Topics

Hello from newcomer Fabio Trevisan - My first Tek Scope 464 + DM44

Brad Thompson <brad.thompson@...>
 

On 11/4/2016 1:28 PM, Fabio Trevisan fabio.tr3visan@... [TekScopes] wrote:


Hi there,

John.. thanks for the reminder. The new resistor is of the same power
rating as the original. In fact, although we're talking about the H.V.
inverter and a 2N3055 transistors, which is a power device, this resistor
itself is a small one, a quarter or half a watt if much.

David, see my comments next to your replies... thanks again... very helpful.

Rgrds,

Fabio

<snip>

We have discussed it in the past and not reached any conclusion about
what Tektronix selected for on these transistors; I suspect it was
either current gain or Ft (current gain-bandwidth product). Tektronix
used at least three different 2N3055 variations and sometimes even
used all three in the same oscilloscope like the 76x3 series.
The major issue appears to be that if the transistor Ft is too high,
then the high voltage inverter will suffer from spurious oscillation
at a much higher frequency. A modern direct replacement which is
readily available is the 2N3771G (40V 30A 150W) or 2N3772G (60V 20A
150W) which both have an Ft of 200 kH which is very close to the 300
kHz Ft of the original 2N3055 that Tektronix used and they are tougher
also. The later 2N3055s had a 800 kHz Ft and modern ones are 1.5 MHz
which can definitely be a problem in this circuit.
Wow... I didn't know the fT for the older 2N3055 to be so low... Never
mention knowing
that they improved it over the years keeping the same part identifier...
That's confusing!
Hello--

I posted this a couple of years ago by way of explaining the 2N3055's
variations.

As registered with JEDEC(*) in 1964, the 2N3055's original electrical specifications
were relatively relaxed (some may prefer the adjective "sloppy"). Some
manufacturers compounded the problem by packaging rejected dice from
production lots of faster-technology devices and labeling these as 2N3055s.

So, if you designed a circuit around the original 2N3055, a single-diffused mesa
device with an fT around 800 KHz, and your company's purchasing department
found a source of inexpensive 2N3055s built around triple-diffused rejects with
a much higher fT, your circuit may behave unpredictably.

Also, the 2N3055's original registered characteristics underwent several
revisions, culminating with revision E in 1968 which increased the upper hFE
limit from 60 to 120 at a collector current of 1 ampere.

And once your purchasing department wanders away from the JEDEC-registered
spec into the wild world of plastic-encapsulated "3055s", house-marked 3055s
and MOSFETs labeled "3055", your painstakingly-engineered circuit may not work at all
when the weird parts hit the production floor ("But it says 3055 on the label!").

(At this point, design and components engineers sharpen their pitchforks,
light their pine-knot torches and descend upon the purchasing department in
search of a buyer named Frankenstein....)

73--

Brad AA1IP

(*) JEDEC: Joint Electron Device Engineering Councils

 

On Fri, 4 Nov 2016 15:28:45 -0200, you wrote:

John.. thanks for the reminder. The new resistor is of the same power
rating as the original. In fact, although we're talking about the H.V.
inverter and a 2N3055 transistors, which is a power device, this resistor
itself is a small one, a quarter or half a watt if much.

Fabio
Resistor R1483 sets the minimum bias current using the +5 volt supply
into the base of the 2N3055 and is relatively high in value so power
dissipation even under adverse conditions is only like 2 milliamps for
20 milliwatts. Increasing its value just lowers the power even more.

2016-11-04 4:14 GMT-02:00 David @DWH [TekScopes] <
TekScopes@...>:

...

We have discussed it in the past and not reached any conclusion about
what Tektronix selected for on these transistors; I suspect it was
either current gain or Ft (current gain-bandwidth product). Tektronix
used at least three different 2N3055 variations and sometimes even
used all three in the same oscilloscope like the 76x3 series.
The major issue appears to be that if the transistor Ft is too high,
then the high voltage inverter will suffer from spurious oscillation
at a much higher frequency. A modern direct replacement which is
readily available is the 2N3771G (40V 30A 150W) or 2N3772G (60V 20A
150W) which both have an Ft of 200 kH which is very close to the 300
kHz Ft of the original 2N3055 that Tektronix used and they are tougher
also. The later 2N3055s had a 800 kHz Ft and modern ones are 1.5 MHz
which can definitely be a problem in this circuit.
Wow... I didn't know the fT for the older 2N3055 to be so low... Never
mention knowing
that they improved it over the years keeping the same part identifier...
That's confusing!

