I am looking for an output amplifier for a signal generator

[Ken Smith]

Way back, I'd drooled over using one of National's LH0033s. Most probably
because they'd cheekily advertised it as "Damned fast". Never got the chance
as at the time they cost way too much.

The LH0063 IIRC ws even faster. I actually did get to use LH0033s. They
were quite nice about improperly terminated lines.

 

[mmm]

Harry Dellamano wrote...



Use a better current-source, e.g., Wilson mirror, etc.

nice piece of hardware , but is not present in your book

Mr Hill , many thanks for your work, it's an enourmuos collection of
info on electronic technology ( I have some doubt about the WOM memory
..... )

I think to reserve this design for another application where I probably
need a bit more output voltage and current

but I need some clarification

1) can I use non SMD parts, good old 2n2222 etc. ?

2) can I modifie the feedback network ( 499 - 499 resistor ) or even
consider the circuit as a discrete component OA ?

3) can I parallel some more output transistor ?

at +-25 volt supply can be a very nice beast

 

[mmm]

John Larkin wrote...



Yes, very nice. John, can you tell us your recommendations for
effectively using TI's PowerPad thermal path in prototype assembly?

I am bit curious too , the power pad is around 2mm x 2.5 mmm

perhaps I will use a small strip of copper but .....

I the meanwhile I found two interesting chip from texas 'improved AD815'

ths6032
ths6012

so-powerpad too but a bit bigger ....

do the paralleling of op amps is a routine ( with output resistors ) or
not ?

 

[Winfield Hill]

mmm wrote...

nice piece of hardware , but is not present in your book

See figure 2.48

Mr Hill , many thanks for your work, it's an enourmuos collection
of info on electronic technology ( I have some doubt about the
WOM memory .... )

I think to reserve this design for another application where I
probably need a bit more output voltage and current

but I need some clarification

1) can I use non SMD parts, good old 2n2222 etc. ?

2) can I modifie the feedback network ( 499 - 499 resistor )
or even consider the circuit as a discrete component OA ?

3) can I parallel some more output transistor ?

at +-25 volt supply can be a very nice beast

Yes.

 

[Joerg]

Hello Ken,

The LH0063 IIRC ws even faster. ...

They are great. But I guess the market for those was too small.

Regards, Joerg

 

[mmm]

mmm wrote...



See figure 2.48

not the Wilson current mirror .... but the high speed amplifier as a whole.

I was already aware of the presence in the book of the current mirrors
as 'programmable' current source.

 

[Ken Smith]

Something worries me about this circuit. I'm going to add voltages
Modified:
It looks to me like the 9.3V on the collector of the PNP transistor would
be forward biasing it. Did I miss something?

 

[Winfield Hill]

Ken Smith wrote...

Winfield Hill wrote: [ circuit snipped ]

Something worries me about this circuit. I'm going to add voltages
Modified:
It looks to me like the 9.3V on the collector of the PNP transistor
would be forward biasing it. Did I miss something?

Which PNP are you referring to, the input PNP long-tail differential
pair? It's base would be at +5V. Oops! That doesn't work at all!

 

[Winfield Hill]

mmm wrote...

not the Wilson current mirror .... but the high speed amplifier
as a whole. I was already aware of the presence in the book of
the current mirrors as 'programmable' current source.

Well, they make the best current sources. I get 20 transistors,
unless one wants to add two more for reliable performance, then
it's 22 transistors total. Not so many in the IC world, nor in
our new SMT machine-assembly "parts-are-free" world, either.

 

[John Larkin]

John Larkin wrote...

Yes, very nice. John, can you tell us your recommendations for
effectively using TI's PowerPad thermal path in prototype assembly?

You really need a pc board to test parts like this. Given that, a
copper island under the chip, peppered with thermal vias to the ground
plane (if a multilayer) and also to a 'thermal antenna' pad on the
flip side. It doesn't take a lot of copper to cut the theta of an SO-8
by a factor of 3 or so.

Like most fast amps, these run out of swing at high currents and close
to the rails, so you get droopy corners. If you back off on current
enough to get crisp edges, the power dissipation per part is reduced,
too. But then, the THS3062 gets pretty hot at Iq, so needs some
heatsink help even of you don't load them hard. The 3-in-parallel
situation is a pretty good compromise.

Somebody should make a universal pin driver: +-10 swing at least, LVDS
input, programmable Vh:Vl, thermal limiting, fast/clean edges in the
5v/ns sort of range.

John

 

[Winfield Hill]

John Larkin wrote...

You really need a pc board to test parts like this. Given that, a
copper island under the chip, peppered with thermal vias to the ground
plane (if a multilayer) and also to a 'thermal antenna' pad on the
flip side. It doesn't take a lot of copper to cut the theta of an SO-8
by a factor of 3 or so.

Can you give us a typical R-theta for the setup you describe above?

 

[John Larkin]

John Larkin wrote...

Can you give us a typical R-theta for the setup you describe above?

Nope.

John

 

[Ken Smith]

How about we remove the gain of 2:

Now the PNP still has a reverse bias on its collector but I'm still
worried about it a bit. Many bipolars have a "dynamic saturation" effect
where they slow down as the collector voltage gets down below about 2V. I
don't off hand know to what degree XXX2907s have this effect.

