Voltage Gain Switch Troubleshooting

[[email protected]]

***MANUAL QUOTE***

.. Gain switching the feedback dividers eliminates effects
.. __ from the CMOS switch's series resistance.
.. ------|+ \
.. | >-----+----+----+---
.. .--|-_/ R1 | |
.. | | R3 | non-inverting gain
.. '---o-->o----+ | R5
.. o--- | ---+ |
.. o--- | -- | ---+
.. | | |
.. R2 R4 R6
.. | | |
.. GND --+----+----'

And if Gmin = 2, i.e., R1 => R2, etc., then a +/-5V switch
can be used with +/-10V output signals.


--
Thanks,
- Win

***END QUOTE***

Hi, if you recall from an old thread I had to design a voltage gain
switch using solid state parts. I finally made the circuit, and chose
the above configuration for the switch which is a 4051BP(stage 2). The
first stage is through a discrete in-amp(3 op-amp config, 10k
resistors, 200ohm gain select resistor) with a gain of 100. Stage 2 has
gains of 1, 5, 100 and 500, and Stage 3 is a follower. All Op-Amps are
OP-07.

Before assembly I first tested the ckt in a protoboard with DC as input
signal. It worked fine. When I finished the PCB I tested it and the
output won`t swing past +-6V. The DC power supply is +-12V (OP-07s),
and the 4051BP is powered with 5V1 zenners from positive and negative
supply respectibly. Supply power to ICs is checked. Output at Stage 2
Input is aprox. +-10V. Output without 4051BP and manual feedback (ie
short the pins) is +-10V. Always DC input signal. I measured with a DVM
the drop across the switch when any gain is selected and it equals the
output voltage (+-6V). When the output doesn`t saturate the circuit
works fine.

I don`t know what's the deal with this, I`m sure I did it on protoboard
and it worked fine. Tomorrow I`m gonna assemble it again and see what
happens, but I`m posting here because I assume the problem is something
else. I`m not sure why the output saturates at 6V, if anything I would
expect 5V.

Thanks in advance.

 

[Genome]

***MANUAL QUOTE***

. Gain switching the feedback dividers eliminates effects
. __ from the CMOS switch's series resistance.
. ------|+ \
. | >-----+----+----+---
. .--|-_/ R1 | |
. | | R3 | non-inverting gain
. '---o-->o----+ | R5
. o--- | ---+ |
. o--- | -- | ---+
. | | |
. R2 R4 R6
. | | |
. GND --+----+----'

And if Gmin = 2, i.e., R1 => R2, etc., then a +/-5V switch
can be used with +/-10V output signals.


--
Thanks,
- Win

***END QUOTE***

Hi, if you recall from an old thread I had to design a voltage gain
switch using solid state parts. I finally made the circuit, and chose
the above configuration for the switch which is a 4051BP(stage 2). The
first stage is through a discrete in-amp(3 op-amp config, 10k
resistors, 200ohm gain select resistor) with a gain of 100. Stage 2 has
gains of 1, 5, 100 and 500, and Stage 3 is a follower. All Op-Amps are
OP-07.

Before assembly I first tested the ckt in a protoboard with DC as input
signal. It worked fine. When I finished the PCB I tested it and the
output won`t swing past +-6V. The DC power supply is +-12V (OP-07s),
and the 4051BP is powered with 5V1 zenners from positive and negative
supply respectibly. Supply power to ICs is checked. Output at Stage 2
Input is aprox. +-10V. Output without 4051BP and manual feedback (ie
short the pins) is +-10V. Always DC input signal. I measured with a DVM
the drop across the switch when any gain is selected and it equals the
output voltage (+-6V). When the output doesn`t saturate the circuit
works fine.

I don`t know what's the deal with this, I`m sure I did it on protoboard
and it worked fine. Tomorrow I`m gonna assemble it again and see what
happens, but I`m posting here because I assume the problem is something
else. I`m not sure why the output saturates at 6V, if anything I would
expect 5V.

Thanks in advance.

Hi yourself.