I almost bought a 2N3773 instead of the MJ15015 but, as usual, I was
looking at a part
that could match or exceed the 2N3055 absolute maximum values... It
wouldn't cross
my mind replacing it by the 3771, which is 40V (and not 60)... although I
admit that in this
circuit, 40V is enough (I actually measured the Vce on the oscilloscope and
it goes from
-10V to +29V)
The 2N3773 used to be my recommendation to replace old and slow
2N3055s but the modern 2N3773 is 4 MHz instead of the 200 kHz that it
used to be.

I agree that the 2N3772 is a better choice because of its higher Vceo
but for completeness I also listed the 2N3771 which I think will also
work fine.

Indeed, I noticed that the picked-up interference that I see on the screen
(while having the
H.V. cages opened, the ~45Khz signal is clearly visible without any probe),
is less clean
than it used to be... it may be due to spurious high frequency
oscillations...
I will try to look into it with more detail... If I find evidence of higher
harmonics, I will try to
get a 2N3772 (which is 60V)...
That is probably normal without the shielding provided by the cage.

What I worry about is that even without spurious oscillation, as the
transistor heats up its gain will rise allowing the inverter to
saturate once the whole unit is back together. Luckily that should
not be fatal and it will be readily apparent when the horizontal and
vertical deflection are both reduced.

Having the horizontal and vertical deflection change indicates that
the cathode voltage is not being regulated. Lower cathode voltages
means lower electron velocity which gives the deflection plates more
time to deflect the beam yielding greater deflection.
Ack to that! Vertical sensitivity was low by some 50% while the cathode was
at about -1800V.
After I got regulation back, it came down to the correct levels.
I immediately thought of regulation when I noticed that it was "breathing"
as I
changed Intensity control, but my first thoughts (and fears) was that
something
could have blown in the H.V. feedback divider (the large custom film
resistor).
I was relieved when I noticed that feedback summing point was negative.
The custom thick film resistor networks are very reliable and in
theory can be replaced if necessary using a couple of modern thick
film high voltage resistors.

* Cathode voltage was too high and out of regulation
* Changed R1483 to lower it indicating too much drive to 1486.
* Gain too high?
Well, I reached to that conclusion from the theory of operation in the S.M.
in where it says that the feedback controls how much bias is fed into Q1486
bringing it closer or farther from the conduction point and therefore
making it
conduct for more or less time and, therefore, putting more or or less energy
into the transformer which translates into higher or lower voltage.
Sorry, I meant that as a summary to reduce the amount of quoting. I
was not questioning what you found or your conclusions.

I would like to know why the circuit is designed the way it is. Why
didn't Tektronix make it so that it would not be as sensitive to
transistor characteristics? It would be nice to have a general
solution so modern high Ft 2N3055s could be used without issues.
I confess that transformer dependent oscillator theory has never
been "intuitive" to me... I understand the concepts involved, but not enough
that I could design one from the paper up...and be sure that practice would
meet theory at the fiberglass.
I can think of building one empirically, but that doesn't count to know or
understand
why folks at Tek did it that way and not otherwise.
There are a lot of unknowns in the circuit because we do not know the
parasitic elements of the transformer which are what allow spurious
oscillation. In the past when working with circuits like this, I have
fiddled with the local frequency response of the transistor while
observing the results and then used that to make a model of the
parasitics. If convenient, I might then confirm with a direct
measurement to see if the two agree.

Of course if I have access to the transformer without uninstalling it,
then I would make some measurements first.

In their defense I can say that my Kenwood's H.V. oscillator seems quite
similar. An NPN
driving transistor with emitter connected directly (no emitter resistor) to
-12V,
collector pulling the transformer down, base connected to the feedback
winding
of the transformer which is biased, from the other side, by the feedback
circuitry.
Only difference is that it has a 100R base stopper resistor, probably
inserted there
to tame H.F. spurious oscillation.
In the past I have recommended adding a low value of resistance in
series with the base and some of the Tektronix designs do include
this.

