Re: lateral mosfets vs. bjts in audio amplifier design

[Eeyore]

Shit....
Any one that can't do higher precision in Uc's other than
what is supplied natively, does not belong coding in uc's

you're hitting bottom!..

and for you info, that was a 8 bit Uc but the FFT was 16 bit,.
more than enough.

FFTs have nothing to do with it.

Graham

 

[Eeyore]

What a pig shitting BS'er you are..

Go away!...

You're treaded on thin ice you old crony!

You're the fool. You are nothing more than a disgusting piece of pig
excrement.

You're just jealous because you can't do it. Bloody typical of idiots.

 

[Jamie]

Jamie wrote:




FFTs have nothing to do with it.

Graham

What a pig shitting BS'er you are..

Go away!...

You're treaded on thin ice you old crony!


http://webpages.charter.net/jamie_5"

 

[Eeyore]

You have also now shown you don't know what a live performance is.

Jolly well done I say. Even orchestras have amplification these days.

Graham

 

[Eeyore]

---
Whoosh...

He nailed you and you don't even know why.

Delicious!

IDIOT

 

[Eeyore]

He's too stupid to understand that concept and too childish to admit
it..

You and Fields really are utterly out of touch with pro-audio. You're in no
position even to comment on it.

Graham

 

[Jasen Betts]

One of the few truly vile people round here is you.

Did you even bother following the links ?

he only posts links, he never even checks them.

Bye.
Jasen

 

[Kevin Aylward]

Cough, Cough.................................

Jesus Christ...

Who the hell would want your reverb code.

Do you honestly believe people out there can not do their
own DSP coding?

Choosing the right co-efficient could take a month of Sundays though, so not
surprising one don't want to tell anyone what they are.

Like, the basics of a nuclear bomp are pretty trivial, but its the final
details that matter.

I have tried quite a lot of digital delays, and it is somewhat surprising
that many, many of them, still sound like shit. Not one usable setting.
Trust me.

Like, do you know the one about the Coke A Cola formula....that's not rocket
science either.

I don't have the time to read most of this thread, Graham, but...

Kevin Aylward

www.kevinaylward.co.uk

 

[Kevin Aylward]

Simply untrue, the market is VERY competitive.

Indeed. Its just about impossible to make a profit in "pro audio" now adays.
Too much stuff on the market.

Kevin Aylward

www.kevinaylward.co.uk

 

[Kevin Aylward]

AMS-Neve don't seem to have that problem. Read the client list.

Graham

I am *really* loving my new tascam digital multitrack. Its a joy to record
and randomly seek to any point in the song. The digital saving and operation
of tuneable tone controls are magic. Now I want a digital desk, with
infinite rotate pots with position lights on each channel. Like, there is no
going back once one has the digital controllability.

Kevin Aylward
www.kevinaylward.co.uk

 

[Kevin Aylward]

It did one my models by an arm and a leg. Exactly where I wanted it.



Speed is good.

Yes. I always use emitter resister now. It does help the matching problem
tremendously. I have some sim results of diff pair distortion with mismatch
here, http://www.kevinaylward.co.uk/ee/micampdesign/MicAmpDesign.html

Who said it had to be in the same overall loop ?

Not sure I understand what you are saying here.

Well, I made a QSC-lokalike design perform far better by applying
those techniques of mine.

Including 2 stages of pole zero compensation.

Yes, if one has N rolloffs, then take them all but one out at the UGF, then
these higher order systems can be compensated. Discretely, this is not too
difficult, in ic's it can be really hard to get the compensation due to cap
size. But note, thes higher order systems still have a net lower BW. You get
all the loop gain up to a frequency point, then kill it all very quickly.
Its still a trade off of speed verses accuracy in the chosen BW.

www.kevinaylward.co.uk

 

[Kevin Aylward]

I don't disagree with you overall Kevin. I keep a little trick up my
sleeve with power amps that changes the game somewhat and eliminates
any such 'expenses'. However it's absurdly obvious (when you think
about it) and consequently couldn't be patented I'm sure and the
moment the 'secret' gets out everyone will be doing it. It's insanely
simple too. You just have to have the wits to think of it.

