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 4142434445 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: 27 Jan 2019 at 8:18am
Apart from the basics, nobody has written anything about these amplifiers. Complexity is the more boastful for discussion (unless it's valves, where simplicity is the more boastful for discussion).

This topic is not about boasting (bragging). It is an exploration about a loose end. "They" just left it there to rust: a future generation might make some use out of it, that sort of thing.

My ears must have been playing tricks with me, that or other substances, because the Lin (RCA)/Tobey/Dinsdale/Mitsubishi/Leak/etc proposition is actually flawed. The Sziklai pair emitter resistor is in the wrong place.

As signal voltage goes negative the Sziklai pair conducts more, which is opposed by the emitter resistor conducting more, making the Sziklai pair conduct less.

Those of you who have kept up (those I haven't lost) will say "but you've been there before", and true, I have, but my last diagnosis was wrong (I can be wrong), so I reverted to the original emitter resistor position. But now my diagnosis makes sense, the emitter resistor is now firmly planted above, not below, the Sziklai pair, and there it will remain.

Last post regarding stability talked about placing a capacitor from T1 collector to ground, but all this did was slow the voltage amplifier which therefore sounded more distorted. Remember the output stage Baxendall diode? Well, forget it, but placing a resistor in its place (47 - 68 ohms) does the same as the capacitor from T1 collector to ground, and without slowing the voltage amplifier.

You see, there are parasitic capacitances all over an amplifier circuit and stability is about tracking them down and dealing with them. I repeat "nobody has written anything about these amplifiers", so for me this is a difficult task.

I am also retracing my steps since ditching the snap-in electrolytics. Nothing could render a good sound using them, and so I could have been on the right path weeks ago, but these things have to be found out.

I can also announce that I found a relationship between the input stage and dominant pole capacitor that will have "the experts" in stitches (because I didn't realise it sooner). T1 emitter resistance sets the gm (mutual conductance) of the stage, and if it were only little-r, would be 39 milli-siemens.

The whole of the emitter resistances must be taken into consideration, not only the degeneration resistor, which means the NFB divider resistor plus the degeneration resistor make up Re. This differs from a typical LTP voltage feedback amplifier.

The only contribution the degeneration resistor makes is to input slew-rate and/or input overload, not amplifier slew rate. The total of Re's contribute to everything else.

So, where the Proprius and this design featured the same emitter degeneration, the dominant pole capacitor will be different because T1's gm is different.

The reason for T1's gm being different is to obtain more voltage amplifier gain so that we can have 50 watts (or thereabouts) instead of 25 watts. And also to boast good distortion figures (there being more NFB).

The difference of gm's between the Proprius and this is roughly 4.5 - this having more gm - and so having to use a 100pf capacitor where there was 22pf before makes perfect sense, and results in the same gain-bandwidth product. So now we have the dominant pole capacitor set in stone (it is 100pf and not 150pf like last time).

So with a stabilised amplifier music should sound right? Actually it is a trade-off between harmonic distortion and slew-rate distortion. Currently we have low THD with roughly 20V/uS slew-rate. The Proprius calculates at over 80V/uS, but has higher THD.

I think I'd prefer a bit more inaudible THD than the exaggerated squeaky leading edge of metallic acoustic guitar strings, and the only way around that is to reduce gm, increasing THD but increasing slew rate. I am told that what I am hearing cannot be heard.
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: 30 Jan 2019 at 1:49pm
Anyone would think there's no need to deal with frequencies beyond our hearing when designing a power amplifier... as long as it is stable enough not to overheat or damage itself.

That must be the conclusion many amplifier designers come to. But it cannot result in sufficient transparency to be able to hear the correct tonality of music.

Keith Armstrong (the EMC specialist) says that if the EMC is right an amplifier usually sounds better. I agree: if it is not busy producing spurious ultra high frequencies then it only has to do its job with music frequencies.

We had a point in this design where the amplifier could be judged to be stable by all accepted methods/standards/recommendations. However, it did not sound all that good to my ears.

The knee in the response was much below the unity gain crossing so should not affect anything, and the "kick-back" at frequencies 100MHz upwards were below -10dB... better than many an op-amp ...and nothing is written about this, so I suspect it is considered alright by the "respected authors"?

Well, it isn't right at all!

Anyone would think that components cannot do radio! Well, they do, and that's why radio sets work!

I once managed to transmit audio a distance of 300 metres on the shortwave radio band using little more than a BD139 transistor taking the carrier frequency modulated audio onto a rudimentary antenna. It could be tuned at its fundamental frequency and each harmonic until I ran out of shortwave bandwidth.

