Why Is High Feedback Considered Bad In Audio? In Simple Terms

[Eeyore]

Several of my own P.A. designs (all transistor) have soft clipping in
the driver stages. On the occasions where I have had to run them
overloaded[1], they sounded much louder than their numerical wattage
would have suggested and there were no complaints about distortion.

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
A typical audiofool ignorance.

The amount of nonlinear distortion = deviation of the transfer curve
from the straight line. Soft clipping = more distortion if compared to
hard clipping to the same ceiling.

If the amp has to enter the area where it clips the signal, one has to
limit the windup of the integrators in the feedback path. So the
recovery from the saturation will be quick.

Baker Clamp.

Graham

 

[Eeyore]

BTW, the main push towards class D is because of the material savings
due to the reduced size of the heat sinks.

And power supply.

But watch out for that 'LF pumping' ! A Class D output stage can make a nice
charge pump.

Graham

 

[Rich Grise]

krw wrote:

Some products have that stated purpose, it's true, but a great many make
claims (usually expressed alongside weasel words) as to benefits that
are not supportable.

The beer folks do the same thing - "Drink our beer, get this hot babe!" ;-)

Cheers!
Rich

 

[Rich Grise]

The goldenearies seem to love the 6SN7. I always hated 6SN7s when I
was a kid. Rotten transconductance, lots of plate capacitance, a
klunker in general.

I really liked the frame-grid tubes, 6DJ8 sorts of things, with
screaming gain-bandwidth. Nuvistors were fun, too.

Remember the RCA "New Vista" TVs? ;-)

Cheers!
Rich

 

[[email protected]]

Sadly the excellent Hitachi lateral mosfets have now gone, but you can get
approx equivalents from Semelab/Magantec and Exicon.

I built an amp module using the Hitachi parts with 0.0005% THD ( yes -106dB
SINAD) @ 350W into 8 ohms btw. The final product THD was determined by
op-amp buffer stage performance @ 0.003%. You can get better op-amps now
though. It would also drive 600W into 4 ohms and around 900W into 2 ohms.

That used tons of feedack both local, overall and pole-zero. It also
sounded lovely. Quite a few made their way into recording studios as
monitor amps.

Graham

http://tinyurl.com/c8ydwm
for the Renasys parts. It is the semi spin off from Hitachi.

It must be a small market for the lateral current flow mosfets. They
are pretty simple to build, so technology isn't an issue. I didn't see
a die photo on the Renasys part, but I assume the fets are just waffle
pattern.

 

[Eeyore]

http://tinyurl.com/c8ydwm
for the Renasys parts. It is the semi spin off from Hitachi.

Yes, I was aware of that but thought they had deleted them. I'm quite astonished
! Thanks for the info.

It must be a small market for the lateral current flow mosfets. They
are pretty simple to build, so technology isn't an issue. I didn't see
a die photo on the Renasys part, but I assume the fets are just waffle
pattern.

Shame the TO-3s have gone though. They were phenomenally durable.

Graham

 

[MooseFET]

Chopper type opamps don't have 1/F noise, and have drifts measured in
nanovolts per degree C. They are good for some low-level stuff.

Chopper opamps do sort of have a 1/F noise. It is displaced from the
zero frequency up to the Nyquist. This can lead to an increasing
noise with frequency character

I can't recall any cases where tubes are still sensible, except for
very high power, microwave, and optical applications (PMTs,
microchannels, night vision stuff.) Any more?

Do GDT lightning arresters count?

 

[[email protected]]

Mostly hexagonal now, slightly better HF corner.  Say thanx to IR.

The IR hexfet is a vertical flow device. I did a back of the envelope
sketch of a lateral current flow with hex shaped poly and it doesn't
pack as well as the waffle. Do you a particular device with a diagram
showing the layout?

 

[Jasen Betts]

The beer folks do the same thing - "Drink our beer, get this hot babe!" ;-)

but they deliver, if you keep drinking she stays hot

 

[Eeyore]

The IR hexfet is a vertical flow device.

Correct. Utterly different and try matching Vgs thr.

I did a back of the envelope
sketch of a lateral current flow with hex shaped poly and it doesn't
pack as well as the waffle. Do you a particular device with a diagram
showing the layout?

If you can find an old Hitachi application databook it shows the thermal
distribution in the die. It's very very well controlled.

Graham

 

[[email protected]]

Correct. Utterly different and try matching Vgs thr.


If you can find an old Hitachi application databook it shows the thermal
distribution in the die. It's very very well controlled.

Graham

But that doesn't explain the current flow. These lateral current flow
devices don't need epi, i.e. there isn't a low resistance path to the
back side, so they should be pretty cheap to build. I'm not a process
guru, but I've been told vertical flow is more likely to effected by
wafer defects than lateral flow, so the linear MOS device should be
high yield.

