开云体育

4. I'd do what tek did, and change that value, leaving the 3300 pf in
circuit.? In splitting the atom (or hairs), it is too easy to end up
with a bunch of loose subatomic particles rolling around under foot.


Harvey



On 2/3/2024 9:52 AM, n4buq wrote:

Well, after a bit more checking things out, I'm now thinking that the problem
may not necessarily have been with the original C32119. When I placed a
substitute capacitor in its place, I was using 3300pF which made the circuit
very stable. While I don't have a 2200pF replacement just yet, I did try two
1000pF caps in parallel and that didn't work. The scope would come up okay,
but pressing the BEAM FINDER would destabilize it just like before. Adding a
third 1000pF to get 3000pF works but that's still over the design value of
2200pF, 20%.

Tek started with C32119 at 1000pF, increased it to 5000pF, and then backed that
down to 2200pF. I'm not sure why they made that last change but I wonder if
having too much C is not desirable there. I could just go with 3300pF and call
it good but I think it's going to bug me.

I did a lot of side-by-side comparisons and found something that I cannot
explain. The base of Q32115 is fed from +17V through R32117, a 120k. If I
check across that resistor in circuit, it checks at 114 ohms. Thinking it
could have drifted low, I lifted one leg and it checks at 118 ohms. Not spot
on but at least in tolerance. The base resistor for Q3285, R4377, another 120k
ohm, checks low in-circuit the same way as R32117. I haven't lifted a leg on
that one to see if it may be low but I plan to do that.

I don't know if that could have anything to do with the instability going on in
the -50V circuit but it seems rather wrong and I can't explain why.

Thanks,
Barry - N4BUQ

I FOUND IT.

C32119 was the culprit. I had a 0.01uF ceramic handy and when I placed it
across C32119, the scope came to life. I then clipped a 0.0033uF in parallel
with it and it still worked perfectly. I desoldered the existing cap and
placed the 0.0033uF in the holes (I didn't solder it in place as I'd like to
get a 0.0022uF to stick with the book) and that also worked.

Now that I had it out of the circuit, I tested the old cap and, oddly, it tested
very close to 0.0022uF. That made me really wonder what was going on so I
stuck it back in the holes and it didn't work - at least not right away. I
wiggled it a bit and could get the scope to work but I think it would still go
haywire when I pressed the BEAM FINDER. I don't know but it could have
something to do with the way the leads were bent because before I tested it, I
straightened the leads so it would fit better in the ZIF socket. Still, I
think it's flaky.

I want to thank everyone who has given me help and suggestions. I probably
should apologize to the list for droning on about this because several weeks
ago, Prof. Lee had suggested I do exactly what I did tonight and, if I had, it
would have saved a lot of "bandwidth". Ed's email earlier today further pushed
me to look at those caps again and I'm glad he did. Thanks also to Harvey,
Mark, Renée, Dave, G?ran, Renaud. John, and Andreas (and, out of 133 messages,
probably others that I'm missing) for the responses and suggestions.

Barry - N4BUQ

Hi Ed,

Thank you SO MUCH for that analysis.

C32119 was one of the first suspects and I need to go ahead and look at either
replacing it or adding some capacitance in parallel with it. I really hadn't
considered C32134 very much at all but perhaps I don't need to cross it off too
quickly. I've replaced C32141 and C3289 to no avail and, truthfully, I didn't
much expect it would. C32119 and C3278 might have much more effect than either
of C32141 or C3289. I believe C32119 and C32134 are both the boxed values in
both regulator boards.

I wasn't considering the extra load on the -50V supply when the BEAM FINDER is
pressed but, if so, then that would make more sense. I was thinking that the
small changes on the -15V supply when that was pressed was possibly the
triggering event but maybe not.

I plan to swap back in the original board and try some of the things you've
suggested as that's a fairly easy thing to do now. I'll report back what I
find.

Thank you again,
Barry - N4BUQ

I looked up the manual, and I think the -50V regulator is marginally stable.
Your observations about relation to the approx 200 mV drop on the current sense
resistor, and the clue that the beam-finder trips it too, indicate that when
the DC load is low enough, the overall open-loop voltage gain rises enough to
go unstable. When the beam-finder is activated, the deflection amplifiers have
marked change in gain, which also reflects in the -50V load current (likely
goes down). Once oscillation starts, the DC bias levels throughout the
regulator loop may change enough form a bi-stable system, which could explain
having to shut it down to reset it.