If it was "powerful" enough to do that, what will spurious low level bursts in the high MHz to GHz do within millimetres of the rest of the circuitry? Which can detect and/or intermodulate resulting in heterodyning (the arch-enemy of audio).

I am sure that will give you "richness" and other hi-fi marketing BS terminology designed to pull positives out of a disastrous design.

But since at least 72 hours continuously-on have elapsed from when I implemented what the simulator indicated as 1. being stable; and 2. taking right down the radio spectrum frequencies to acceptable EMC levels, the tone of the music has been considerably better.

I will report again around the 100 hour point.
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: 31 Jan 2019 at 7:14am
This rather lengthy discussion really ought to now try and demystify this configuration, and if it does, could be a world first, at least in this century.

I think I've shown it capable of meeting similar performance, specification wise, to much more complex modern designs.

I suppose the downside (if you can call it that) being low negative feedback, is that it cannot disguise the bad contribution of particular components a high NFB design might be capable of. Here I mean the awful contribution of the power supply reservoir capacitors should they be of the snap-in variety. It would seem that radial types where the foils are held in situ and sealed into the aluminium can by a rubber bung, impart a much less distorted sound subjectively. Use of the snap-ins certainly set back the progress of development, so I think we can derive a design rule...

1. Do not use snap-in electrolytics

Another very important design rule has to be the voltage amplifier transistor emitter voltage. As supply (HT) voltage varies with load current a small emitter voltage can throw the output operating point (roughly half HT) well away from where it should be, and on large signal excursions results in a distortion variance which is measurable. However, the emitter voltage steals from output voltage such that possible power going on the HT voltage, cannot be reached, and so some is wasted.

2. VAS emitter voltage should equal 1/10th (or not very much lower) of the HT voltage

In our design so far we have approx. 6 volts VAS emitter voltage using a 68 volt on-load HT. We also need to make this voltage stable as it provides the bias voltage for T1 and any signal appearing on its emitter would be fed back reducing open-loop gain. Therefore...

3. The VAS emitter should be AC bypassed (using a suitable value capacitor)

And as DC feedback sets the DC stability of the circuit...

4. T1 base bias feed resistor should be taken off nearer to T2 emitter than ground.

(this means tight control of T1's base bias resistor. As a rule of thumb we only want to drop 0.1 - 0.2 volts here, and this could mean AOT, adjust on test)

The most mysterious part in my opinion is the role of the IPS (T1) emitter as regards to HF stability and slew-rate.

So far we have split the required (T1) emitter resistance so that the "earthy" end is the negative feedback divider resistor, and the upper end is used to multiply the "small Re", or intrinsic emitter resistance.

We multiply the "small Re" to get a higher input margin: for example with a DC current of 2mA flowing through the transistor (collector to emitter), we know from existing data that its "small Re" is 26mV/2mA = 13 ohms. By making "big Re" (the real resistor) around 10 times bigger we increase its margin tenfold.

But doesn't the NFB divider resistor also contribute to that? Well yes, and no. At low frequency, yes, but at high frequency where parasitic capacitance held in all parts of the circuit exists, there will be a "delay" in the output signal reaching the NFB divider resistor.

But then, from a stability point of view, both resistors are "seen" as acting in union. And open-loop the series value of both in union decide the IPS voltage gain. Therefore we could simply make it one resistor because we get the same measured result. It is difficult to measure true input slew-rate because we would have to go much higher in frequency to see it, and therefore the upper resistor value's effect has to be thought of as imaginary.

However, to save imagination playing tricks, the saving grace appears in a description of how slew-rate works in a book called "The Art of Electronics".

But here it only applies after the "break frequency" where spurious parasitics kick in thus delaying the fed back signal.

It is doubtful we will ever be able to measure the effect of having the margin suggested by the upper emitter resistor, and so this design rule is too difficult to write down at this point in the discussion.

And here I will have to leave it before we get too bogged-down, but I will continue next post with the correlation of emitter resistance with VAS compensation.
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: 03 Feb 2019 at 4:25am
simplified voltage amplifier stage
(a much modified voltage stage)

Taking a look at the simplified voltage stage of the power amplifier (above) I can tell you it is far from simple, and therefore difficult or nigh-on impossible to create any further design rules.

It however does obey all its characteristics (physics does) but the number of characteristics is colossal, making it a mathematical nightmare to predict its outcome. And so the only way to learn anything about it is with the processing power of a simulator (which its original designers would not have had).

The simplicity of the series negative feedback afforded by T1's emitter resistor(s) is complicated no end by the voltage stage it drives.