If you recall Win Hill died about two months ago but I suppose that since my
newsreader shows 57034 messages and I don't know who the **** you are it is
quite likely that, in as much as I don't know who the **** you are, you
missed the obituary as well.

Having said that I notice that your description of the circuit you are
implementing seems to differ from the one that the now Defunct Win Hill
suggested in the first place.

Although he may now be a dead bloke there is a good chance that you have
fucked something up and I really don't think dead people are sort of
bothered about sorting out your **** ups.

'Thanks in advance' doesn't really work now, does it?

DNA

 

[[email protected]]

Genome ha escrito:

Hi yourself.

If you recall Win Hill died about two months ago but I suppose that since my
newsreader shows 57034 messages and I don't know who the **** you are it is
quite likely that, in as much as I don't know who the **** you are, you
missed the obituary as well.

Having said that I notice that your description of the circuit you are
implementing seems to differ from the one that the now Defunct Win Hill
suggested in the first place.

Although he may now be a dead bloke there is a good chance that you have
fucked something up and I really don't think dead people are sort of
bothered about sorting out your **** ups.

'Thanks in advance' doesn't really work now, does it?

DNA

I`m sorry to hear that. I didn`t have a clue, I try to check the group
once in a while, but I missed that. There's thousands of posts as you
pointed out.

I assure you that the fact that I quoted Win is because his schematic
was convenient for illustrating my circuit, nothing more.

I`ve never met Win personally, but he's always been very kind and
helpful. I shall remember him through what I`ve read from him, and
that's no small thing.

Again I`m sorry, but I guess that like "Thanks in Advance" this doesn`t
work.

 

[Robert Latest]

On 19 Dec 2006 15:27:53 -0800,

Genome ha escrito:

If you recall Win Hill died about two months ago but I suppose that since my
newsreader shows 57034 messages and I don't know who the **** you are it is
quite likely that, in as much as I don't know who the **** you are, you
missed the obituary as well.

I`m sorry to hear that. I didn`t have a clue, I try to check the group
once in a while, but I missed that. There's thousands of posts as you
pointed out.
[...]

I`ve never met Win personally, but he's always been very kind and
helpful. I shall remember him through what I`ve read from him, and
that's no small thing.

Don't let Genome pull your leg *too* hard. Win Hill is alive and kicking
unless it was an impostor who posted something only yesterday on this
newgroup, imitating his writing style.

But, like so often, it is a Genome post that can point your thinking in
the right direction. In this case it's that Win isn't getting younger, and
we're still holding our breath for AoE 3rd ed.

robert

 

[[email protected]]

Robert Latest ha escrito:

On 19 Dec 2006 15:27:53 -0800,

Genome ha escrito:

If you recall Win Hill died about two months ago but I suppose that since my
newsreader shows 57034 messages and I don't know who the **** you are it is
quite likely that, in as much as I don't know who the **** you are, you
missed the obituary as well.

I`m sorry to hear that. I didn`t have a clue, I try to check the group
once in a while, but I missed that. There's thousands of posts as you
pointed out.
[...]

I`ve never met Win personally, but he's always been very kind and
helpful. I shall remember him through what I`ve read from him, and
that's no small thing.

Don't let Genome pull your leg *too* hard. Win Hill is alive and kicking
unless it was an impostor who posted something only yesterday on this
newgroup, imitating his writing style.

But, like so often, it is a Genome post that can point your thinking in
the right direction. In this case it's that Win isn't getting younger, and
we're still holding our breath for AoE 3rd ed.

robert

Well, I guess that explains why I couldn`t find any Winfiel Hill
obituary post. Genome got me 100% convinced, I`ll leave my eyes open
from now on.



Back on topic. Looking back at Win's original post he states that
minimum gain is 2, and I used a gain of 1 without feedback resistor.
Also the stage 2 gains that I stated were 100 and 500 are in fact 10
and 50, sorry for that.