My gut feeling is that a series RC circuit needs to be added between
the collector and base to tailor the medium frequency gain.

At the point that I am right now, I am in position of doing some
experimentation...
I don't pretend to ever reach mathematical conclusions but I can look at
the circuit
as it stands now and, if it's presenting unwanted H.F. artifacts, I can
play around
the transistor and search for a reliable and consistent way of taming it.
One that wouldn't compromise the phase margins (or the lack thereof) at the
nominal
frequency, which is mandatory to assure oscillation.
That part, to make it less dependent on transistor's fT.
About the dependency on the transistor's beta, I think that a common emitter
amplifier without any emitter degeneration as it is, and without any DC
negative feedback
to its base, indeed leaves a lot to desire...
At some expense of gain (which is not plentiful in the 2N3055 but much
better in
modern devices), I think an emitter degeneration resistor could improve the
circuit's susceptibility to change of device's beta (not just from device
to device,
but also from temperature change).

As soon as I get done with the CRT rejuvenation, which is my next chapter
at this
point, I can dedicate some time to research on those improvements.
I might not try it without at least a x100 wide bandwidth probe and an
analog or digital storage oscilloscope. Besides spurious oscillation,
it is important that the output voltage not spike during startup which
is part of the frequency compensation for the relatively slow
regulation loop.

Note that while an inexpensive x100 high voltage probe will work to
measure the +600 volt secondary, it must not be used with AC coupling!
These probes rely on the 1 megohm input resistance of the oscilloscope
so when AC in coupling is used, the coupling capacitor will charge to
the average DC value which can be well beyond the input specifications
of the oscilloscope.

Of course a primary side measurement may be just as good for this.

Increasing the value of R1483 is a fine solution if the problem is
just too much current gain but can you measure the transformer voltage
using an oscilloscope to see if it is oscillating properly? Do you
have a high voltage oscilloscope probe? We know from a previous
discussion that spurious oscillation is possible.
I don't have an H.V. oscilloscope probe. I have a home-made H.V. probe made
of a string of about 30 x 3.3Mohms resistors, all insulated by plastic
spacers,
inserted on a 16mm dia acrylic knitting needle. I tried... but it's not
good for AC.
Well, that will not work and floating the 464 device under test or the
measuring oscilloscope to measure the waveform across the heater
winding is an even worse idea.

But, as I mentioned earlier, the vertical pre-amplifier picks-up the
oscillator frequency
pretty well (with the H.V. cages removed) so...Can I call it an H.V. scope
probe?
It is only a problem if it continues to do so with the high voltage
cages removed.

And I can measure the signal at the H.V. transformer's primary... which
must be
good enough.
To my understanding whatever H.F. artifacts that may be arising, must be
present
on the primary side, if we are to fix it from the primary side or, in other
words, if it can't
be probed on the primary, it can't be fixed from the primary.

I have pictures of how those primary waves were before I changed the
transformer
and transistor... I can look at them now and see if there are differences.

I'll get back to you with the results.
Measuring the voltage across the primary should be good enough and is
probably the best option anyway. If there is parasitic oscillation,
it will be present there.

I wonder why R1483 was needed at all. The AC impedance at the emitter
of Q1484 is low because of C1483 so R1483 only provides a minimum
operating current. I would have expected R1483 to go to ground and
Q1484 to provide all of the base current but maybe R1483 has something
to do with the startup characteristics like preventing the output
voltage from overshooting which could damage something.
I don't think it's a matter of being REALLY needed... I think it's more
like a practical circuit's approach.
They probably departed from a working circuit (an oscillator without
voltage level feedback)
and added what was necessary so that it self regulated its output voltage.
As it stands, the control circuit is not responsible for making it work,
but just responsible
for adding a variable (controlled) amount of bias, enough to achieve just
the desired line
and load regulation.
And you can be right as well regarding the startup... oscillators are more
or less
easy to understand while under "regime"... but how they startup is often
guesswork
(especially before Spice simulation... but even though, circuits that
doesn't oscillate in Spice
do oscillate in real world and vice-versa).