T2's input resistance rises toward the lowest bass due to its emitter AC coupling capacitor allowing emitter resistance to influence its mutual conductance (lowering gm). But then, this is necessary for low frequency stability.

The effect of the dominant pole capacitor coupled with T2's gain presents a much larger (and non-linear) capacitive load on T1's collector.

The gain profile of T1 is therefore high (around 34dB) at bass frequencies, falling to low (around -34dB) at beyond-treble frequencies, and nowhere as constant as my simple earlier investigations. In fact, it only produces a flat output from 100kHz up to 10MHz before parasitics take over.

T2 normalises it to produce the regular 90 degree downward slope typical of other open-loop responses, and negative feedback produces an admirably flat response.

The only thing we can glean really is that compensation capacitance is related to T1's emitter degeneration, and that degeneration includes the NFB divider (or grounding) resistor.

The smaller the emitter resistor the larger Cdom needs to be, and vice versa. So if 100pF is stable with 270 ohms (total emitter resistance) then 22pF is stable with 1.2k.

We end up with the same bandwidth product in each event. The thing which changes is slew-rate, and with it distortion. The smaller we can make Cdom, the faster the slew-rate, but we have to make the emitter resistance higher to make Cdom smaller, and higher emitter resistance lowers open-loop gain, which increases distortion.

(slew rate is determined by T1's collector current divided by Cdom, which it has to drive)

But how important is slew-rate? Comparing this amplifier with the similar Proprius design which uses the same T1 current, but Cdom of 22pf, the Proprius does above 80V/uS (actually 2mA/22pf = 90.9). This amplifier does 20V/uS.

It doesn't sound as good as a Proprius in my opinion. The highs don't sound right. This suggests the slew-rate should be increased, but that would require a reduction in T1's gain because of increasing its emitter resistance. That will result directly in increased THD. If THD was the problem then this amplifier would sound (to me) better than a Proprius, but it clearly does not, and so as I've said so many times before, the grail is not ultra-low distortion, it is speed.

Perhaps a modest increase to 40V/uS, which incidentally will double the THD, might be the answer?

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: 03 Feb 2019 at 12:31pm
Originally posted by Graham Slee Graham Slee wrote:


Perhaps a modest increase to 40V/uS, which incidentally will double the THD, might be the answer?



And indeed it is. Well, so far. As everybody should know, components reveal problems only after they have bedded-in. So here I embark on another 100 hour proving session.

So far the highs are more well defined, and so is everything else.
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: 03 Feb 2019 at 4:03pm
The voltage across R1 (16k) is 32 volts and so 2mA flows (simplified voltage stage).

T1 can swing 4 volts negative which places 36V across R1 and so 2.25mA flows.

Having changed IPS gm to unity by increasing its emitter resistance to 560 ohms so I can use a smaller Cdom of 47pF with no change to stability, then if slew rate is IPS current/Cdom, its negative slew rate is 47.87V/uS.

If T1 swings positive by the same 4 volts the voltage across R1 is 28 volts so 1.75mA flows, and so 1.75/0.047 gives 37.23V/uS.

This simple calculation indicates we have an average 40V/uS, and its minimum is 37.23V/uS positive-going, but will it really be that fast?

It assumes that the current on the collector end of Cdom doesn't move. In reality it will and so slew rate will not be as we think. But rather than going into long calculations we can assume that positive slew rate will be even-lower, and negative slew rate even-higher.

As long as it is no worse than 20V/uS we have a fast-enough amplifier, and I think it will be no worse.
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: 03 Feb 2019 at 5:29pm
Following your design closely, I like your methods.
 As you stated in your last post...
 
Originally posted by Graham Slee Graham Slee wrote:

Originally posted by Graham Slee Graham Slee wrote:

Perhaps a modest increase to 40V/uS, which incidentally will double the THD, might be the answer?
 

And indeed it is. Well, so far. As everybody should know, components reveal problems only after they have bedded-in. So here I embark on another 100 hour proving session.

So far the highs are more well defined, and so is everything else.

 I believe you are correct that "the grail is not ultra-low distortion, it is speed."
Of course that speed is balanced against THD as in a compromise.
The system designer faces compromises at every critical part of a design.

 I also believe that in the final stage of amplification (as in a power amplifier)
the THD needs to be kept below some pre-determined maximum (as a preamp may need to be lower THD so its contribution to the whole is not passed on) and can be placed in the design process at a "lower priority" to achieve another characteristic having better results...
 in this case; better slew rate and better stability.

 The better slew rate should result in better sound quality,
and the better stability will give longer life.

 Patiently waiting to hear of your next 100 hour proving session results.
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 4142434445 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 0.094 seconds.