I'd suggest doing some experiments to swamp the HF gain, mostly in the output
Darlington pair Q32143 and Q32139. Note that in the +50V supply, there are only
two voltage gain stages, while in the -50V one, there are three - the last
being the Darlington operating in common-emitter mode, versus the one in the
+50V being an emitter follower only. The various scaling for the regulation
voltages and stages should make their overall gains roughly the same, but the
extra stage might make it more sensitive to part variations in the minus
supply.

There are some spots where compensation is done, namely involving C32119,
C32134, and C32141. Note that C32119 and C32134 are "boxed" (at least in the
manual version I found), indicating that the values or parts were changed at
some points, or may be selected. Note that in the -15V regulator, C3278 is also
boxed. It has the same topology as the -50V one, so similar issues due to that
"extra" gain stage. The positive regulators all use the same type, with NPN
emitter follower output, and no boxing of the compensation caps. To have the
same (symmetric) topology, negative supplies would normally want PNPs for the
output, but in many designs, the NPN circuits are used for various reasons like
making all the same part etc. This is very common in Tek, HP, and other designs
- HP has some very strange arrangements including stacking raw supplies on top
of ground-referenced pass transistors, for instance.

Anyway, it's not a bad thing, but it makes it a little more complicated to
compensate and handle part variations. Again, the boxed parts in the minus
regulators here indicate that changes have been needed. So, you may want to box
certain ones yourself to custom-fix the unit. First though, you'd want to
verify the (in)stability situation, by grossly swamping things out to see what
happens. You could try upping the value of those caps mentioned - not by
extreme amounts, but say maybe two up to ten times, by tacking in extra caps in
parallel (not replacing). I'd start with C32141, since it's the one most
associated with the extra gain. You could try a brute force approach putting a
Miller C (C-B) right on Q32139, but it could be tricky since the base impedance
is very low, so the C could need to be pretty big. Swamping it here would come
close to simulating a slower transistor in this spot, without actually changing
the part. Next I would look at C32119, and lastly C32134. With all this sort of
stuff you have to be careful to not go too far, or it could aggravate the
oscillation instead. The main thing is to see if simple small changes can get
you enough phase margin. I would think so.

I think with some experiments you can figure out a workable fix. Good luck.

Ed

Some factors to consider:

1. Transistors of that era (!) may have used a different process.? As
time went on, the methods of making transistors improved (generally,
I think, for higher voltages and higher frequency.? In the TM500
series and perhaps the 7000 series, the increased frequency response
of the transistors could lead to unwanted oscillations.
2. think integrator (as in op-amp) controlling the transient response.
With too much capacitance, the feedback circuit is slow, and takes a
bit to get there.? With too little capacitance, the circuit can
either overshoot and correct, (which may lead to ringing), or might
oscillate.? What you want is a circuit that is critically damped,
which means a bit of overshoot and then settles down.
3. regarding the capacitors:? One thing to consider is that they may
have had transistors with different processes that changed the
circuit, same transistor number or not.? A second thing is that
other parts tolerances may have changed the response of the whole
circuit.? so 5000 pf might have been overkill.? 2200 pf was their
best guess, but if any transistors were of a different process, or
had a higher beta, then that 3300 pf may be ideal.
4. I'd do what tek did, and change that value, leaving the 3300 pf in
circuit.? In splitting the atom (or hairs), it is too easy to end up
with a bunch of loose subatomic particles rolling around under foot.


Harvey



On 2/3/2024 9:52 AM, n4buq wrote:

Well, after a bit more checking things out, I'm now thinking that the problem
may not necessarily have been with the original C32119. When I placed a
substitute capacitor in its place, I was using 3300pF which made the circuit
very stable. While I don't have a 2200pF replacement just yet, I did try two
1000pF caps in parallel and that didn't work. The scope would come up okay,
but pressing the BEAM FINDER would destabilize it just like before. Adding a
third 1000pF to get 3000pF works but that's still over the design value of
2200pF, 20%.

