 |
|
| Science Forum Index » Electronics - Design Forum » D at (no spam) mn ground issues... again...... |
|
Page 1 of 2 Goto page 1, 2 Next |
|
| Author |
Message |
| George Herold... |
 Posted: Fri Nov 06, 2009 4:00 pm |
|
|
|
Guest
|
A proto PCB arrived earlier this week. This is the 2nd generation
board for the High Level electronics to measure noise. This
instrument is for teaching physics students about noise and is
modular. The students must make connections between the modules to
make it all work. Something like this.
http://www.teachspin.com/instruments/signal_processor/index.shtml
It has two switch selectable SV filter sections (10, 30... 3k and
300, 1k,...100k Hz) A gain section, Gain of 10 to 10^4 with switch
selectable fine control. And a multiplier section. (The first
generation board had some conceptual errors that lead to a design with
gain peaking at the highest gain settings.)
I populated the gain section and it looked great. There was no gain
peaking and at the highest gain (10^4) the 3dB point was at about
1.5MHz. I should mention here that the output from the final opamp is
coupled to the output BNC connector with coax. With bare wire
capacitive coupling from the output to input leads to gain peaking.
The output BNC is also isolated from the front panel, with the return
currents flowing back along the outer braid of the coax, and grounded
back at the output opamp. If the output is not isolated there is
again gain peaking, due, I assume, to the return currents flowing back
along the shield and making the input and output connectors have
slightly different voltages. (The only ground connection between the
shield and pcb is made at the input to the gain stage.)
So far, so good. I then powered up the filter section and plugged it
into the gain stage. Crap!!! Gain peaking at the highest gains. And
if I run the High Pass output from the filter into the gain stage it
breaks into oscillations. (Gain peaking is at about 1.5MHz as you
might expect and the oscillation frequency is 1MHz.)
I've spent the past few days banging my head against this PCB.
The filters and gain section were made with a common ground plane.
(The multiplier section aready has a separate ground plane and power
filter.) I dremeled a slot between the planes and ran wires back to
the power supply ground input connection. (No real change.) I won't
bore you with the rest of the gory details, except to say that I did
learn that capacitive loading of the output made matters worse. This
was not capacitive loading of the final opamp. If I grounded the
input of the gain stage or drove it from some external generator there
was no gain peaking when I added more capacitance to the output. I was
adding 220 pF to the already existing 3 foot coax cable running to the
scope
Today I powered up the filter section from a pair of 9V batteries.
When I used the shield as the ground for the filter section every
thing was great. No gain peaking. When I used the power supply
ground from the gain stage as ground it was the same old oscillating
story.
So I then built this configuration
+-------+ |Shield
(V+)----R1R1---+---+ Vin | |
| | | |
C |filter | |
C | | |
+---+ GND---+------+
| | |
+-------+
Where R1 was 100 ohms and C was a parallel combination of 47uF tant,
1uF ceramic and 0,1uF ceramic.
This worked fine, but I'm not sure I like the return current from the
filter section flowing through the shield. (And then back through the
gain stage.)
So then I made this,
+-------+ |Shield
(V+)----R1R1---+---+ Vin | |
| | | |
C |filter | |
C | | |
PGND---LLL----+---+ GND---+-R2R2-+
| +-C2C2-+
+-------+ |
Where R1 and C are as before, PGND is the power supply ground, L is
100mH, R2 is 10 ohms and C2 is 1 uF.
This worked also. (Though I still have the high frequency currents
flowing through the shield.)
The values were picked from what was lying around, It was late Friday
and I decided to declare victory and go home.
There is a symmetrical R and C on the negative supply line.
If you've read this far, Thanks.
Is there something obvious I might have missed? Some way to make it
better?
I'm not sure which of the above configurations to use.... My fear is
that neither may end up working. I worry about sending the ground
currents from the filter section through the gain section.
Thanks again,
George H. |
|
|
| Back to top |
|
|
|
| George Herold... |
| Posted: Fri Nov 06, 2009 5:13 pm |
|
|
|
Guest
|
On Nov 6, 9:54 pm, "Jon Slaughter" <Jon_Slaugh... at (no spam) Hotmail.com> wrote:
[quote]George Herold wrote:
A proto PCB arrived earlier this week. This is the 2nd generation
board for the High Level electronics to measure noise. This
instrument is for teaching physics students about noise and is
modular. The students must make connections between the modules to
make it all work. Something like this.
http://www.teachspin.com/instruments/signal_processor/index.shtml
It has two switch selectable SV filter sections (10, 30... 3k and
300, 1k,...100k Hz) A gain section, Gain of 10 to 10^4 with switch
selectable fine control. And a multiplier section. (The first
generation board had some conceptual errors that lead to a design with
gain peaking at the highest gain settings.)
I populated the gain section and it looked great. There was no gain
peaking and at the highest gain (10^4) the 3dB point was at about
1.5MHz. I should mention here that the output from the final opamp is
coupled to the output BNC connector with coax. With bare wire
capacitive coupling from the output to input leads to gain peaking.
The output BNC is also isolated from the front panel, with the return
currents flowing back along the outer braid of the coax, and grounded
back at the output opamp. If the output is not isolated there is
again gain peaking, due, I assume, to the return currents flowing back
along the shield and making the input and output connectors have
slightly different voltages. (The only ground connection between the
shield and pcb is made at the input to the gain stage.)
So far, so good. I then powered up the filter section and plugged it
into the gain stage. Crap!!! Gain peaking at the highest gains. And
if I run the High Pass output from the filter into the gain stage it
breaks into oscillations. (Gain peaking is at about 1.5MHz as you
might expect and the oscillation frequency is 1MHz.)
I've spent the past few days banging my head against this PCB.
The filters and gain section were made with a common ground plane.
(The multiplier section aready has a separate ground plane and power
filter.) I dremeled a slot between the planes and ran wires back to
the power supply ground input connection. (No real change.) I won't
bore you with the rest of the gory details, except to say that I did
learn that capacitive loading of the output made matters worse. This
was not capacitive loading of the final opamp. If I grounded the
input of the gain stage or drove it from some external generator there
was no gain peaking when I added more capacitance to the output. I was
adding 220 pF to the already existing 3 foot coax cable running to the
scope
Today I powered up the filter section from a pair of 9V batteries.
When I used the shield as the ground for the filter section every
thing was great. No gain peaking. When I used the power supply
ground from the gain stage as ground it was the same old oscillating
story.
So I then built this configuration
+-------+ |Shield
(V+)----R1R1---+---+ Vin | |
| | | |
C |filter | |
C | | |
+---+ GND---+------+
| | |
+-------+
Where R1 was 100 ohms and C was a parallel combination of 47uF tant,
1uF ceramic and 0,1uF ceramic.
This worked fine, but I'm not sure I like the return current from the
filter section flowing through the shield. (And then back through the
gain stage.)
So then I made this,
+-------+ |Shield
(V+)----R1R1---+---+ Vin | |
| | | |
C |filter | |
C | | |
PGND---LLL----+---+ GND---+-R2R2-+
| +-C2C2-+
+-------+ |
Where R1 and C are as before, PGND is the power supply ground, L is
100mH, R2 is 10 ohms and C2 is 1 uF.
