucky, USA
Status: Offline
Points: 1744
Post Options Post Options   Thanks (0) Thanks(0)   Quote BAK Quote  Post ReplyReply Direct Link To This Post Posted: 13 Jan 2019 at 12:11am
Originally posted by Graham Slee Graham Slee wrote:

This begs the question: do we want distortion or ringing? Because that's the compromise with power amplifiers. By compensating for stability as in the dashed curve we have less NFB to combat 20kHz distortion (16dB instead of 36dB), but ringing is substantially reduced.

We can design an amplifier with little chance of emitter follower oscillation which can improve the situation, but that would not be in the spirit of this 70s design topic. And that is to use complimentary feedback pairs in both top and bottom halves of the output stage.

But back to the question: do we want distortion or ringing? Ringing has a bad habit of "wrecking" electrolytic capacitors, and for this design to be authentic it is AC-coupled meaning it uses electrolytic capacitors in the signal path. Therefore distortion is the correct answer!

 
1. So, is the VAS local feedback capacitor (on Q2 collector-to-base) like the "fulcrum" of this balancing between ringing and/or distortion? ... Hence the compromise?
2. And, is the input filtering capacitor (across the input signal) helping to "curb" the slew rate limiting (in addition to EMI blocking) to reduce the slew rate induced distortion?
These 2 maintain the closed-loop gain response to give the stability of 20dB gain margin and ~90deg phase margin.

3. Then, the zobel network (and the 5uH coil in series w/ the speaker) just "conditions" the power output to remove oscillation tendancies and to be able to drive complex impedances... 
including frequency-variable speaker impedances and speaker cable capacitances?

 Interesting that Tandberg put the tone controls in their power amp's feedback loop. As the tone controls were adjusted, this must have had a direct effect on the amp's stability.
Bruce
AT-14SA, Pickering XV-15, Hana EL, Technics SL-1600MK2, Lautus, Majestic DAC, Technics SH-8055 spectrum analyzer, Eminence Beta8A custom cabs; Proprius & Reflex M or C, Enjoy Life your way!
Back to Top
 Post Reply Post Reply Page  <1 3940414243 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 0.109 seconds.