I re-assembled the ckt in a protoboard and found out that gain of 1 is
a no go, but if I put a feedback resistor of 100k the circuit works
normally. Also if I use other gains like 2, 5 or 10 they work, but I`ve
noticed that as the gain-setting resistors become smaller the output
swing also becomes smaller. The PCB ckt remains unchanged, could it be
that the gain of 1 path is affecting the other path's output swing? I
mean they should be isolated and independent, but that's my only guess.
I don`t wanna cut the track and later find out that it didn`t work.

Can someone explain why the gain of 1 config. doesn`t work with the
4051 on the feeback path ?

 

[Fred Bartoli]

[email protected] a écrit :

Robert Latest ha escrito:

On 19 Dec 2006 15:27:53 -0800,

Genome ha escrito:
If you recall Win Hill died about two months ago but I suppose that since my
newsreader shows 57034 messages and I don't know who the **** you are it is
quite likely that, in as much as I don't know who the **** you are, you
missed the obituary as well.
I`m sorry to hear that. I didn`t have a clue, I try to check the group
once in a while, but I missed that. There's thousands of posts as you
pointed out.
[...]
I`ve never met Win personally, but he's always been very kind and
helpful. I shall remember him through what I`ve read from him, and
that's no small thing.

Don't let Genome pull your leg *too* hard. Win Hill is alive and kicking
unless it was an impostor who posted something only yesterday on this
newgroup, imitating his writing style.

But, like so often, it is a Genome post that can point your thinking in
the right direction. In this case it's that Win isn't getting younger, and
we're still holding our breath for AoE 3rd ed.

robert

Well, I guess that explains why I couldn`t find any Winfiel Hill
obituary post. Genome got me 100% convinced, I`ll leave my eyes open
from now on.



Back on topic. Looking back at Win's original post he states that
minimum gain is 2, and I used a gain of 1 without feedback resistor.
Also the stage 2 gains that I stated were 100 and 500 are in fact 10
and 50, sorry for that.

I re-assembled the ckt in a protoboard and found out that gain of 1 is
a no go, but if I put a feedback resistor of 100k the circuit works
normally. Also if I use other gains like 2, 5 or 10 they work, but I`ve
noticed that as the gain-setting resistors become smaller the output
swing also becomes smaller. The PCB ckt remains unchanged, could it be
that the gain of 1 path is affecting the other path's output swing? I
mean they should be isolated and independent, but that's my only guess.
I don`t wanna cut the track and later find out that it didn`t work.

Can someone explain why the gain of 1 config. doesn`t work with the
4051 on the feeback path ?

because the 4051 will clamp the opamp output signal to +/-(5V+Vd) with
its internal... clamp diodes.

 

[[email protected]]

Fred Bartoli ha escrito:

[email protected] a écrit :

Robert Latest ha escrito:

On 19 Dec 2006 15:27:53 -0800,
in Msg. <[email protected]>
Genome ha escrito:
If you recall Win Hill died about two months ago but I suppose that since my
newsreader shows 57034 messages and I don't know who the **** you are it is
quite likely that, in as much as I don't know who the **** you are, you
missed the obituary as well.
I`m sorry to hear that. I didn`t have a clue, I try to check the group
once in a while, but I missed that. There's thousands of posts as you
pointed out.

[...]
I`ve never met Win personally, but he's always been very kind and
helpful. I shall remember him through what I`ve read from him, and
that's no small thing.
Don't let Genome pull your leg *too* hard. Win Hill is alive and kicking
unless it was an impostor who posted something only yesterday on this
newgroup, imitating his writing style.

But, like so often, it is a Genome post that can point your thinking in
the right direction. In this case it's that Win isn't getting younger,and
we're still holding our breath for AoE 3rd ed.

robert

Well, I guess that explains why I couldn`t find any Winfiel Hill
obituary post. Genome got me 100% convinced, I`ll leave my eyes open
from now on.



Back on topic. Looking back at Win's original post he states that
minimum gain is 2, and I used a gain of 1 without feedback resistor.
Also the stage 2 gains that I stated were 100 and 500 are in fact 10
and 50, sorry for that.