This worked also. (Though I still have the high frequency currents
flowing through the shield.)
The values were picked from what was lying around, It was late Friday
and I decided to declare victory and go home.
There is a symmetrical R and C on the negative supply line.
If you've read this far, Thanks.
Is there something obvious I might have missed? Some way to make it
better?
I'm not sure which of the above configurations to use.... My fear is
that neither may end up working. I worry about sending the ground
currents from the filter section through the gain section.
Thanks again,
Do what most engineers do... Stamp it approved and get it sold...- Hide quoted text -
- Show quoted text -
[/quote]
Ahh, Well it's a small company, I have to test it before it leaves and
fix it if it comes back. Better to get it right before it ships.
George H. |
|
|
| Back to top |
|
|
|
| Jon Slaughter... |
| Posted: Fri Nov 06, 2009 9:54 pm |
|
|
|
Guest
|
George Herold wrote:
[quote]A proto PCB arrived earlier this week. This is the 2nd generation
board for the High Level electronics to measure noise. This
instrument is for teaching physics students about noise and is
modular. The students must make connections between the modules to
make it all work. Something like this.
http://www.teachspin.com/instruments/signal_processor/index.shtml
It has two switch selectable SV filter sections (10, 30... 3k and
300, 1k,...100k Hz) A gain section, Gain of 10 to 10^4 with switch
selectable fine control. And a multiplier section. (The first
generation board had some conceptual errors that lead to a design with
gain peaking at the highest gain settings.)
I populated the gain section and it looked great. There was no gain
peaking and at the highest gain (10^4) the 3dB point was at about
1.5MHz. I should mention here that the output from the final opamp is
coupled to the output BNC connector with coax. With bare wire
capacitive coupling from the output to input leads to gain peaking.
The output BNC is also isolated from the front panel, with the return
currents flowing back along the outer braid of the coax, and grounded
back at the output opamp. If the output is not isolated there is
again gain peaking, due, I assume, to the return currents flowing back
along the shield and making the input and output connectors have
slightly different voltages. (The only ground connection between the
shield and pcb is made at the input to the gain stage.)
So far, so good. I then powered up the filter section and plugged it
into the gain stage. Crap!!! Gain peaking at the highest gains. And
if I run the High Pass output from the filter into the gain stage it
breaks into oscillations. (Gain peaking is at about 1.5MHz as you
might expect and the oscillation frequency is 1MHz.)
I've spent the past few days banging my head against this PCB.
The filters and gain section were made with a common ground plane.
(The multiplier section aready has a separate ground plane and power
filter.) I dremeled a slot between the planes and ran wires back to
the power supply ground input connection. (No real change.) I won't
bore you with the rest of the gory details, except to say that I did
learn that capacitive loading of the output made matters worse. This
was not capacitive loading of the final opamp. If I grounded the
input of the gain stage or drove it from some external generator there
was no gain peaking when I added more capacitance to the output. I was
adding 220 pF to the already existing 3 foot coax cable running to the
scope
Today I powered up the filter section from a pair of 9V batteries.
When I used the shield as the ground for the filter section every
thing was great. No gain peaking. When I used the power supply
ground from the gain stage as ground it was the same old oscillating
story.
So I then built this configuration
+-------+ |Shield
(V+)----R1R1---+---+ Vin | |
| | | |
C |filter | |
C | | |
+---+ GND---+------+
| | |
+-------+
Where R1 was 100 ohms and C was a parallel combination of 47uF tant,
1uF ceramic and 0,1uF ceramic.
This worked fine, but I'm not sure I like the return current from the
filter section flowing through the shield. (And then back through the
gain stage.)
So then I made this,
+-------+ |Shield
(V+)----R1R1---+---+ Vin | |
| | | |
C |filter | |
C | | |
PGND---LLL----+---+ GND---+-R2R2-+
| +-C2C2-+
+-------+ |
Where R1 and C are as before, PGND is the power supply ground, L is
100mH, R2 is 10 ohms and C2 is 1 uF.
This worked also. (Though I still have the high frequency currents
flowing through the shield.)
The values were picked from what was lying around, It was late Friday
and I decided to declare victory and go home.
There is a symmetrical R and C on the negative supply line.
If you've read this far, Thanks.
Is there something obvious I might have missed? Some way to make it
better?
I'm not sure which of the above configurations to use.... My fear is
that neither may end up working. I worry about sending the ground
currents from the filter section through the gain section.
Thanks again,
[/quote]
Do what most engineers do... Stamp it approved and get it sold... |
|
|
| Back to top |
|
|
|
| ... |
| Posted: Sun Nov 08, 2009 2:10 pm |
|
|
|
Guest
|
On Nov 6, 9:00 pm, George Herold <ggher... at (no spam) gmail.com> wrote:
[quote]A proto PCB arrived earlier this week. This is the 2nd generation
board for the High Level electronics to measure noise. This
instrument is for teaching physics students about noise and is
modular. The students must make connections between the modules to
make it all work. Something like this.http://www.teachspin.com/instruments/signal_processor/index.shtml
It has two switch selectable SV filter sections (10, 30... 3k and
300, 1k,...100k Hz) A gain section, Gain of 10 to 10^4 with switch
selectable fine control. And a multiplier section. (The first
generation board had some conceptual errors that lead to a design with
gain peaking at the highest gain settings.)
I populated the gain section and it looked great. There was no gain
peaking and at the highest gain (10^4) the 3dB point was at about
1.5MHz. I should mention here that the output from the final opamp is
coupled to the output BNC connector with coax. With bare wire
capacitive coupling from the output to input leads to gain peaking.
The output BNC is also isolated from the front panel, with the return
currents flowing back along the outer braid of the coax, and grounded
back at the output opamp. If the output is not isolated there is
again gain peaking, due, I assume, to the return currents flowing back
along the shield and making the input and output connectors have
slightly different voltages. (The only ground connection between the
shield and pcb is made at the input to the gain stage.)
So far, so good. I then powered up the filter section and plugged it
into the gain stage. Crap!!! Gain peaking at the highest gains. And
if I run the High Pass output from the filter into the gain stage it
breaks into oscillations. (Gain peaking is at about 1.5MHz as you
might expect and the oscillation frequency is 1MHz.)
I've spent the past few days banging my head against this PCB.
The filters and gain section were made with a common ground plane.
(The multiplier section aready has a separate ground plane and power
filter.) I dremeled a slot between the planes and ran wires back to
the power supply ground input connection. (No real change.) I won't
bore you with the rest of the gory details, except to say that I did
learn that capacitive loading of the output made matters worse. This
was not capacitive loading of the final opamp. If I grounded the
input of the gain stage or drove it from some external generator there
was no gain peaking when I added more capacitance to the output. I was
adding 220 pF to the already existing 3 foot coax cable running to the
scope
Today I powered up the filter section from a pair of 9V batteries.
When I used the shield as the ground for the filter section every
thing was great. No gain peaking. When I used the power supply
ground from the gain stage as ground it was the same old oscillating
story.
So I then built this configuration
+-------+ |Shield
(V+)----R1R1---+---+ Vin | |
| | | |
C |filter | |
C | | |
+---+ GND---+------+
| | |
+-------+
Where R1 was 100 ohms and C was a parallel combination of 47uF tant,
1uF ceramic and 0,1uF ceramic.