Graham

Note, as in the mosfet 1000, basically a stolen idea from Hitachi. The
second stage is a diff pair. This diff pair has less distortion inherently.
Don't know why D.Self don't use it in his Blameless amp. Keep everything
differential is one of my mottos. The current mirror load gives the prior
post mentioned push/pull drive to the output buffer. The current source
loads (CMFB) on the 1st stage, rather than resisters, makes the total LF
loop gain, truly huge, making LF distortion, vanishing small.

Kevin Aylward

www.kevinaylward.co.uk

 

[Kevin Aylward]

---
What I was responding to was Graham's:

"I confess I do that kind of thing. Just not put op-amps round the
actual output devices myself so far but it sounds interesting. They'd
have to be damn fast though."

The way I see it is that the MOSFETs are one amplifier comprising
complementary emitter followers surrounded by another amplifier (the
opamp) which sets the closed loop voltage gain at 5.

Obviously some genuine confusion here. This is not what err..us amp
designers mean. We assume from the outset there is a standard main gain
stage already in the system. I think I can speak for Graham and myself on
this one. A proposed amp around the output devices is an additional amp.

One only wants at most two gain stages. Having more, even in disguise as a
unity gain buffer, is a major problem in stability. Of course, multiple
rollofs can be handled in principle, but its another complication.


Kevin Aylward
www.kevinaylward.co.uk

 

[Kevin Aylward]

On Mon, 20 Oct 2008 19:34:52 GMT, "Kevin Aylward"

Well, neither. I was suggesting a fast opamp *per fet*, with feedback
from the source, to make each fet look like an ideal transconductance
device, perfectly linear, no offset or threshold, all exactly matched,
with very low input capacitance.

But how does improving and parallelizing gate drives cost speed? It

I made no criteque on parallelising gate drives. Bigger drive is usually
better.

makes my amps faster and a lot more stable. Your amp (the one you
never built) has a couple of wimpy current sources driving 10 fets in
parallel;

No its doesn't. Its a push/pull class AB current drive to the mosfets, and
secondly, you don't need much. 20ma class drive is way more than enough for
at least 500w at 200Khz power BW.

I'm suggesting a beefy voltage source per fet gate, with
local feedback.

Its the local feedback, within another overall loop that is the problem.
Despite the apparent closed loop nature of the second buffer, it is still an
opamp with an inherent main pole, plus higher order poles within the main
loop. All these poles matter, unless you seriously clobber the whole
response. That is you get better accuracy at LF, but the amplifier will
always be slower than if the amp dd not have that extra op-amp. well not
unless you 100Ghz opamps or such like.

I am addressing the basic concept here. Things are not what they might seem
from an initial examination.

Kevin Aylward
www.kevinaylward.co.uk

 

[Kevin Aylward]

How the heck do they separate them when they're 1 x 1 mm ?

Not when drunk

What's the
wafer thickness ?

Not a lot.

Is testing done at the die level ?

Yes. Wafer probe.

Graham

I will leave this one. Running out of time...

Kevin Aylward

www.kevinaylward.co.uk

 

[Kevin Aylward]

So, in terms of THD, s/n ratio and response flatness where would you
draw the line where it becomes overkill ?

Graham

0.01% IMD @ 20Khz/19Khz at all power levels. 1db flatness to 50Khz power BW.

I don't claim that this is necessary, but my take is that it is sufficient.
I won't buy a PA amp if it is say, 0.1% at 20khz.

As far as what I am interested in as a personal achievement goal, its
probably 0.001% IMD at 20khz, 1db flatness to 1Mhz power BW.

Kevin Aylward
www.kevinaylward.co.uk

 

[Kevin Aylward]

"Pro" audio design seems like an infinite chain of stolen circuits.