We shall see... Or not!
Tektronix used variations of this circuit without the resistor or with
a voltage divider providing a lower impedance at a lower voltage;
check out the 7603 and 7403N schematics for examples. Like I said, my
suspicion is that pre biasing the oscillator to turn on prevents high
voltage overshoot during startup. The 7603 and 7403N examples show a
variation in the frequency compensation of the error amplifier with
more complex compensation used where the pre bias resistor was not
included. Note that these are push-pull designs but they still rely
on saturation of the transformer to control the oscillation frequency.

The most advanced version of this inverter design may be in the 465M
and 455. Ignoring the energy recovery circuit there which is not
required for operation, they include no pre biasing to turn on the
oscillator, use a low value of base series resistor (to prevent
parasitic oscillation?), and includes a low value base-emitter shunt
resistor to reduce storage time improving the performance of the
output transistor. Bias control from the regulator is a current
instead of a voltage. However the design also includes fast feedback
from the secondary to the base of the output transistor bypassing the
error amplifier which probably prevents overshoot at turn on. The
output transistor is a selected high Ft triple diffused type but they
changed it at least 3 times so maybe there were issues.

Given the above three examples, I suspect Tektronix continually
grappled with turn on transients in these designs.

Note that the issue with spurious oscillation is completely separate
from the transient response requirements.

 

On Fri, 4 Nov 2016 15:10:47 -0400, you wrote:

...

So, if you designed a circuit around the original 2N3055, a
single-diffused mesa
device with an fT around 800 KHz, and your company's purchasing department
found a source of inexpensive 2N3055s built around triple-diffused
rejects with
a much higher fT, your circuit may behave unpredictably.

...

Brad AA1IP
The version of the 2N3055 which Tektronix used in some of the 76x3
oscilloscopes for the high voltage inverter was specified to have an
Ft of 300 kHz. At the same time, they were also using the 800 kHz
2N3055s (a normal one and a "selected" one!) in other areas of the
same oscilloscope. So there were three "different" 2N3055s in the
same oscilloscope.

They also used a 200 kHz 2N3773 which today is a 4 MHz part if you buy
it from the wrong source. The current On Semiconductor ones are still
200 kHz though.

Fabio Trevisan
 

Hello David / Brad,

I won't answer next to your comments this time...
I posted pictures on the yahoo Tekscopes.
https://groups.yahoo.com/neo/groups/TekScopes/photos/albums/1713030301.

1. There's one picture of the scope (for reference)

2. There are 2 pictures of Q1486 collector and base voltages that serves as
a reference, before I replaced the HV transformer (which was defective) and
before I managed to blew Q1486. Although, back then, the H.V. transformer
was suffering from H.V. leakage (only after warm-up), it can be seen in the
picture that the waveform of the collector is very clean, almost a perfect
sinusoid. Trace "Readouts" are in each picture's comments.

3. There are 2 pictures of the 1st attempt to replace Q1486, by an MJ15015.
A picture of the transistor itself and a dual trace picture of the
collector voltage and emitter current (sensed with a 0.22R resistor).
Waveform is not as clean as the original and it has an approx. 520KHz
ringing, more visible after after the lower collector voltage peak.
With this transistor, circuit only achieves regulation when R1483 is raised
to 10kOhm. Picture comments have the traces' "readouts".

On this replacement, I tried to investigate why there's an apparent
clipping on the low side of the collector's waveform (the 500KHz ringing
seems to be triggered by this abrupt transition)
It's not saturation because Vce here is about 9V and current is about
750mA... The clipping seems to come from base current being limited at a
certain level and I tried then to lower the value of Q1486 (to about 3K)
and, since it doesn't make any difference, I conclude that the base current
is being limited by the action of the feedback itself.
Since this transistor apparently have a larger beta, lowering the current
from R1483 was apparently enough to put the circuit in "closed-loop", but
since the (AC) gain is still high, the regulation occurs at a point at
which base current is limited by the control circuit.
I attempted to insert (up to) 220R resistor in series with the base and no
change in behavior... Then I tried to make it a miller integrator, to add a
dominant pole to this amplifier... and added a capacitor of as much as 1nF
from the collector to the base. As per my calculations, 220R and 1nF should
have been enough to eliminate the 520KHz ringing, but there was no change
to that as well.
Although I`m already sold-out in my theories, I conclude that this ringing
is just resonation from the transformer due to the abrupt current
"clipping", and not being amplified by the transistor.