This worked fine, but I'm not sure I like the return current from the
filter section flowing through the shield. (And then back through the
gain stage.)
So then I made this,
+-------+ |Shield
(V+)----R1R1---+---+ Vin | |
| | | |
C |filter | |
C | | |
PGND---LLL----+---+ GND---+-R2R2-+
| +-C2C2-+
+-------+ |
Where R1 and C are as before, PGND is the power supply ground, L is
100mH, R2 is 10 ohms and C2 is 1 uF.
This worked also. (Though I still have the high frequency currents
flowing through the shield.)
The values were picked from what was lying around, It was late Friday
and I decided to declare victory and go home.
There is a symmetrical R and C on the negative supply line.
If you've read this far, Thanks.
Is there something obvious I might have missed? Some way to make it
better?
I'm not sure which of the above configurations to use.... My fear is
that neither may end up working. I worry about sending the ground
currents from the filter section through the gain section.
Thanks again,
George H.
[/quote]
I'm with Nemo's Thought #2: gain-peaking suggests the op-amp's annoyed
by its capacitive loading. CL=220pf is about 500 ohms at (no spam) 1.5MHz. Even
lousy grounding's impedance at (no spam) 1.5MHz should be beat that by so much
that ground-loop feedback currents shouldn't matter. They might, but
they shouldn't.
Nemo's paralleled feedback cap is the cure, or classic op-amp
capacitive load isolation techniques, or a different, huskier op-amp.
Nice ASCII art.
--
Cheers,
James Arthur |
|
|
| Back to top |
|
|
|
| ... |
| Posted: Sun Nov 08, 2009 5:26 pm |
|
|
|
Guest
|
On Nov 8, 7:56 pm, John Larkin
<jjlar... at (no spam) highNOTlandTHIStechnologyPART.com> wrote:
[quote]And I'd recommend not connecting an opamp output to an output
connector; you never know what the load may be.
John
[/quote]
Yep. At a minimum isolate the feedback node from the load:
.-----R2-------------.
| |
+-----C1-----. |
| | |
| |\ | |
| | \ | |
Vin >---R1--+---|- \ | |
| >----+---R3--+----> Vout
.---|+ /
| | /
| |/
===
GND
I've done that. Or use a buffer.
Which op-amp are we talking about here George?
--
Cheers,
James Arthur |
|
|
| Back to top |
|
|
|
| Nemo... |
| Posted: Sun Nov 08, 2009 6:32 pm |
|
|
|
Guest
|
Possibility #1: Are any of the other BNC connectors between each stage
isolated, or are they metal shelled types shorting through the metal
panel, giving you unexpected return currents?
Thought #2: The fact that the signal gets oscillations on it, and all
this gain peaking, when you connect the filter to the input of the amp
implies you are putting too much capacitance on the input to the
amplifier. Solutions are usually to either add say 10pF in the feedback
loop of the amp (assuming it's as simple as an inverting op amp); or
change to a non inverting amp, which is often better behaved; or just
pay some attention to the physical connection between these two stages
to reduce the capacitance (maybe the filter itself is too capacitive);
or add a unity gain buffer between them; or add a cascode stage in front
of the amp to isolate incoming capacitance, see here -
http://www.electrooptical.net/www/frontends/frontends.pdf
Thought #3: 1.5MHz isn't very fast. Why use coax? Might be simpler to
use twisted pair, easier to manufacture than all those fiddly braid
stripping & connections.
I suspect I've pitched that explanation below your level, but I hope it
helps spark some ideas.
Nemo
In article
<60109f90-2ec9-45cd-80cf-1b888d4cceec at (no spam) u16g2000pru.googlegroups.com>,
George Herold writes
[quote]
A proto PCB arrived earlier this week. This is the 2nd generation
board for the High Level electronics to measure noise. This
instrument is for teaching physics students about noise and is
modular. The students must make connections between the modules to
make it all work. Something like this.
http://www.teachspin.com/instruments/signal_processor/index.shtml
It has two switch selectable SV filter sections (10, 30... 3k and
300, 1k,...100k Hz) A gain section, Gain of 10 to 10^4 with switch
selectable fine control. And a multiplier section. (The first
generation board had some conceptual errors that lead to a design with
gain peaking at the highest gain settings.)
I populated the gain section and it looked great. There was no gain
peaking and at the highest gain (10^4) the 3dB point was at about
1.5MHz. I should mention here that the output from the final opamp is
coupled to the output BNC connector with coax. With bare wire
capacitive coupling from the output to input leads to gain peaking.
The output BNC is also isolated from the front panel, with the return
currents flowing back along the outer braid of the coax, and grounded
back at the output opamp. If the output is not isolated there is
again gain peaking, due, I assume, to the return currents flowing back
along the shield and making the input and output connectors have
slightly different voltages. (The only ground connection between the
shield and pcb is made at the input to the gain stage.)
So far, so good. I then powered up the filter section and plugged it
into the gain stage. Crap!!! Gain peaking at the highest gains. And
if I run the High Pass output from the filter into the gain stage it
breaks into oscillations. (Gain peaking is at about 1.5MHz as you
might expect and the oscillation frequency is 1MHz.)
I've spent the past few days banging my head against this PCB.
The filters and gain section were made with a common ground plane.
(The multiplier section aready has a separate ground plane and power
filter.) I dremeled a slot between the planes and ran wires back to
the power supply ground input connection. (No real change.) I won't
bore you with the rest of the gory details, except to say that I did
learn that capacitive loading of the output made matters worse. This
was not capacitive loading of the final opamp. If I grounded the
input of the gain stage or drove it from some external generator there
was no gain peaking when I added more capacitance to the output. I was
adding 220 pF to the already existing 3 foot coax cable running to the
scope
Today I powered up the filter section from a pair of 9V batteries.
When I used the shield as the ground for the filter section every
thing was great. No gain peaking. When I used the power supply
ground from the gain stage as ground it was the same old oscillating
story.
So I then built this configuration
+-------+ |Shield
(V+)----R1R1---+---+ Vin | |
| | | |
C |filter | |
C | | |
+---+ GND---+------+
| | |
+-------+
Where R1 was 100 ohms and C was a parallel combination of 47uF tant,
1uF ceramic and 0,1uF ceramic.
This worked fine, but I'm not sure I like the return current from the
filter section flowing through the shield. (And then back through the
gain stage.)
So then I made this,
+-------+ |Shield
(V+)----R1R1---+---+ Vin | |
| | | |
C |filter | |
C | | |
PGND---LLL----+---+ GND---+-R2R2-+
| +-C2C2-+
+-------+ |
Where R1 and C are as before, PGND is the power supply ground, L is
100mH, R2 is 10 ohms and C2 is 1 uF.
This worked also. (Though I still have the high frequency currents
flowing through the shield.)
The values were picked from what was lying around, It was late Friday
and I decided to declare victory and go home.
There is a symmetrical R and C on the negative supply line.
If you've read this far, Thanks.
Is there something obvious I might have missed? Some way to make it
better?
I'm not sure which of the above configurations to use.... My fear is
that neither may end up working. I worry about sending the ground
currents from the filter section through the gain section.
