Crossover distortion and NFB

[Winfield Hill]

Kevin Aylward wrote...

The proof is in the pudding. All of the plots on Self's amp page show
THD at 20khz for his reference amps at over 0.01%. Without trying to
blow my own trumpet here, a 500W per chan mosfet amp I designed (i.e.
small modifications of standard circuits,
http://www.studiomaster.com/hp5.html) was < 0.005% at similar 8 ohm
loads.

Kevin, why don't you create a little web article with the details
of your amplifier design? It's been some time now, so surely the
proprietary aspects of your original design are no longer an issue?

 

[Kevin Aylward]

Kevin Aylward wrote...

Kevin, why don't you create a little web article with the details
of your amplifier design? It's been some time now, so surely the
proprietary aspects of your original design are no longer an issue?

Its not *really* my design in the sense that anything new is there. Its
what what we all do, a few simple mods to existing, standard designs.
Its my design in the sense that I had full *ownership* for all
electronic parts of the design. I was free to steal whatever I needed.

The basic design is a diff pair feeding a diff pair with current mirror,
as shown in the Hitachi apps docs for the 2sk135/2sj50 pair. My extra
bit is simply adding in a push pull driver between the second stage and
the mosfets. This was an absolute must to get the distortion down when
driving many || fets. The reason for this is actually a bit subtle. The
main capacitance is Cgs, which is about 600p. If we || up, say two
devices, the capacitance doubles, but so does the gm, so the reflected
input capacitance gmRl/(1+gmRl), say 100p, should remain constant.
However, what one initially neglected is the Cgd at 40p. This starts to
mount up with many devices.

I did correct one glaring error in the Hitachi design. It had 10ma at
50V for the diff/current mirror using 625mW devices. A no-no. Indeed,
when I first arrived at Studiomaster I pointed out that their existing
directly copied circuit, out in the field would all fail. They stared
coming back in on the Wednesday. I had already analysed the design way
before hand

A main attraction of mosfets is that a really trivial amp design is
possible that achieves really low distortion. Secondly, no secondary
breakdown. A simple zenor clamped gate drive current limit can protect
the devices for minutes. A simple limit like this on bipolar is not
usually possible. They simply wont take 75V at 10 amps for long. The
alternative VI limiting can get into all sorts of issues on reactive
loads.

Kevin Aylward

[email protected]
http://www.anasoft.co.uk
SuperSpice, a very affordable Mixed-Mode
Windows Simulator with Schematic Capture,
Waveform Display, FFT's and Filter Design.

 

[Winfield Hill]

Kevin Aylward wrote...

Its not *really* my design in the sense that anything new is there. Its
what what we all do, a few simple mods to existing, standard designs.
Its my design in the sense that I had full *ownership* for all
electronic parts of the design. I was free to steal whatever I needed.

The basic design is a diff pair feeding a diff pair with current
mirror, as shown in the Hitachi apps docs for the 2sk135/2sj50 pair.

Not everyone has access to the old Hitachi app notes.

My extra bit is simply adding in a push pull driver between the second
stage and the mosfets. This was an absolute must to get the distortion
down when driving many || fets. The reason for this is actually a bit
subtle. The main capacitance is Cgs, which is about 600p. If we || up,
say two devices, the capacitance doubles, but so does the gm, so the
reflected input capacitance gmRl/(1+gmRl), say 100p, should remain
constant. However, what one initially neglected is the Cgd at 40p.
This starts to mount up with many devices.

An excellent point. I have often puzzled at the wimpy MOSFET gate
drive circuits in most hi-fi amps. Hmm, any issue with crossover
distortion in your push pull driver?

I still think you should do a nice writeup of your design: schematic,
discusssion (as above), a good photo of one of your early production
amps still in use someplace, etc.

 

[Kevin Aylward]

Kevin Aylward wrote...

Not everyone has access to the old Hitachi app notes.


An excellent point.
I have often puzzled at the wimpy MOSFET gate
drive circuits in most hi-fi amps. Hmm, any issue with crossover
distortion in your push pull driver?

No. I ran it in class A at 20ma. Each mosfet (5 in ||) had its own 1k
gate resistor.

I still think you should do a nice writeup of your design: schematic,
discusssion (as above),

It had +/-75V for the main output, and an independent +/-90V supply for
the actual amp. Not only does this avoid the mosfet gate drive issue,
but avoids having volts of triangle ripple on the small signal power
lines.

a good photo of one of your early production
amps still in use someplace, etc.

I have one at home. A bit battered though as it over 20 years old.