Tek started with C32119 at 1000pF, increased it to 5000pF, and then backed that
down to 2200pF. I'm not sure why they made that last change but I wonder if
having too much C is not desirable there. I could just go with 3300pF and call
it good but I think it's going to bug me.

I did a lot of side-by-side comparisons and found something that I cannot
explain. The base of Q32115 is fed from +17V through R32117, a 120k. If I
check across that resistor in circuit, it checks at 114 ohms. Thinking it
could have drifted low, I lifted one leg and it checks at 118 ohms. Not spot
on but at least in tolerance. The base resistor for Q3285, R4377, another 120k
ohm, checks low in-circuit the same way as R32117. I haven't lifted a leg on
that one to see if it may be low but I plan to do that.

I don't know if that could have anything to do with the instability going on in
the -50V circuit but it seems rather wrong and I can't explain why.

Thanks,
Barry - N4BUQ

I FOUND IT.

C32119 was the culprit. I had a 0.01uF ceramic handy and when I placed it
across C32119, the scope came to life. I then clipped a 0.0033uF in parallel
with it and it still worked perfectly. I desoldered the existing cap and
placed the 0.0033uF in the holes (I didn't solder it in place as I'd like to
get a 0.0022uF to stick with the book) and that also worked.

Now that I had it out of the circuit, I tested the old cap and, oddly, it tested
very close to 0.0022uF. That made me really wonder what was going on so I
stuck it back in the holes and it didn't work - at least not right away. I
wiggled it a bit and could get the scope to work but I think it would still go
haywire when I pressed the BEAM FINDER. I don't know but it could have
something to do with the way the leads were bent because before I tested it, I
straightened the leads so it would fit better in the ZIF socket. Still, I
think it's flaky.

I want to thank everyone who has given me help and suggestions. I probably
should apologize to the list for droning on about this because several weeks
ago, Prof. Lee had suggested I do exactly what I did tonight and, if I had, it
would have saved a lot of "bandwidth". Ed's email earlier today further pushed
me to look at those caps again and I'm glad he did. Thanks also to Harvey,
Mark, Renée, Dave, G?ran, Renaud. John, and Andreas (and, out of 133 messages,
probably others that I'm missing) for the responses and suggestions.

Barry - N4BUQ

Hi Ed,

Thank you SO MUCH for that analysis.

C32119 was one of the first suspects and I need to go ahead and look at either
replacing it or adding some capacitance in parallel with it. I really hadn't
considered C32134 very much at all but perhaps I don't need to cross it off too
quickly. I've replaced C32141 and C3289 to no avail and, truthfully, I didn't
much expect it would. C32119 and C3278 might have much more effect than either
of C32141 or C3289. I believe C32119 and C32134 are both the boxed values in
both regulator boards.

I wasn't considering the extra load on the -50V supply when the BEAM FINDER is
pressed but, if so, then that would make more sense. I was thinking that the
small changes on the -15V supply when that was pressed was possibly the
triggering event but maybe not.

I plan to swap back in the original board and try some of the things you've
suggested as that's a fairly easy thing to do now. I'll report back what I
find.

Thank you again,
Barry - N4BUQ

I looked up the manual, and I think the -50V regulator is marginally stable.
Your observations about relation to the approx 200 mV drop on the current sense
resistor, and the clue that the beam-finder trips it too, indicate that when
the DC load is low enough, the overall open-loop voltage gain rises enough to
go unstable. When the beam-finder is activated, the deflection amplifiers have
marked change in gain, which also reflects in the -50V load current (likely
goes down). Once oscillation starts, the DC bias levels throughout the
regulator loop may change enough form a bi-stable system, which could explain
having to shut it down to reset it.

I'd suggest doing some experiments to swamp the HF gain, mostly in the output
Darlington pair Q32143 and Q32139. Note that in the +50V supply, there are only
two voltage gain stages, while in the -50V one, there are three - the last
being the Darlington operating in common-emitter mode, versus the one in the
+50V being an emitter follower only. The various scaling for the regulation
voltages and stages should make their overall gains roughly the same, but the
extra stage might make it more sensitive to part variations in the minus
supply.