Thanks again,
George H.[/quote] |
|
|
| Back to top |
|
|
|
| John Larkin... |
| Posted: Sun Nov 08, 2009 7:56 pm |
|
|
|
Guest
|
On Fri, 6 Nov 2009 18:00:23 -0800 (PST), George Herold
<ggherold at (no spam) gmail.com> wrote:
[quote]
A proto PCB arrived earlier this week. This is the 2nd generation
board for the High Level electronics to measure noise. This
instrument is for teaching physics students about noise and is
modular. The students must make connections between the modules to
make it all work. Something like this.
http://www.teachspin.com/instruments/signal_processor/index.shtml
It has two switch selectable SV filter sections (10, 30... 3k and
300, 1k,...100k Hz) A gain section, Gain of 10 to 10^4 with switch
selectable fine control. And a multiplier section. (The first
generation board had some conceptual errors that lead to a design with
gain peaking at the highest gain settings.)
I populated the gain section and it looked great. There was no gain
peaking and at the highest gain (10^4) the 3dB point was at about
1.5MHz. I should mention here that the output from the final opamp is
coupled to the output BNC connector with coax. With bare wire
capacitive coupling from the output to input leads to gain peaking.
The output BNC is also isolated from the front panel, with the return
currents flowing back along the outer braid of the coax, and grounded
back at the output opamp. If the output is not isolated there is
again gain peaking, due, I assume, to the return currents flowing back
along the shield and making the input and output connectors have
slightly different voltages. (The only ground connection between the
shield and pcb is made at the input to the gain stage.)
So far, so good. I then powered up the filter section and plugged it
into the gain stage. Crap!!! Gain peaking at the highest gains. And
if I run the High Pass output from the filter into the gain stage it
breaks into oscillations. (Gain peaking is at about 1.5MHz as you
might expect and the oscillation frequency is 1MHz.)
I've spent the past few days banging my head against this PCB.
The filters and gain section were made with a common ground plane.
(The multiplier section aready has a separate ground plane and power
filter.) I dremeled a slot between the planes and ran wires back to
the power supply ground input connection. (No real change.) I won't
bore you with the rest of the gory details, except to say that I did
learn that capacitive loading of the output made matters worse. This
was not capacitive loading of the final opamp. If I grounded the
input of the gain stage or drove it from some external generator there
was no gain peaking when I added more capacitance to the output. I was
adding 220 pF to the already existing 3 foot coax cable running to the
scope
Today I powered up the filter section from a pair of 9V batteries.
When I used the shield as the ground for the filter section every
thing was great. No gain peaking. When I used the power supply
ground from the gain stage as ground it was the same old oscillating
story.
So I then built this configuration
+-------+ |Shield
(V+)----R1R1---+---+ Vin | |
| | | |
C |filter | |
C | | |
+---+ GND---+------+
| | |
+-------+
Where R1 was 100 ohms and C was a parallel combination of 47uF tant,
1uF ceramic and 0,1uF ceramic.
This worked fine, but I'm not sure I like the return current from the
filter section flowing through the shield. (And then back through the
gain stage.)
So then I made this,
+-------+ |Shield
(V+)----R1R1---+---+ Vin | |
| | | |
C |filter | |
C | | |
PGND---LLL----+---+ GND---+-R2R2-+
| +-C2C2-+
+-------+ |
Where R1 and C are as before, PGND is the power supply ground, L is
100mH, R2 is 10 ohms and C2 is 1 uF.
This worked also. (Though I still have the high frequency currents
flowing through the shield.)
The values were picked from what was lying around, It was late Friday
and I decided to declare victory and go home.
There is a symmetrical R and C on the negative supply line.
If you've read this far, Thanks.
Is there something obvious I might have missed? Some way to make it
better?
I'm not sure which of the above configurations to use.... My fear is
that neither may end up working. I worry about sending the ground
currents from the filter section through the gain section.
Thanks again,
George H.
[/quote]
Life is much simpler if you ground everything... the PCB ground plane,
all the connector shells, the power supply common, everything, to the
metal chassis. Your single input-related ground may be flailing as a
result of output current.
And I'd recommend not connecting an opamp output to an output
connector; you never know what the load may be.
John |
|
|
| Back to top |
|
|
|
| George Herold... |
| Posted: Mon Nov 09, 2009 4:12 am |
|
|
|
Guest
|
On Nov 8, 6:32 pm, Nemo <z... at (no spam) nospam.nospam.nospam.nospam.co.uk>
wrote:
[quote]Possibility #1: Are any of the other BNC connectors between each stage
isolated, or are they metal shelled types shorting through the metal
panel, giving you unexpected return currents?
Thought #2: The fact that the signal gets oscillations on it, and all
this gain peaking, when you connect the filter to the input of the amp
implies you are putting too much capacitance on the input to the
amplifier. Solutions are usually to either add say 10pF in the feedback
loop of the amp (assuming it's as simple as an inverting op amp); or
change to a non inverting amp, which is often better behaved; or just
pay some attention to the physical connection between these two stages
to reduce the capacitance (maybe the filter itself is too capacitive);
or add a unity gain buffer between them; or add a cascode stage in front
of the amp to isolate incoming capacitance, see here -http://www.electrooptical.net/www/frontends/frontends.pdf
Thought #3: 1.5MHz isn't very fast. Why use coax? Might be simpler to
use twisted pair, easier to manufacture than all those fiddly braid
stripping & connections.
I suspect I've pitched that explanation below your level, but I hope it
helps spark some ideas.
Nemo
In article
60109f90-2ec9-45cd-80cf-1b888d4cc... at (no spam) u16g2000pru.googlegroups.com>,
George Herold writes
A proto PCB arrived earlier this week. This is the 2nd generation
board for the High Level electronics to measure noise. This
instrument is for teaching physics students about noise and is
modular. The students must make connections between the modules to
make it all work. Something like this.
http://www.teachspin.com/instruments/signal_processor/index.shtml
It has two switch selectable SV filter sections (10, 30... 3k and
300, 1k,...100k Hz) A gain section, Gain of 10 to 10^4 with switch
selectable fine control. And a multiplier section. (The first
generation board had some conceptual errors that lead to a design with
gain peaking at the highest gain settings.)
I populated the gain section and it looked great. There was no gain
peaking and at the highest gain (10^4) the 3dB point was at about
1.5MHz. I should mention here that the output from the final opamp is
coupled to the output BNC connector with coax. With bare wire
capacitive coupling from the output to input leads to gain peaking.
The output BNC is also isolated from the front panel, with the return
currents flowing back along the outer braid of the coax, and grounded
back at the output opamp. If the output is not isolated there is
again gain peaking, due, I assume, to the return currents flowing back
along the shield and making the input and output connectors have
slightly different voltages. (The only ground connection between the
shield and pcb is made at the input to the gain stage.)
So far, so good. I then powered up the filter section and plugged it
into the gain stage. Crap!!! Gain peaking at the highest gains. And
if I run the High Pass output from the filter into the gain stage it
breaks into oscillations. (Gain peaking is at about 1.5MHz as you
might expect and the oscillation frequency is 1MHz.)
I've spent the past few days banging my head against this PCB.
The filters and gain section were made with a common ground plane.