Kevin Aylward
[email protected]
http://www.anasoft.co.uk
SuperSpice, a very affordable Mixed-Mode
Windows Simulator with Schematic Capture,
Waveform Display, FFT's and Filter Design.

 

[Roy McCammon]

ChrisGibboGibson wrote:

On audio, take the output of an opamp and feed it into dual emitter followers
with no overall NFB.

The distortion looks horrendous on a scope and sounds it. Take the NFB from
outside the emitter followers and it looks great on a scope, sounds better, but
the distortion as measured by a THD meter is the same. The harmonics have moved
(much higher in frequency) but the total %age is the same.

My own limited experience is the opposite. I started
with lousy cross-over. Added NFB. The THD got lower,
but it still sounded lousy.

 

[Ken Smith]

My own limited experience is the opposite. I started
with lousy cross-over. Added NFB. The THD got lower,
but it still sounded lousy.[/QUOTE]

There are posible reasons:

If you only suppress the 3rd harmonic, the THD goes down but the result
doesn't sound much better.

If you have too little phase margin in the feedback loop, the peaking at
the high end can make for some funny artifacts. If the gain crossover
point is too close to the human hearing range, these artifacts can be
heard.

Intermodulation distortiong and THD don't always track each other. If you
really work at it, you can make a circuit where increasing the NFB leaves
a high IM distortion. Perhaps you (un)lucked onto one of those
situations.

 

[John Larkin]

I build amps that need a few PPM noise+distortion up to, say, 50 KHz
bw and peak powers up to 17 KW. My first rule is to make the amp as
linear as possible *before* applying overall nfb. The issue is that,
near the bw limit, there's not a lot of excess gain available, so the
feedback doesn't help much.

John

 

[Tim Shoppa]

In it he argues that if the output transistors are not conducting half the
time for a sine wave, then it shouldn't be called "class B." He reserves
the term "class B" for when there is an essentially seamless transition
from the upper transistor to the lower one, but no period where they
both conduct. So if the bases of the output transistors are connected
together, it is NOT class B, according to Self. This makes some sense.

It's more of a tautology. If you define class B that exact way, then
no real circuit is ever class B. It's like asking for all people who
are exactly 6 feet tall: nobody is exactly 6 feet tall. There will be
some people who are between 5.99999 and 6.00001 feet tall, though.

But the interesting thing is that he measures the THD and residuals for
class B (his definition)

No real or imagined circuit meets his class B definition, so how can me
make a measurement?

Real semiconductors don't "conduct" and "not conduct", thank god, and
real designers use these real parts .

Tim.

 

[Rich Grise]

It's more of a tautology. If you define class B that exact way, then
no real circuit is ever class B. It's like asking for all people who
are exactly 6 feet tall: nobody is exactly 6 feet tall. There will be
some people who are between 5.99999 and 6.00001 feet tall, though.

Most will be 6' 1/2" in the morning and 5' 11 1/2" in the evening.

;-)
Rich

 

[Mike Monett]

I build amps that need a few PPM noise+distortion up to, say, 50 KHz
bw and peak powers up to 17 KW. My first rule is to make the amp as
linear as possible *before* applying overall nfb. The issue is that,
near the bw limit, there's not a lot of excess gain available, so the
feedback doesn't help much.

John

John,

How do you measure the distortion? Notch the fundamental, use a
spectrum analyzer, or is there a commercial distortion analyzer that
will do the job?

If you use a spectrum analyzer, which one?

Regards,

Mike

 

[Terry Given]

Most will be 6' 1/2" in the morning and 5' 11 1/2" in the evening.

;-)
Rich

and lets not forget that yoga practitioners tend to get taller with age,
as opposed to the rest of the population, who shrink (many try to
compensate for this by expanding sideways)

Cheers
Terry

 

[John Larkin]

John,

How do you measure the distortion? Notch the fundamental, use a
spectrum analyzer, or is there a commercial distortion analyzer that
will do the job?

If you use a spectrum analyzer, which one?

Regards,

Mike

I work in time domain, and measure error as PPM deviation from
absolute accuracy as a function of time after a signal (usually a
pulse) is applied. These are transconductance amps, so what I have to
measure is current, and the test sets are nasty; you can't buy a
current shunt or a current sensor that settles to PPM precision in
microseconds, so we have to make them.

John

 

[Rich Grise]

and lets not forget that yoga practitioners tend to get taller with age,
as opposed to the rest of the population, who shrink (many try to
compensate for this by expanding sideways)

Well, yeah. Everybody knows about middle age, when you stop growing at
the ends and start growing in the middle. ;-)

(BTW, I tried that Yoga stuff. Doesn't work. Or there's a high-pass filter
at the gates of Nirvana or something.)