There are some spots where compensation is done, namely involving C32119,
C32134, and C32141. Note that C32119 and C32134 are "boxed" (at least in the
manual version I found), indicating that the values or parts were changed at
some points, or may be selected. Note that in the -15V regulator, C3278 is also
boxed. It has the same topology as the -50V one, so similar issues due to that
"extra" gain stage. The positive regulators all use the same type, with NPN
emitter follower output, and no boxing of the compensation caps. To have the
same (symmetric) topology, negative supplies would normally want PNPs for the
output, but in many designs, the NPN circuits are used for various reasons like
making all the same part etc. This is very common in Tek, HP, and other designs
- HP has some very strange arrangements including stacking raw supplies on top
of ground-referenced pass transistors, for instance.

Anyway, it's not a bad thing, but it makes it a little more complicated to
compensate and handle part variations. Again, the boxed parts in the minus
regulators here indicate that changes have been needed. So, you may want to box
certain ones yourself to custom-fix the unit. First though, you'd want to
verify the (in)stability situation, by grossly swamping things out to see what
happens. You could try upping the value of those caps mentioned - not by
extreme amounts, but say maybe two up to ten times, by tacking in extra caps in
parallel (not replacing). I'd start with C32141, since it's the one most
associated with the extra gain. You could try a brute force approach putting a
Miller C (C-B) right on Q32139, but it could be tricky since the base impedance
is very low, so the C could need to be pretty big. Swamping it here would come
close to simulating a slower transistor in this spot, without actually changing
the part. Next I would look at C32119, and lastly C32134. With all this sort of
stuff you have to be careful to not go too far, or it could aggravate the
oscillation instead. The main thing is to see if simple small changes can get
you enough phase margin. I would think so.

I think with some experiments you can figure out a workable fix. Good luck.

Ed

Some factors to consider:

1. Transistors of that era (!) may have used a different process. As
time went on, the methods of making transistors improved (generally,
I think, for higher voltages and higher frequency. In the TM500
series and perhaps the 7000 series, the increased frequency response
of the transistors could lead to unwanted oscillations.
2. think integrator (as in op-amp) controlling the transient response.
With too much capacitance, the feedback circuit is slow, and takes a
bit to get there. With too little capacitance, the circuit can
either overshoot and correct, (which may lead to ringing), or might
oscillate. What you want is a circuit that is critically damped,
which means a bit of overshoot and then settles down.
3. regarding the capacitors: One thing to consider is that they may
have had transistors with different processes that changed the
circuit, same transistor number or not. A second thing is that
other parts tolerances may have changed the response of the whole
circuit. so 5000 pf might have been overkill. 2200 pf was their
best guess, but if any transistors were of a different process, or
had a higher beta, then that 3300 pf may be ideal.
4. I'd do what tek did, and change that value, leaving the 3300 pf in
circuit. In splitting the atom (or hairs), it is too easy to end up
with a bunch of loose subatomic particles rolling around under foot.


Harvey



On 2/3/2024 9:52 AM, n4buq wrote:

Well, after a bit more checking things out, I'm now thinking that the

problem may not necessarily have been with the original C32119. When I
placed a substitute capacitor in its place, I was using 3300pF which made
the circuit very stable. While I don't have a 2200pF replacement just yet,
I did try two 1000pF caps in parallel and that didn't work. The scope
would come up okay, but pressing the BEAM FINDER would destabilize it just
like before. Adding a third 1000pF to get 3000pF works but that's still
over the design value of 2200pF, 20%.

Tek started with C32119 at 1000pF, increased it to 5000pF, and then

backed that down to 2200pF. I'm not sure why they made that last change
but I wonder if having too much C is not desirable there. I could just go
with 3300pF and call it good but I think it's going to bug me.

I did a lot of side-by-side comparisons and found something that I

cannot explain. The base of Q32115 is fed from +17V through R32117, a
120k. If I check across that resistor in circuit, it checks at 114 ohms.
Thinking it could have drifted low, I lifted one leg and it checks at 118
ohms. Not spot on but at least in tolerance. The base resistor for Q3285,
R4377, another 120k ohm, checks low in-circuit the same way as R32117. I
haven't lifted a leg on that one to see if it may be low but I plan to do
that.

I don't know if that could have anything to do with the instability

going on in the -50V circuit but it seems rather wrong and I can't explain
why.