How about:
Vcc Vcc Vcc
! ! !
\ \ /
/ / \
\ \ /
/ / \
! ! !
V ! !
--- ! !
! !/ e !/ e
+-------!----------!------ .... repeat as needed
! !\ !\
!/ !
in-! !
!\ e !
! ! ........
+--- +---------+------- Out
! ! ! . !
\ \ ! . /
/ / ! . \ 50R
\ \ ! . /
/ / ! . \
! ! ! . !
gnd Vee ! . Gnd
! .
----- Mirror line -----------

 

[John Larkin]

Very nice Ban and Win. I count 20 transistors with the current sources
added. What have I missed? Please don't answer, "everything since the third
grade".
Regards,
Harry

This is the kind of circuit we refer to as "component rich."

John

 

[Jim Thompson]

This is the kind of circuit we refer to as "component rich."

John

Seems pretty SPARSE to me ;-)

The PORTION of a WiFi chip that I'm currently working on (just the
limiting/log RSSI IF strip) has....

PMOS 6
NMOS 54
NPN 296 (SiGe)
R 288
C 318

Counts in the 20's are simply farts in a strong breeze ;-)

...Jim Thompson

 

[John Larkin]

Seems pretty SPARSE to me ;-)

The PORTION of a WiFi chip that I'm currently working on (just the
limiting/log RSSI IF strip) has....

PMOS 6
NMOS 54
NPN 296 (SiGe)
R 288
C 318

Counts in the 20's are simply farts in a strong breeze ;-)

...Jim Thompson

Sometimes an IC datasheet will include a component count. It's amazing
to see (it's getting hard to type, with two new kittens walking on my
keyboard. The Brat named them Ajax and Comet) numbers like "278
equivalent transistors" for an opamp or a simple logic part.

That's one reason that discrete circuit design (ie, not IC) is a dying
art: most of the time, it makes more sense to buy ICs than to design
with a bunch of discretes. So us circuit designers have to move up the
abstraction stack, buy lots of cheap chips, and combine them into more
complex stuff. I'm doing small PCBs that would have been racks full of
stuff a few years ago. A few exceptions remain: high power, high
speed, high voltage, RF.

Simple = elegant.

John

 

[Jim Thompson]

On Tue, 28 Jun 2005 08:03:49 -0700, John Larkin
[snip]

This is the kind of circuit we refer to as "component rich."

John

Seems pretty SPARSE to me ;-)

The PORTION of a WiFi chip that I'm currently working on (just the
limiting/log RSSI IF strip) has....

PMOS 6
NMOS 54
NPN 296 (SiGe)
R 288
C 318

Counts in the 20's are simply farts in a strong breeze ;-)

...Jim Thompson

Sometimes an IC datasheet will include a component count. It's amazing
to see (it's getting hard to type, with two new kittens walking on my
keyboard. The Brat named them Ajax and Comet) numbers like "278
equivalent transistors" for an opamp or a simple logic part.

That's one reason that discrete circuit design (ie, not IC) is a dying
art: most of the time, it makes more sense to buy ICs than to design
with a bunch of discretes. So us circuit designers have to move up the
abstraction stack, buy lots of cheap chips, and combine them into more
complex stuff. I'm doing small PCBs that would have been racks full of
stuff a few years ago. A few exceptions remain: high power, high
speed, high voltage, RF.

Simple = elegant.

John

I've been at IC design for so long, ~43 years now, that I have trouble
doing cost-efficient discrete designs... I tend to use too many
components... of course everything works textbook since, in my world,
there are NO tweaks allowed.

...Jim Thompson

 

[Joerg]

Hello John,

That's one reason that discrete circuit design (ie, not IC) is a dying
art: most of the time, it makes more sense to buy ICs than to design
with a bunch of discretes. So us circuit designers have to move up the
abstraction stack, buy lots of cheap chips, and combine them into more
complex stuff. I'm doing small PCBs that would have been racks full of
stuff a few years ago. A few exceptions remain: high power, high
speed, high voltage, RF.

Not always. It is amazing how much of the mundane stuff is still done in
discrete. I was looking whether the time had come, after 10 years or so,
to move one of the designs I had done for a client to a uC or maybe an
ASIC. So I asked them what the current cost per board was. The answer
blew me away. Maybe I'll look at the circuit again in another ten years.

Simple = elegant.

Sometimes it is "cheap = better than elegant" ;-)

Regards, Joerg

 

[Jim Thompson]

Here is a discrete O/P stage, you can parallel more transistors. I also have
a gain stage somewhere, but do not remember the file name. :-(

+---+-----------+----+-----
| | | |
| | | |
current sources (I) (I) | |
| | _ |/ |/
| +------U-------| BF720
| | bead |> |>
| | | |
| | | |
o +---)---)----+ | |
| |/ | \| | .-. .-.
+-| | |- + | | | |
| |> | <| | | | | |
| +-+-)---+ | '-' '-'
| | | BF720 | | |
| | | | | |
| | | +------+----+-o
| | | BF721 | | |
| +-)-+---+ | .-. .-.
| |< | >| | | | | |
+-| | |- + | | | |
|\ | /| | '-' '-'
+-)-----)----+ | |
| | | |
| | _ |< |<
| +------U-------| BF721
| | bead |\ |\
current sources (I) (I) | |
| | | |
| | | |
+-----+-----------+----+----
(created by AACircuit v1.28 beta 10/06/04 www.tech-chat.de)
view\fixed font

What subtlety am I missing that keeps it from having big rail-to-rail
current draw?

...Jim Thompson

 

[John Larkin]

John Larkin wrote...

Can you give us a typical R-theta for the setup you describe above?

Take a look at the board in a.b.s.e. It has two GigaLogic chips and
two GaAs distributed amps, all power-padded and heatsunk. Ghastly.

John