Cheers!
Rich

 

[Mike Monett]

On 26 Oct 2004 23:58:45 -0700, [email protected] (Mike Monett) wrote:

I work in time domain, and measure error as PPM deviation from
absolute accuracy as a function of time after a signal (usually a
pulse) is applied. These are transconductance amps, so what I have to
measure is current, and the test sets are nasty; you can't buy a
current shunt or a current sensor that settles to PPM precision in
microseconds, so we have to make them.

John

Thanks for the reply, John. That sounds even more difficult than
measuring harmonic distortion

Regards,

Mike

 

[N. Thornton]

Hi Chris,

Well, how could that be HiFi? In 1930 there was no FM so your radio will
only feature the AM band, since you live in Europe probably also long
wave and then maybe a portion of short wave. I have one of those as
well, an old 'Sachsenwerk'. It states the dial position in meters
instead of kilohertz. Most of the stations listed on that dial are long
gone. The main problem is that the black paint of which the lettering
behind the dial glass consists is starting to flake off.

In 1930 it would also most likely be a dirty detector, such as a grid
leak, giving a non linear output. And fairly likely driving a non
linear speaker as well. By late 30s radios were good enough to pass
for something modern, if you couldnt actually see it, but 1930 is
crude territory with A B and C batteries, typically unstable pfb
reaction for financial reasons, grid leak detector, 2 tuning knobs,
moving iron speakers, fairly often the ability to receive only a few
stations, etc.

NT

 

[N. Thornton]

[...]

Tubes? Well, they just sound great. Then there is that glow, a little
hum, a crackle now and then. Anyway, feedback also works great with
tubes. For some reason it wasn't done a lot in their days.

When you have to pay an hours wages for a gain of 10, you are a lot less
willing to give it up than when it costs you less than 1 seconds wages.
Tubes cost a lot to make even back then.

Yes, though even more than that in fact. One stage of amplification
would normally cost well over a weeks wages, and those valves didnt
last forever. Pfb was big business, just about no-one took nfb
seriously. Hence the development of circuits that amplified the rf,
detected it, then fed it back as audio through the same set of valves
a 2nd time round


NT

 

[N. Thornton]

Hi Chris,

It would be next to impossible to obtain reasonable audio quality with
that. Maybe ok for a noisy pub but not at home if you want to listen to
Mozart or Tchaikovsky.

True nuff, unless youre willing to get silly Run that output pair
on +/- 200v rails and it would probably sound ok.

NT

 

[Ken Smith]

would normally cost well over a weeks wages, and those valves didnt
last forever. Pfb was big business, just about no-one took nfb
seriously. Hence the development of circuits that amplified the rf,
detected it, then fed it back as audio through the same set of valves
a 2nd time round

In my younger days, I built a transistor version of a refex super regen
receiver. It worked surprisingly well. The audio could drive a small
speaker with just 2 transistors and a modest antenna.

 

[N. Thornton]

[...]

would normally cost well over a weeks wages, and those valves didnt
last forever. Pfb was big business, just about no-one took nfb
seriously. Hence the development of circuits that amplified the rf,
detected it, then fed it back as audio through the same set of valves
a 2nd time round

In my younger days, I built a transistor version of a refex super regen
receiver. It worked surprisingly well. The audio could drive a small
speaker with just 2 transistors and a modest antenna.

Lot of sets back then were 2 valvers - and valve gain was pretty tame
in 1930. I built a reaction set with 2 trs that ran a speaker, with
only one rf tr it needed a big long aerial and only got 2 stations,
but sounded surprisingly good. But stabilty... jeez.

NT

 

[Ken Smith]

Lot of sets back then were 2 valvers - and valve gain was pretty tame
in 1930. I built a reaction set with 2 trs that ran a speaker, with
only one rf tr it needed a big long aerial and only got 2 stations,
but sounded surprisingly good. But stabilty... jeez.

One of the advantages of super-regen. is that it tends to self servo to
swinging just above and below the point of oscillation. Such designs have
two time constants in the self-squegging circuit. The short TC sets the
frequency of the sqegging action and the longer one rides along at the
point of oscillation.

IIRC:
In the one I made, the first transistor was a RF amplifier and also served
as the audio output amplifier. The second transistor was the regen stage.
The first stage is the one that provides almost all of the selectivity.
The second stage is basically sampling the RF at a 30KHz rate.