Thanks,
Barry - N4BUQ

I FOUND IT.

C32119 was the culprit. I had a 0.01uF ceramic handy and when I placed

it

across C32119, the scope came to life. I then clipped a 0.0033uF in

parallel

with it and it still worked perfectly. I desoldered the existing cap

and

placed the 0.0033uF in the holes (I didn't solder it in place as I'd

like to

get a 0.0022uF to stick with the book) and that also worked.

Now that I had it out of the circuit, I tested the old cap and, oddly,

it tested

very close to 0.0022uF. That made me really wonder what was going on

so I

stuck it back in the holes and it didn't work - at least not right

away. I

wiggled it a bit and could get the scope to work but I think it would

still go

haywire when I pressed the BEAM FINDER. I don't know but it could have
something to do with the way the leads were bent because before I

tested it, I

straightened the leads so it would fit better in the ZIF socket.

Still, I

think it's flaky.

I want to thank everyone who has given me help and suggestions. I

probably

should apologize to the list for droning on about this because several

weeks

ago, Prof. Lee had suggested I do exactly what I did tonight and, if I

had, it

would have saved a lot of "bandwidth". Ed's email earlier today

further pushed

me to look at those caps again and I'm glad he did. Thanks also to

Harvey,

Mark, Renée, Dave, G?ran, Renaud. John, and Andreas (and, out of 133

messages,

probably others that I'm missing) for the responses and suggestions.

Barry - N4BUQ

Hi Ed,

Thank you SO MUCH for that analysis.

C32119 was one of the first suspects and I need to go ahead and look

at either

replacing it or adding some capacitance in parallel with it. I really

hadn't

considered C32134 very much at all but perhaps I don't need to cross

it off too

quickly. I've replaced C32141 and C3289 to no avail and, truthfully,

I didn't

much expect it would. C32119 and C3278 might have much more effect

than either

of C32141 or C3289. I believe C32119 and C32134 are both the boxed

values in

both regulator boards.

I wasn't considering the extra load on the -50V supply when the BEAM

FINDER is

pressed but, if so, then that would make more sense. I was thinking

that the

small changes on the -15V supply when that was pressed was possibly the
triggering event but maybe not.

I plan to swap back in the original board and try some of the things

you've

suggested as that's a fairly easy thing to do now. I'll report back

what I

find.

Thank you again,
Barry - N4BUQ

I looked up the manual, and I think the -50V regulator is marginally

stable.

Your observations about relation to the approx 200 mV drop on the

current sense

resistor, and the clue that the beam-finder trips it too, indicate

that when

the DC load is low enough, the overall open-loop voltage gain rises

enough to

go unstable. When the beam-finder is activated, the deflection

amplifiers have

marked change in gain, which also reflects in the -50V load current

(likely

goes down). Once oscillation starts, the DC bias levels throughout the
regulator loop may change enough form a bi-stable system, which could

explain

having to shut it down to reset it.

I'd suggest doing some experiments to swamp the HF gain, mostly in

the output

Darlington pair Q32143 and Q32139. Note that in the +50V supply,

there are only

two voltage gain stages, while in the -50V one, there are three - the

last

being the Darlington operating in common-emitter mode, versus the one

in the

+50V being an emitter follower only. The various scaling for the

regulation

voltages and stages should make their overall gains roughly the same,

but the

extra stage might make it more sensitive to part variations in the

minus

supply.

There are some spots where compensation is done, namely involving

C32119,

C32134, and C32141. Note that C32119 and C32134 are "boxed" (at least

in the

manual version I found), indicating that the values or parts were

changed at

some points, or may be selected. Note that in the -15V regulator,

C3278 is also

boxed. It has the same topology as the -50V one, so similar issues

due to that

"extra" gain stage. The positive regulators all use the same type,

with NPN

emitter follower output, and no boxing of the compensation caps. To

have the

same (symmetric) topology, negative supplies would normally want PNPs

for the

output, but in many designs, the NPN circuits are used for various

reasons like

making all the same part etc. This is very common in Tek, HP, and

other designs

- HP has some very strange arrangements including stacking raw

supplies on top

of ground-referenced pass transistors, for instance.