(The multiplier section aready has a separate ground plane and power
filter.) I dremeled a slot between the planes and ran wires back to
the power supply ground input connection. (No real change.) I won't
bore you with the rest of the gory details, except to say that I did
learn that capacitive loading of the output made matters worse. This
was not capacitive loading of the final opamp. If I grounded the
input of the gain stage or drove it from some external generator there
was no gain peaking when I added more capacitance to the output. I was
adding 220 pF to the already existing 3 foot coax cable running to the
scope
Today I powered up the filter section from a pair of 9V batteries.
When I used the shield as the ground for the filter section every
thing was great. No gain peaking. When I used the power supply
ground from the gain stage as ground it was the same old oscillating
story.
So I then built this configuration
+-------+ |Shield
(V+)----R1R1---+---+ Vin | |
| | | |
C |filter | |
C | | |
+---+ GND---+------+
| | |
+-------+
Where R1 was 100 ohms and C was a parallel combination of 47uF tant,
1uF ceramic and 0,1uF ceramic.
This worked fine, but I'm not sure I like the return current from the
filter section flowing through the shield. (And then back through the
gain stage.)
So then I made this,
+-------+ |Shield
(V+)----R1R1---+---+ Vin | |
| | | |
C |filter | |
C | | |
PGND---LLL----+---+ GND---+-R2R2-+
| +-C2C2-+
+-------+ |
Where R1 and C are as before, PGND is the power supply ground, L is
100mH, R2 is 10 ohms and C2 is 1 uF.
This worked also. (Though I still have the high frequency currents
flowing through the shield.)
The values were picked from what was lying around, It was late Friday
and I decided to declare victory and go home.
There is a symmetrical R and C on the negative supply line.
If you've read this far, Thanks.
Is there something obvious I might have missed? Some way to make it
better?
I'm not sure which of the above configurations to use.... My fear is
that neither may end up working. I worry about sending the ground
currents from the filter section through the gain section.
Thanks again,
George H.- Hide quoted text -
- Show quoted text -
[/quote]
Thanks Nemo, This consumed some of my thoughts over the weekend and I
believe one thing I had been missing.. (ignoring) was proper isolation
of the power supplies for the two different circuits. I'd been F'ing
around with the ground connections, but had the power supplies were
connected together. (Except for the last circuit hack which I
poseted.) At 1 MHz the 0.1 uF caps I have sprinkled everywhere are a
nice low impedance path to ground... and if there's a path some of the
current is going to flow.
George H. |
|
|
| Back to top |
|
|
|
| George Herold... |
| Posted: Mon Nov 09, 2009 4:59 am |
|
|
|
Guest
|
On Nov 8, 7:10 pm, dagmargoodb... at (no spam) yahoo.com wrote:
[quote]On Nov 6, 9:00 pm, George Herold <ggher... at (no spam) gmail.com> wrote:
A proto PCB arrived earlier this week. This is the 2nd generation
board for the High Level electronics to measure noise. This
instrument is for teaching physics students about noise and is
modular. The students must make connections between the modules to
make it all work. Something like this.http://www.teachspin.com/instruments/signal_processor/index.shtml
It has two switch selectable SV filter sections (10, 30... 3k and
300, 1k,...100k Hz) A gain section, Gain of 10 to 10^4 with switch
selectable fine control. And a multiplier section. (The first
generation board had some conceptual errors that lead to a design with
gain peaking at the highest gain settings.)
I populated the gain section and it looked great. There was no gain
peaking and at the highest gain (10^4) the 3dB point was at about
1.5MHz. I should mention here that the output from the final opamp is
coupled to the output BNC connector with coax. With bare wire
capacitive coupling from the output to input leads to gain peaking.
The output BNC is also isolated from the front panel, with the return
currents flowing back along the outer braid of the coax, and grounded
back at the output opamp. If the output is not isolated there is
again gain peaking, due, I assume, to the return currents flowing back
along the shield and making the input and output connectors have
slightly different voltages. (The only ground connection between the
shield and pcb is made at the input to the gain stage.)
So far, so good. I then powered up the filter section and plugged it
into the gain stage. Crap!!! Gain peaking at the highest gains. And
if I run the High Pass output from the filter into the gain stage it
breaks into oscillations. (Gain peaking is at about 1.5MHz as you
might expect and the oscillation frequency is 1MHz.)
I've spent the past few days banging my head against this PCB.
The filters and gain section were made with a common ground plane.
(The multiplier section aready has a separate ground plane and power
filter.) I dremeled a slot between the planes and ran wires back to
the power supply ground input connection. (No real change.) I won't
bore you with the rest of the gory details, except to say that I did
learn that capacitive loading of the output made matters worse. This
was not capacitive loading of the final opamp. If I grounded the
input of the gain stage or drove it from some external generator there
was no gain peaking when I added more capacitance to the output. I was
adding 220 pF to the already existing 3 foot coax cable running to the
scope
Today I powered up the filter section from a pair of 9V batteries.
When I used the shield as the ground for the filter section every
thing was great. No gain peaking. When I used the power supply
ground from the gain stage as ground it was the same old oscillating
story.
So I then built this configuration
+-------+ |Shield
(V+)----R1R1---+---+ Vin | |
| | | |
C |filter | |
C | | |
+---+ GND---+------+
| | |
+-------+
Where R1 was 100 ohms and C was a parallel combination of 47uF tant,
1uF ceramic and 0,1uF ceramic.
This worked fine, but I'm not sure I like the return current from the
filter section flowing through the shield. (And then back through the
gain stage.)
So then I made this,
+-------+ |Shield
(V+)----R1R1---+---+ Vin | |
| | | |
C |filter | |
C | | |
PGND---LLL----+---+ GND---+-R2R2-+
| +-C2C2-+
+-------+ |
Where R1 and C are as before, PGND is the power supply ground, L is
100mH, R2 is 10 ohms and C2 is 1 uF.
This worked also. (Though I still have the high frequency currents
flowing through the shield.)
The values were picked from what was lying around, It was late Friday
and I decided to declare victory and go home.
There is a symmetrical R and C on the negative supply line.
If you've read this far, Thanks.
Is there something obvious I might have missed? Some way to make it
better?
I'm not sure which of the above configurations to use.... My fear is
that neither may end up working. I worry about sending the ground
currents from the filter section through the gain section.
Thanks again,
George H.
I'm with Nemo's Thought #2: gain-peaking suggests the op-amp's annoyed
by its capacitive loading. CL=220pf is about 500 ohms at (no spam) 1.5MHz. Even
lousy grounding's impedance at (no spam) 1.5MHz should be beat that by so much
that ground-loop feedback currents shouldn't matter. They might, but
they shouldn't.
Nemo's paralleled feedback cap is the cure, or classic op-amp
capacitive load isolation techniques, or a different, huskier op-amp.
Nice ASCII art.
--
Cheers,
James Arthur- Hide quoted text -
- Show quoted text -
[/quote]
James, Thanks for the response. The circuit will typically just
drive a piece of coax to a 'scope input. I notcied that adding the
coax made the gain peaking a bit worse. Working on the premiss that
if you can't make something better you can learn something by making
it worse, I tried adding more capacitance to the output. The amp
could drive the extra capacitance just fine when the input to the amp
stage was external... some signal generator. But crapped out when
driven by the 'internal' circuit. I think the added cap was causing
more current that was out of phase with the input. And this feeds
back to the interanl circuit.. The out of phase stuff is what it
'needs' to oscillate. (Loading down the output with a resistor did
nothing.)