I re-assembled the ckt in a protoboard and found out that gain of 1 is
a no go, but if I put a feedback resistor of 100k the circuit works
normally. Also if I use other gains like 2, 5 or 10 they work, but I`ve
noticed that as the gain-setting resistors become smaller the output
swing also becomes smaller. The PCB ckt remains unchanged, could it be
that the gain of 1 path is affecting the other path's output swing? I
mean they should be isolated and independent, but that's my only guess.
I don`t wanna cut the track and later find out that it didn`t work.

Can someone explain why the gain of 1 config. doesn`t work with the
4051 on the feeback path ?

because the 4051 will clamp the opamp output signal to +/-(5V+Vd) with
its internal... clamp diodes.

Ok thanks, I think that's it. The unity-gain path is clamping the
output regardless of the feedback path selected. I`ll cut the track and
add a resistor. Problem solved

 

[Winfield Hill]

[email protected] wrote...

Fred Bartoli ha escrito:

Ok thanks, I think that's it. The unity-gain path is clamping
the output regardless of the feedback path selected. I`ll cut
the track and add a resistor. Problem solved

I doubt that'll solve the problem. Normally we have to use
high-voltage cmos switches powered from the supply rails for
general purpose full-swing circuits, but my post suggested a
hack allowing the use of cheaper low-voltage logic-family
switches, provided your gain was at least 2x. Viz,

"Gain switching the feedback dividers eliminates effects
.. __ from the CMOS switch's series resistance.
.. ------|+ \
.. | >-----+----+----+---
.. .--|-_/ R1 | |
.. | | R3 | non-inverting gain
.. '---o-->o----+ | R5
.. o--- | ---+ |
.. o--- | -- | ---+
.. | | |
.. R2 R4 R6
.. | | |
.. GND --+----+----'

And if Gmin = 2, i.e., R1 => R2, etc., then a +/-5V switch
can be used with +/-10V output signals."

I don't see how you can cut a trace and add a resistor and
thereby magically get around this principle for G = 1.

But there is a trick you can use to get full output swing
with half-swing capable switches such as the hc4041 family:
first attenuate the signal by 2x, and follow with a gain of
2x for the G=1 overall-gain position, etc., and so on.

 

[[email protected]]

Winfield Hill ha escrito:

[email protected] wrote...

I doubt that'll solve the problem. Normally we have to use
high-voltage cmos switches powered from the supply rails for
general purpose full-swing circuits, but my post suggested a
hack allowing the use of cheaper low-voltage logic-family
switches, provided your gain was at least 2x. Viz,

"Gain switching the feedback dividers eliminates effects
. __ from the CMOS switch's series resistance.
. ------|+ \
. | >-----+----+----+---
. .--|-_/ R1 | |
. | | R3 | non-inverting gain
. '---o-->o----+ | R5
. o--- | ---+ |
. o--- | -- | ---+
. | | |
. R2 R4 R6
. | | |
. GND --+----+----'

And if Gmin = 2, i.e., R1 => R2, etc., then a +/-5V switch
can be used with +/-10V output signals."

I don't see how you can cut a trace and add a resistor and
thereby magically get around this principle for G = 1.

But there is a trick you can use to get full output swing
with half-swing capable switches such as the hc4041 family:
first attenuate the signal by 2x, and follow with a gain of
2x for the G=1 overall-gain position, etc., and so on.

I have found different results, I tested the following circuit:


From Stage 1
|\ To voltage follower
o--------|+\
| >--------------o-----o-o
.-----|-/ | |
| |/ | |
| | .-.
| | | |
| 100k | | | 100k
| ___ | '-'
| o-----|___|---' |
'-----o--__ |
o-------------------o
|
TC4051BP |
.-.
| | 100k
| |
'-'
|
|
===
GND
(created by AACircuit v1.28.6 beta 04/19/05 www.tech-chat.de)

The output is able to swing fully with this configuration, regardless
of the switch's current position. Are there any problems that I`m not
accounting for ?

I wouldn`t have had any problems had I not overlooked the Gmin=2. If
this workaround doesn`t work I`ll have to redesign the whole thing.