I know from ESD testing, the waffle pattern seemed less rugged than
plain stripes.

 

[[email protected]]

And since lots of lowpass filters don't overshoot, they don't ring, so
there must be highpass filters that don't ring.

This makes less and less sense all the time.

John

I know you can't make a high pass that is linear phase. [Highpass in
the true sense, i.e. zeros at DC/] A highpass filter has no delay at
infinite frequency, but has delay near it's corner. There may be a
highpass/allpass combo that makes a net filter that is linear phase.
Of course, linear phase doesn't always translate into no ringing.

One thing to investigate would the Linkwitz-Riley (sp) class of
highpass filters. My recollection is they take an optimal delay and
subtract a Bessel to get a highpass. The idea was to get a smooth
response at the crossover.

 

[JosephKK]

The IR hexfet is a vertical flow device. I did a back of the envelope
sketch of a lateral current flow with hex shaped poly and it doesn't
pack as well as the waffle. Do you a particular device with a diagram
showing the layout?

No. It is a horizontal device with a hexagonal channel pattern. See:
http://www.irf.com/technical-info/guide/device.html

 

[[email protected]]

No.  It is a horizontal device with a hexagonal channel pattern.  See:http://www.irf.com/technical-info/guide/device.html

I don't see a lateral current flow device on that page. You can see
the epi layer on the hexfet with drain contact labeled on the bottom.
Here is the text from that page:
"As shown in Figure 17 above, the HEXFET® power MOSFET is named for
the hexagonal shape of its individual cells. Current flows from the
source metallization down through the device, and out through the
drain contact. Vertical current flow is the reason the HEXFET is also
called a vertical MOSFET. Nearly all power MOSFETs on the market
employ this vertical structure."

It says vertical current flow, which matches the diagram.

 

[JosephKK]

I don't see a lateral current flow device on that page. You can see
the epi layer on the hexfet with drain contact labeled on the bottom.
Here is the text from that page:
"As shown in Figure 17 above, the HEXFET® power MOSFET is named for
the hexagonal shape of its individual cells. Current flows from the
source metallization down through the device, and out through the
drain contact. Vertical current flow is the reason the HEXFET is also
called a vertical MOSFET. Nearly all power MOSFETs on the market
employ this vertical structure."

It says vertical current flow, which matches the diagram.

The channel flow is horizontal across the near surface the hexagonal p
regions. It is lateral, as opposed to the earlier VMOS parts. Also
look up TMOS (TI IIRC).

 

[[email protected]]

The channel flow is horizontal across the near surface the hexagonal p
regions.  It is lateral, as opposed to the earlier VMOS parts.  Also
look up TMOS (TI IIRC).

I just don't see how the current flow is lateral without drain
contacts on top. I suppose they could do an implant to hit epi, but
that would not be efficient. Anyway, the idea is to use vertical flow
to make the device denser for the given on-resistance. And then how do
you explain their own datasheet saying the flow is vertical?

No luck on TMOS. Lots of part numbers, but no profile.

 

[Eeyore]

But that doesn't explain the current flow. These lateral current flow
devices don't need epi, i.e. there isn't a low resistance path to the
back side, so they should be pretty cheap to build. I'm not a process
guru, but I've been told vertical flow is more likely to effected by
wafer defects than lateral flow, so the linear MOS device should be
high yield.

I know from ESD testing, the waffle pattern seemed less rugged than
plain stripes.

From the photographs I recall, the Hitachi types were plain stripes. They were
damn difficult to kill but C-Audio managed it in their designs (insufficient
cooling). Think heatsink temp > 150C !

Graham

 

[Eeyore]

No. It is a horizontal device with a hexagonal channel pattern. See:
http://www.irf.com/technical-info/guide/device.html

News to me after about several decades !

Graham

 

[Eeyore]

I just don't see how the current flow is lateral without drain
contacts on top. I suppose they could do an implant to hit epi, but
that would not be efficient. Anyway, the idea is to use vertical flow
to make the device denser for the given on-resistance. And then how do
you explain their own datasheet saying the flow is vertical?

Precisely my understanding too. The consequence of using lateral mosfets is a relatively high on-resistance.

Graham

 

[JosephKK]

News to me after about several decades !

Quoting from the cited device.html page:

"As shown in Figure 17 above, the HEXFET® power MOSFET is named for
the hexagonal shape of its individual cells. Current flows from the
source metallization down through the device, and out through the
drain contact. Vertical current flow is the reason the HEXFET is also
called a vertical MOSFET. Nearly all power MOSFETs on the market
employ this vertical structure."[/QUOTE]

Just the same, where is the channel? What is the orientation?