George H. |
|
|
| Back to top |
|
|
|
| George Herold... |
| Posted: Mon Nov 09, 2009 5:05 am |
|
|
|
Guest
|
On Nov 8, 7:56 pm, John Larkin
<jjlar... at (no spam) highNOTlandTHIStechnologyPART.com> wrote:
[quote]On Fri, 6 Nov 2009 18:00:23 -0800 (PST), George Herold
ggher... at (no spam) gmail.com> wrote:
A proto PCB arrived earlier this week. This is the 2nd generation
board for the High Level electronics to measure noise. This
instrument is for teaching physics students about noise and is
modular. The students must make connections between the modules to
make it all work. Something like this.
http://www.teachspin.com/instruments/signal_processor/index.shtml
It has two switch selectable SV filter sections (10, 30... 3k and
300, 1k,...100k Hz) A gain section, Gain of 10 to 10^4 with switch
selectable fine control. And a multiplier section. (The first
generation board had some conceptual errors that lead to a design with
gain peaking at the highest gain settings.)
I populated the gain section and it looked great. There was no gain
peaking and at the highest gain (10^4) the 3dB point was at about
1.5MHz. I should mention here that the output from the final opamp is
coupled to the output BNC connector with coax. With bare wire
capacitive coupling from the output to input leads to gain peaking.
The output BNC is also isolated from the front panel, with the return
currents flowing back along the outer braid of the coax, and grounded
back at the output opamp. If the output is not isolated there is
again gain peaking, due, I assume, to the return currents flowing back
along the shield and making the input and output connectors have
slightly different voltages. (The only ground connection between the
shield and pcb is made at the input to the gain stage.)
So far, so good. I then powered up the filter section and plugged it
into the gain stage. Crap!!! Gain peaking at the highest gains. And
if I run the High Pass output from the filter into the gain stage it
breaks into oscillations. (Gain peaking is at about 1.5MHz as you
might expect and the oscillation frequency is 1MHz.)
I've spent the past few days banging my head against this PCB.
The filters and gain section were made with a common ground plane.
(The multiplier section aready has a separate ground plane and power
filter.) I dremeled a slot between the planes and ran wires back to
the power supply ground input connection. (No real change.) I won't
bore you with the rest of the gory details, except to say that I did
learn that capacitive loading of the output made matters worse. This
was not capacitive loading of the final opamp. If I grounded the
input of the gain stage or drove it from some external generator there
was no gain peaking when I added more capacitance to the output. I was
adding 220 pF to the already existing 3 foot coax cable running to the
scope
Today I powered up the filter section from a pair of 9V batteries.
When I used the shield as the ground for the filter section every
thing was great. No gain peaking. When I used the power supply
ground from the gain stage as ground it was the same old oscillating
story.
So I then built this configuration
+-------+ |Shield
(V+)----R1R1---+---+ Vin | |
| | | |
C |filter | |
C | | |
+---+ GND---+------+
| | |
+-------+
Where R1 was 100 ohms and C was a parallel combination of 47uF tant,
1uF ceramic and 0,1uF ceramic.
This worked fine, but I'm not sure I like the return current from the
filter section flowing through the shield. (And then back through the
gain stage.)
So then I made this,
+-------+ |Shield
(V+)----R1R1---+---+ Vin | |
| | | |
C |filter | |
C | | |
PGND---LLL----+---+ GND---+-R2R2-+
| +-C2C2-+
+-------+ |
Where R1 and C are as before, PGND is the power supply ground, L is
100mH, R2 is 10 ohms and C2 is 1 uF.
This worked also. (Though I still have the high frequency currents
flowing through the shield.)
The values were picked from what was lying around, It was late Friday
and I decided to declare victory and go home.
There is a symmetrical R and C on the negative supply line.
If you've read this far, Thanks.
Is there something obvious I might have missed? Some way to make it
better?
I'm not sure which of the above configurations to use.... My fear is
that neither may end up working. I worry about sending the ground
currents from the filter section through the gain section.
Thanks again,
George H.
Life is much simpler if you ground everything... the PCB ground plane,
all the connector shells, the power supply common, everything, to the
metal chassis. Your single input-related ground may be flailing as a
result of output current.
And I'd recommend not connecting an opamp output to an output
connector; you never know what the load may be.
John- Hide quoted text -
- Show quoted text -
[/quote]
Thanks John, I was reading the 'useful hints' on Joerg's website
http://www.analogconsultants.com/hints.htm
and he suggests the exact same thing. (attaching sheild to ground
plane at as many points a possible.) I worry about ground loops and
magnetic pickup.. But I'll give it a try.
George H. |
|
|
| Back to top |
|
|
|
| George Herold... |
| Posted: Mon Nov 09, 2009 5:24 am |
|
|
|
Guest
|
On Nov 8, 10:26 pm, dagmargoodb... at (no spam) yahoo.com wrote:
[quote]On Nov 8, 7:56 pm, John Larkin
jjlar... at (no spam) highNOTlandTHIStechnologyPART.com> wrote:
And I'd recommend not connecting an opamp output to an output
connector; you never know what the load may be.
John
Yep. At a minimum isolate the feedback node from the load:
.-----R2-------------.
| |
+-----C1-----. |
| | |
| |\ | |
| | \ | |
Vin >---R1--+---|- \ | |
| >----+---R3--+----> Vout
.---|+ /
| | /
| |/
==> GND
I've done that. Or use a buffer.
Which op-amp are we talking about here George?
--
Cheers,
James Arthur
[/quote]
Yeah I could try that too. (I did add a bit of series resistance to
the output, but didn't include it in the feed back path as you've
shown.
"> Which op-amp are we talking about here George?"
The gain stage starts with two OPA228's each inverting with a gain of
10, 10kOhm feed back resistors with 3.3pF of parallel C, gain 3dB
point is about 3MHz. Next is a OPA228 again inverting with Gain of 5.
3dB = 4MHz There is then a resistor divider chain that sets the fine
gain control... "gain" from 1 to 0.1. This drives a dual OPA2134
opamp. The first stage is non-inverting with a gain of 4 and the
second opamp is inverting with a gain of 5. The output drives a
multiplier stage that has the highest accuracy when the peak voltage
is near 10 volts, The OPA2134's give me a 'full power' bandwidth of
something near 900kHz. (I haven't measured this for the latest
configuration yet.) The small signal 3dB point of the two opa2134's is
2.1MHz. I can throw away a bit of this bandwidth if that will help
make the whole circuit more stable.
George H. |
|
|
| Back to top |
|
|
|
| ... |
| Posted: Mon Nov 09, 2009 7:34 am |
|
|
|
Guest
|
On Nov 9, 10:24 am, George Herold <ggher... at (no spam) gmail.com> wrote:
[quote]On Nov 8, 10:26 pm, dagmargoodb... at (no spam) yahoo.com wrote:
On Nov 8, 7:56 pm, John Larkin
jjlar... at (no spam) highNOTlandTHIStechnologyPART.com> wrote:
And I'd recommend not connecting an opamp output to an output
connector; you never know what the load may be.