Anyway, it's not a bad thing, but it makes it a little more

complicated to

compensate and handle part variations. Again, the boxed parts in the

minus

regulators here indicate that changes have been needed. So, you may

want to box

certain ones yourself to custom-fix the unit. First though, you'd

want to

verify the (in)stability situation, by grossly swamping things out to

see what

happens. You could try upping the value of those caps mentioned - not

by

extreme amounts, but say maybe two up to ten times, by tacking in

extra caps in

parallel (not replacing). I'd start with C32141, since it's the one

most

associated with the extra gain. You could try a brute force approach

putting a

Miller C (C-B) right on Q32139, but it could be tricky since the base

impedance

is very low, so the C could need to be pretty big. Swamping it here

would come

close to simulating a slower transistor in this spot, without

actually changing

the part. Next I would look at C32119, and lastly C32134. With all

this sort of

stuff you have to be careful to not go too far, or it could aggravate

the

oscillation instead. The main thing is to see if simple small changes

can get

you enough phase margin. I would think so.

I think with some experiments you can figure out a workable fix. Good

luck.

Ed

Well, after a bit more checking things out, I'm now thinking that the problem may not necessarily have been with the original C32119. When I placed a substitute capacitor in its place, I was using 3300pF which made the circuit very stable. While I don't have a 2200pF replacement just yet, I did try two 1000pF caps in parallel and that didn't work. The scope would come up okay, but pressing the BEAM FINDER would destabilize it just like before. Adding a third 1000pF to get 3000pF works but that's still over the design value of 2200pF, 20%.

Tek started with C32119 at 1000pF, increased it to 5000pF, and then backed that down to 2200pF. I'm not sure why they made that last change but I wonder if having too much C is not desirable there. I could just go with 3300pF and call it good but I think it's going to bug me.

I did a lot of side-by-side comparisons and found something that I cannot explain. The base of Q32115 is fed from +17V through R32117, a 120k. If I check across that resistor in circuit, it checks at 114 ohms. Thinking it could have drifted low, I lifted one leg and it checks at 118 ohms. Not spot on but at least in tolerance. The base resistor for Q3285, R4377, another 120k ohm, checks low in-circuit the same way as R32117. I haven't lifted a leg on that one to see if it may be low but I plan to do that.

I don't know if that could have anything to do with the instability going on in the -50V circuit but it seems rather wrong and I can't explain why.

Thanks,
Barry - N4BUQ

I FOUND IT.

C32119 was the culprit. I had a 0.01uF ceramic handy and when I placed it
across C32119, the scope came to life. I then clipped a 0.0033uF in parallel
with it and it still worked perfectly. I desoldered the existing cap and
placed the 0.0033uF in the holes (I didn't solder it in place as I'd like to
get a 0.0022uF to stick with the book) and that also worked.

Now that I had it out of the circuit, I tested the old cap and, oddly, it tested
very close to 0.0022uF. That made me really wonder what was going on so I
stuck it back in the holes and it didn't work - at least not right away. I
wiggled it a bit and could get the scope to work but I think it would still go
haywire when I pressed the BEAM FINDER. I don't know but it could have
something to do with the way the leads were bent because before I tested it, I
straightened the leads so it would fit better in the ZIF socket. Still, I
think it's flaky.

I want to thank everyone who has given me help and suggestions. I probably
should apologize to the list for droning on about this because several weeks
ago, Prof. Lee had suggested I do exactly what I did tonight and, if I had, it
would have saved a lot of "bandwidth". Ed's email earlier today further pushed
me to look at those caps again and I'm glad he did. Thanks also to Harvey,
Mark, Renée, Dave, G?ran, Renaud. John, and Andreas (and, out of 133 messages,
probably others that I'm missing) for the responses and suggestions.

Barry - N4BUQ

Hi Ed,

Thank you SO MUCH for that analysis.

C32119 was one of the first suspects and I need to go ahead and look at either
replacing it or adding some capacitance in parallel with it. I really hadn't
considered C32134 very much at all but perhaps I don't need to cross it off too
quickly. I've replaced C32141 and C3289 to no avail and, truthfully, I didn't
much expect it would. C32119 and C3278 might have much more effect than either
of C32141 or C3289. I believe C32119 and C32134 are both the boxed values in
both regulator boards.