 

[ChrisQuayle]

I have found different results, I tested the following circuit:


From Stage 1
|\ To voltage follower
o--------|+\
| >--------------o-----o-o
.-----|-/ | |
| |/ | |
| | .-.
| | | |
| 100k | | | 100k
| ___ | '-'
| o-----|___|---' |
'-----o--__ |
o-------------------o
|
TC4051BP |
.-.
| | 100k
| |
'-'
|
|
===
GND
(created by AACircuit v1.28.6 beta 04/19/05 www.tech-chat.de)

The output is able to swing fully with this configuration, regardless
of the switch's current position. Are there any problems that I`m not
accounting for ?

I wouldn`t have had any problems had I not overlooked the Gmin=2. If
this workaround doesn`t work I`ll have to redesign the whole thing.

In the top position, ie, gain = 1, have you driven the input with a sine
or ramp and used a scope to check that the output is gain = 1 and that
it's following the input and not just running out of feedback loop once
the output reaches the switch supply voltage ?...

Chris

 

[[email protected]]

ChrisQuayle ha escrito:

In the top position, ie, gain = 1, have you driven the input with a sine
or ramp and used a scope to check that the output is gain = 1 and that
it's following the input and not just running out of feedback loop once
the output reaches the switch supply voltage ?...

Chris

You are right, I tried the circuit in Pspice and it does the strangest
thing. This made me realise that the node of the inv input has to be
able to allow 10v for the loop to work correctly, and that's unbearable
for the switch.

So I guess I`ll reduce the first stage gain to 50, then followed by
second stage gains of 2, 10, 20 and 100, to produce 4 different gains
of 100, 500, 1000 and 5000.

That PCB's going to waste, back at the start now. Can I call it problem
solved now ?

 

[ChrisQuayle]

You are right, I tried the circuit in Pspice and it does the strangest
thing. This made me realise that the node of the inv input has to be
able to allow 10v for the loop to work correctly, and that's unbearable
for the switch.

So I guess I`ll reduce the first stage gain to 50, then followed by
second stage gains of 2, 10, 20 and 100, to produce 4 different gains
of 100, 500, 1000 and 5000.

That PCB's going to waste, back at the start now. Can I call it problem
solved now ?

It's still not an optimum design, since the circuit should be designed
so that device ratings are not exceeded under *any* conditions, not just
those expected in normal operation. While the switch may withstand being
over driven, it still leaves a corner for failure which is not covered
by the device data sheet.

Two possible approaches:

1) Put in a switch that you can run from the op amp supply. The 4051 is
a very old device (1976 ?) now and there are much better devices available.

2) Run the op amp from the +/- switch supply

A third way may be to attenuate the output before applying to the
switch, say 10-1, which brings the switch input well within range, but
this is a more complicated solution. Also, the minimum gain then becomes 10.

Might seem a little obsessive, but I never feel happy unless a circuit
design looks ok from all angles. Do it right to start with and you never
need to look at it again etc ...

Chris

 

[Genome]

[email protected] wrote...

I doubt that'll solve the problem. Normally we have to use
high-voltage cmos switches powered from the supply rails for
general purpose full-swing circuits, but my post suggested a
hack allowing the use of cheaper low-voltage logic-family
switches, provided your gain was at least 2x. Viz,

"Gain switching the feedback dividers eliminates effects
. __ from the CMOS switch's series resistance.
. ------|+ \
. | >-----+----+----+---
. .--|-_/ R1 | |
. | | R3 | non-inverting gain
. '---o-->o----+ | R5
. o--- | ---+ |
. o--- | -- | ---+
. | | |
. R2 R4 R6
. | | |
. GND --+----+----'

And if Gmin = 2, i.e., R1 => R2, etc., then a +/-5V switch
can be used with +/-10V output signals."

I don't see how you can cut a trace and add a resistor and
thereby magically get around this principle for G = 1.

But there is a trick you can use to get full output swing
with half-swing capable switches such as the hc4041 family:
first attenuate the signal by 2x, and follow with a gain of
2x for the G=1 overall-gain position, etc., and so on.

Whoaaaaaaaaa!!!!!! FREAKY.

You're meant to spin in your grave not claw your way out of it and answer
the question.

DNA