Lets not kid ourselves, "all" circuits are stolen, whatever the field. Its
that simple. Show me *any* design, and I will show you a bit of is in a
prior design. Like diff pairs, current mirrors, cascodes, are all the same
building blocks we all use. However, designing say a cmos opamp, even with a
standard topology, can take considerable time in getting just the right
combination of W, L and M to satisfy a particular specification.

The only way to make production designs reliable, is to use what is already
known to work, and only add the minimal of new additions.That's what being
an engineer is. Maximising profit with the minimum of RISK. There are no
brownie points for a novel circuit that achieves no net advantage. All new
design is based on modifications of existing design. That's just the way it
is. Like, how do you proposes a new car engine is "designed"..Like some new
law of thermodynamics is discovered? I don't think so mate...

Anyone that says that they don't use say, 95% of existing work is either a
liar, or seriously deluding themselves.

Kevin Aylward
www.blonddee.co.uk
www.kevinaylward.co.uk
www.anasoft.co.uk - SuperSpice

 

[Eeyore]

Note, as in the mosfet 1000, basically a stolen idea from Hitachi. The
second stage is a diff pair. This diff pair has less distortion inherently.
Don't know why D.Self don't use it in his Blameless amp. Keep everything
differential is one of my mottos.

Ditto. It's crazy to throw away that advantage.

The current mirror load gives the prior
post mentioned push/pull drive to the output buffer. The current source
loads (CMFB) on the 1st stage, rather than resisters, makes the total LF
loop gain, truly huge, making LF distortion, vanishing small.

Indeed so. Mind you, I actually *limited* my LF gain because it was so huge
already ! It didn't need any more. THD didn't start to climb on the 1200B
design until about 2-3 kHz. Knowing what I do now, I'm sure I could do better
than that. This was 20 years ago you know.

The Matti Otala idea of 3 voltage gain stages attracts me. Ever hear of the
Electro-Companiet amplifier ?
http://www.google.com/search?hl=en&rls=en&hs=3lm&q=Companiet+amplifier&btnG=Search

They were hopelessly built (I ended up fixing an early one) but the technical
idea was sound.

Graham

 

[Eeyore]

I am *really* loving my new tascam digital multitrack. Its a joy to record
and randomly seek to any point in the song. The digital saving and operation
of tuneable tone controls are magic. Now I want a digital desk, with
infinite rotate pots with position lights on each channel. Like, there is no
going back once one has the digital controllability.

Horses for courses. The control surface and GUI of a digital desk can make or
break it.

For live situations where many engineers may use the desk, analogue is still
unbeatable for its simplicity and anyone experienced can master one in 5 minutes
even if they never saw that model before (and they're not short on performance
these days either) but even in live, digital is making increasing inroads.

Check out Cadac's S-Digital for example. Just out and I think they hit the sweet
spot.

Graham

 

[Eeyore]

Choosing the right co-efficient could take a month of Sundays though, so not
surprising one don't want to tell anyone what they are.

Like, the basics of a nuclear bomp are pretty trivial, but its the final
details that matter.

I have tried quite a lot of digital delays, and it is somewhat surprising
that many, many of them, still sound like shit. Not one usable setting.
Trust me.

Oh so true. Read the reviews.

Like, do you know the one about the Coke A Cola formula....that's not rocket
science either.

I don't have the time to read most of this thread, Graham, but...

'Designing' those reverbs was a fascinating task, not least because they had to
sound as near identical as possible to some on a withdrawn OEM module we'd used.
I managed a replica as good as anyone could tell. Why ? Because people *loved*
those reverbs and wanted us to keep them.

I read no less than 3 long papers on the subject in the process. And I threw out
some 'holy cows' about how to make them sound right that have always been
assumed - because they actually sound shit.

Just as I do with EQ, I only sign off for production when it *sounds* right as
well as measures right.

Graham