I wasn't considering the extra load on the -50V supply when the BEAM FINDER is
pressed but, if so, then that would make more sense. I was thinking that the
small changes on the -15V supply when that was pressed was possibly the
triggering event but maybe not.

I plan to swap back in the original board and try some of the things you've
suggested as that's a fairly easy thing to do now. I'll report back what I
find.

Thank you again,
Barry - N4BUQ

I looked up the manual, and I think the -50V regulator is marginally stable.
Your observations about relation to the approx 200 mV drop on the current sense
resistor, and the clue that the beam-finder trips it too, indicate that when
the DC load is low enough, the overall open-loop voltage gain rises enough to
go unstable. When the beam-finder is activated, the deflection amplifiers have
marked change in gain, which also reflects in the -50V load current (likely
goes down). Once oscillation starts, the DC bias levels throughout the
regulator loop may change enough form a bi-stable system, which could explain
having to shut it down to reset it.

I'd suggest doing some experiments to swamp the HF gain, mostly in the output
Darlington pair Q32143 and Q32139. Note that in the +50V supply, there are only
two voltage gain stages, while in the -50V one, there are three - the last
being the Darlington operating in common-emitter mode, versus the one in the
+50V being an emitter follower only. The various scaling for the regulation
voltages and stages should make their overall gains roughly the same, but the
extra stage might make it more sensitive to part variations in the minus
supply.

There are some spots where compensation is done, namely involving C32119,
C32134, and C32141. Note that C32119 and C32134 are "boxed" (at least in the
manual version I found), indicating that the values or parts were changed at
some points, or may be selected. Note that in the -15V regulator, C3278 is also
boxed. It has the same topology as the -50V one, so similar issues due to that
"extra" gain stage. The positive regulators all use the same type, with NPN
emitter follower output, and no boxing of the compensation caps. To have the
same (symmetric) topology, negative supplies would normally want PNPs for the
output, but in many designs, the NPN circuits are used for various reasons like
making all the same part etc. This is very common in Tek, HP, and other designs
- HP has some very strange arrangements including stacking raw supplies on top
of ground-referenced pass transistors, for instance.

Anyway, it's not a bad thing, but it makes it a little more complicated to
compensate and handle part variations. Again, the boxed parts in the minus
regulators here indicate that changes have been needed. So, you may want to box
certain ones yourself to custom-fix the unit. First though, you'd want to
verify the (in)stability situation, by grossly swamping things out to see what
happens. You could try upping the value of those caps mentioned - not by
extreme amounts, but say maybe two up to ten times, by tacking in extra caps in
parallel (not replacing). I'd start with C32141, since it's the one most
associated with the extra gain. You could try a brute force approach putting a
Miller C (C-B) right on Q32139, but it could be tricky since the base impedance
is very low, so the C could need to be pretty big. Swamping it here would come
close to simulating a slower transistor in this spot, without actually changing
the part. Next I would look at C32119, and lastly C32134. With all this sort of
stuff you have to be careful to not go too far, or it could aggravate the
oscillation instead. The main thing is to see if simple small changes can get
you enough phase margin. I would think so.

I think with some experiments you can figure out a workable fix. Good luck.

Ed

most will have no patience and just pass over a page of test without any photos or diagram

Suggest posting on eevblog forum for the bus type you need to extend.

j

I FOUND IT.

C32119 was the culprit. I had a 0.01uF ceramic handy and when I placed it
across C32119, the scope came to life. I then clipped a 0.0033uF in parallel
with it and it still worked perfectly. I desoldered the existing cap and
placed the 0.0033uF in the holes (I didn't solder it in place as I'd like to
get a 0.0022uF to stick with the book) and that also worked.

Now that I had it out of the circuit, I tested the old cap and, oddly, it tested
very close to 0.0022uF. That made me really wonder what was going on so I
stuck it back in the holes and it didn't work - at least not right away. I
wiggled it a bit and could get the scope to work but I think it would still go
haywire when I pressed the BEAM FINDER. I don't know but it could have
something to do with the way the leads were bent because before I tested it, I
straightened the leads so it would fit better in the ZIF socket. Still, I
think it's flaky.