John
Yep. At a minimum isolate the feedback node from the load:
.-----R2-------------.
| |
+-----C1-----. |
| | |
| |\ | |
| | \ | |
Vin >---R1--+---|- \ | |
| >----+---R3--+----> Vout
.---|+ /
| | /
| |/
===
GND
I've done that. Or use a buffer.
Which op-amp are we talking about here George?
--
Cheers,
James Arthur
Yeah I could try that too. (I did add a bit of series resistance to
the output, but didn't include it in the feed back path as you've
shown.
[/quote]
You mean taking the feedback from after the resistor? That's
destabilizing.
The problem is this: if the op-amp has a high output impedance,
capacitive
loads introduce a phase delay. Adding series resistance to the output
makes
that worse. The classic dodge I sketched eliminates that delay.
[quote]
"> Which op-amp are we talking about here George?"
The gain stage starts with two OPA228's each inverting with a gain of
10, 10kOhm feed back resistors with 3.3pF of parallel C, gain 3dB
point is about 3MHz. Next is a OPA228 again inverting with Gain of 5.
3dB = 4MHz There is then a resistor divider chain that sets the fine
gain control... "gain" from 1 to 0.1. This drives a dual OPA2134
opamp. The first stage is non-inverting with a gain of 4 and the
second opamp is inverting with a gain of 5. The output drives a
multiplier stage that has the highest accuracy when the peak voltage
is near 10 volts, The OPA2134's give me a 'full power' bandwidth of
something near 900kHz. (I haven't measured this for the latest
configuration yet.) The small signal 3dB point of the two opa2134's is
2.1MHz. I can throw away a bit of this bandwidth if that will help
make the whole circuit more stable.
George H.
[/quote]
Can't look just now--gotta go--but I'll peek back later.
--
Cheers,
James Arthur |
|
|
| Back to top |
|
|
|
| George Herold... |
| Posted: Mon Nov 09, 2009 7:57 am |
|
|
|
Guest
|
On Nov 9, 12:34 pm, dagmargoodb... at (no spam) yahoo.com wrote:
[quote]On Nov 9, 10:24 am, George Herold <ggher... at (no spam) gmail.com> wrote:
On Nov 8, 10:26 pm, dagmargoodb... at (no spam) yahoo.com wrote:
On Nov 8, 7:56 pm, John Larkin
jjlar... at (no spam) highNOTlandTHIStechnologyPART.com> wrote:
And I'd recommend not connecting an opamp output to an output
connector; you never know what the load may be.
John
Yep. At a minimum isolate the feedback node from the load:
.-----R2-------------.
| |
+-----C1-----. |
| | |
| |\ | |
| | \ | |
Vin >---R1--+---|- \ | |
| >----+---R3--+----> Vout
.---|+ /
| | /
| |/
==> > > GND
I've done that. Or use a buffer.
Which op-amp are we talking about here George?
--
Cheers,
James Arthur
Yeah I could try that too. (I did add a bit of series resistance to
the output, but didn't include it in the feed back path as you've
shown.
You mean taking the feedback from after the resistor? That's
destabilizing.
The problem is this: if the op-amp has a high output impedance,
capacitive
loads introduce a phase delay. Adding series resistance to the output
makes
that worse. The classic dodge I sketched eliminates that delay.
"> Which op-amp are we talking about here George?"
The gain stage starts with two OPA228's each inverting with a gain of
10, 10kOhm feed back resistors with 3.3pF of parallel C, gain 3dB
point is about 3MHz. Next is a OPA228 again inverting with Gain of 5..
3dB = 4MHz There is then a resistor divider chain that sets the fine
gain control... "gain" from 1 to 0.1. This drives a dual OPA2134
opamp. The first stage is non-inverting with a gain of 4 and the
second opamp is inverting with a gain of 5. The output drives a
multiplier stage that has the highest accuracy when the peak voltage
is near 10 volts, The OPA2134's give me a 'full power' bandwidth of
something near 900kHz. (I haven't measured this for the latest
configuration yet.) The small signal 3dB point of the two opa2134's is
2.1MHz. I can throw away a bit of this bandwidth if that will help
make the whole circuit more stable.
George H.
Can't look just now--gotta go--but I'll peek back later.
--
Cheers,
James Arthur- Hide quoted text -
- Show quoted text -- Hide quoted text -
- Show quoted text -
[/quote]
"> You mean taking the feedback from after the resistor? That's
[quote]destabilizing."
[/quote]
I don't think so? I just did this,
[quote]
+-----R2-----.
| |
| |\ |
| | \ |
Vin >---R1--+---|- \ |
| >----+---R3--+----> Vout
.---|+ / |
| | / C load
| |/ |
=== ==> > > GND GND
[/quote]
Which I thought was an 'easy' solution. But I'm going to have to
review this.
(I made R3 1k ohm.)
George H. |
|
|
| Back to top |
|
|
|
| George Herold... |
| Posted: Mon Nov 09, 2009 9:01 am |
|
|
|
Guest
|
On Nov 9, 9:12 am, George Herold <ggher... at (no spam) gmail.com> wrote:
[quote]On Nov 8, 6:32 pm, Nemo <z... at (no spam) nospam.nospam.nospam.nospam.co.uk
wrote:
Possibility #1: Are any of the other BNC connectors between each stage
isolated, or are they metal shelled types shorting through the metal
panel, giving you unexpected return currents?
Thought #2: The fact that the signal gets oscillations on it, and all
this gain peaking, when you connect the filter to the input of the amp
implies you are putting too much capacitance on the input to the
amplifier. Solutions are usually to either add say 10pF in the feedback
loop of the amp (assuming it's as simple as an inverting op amp); or
change to a non inverting amp, which is often better behaved; or just
pay some attention to the physical connection between these two stages
to reduce the capacitance (maybe the filter itself is too capacitive);
or add a unity gain buffer between them; or add a cascode stage in front
of the amp to isolate incoming capacitance, see here -http://www.electrooptical.net/www/frontends/frontends.pdf
Thought #3: 1.5MHz isn't very fast. Why use coax? Might be simpler to
use twisted pair, easier to manufacture than all those fiddly braid
stripping & connections.
I suspect I've pitched that explanation below your level, but I hope it
helps spark some ideas.
Nemo
In article
60109f90-2ec9-45cd-80cf-1b888d4cc... at (no spam) u16g2000pru.googlegroups.com>,
George Herold writes
A proto PCB arrived earlier this week. This is the 2nd generation
board for the High Level electronics to measure noise. This
instrument is for teaching physics students about noise and is
modular. The students must make connections between the modules to
make it all work. Something like this.
http://www.teachspin.com/instruments/signal_processor/index.shtml
It has two switch selectable SV filter sections (10, 30... 3k and
300, 1k,...100k Hz) A gain section, Gain of 10 to 10^4 with switch
selectable fine control. And a multiplier section. (The first
generation board had some conceptual errors that lead to a design with
gain peaking at the highest gain settings.)
I populated the gain section and it looked great. There was no gain
peaking and at the highest gain (10^4) the 3dB point was at about
1.5MHz. I should mention here that the output from the final opamp is
coupled to the output BNC connector with coax. With bare wire
capacitive coupling from the output to input leads to gain peaking.