I want to thank everyone who has given me help and suggestions. I probably
should apologize to the list for droning on about this because several weeks
ago, Prof. Lee had suggested I do exactly what I did tonight and, if I had, it
would have saved a lot of "bandwidth". Ed's email earlier today further pushed
me to look at those caps again and I'm glad he did. Thanks also to Harvey,
Mark, Renée, Dave, G?ran, Renaud. John, and Andreas (and, out of 133 messages,
probably others that I'm missing) for the responses and suggestions.

Barry - N4BUQ

Hi Ed,

Thank you SO MUCH for that analysis.

C32119 was one of the first suspects and I need to go ahead and look at either
replacing it or adding some capacitance in parallel with it. I really hadn't
considered C32134 very much at all but perhaps I don't need to cross it off too
quickly. I've replaced C32141 and C3289 to no avail and, truthfully, I didn't
much expect it would. C32119 and C3278 might have much more effect than either
of C32141 or C3289. I believe C32119 and C32134 are both the boxed values in
both regulator boards.

I wasn't considering the extra load on the -50V supply when the BEAM FINDER is
pressed but, if so, then that would make more sense. I was thinking that the
small changes on the -15V supply when that was pressed was possibly the
triggering event but maybe not.

I plan to swap back in the original board and try some of the things you've
suggested as that's a fairly easy thing to do now. I'll report back what I
find.

Thank you again,
Barry - N4BUQ

I looked up the manual, and I think the -50V regulator is marginally stable.
Your observations about relation to the approx 200 mV drop on the current sense
resistor, and the clue that the beam-finder trips it too, indicate that when
the DC load is low enough, the overall open-loop voltage gain rises enough to
go unstable. When the beam-finder is activated, the deflection amplifiers have
marked change in gain, which also reflects in the -50V load current (likely
goes down). Once oscillation starts, the DC bias levels throughout the
regulator loop may change enough form a bi-stable system, which could explain
having to shut it down to reset it.

I'd suggest doing some experiments to swamp the HF gain, mostly in the output
Darlington pair Q32143 and Q32139. Note that in the +50V supply, there are only
two voltage gain stages, while in the -50V one, there are three - the last
being the Darlington operating in common-emitter mode, versus the one in the
+50V being an emitter follower only. The various scaling for the regulation
voltages and stages should make their overall gains roughly the same, but the
extra stage might make it more sensitive to part variations in the minus
supply.

There are some spots where compensation is done, namely involving C32119,
C32134, and C32141. Note that C32119 and C32134 are "boxed" (at least in the
manual version I found), indicating that the values or parts were changed at
some points, or may be selected. Note that in the -15V regulator, C3278 is also
boxed. It has the same topology as the -50V one, so similar issues due to that
"extra" gain stage. The positive regulators all use the same type, with NPN
emitter follower output, and no boxing of the compensation caps. To have the
same (symmetric) topology, negative supplies would normally want PNPs for the
output, but in many designs, the NPN circuits are used for various reasons like
making all the same part etc. This is very common in Tek, HP, and other designs
- HP has some very strange arrangements including stacking raw supplies on top
of ground-referenced pass transistors, for instance.

Anyway, it's not a bad thing, but it makes it a little more complicated to
compensate and handle part variations. Again, the boxed parts in the minus
regulators here indicate that changes have been needed. So, you may want to box
certain ones yourself to custom-fix the unit. First though, you'd want to
verify the (in)stability situation, by grossly swamping things out to see what
happens. You could try upping the value of those caps mentioned - not by
extreme amounts, but say maybe two up to ten times, by tacking in extra caps in
parallel (not replacing). I'd start with C32141, since it's the one most
associated with the extra gain. You could try a brute force approach putting a
Miller C (C-B) right on Q32139, but it could be tricky since the base impedance
is very low, so the C could need to be pretty big. Swamping it here would come
close to simulating a slower transistor in this spot, without actually changing
the part. Next I would look at C32119, and lastly C32134. With all this sort of
stuff you have to be careful to not go too far, or it could aggravate the
oscillation instead. The main thing is to see if simple small changes can get
you enough phase margin. I would think so.

I think with some experiments you can figure out a workable fix. Good luck.

Ed