The output BNC is also isolated from the front panel, with the return
currents flowing back along the outer braid of the coax, and grounded
back at the output opamp. If the output is not isolated there is
again gain peaking, due, I assume, to the return currents flowing back
along the shield and making the input and output connectors have
slightly different voltages. (The only ground connection between the
shield and pcb is made at the input to the gain stage.)
So far, so good. I then powered up the filter section and plugged it
into the gain stage. Crap!!! Gain peaking at the highest gains. And
if I run the High Pass output from the filter into the gain stage it
breaks into oscillations. (Gain peaking is at about 1.5MHz as you
might expect and the oscillation frequency is 1MHz.)
I've spent the past few days banging my head against this PCB.
The filters and gain section were made with a common ground plane.
(The multiplier section aready has a separate ground plane and power
filter.) I dremeled a slot between the planes and ran wires back to
the power supply ground input connection. (No real change.) I won't
bore you with the rest of the gory details, except to say that I did
learn that capacitive loading of the output made matters worse. This
was not capacitive loading of the final opamp. If I grounded the
input of the gain stage or drove it from some external generator there
was no gain peaking when I added more capacitance to the output. I was
adding 220 pF to the already existing 3 foot coax cable running to the
scope
Today I powered up the filter section from a pair of 9V batteries.
When I used the shield as the ground for the filter section every
thing was great. No gain peaking. When I used the power supply
ground from the gain stage as ground it was the same old oscillating
story.
So I then built this configuration
+-------+ |Shield
(V+)----R1R1---+---+ Vin | |
| | | |
C |filter | |
C | | |
+---+ GND---+------+
| | |
+-------+
Where R1 was 100 ohms and C was a parallel combination of 47uF tant,
1uF ceramic and 0,1uF ceramic.
This worked fine, but I'm not sure I like the return current from the
filter section flowing through the shield. (And then back through the
gain stage.)
So then I made this,
+-------+ |Shield
(V+)----R1R1---+---+ Vin | |
| | | |
C |filter | |
C | | |
PGND---LLL----+---+ GND---+-R2R2-+
| +-C2C2-+
+-------+ |
Where R1 and C are as before, PGND is the power supply ground, L is
100mH, R2 is 10 ohms and C2 is 1 uF.
This worked also. (Though I still have the high frequency currents
flowing through the shield.)
The values were picked from what was lying around, It was late Friday
and I decided to declare victory and go home.
There is a symmetrical R and C on the negative supply line.
If you've read this far, Thanks.
Is there something obvious I might have missed? Some way to make it
better?
I'm not sure which of the above configurations to use.... My fear is
that neither may end up working. I worry about sending the ground
currents from the filter section through the gain section.
Thanks again,
George H.- Hide quoted text -
- Show quoted text -
Thanks Nemo, This consumed some of my thoughts over the weekend and I
believe one thing I had been missing.. (ignoring) was proper isolation
of the power supplies for the two different circuits. I'd been F'ing
around with the ground connections, but had the power supplies were
connected together. (Except for the last circuit hack which I
poseted.) At 1 MHz the 0.1 uF caps I have sprinkled everywhere are a
nice low impedance path to ground... and if there's a path some of the
current is going to flow.
George H.- Hide quoted text -
- Show quoted text -
[/quote]
Success! Hi all, Well following the hints from Joerg's website and
John L.'s suggestion I reinstalled the common ground plane between
both havles of the circuit. I added some power line filtering and
powered it up. No Gain peaking! The gain is nice and flat right out
to 1.6MHz...I will have to look for any ground loop issues, but at the
moment it looks like I've found a new circuit approach.
Time for Lunch,
George |
|
|
| Back to top |
|
|
|
| ... |
| Posted: Mon Nov 09, 2009 9:37 am |
|
|
|
Guest
|
On Nov 9, 12:57 pm, George Herold <ggher... at (no spam) gmail.com> wrote:
[quote]On Nov 9, 12:34 pm, dagmargoodb... at (no spam) yahoo.com wrote:
On Nov 9, 10:24 am, George Herold <ggher... at (no spam) gmail.com> wrote:
On Nov 8, 10:26 pm, dagmargoodb... at (no spam) yahoo.com wrote:
On Nov 8, 7:56 pm, John Larkin
jjlar... at (no spam) highNOTlandTHIStechnologyPART.com> wrote:
And I'd recommend not connecting an opamp output to an output
connector; you never know what the load may be.
John
Yep. At a minimum isolate the feedback node from the load:
.-----R2-------------.
| |
+-----C1-----. |
| | |
| |\ | |
| | \ | |
Vin >---R1--+---|- \ | |
| >----+---R3--+----> Vout
.---|+ /
| | /
| |/
===
GND
I've done that. Or use a buffer.
Which op-amp are we talking about here George?
--
Cheers,
James Arthur
Yeah I could try that too. (I did add a bit of series resistance to
the output, but didn't include it in the feed back path as you've
shown.
You mean taking the feedback from after the resistor? That's
destabilizing.
The problem is this: if the op-amp has a high output impedance,
capacitive
loads introduce a phase delay. Adding series resistance to the output
makes
that worse. The classic dodge I sketched eliminates that delay.
"> Which op-amp are we talking about here George?"
The gain stage starts with two OPA228's each inverting with a gain of
10, 10kOhm feed back resistors with 3.3pF of parallel C, gain 3dB
point is about 3MHz. Next is a OPA228 again inverting with Gain of 5.
3dB = 4MHz There is then a resistor divider chain that sets the fine
gain control... "gain" from 1 to 0.1. This drives a dual OPA2134
opamp. The first stage is non-inverting with a gain of 4 and the
second opamp is inverting with a gain of 5. The output drives a
multiplier stage that has the highest accuracy when the peak voltage
is near 10 volts, The OPA2134's give me a 'full power' bandwidth of
something near 900kHz. (I haven't measured this for the latest
configuration yet.) The small signal 3dB point of the two opa2134's is
2.1MHz. I can throw away a bit of this bandwidth if that will help
make the whole circuit more stable.
George H.
Can't look just now--gotta go--but I'll peek back later.
--
Cheers,
James Arthur- Hide quoted text -
- Show quoted text -- Hide quoted text -
- Show quoted text -
"> You mean taking the feedback from after the resistor? That's
destabilizing."
I don't think so? I just did this,
+-----R2-----.
| |
| |\ |
| | \ |
Vin >---R1--+---|- \ |
| >----+---R3--+----> Vout
.---|+ / |
| | / C load
| |/ |
=== ===
GND GND
Which I thought was an 'easy' solution. But I'm going to have to
review this.
(I made R3 1k ohm.)
George H.
[/quote]
That's fine. I thought you meant this:
.-------R2------.
| |
| |\ |
| | \ |
Vin >---R1--+---|- \ |
| >---R3--+----> Vout
.---|+ / |
| | / C load
| |/ |
=== ===
GND GND
which exacerbates the influence of capacitive loading.
--
Cheers,
James Arthur |
|
|
| Back to top |
|
|
|
|
|
All times are GMT - 5 Hours
The time now is Wed Dec 09, 2009 3:51 am
|
|