Since this transistor works apparently well (at least in comparison to the
next one), I appreciate any suggestions on:
A. If this operation regime is already acceptable although with some
ringing (and probably more EMI generation)
B. Suggestions on other ways to modify the circuit in order to compensate
for this transistor's higher gain...

4. There are 2 pictures of the 2nd attempt to replace Q1486, by a 2N3772. A
picture of the transistor itself and a dual trace picture of the collector
voltage and emitter current (sensed with a 0.22R resistor). This time,
there's a remarkable spurious oscillation at approx. 520KHz).
Curiously enough, the 2N3772 should have had a lower fT and therefore less
tendency to oscillate... but in this case, I tend to believe that this
transistor is a fake.
The transistor is an ST part and, although the labeling is weird, I could
see on the net that this labeling over a silver background is normal to
those ST transistors.
Still I think this one might be a fake.

P.S. Regarding startup, curiously enough... (and I can't explain how it is
like it is)... Instead of an initial H.V. overshoot (due to the slow time
constant of the feedback, as earlier mentioned by David), when we look at
the Q1486 collector's waveform while turning-on, we see the AC voltage rise
slowly (in about 1.5 seconds) to about 1/2 of the peak-to-peak voltage of
the normal regime... and it stays at this apmlitude for a about 1/2
second... only then, it increases (quicker this time), to the final
voltage, which is about 30Vpp, with the lower peak at +9V and the higher
peak at +39V (as can be seen on the pictures).
By looking at this, it seems that there's something else at play during the
startup, that actually limits the oscillation amplitude, before the control
circuit can kick-in.

Rgrds,

Fabio



2016-11-04 19:39 GMT-02:00 David @DWH [TekScopes] <
TekScopes@...>:



On Fri, 4 Nov 2016 15:10:47 -0400, you wrote:

...

So, if you designed a circuit around the original 2N3055, a
single-diffused mesa
device with an fT around 800 KHz, and your company's purchasing department
found a source of inexpensive 2N3055s built around triple-diffused
rejects with
a much higher fT, your circuit may behave unpredictably.

...

Brad AA1IP
The version of the 2N3055 which Tektronix used in some of the 76x3
oscilloscopes for the high voltage inverter was specified to have an
Ft of 300 kHz. At the same time, they were also using the 800 kHz
2N3055s (a normal one and a "selected" one!) in other areas of the
same oscilloscope. So there were three "different" 2N3055s in the
same oscilloscope.

They also used a 200 kHz 2N3773 which today is a 4 MHz part if you buy
it from the wrong source. The current On Semiconductor ones are still
200 kHz though.


[Non-text portions of this message have been removed]

 

On Mon, 7 Nov 2016 13:31:40 -0200, you wrote:

Hello David / Brad,

I won't answer next to your comments this time...
I posted pictures on the yahoo Tekscopes.
https://groups.yahoo.com/neo/groups/TekScopes/photos/albums/1713030301.

1. There's one picture of the scope (for reference)

2. There are 2 pictures of Q1486 collector and base voltages that serves as
a reference, before I replaced the HV transformer (which was defective) and
before I managed to blew Q1486. Although, back then, the H.V. transformer
was suffering from H.V. leakage (only after warm-up), it can be seen in the
picture that the waveform of the collector is very clean, almost a perfect
sinusoid. Trace "Readouts" are in each picture's comments.
A shorted or partially shorted transformer might make the waveform
look cleaner than it should be.

3. There are 2 pictures of the 1st attempt to replace Q1486, by an MJ15015.
A picture of the transistor itself and a dual trace picture of the
collector voltage and emitter current (sensed with a 0.22R resistor).
Waveform is not as clean as the original and it has an approx. 520KHz
ringing, more visible after after the lower collector voltage peak.
With this transistor, circuit only achieves regulation when R1483 is raised
to 10kOhm. Picture comments have the traces' "readouts".
Thanks for posting photos of your results and photos of the
transistors you used. They make working on the problem much easier.

On this replacement, I tried to investigate why there's an apparent
clipping on the low side of the collector's waveform (the 500KHz ringing
seems to be triggered by this abrupt transition)
It's not saturation because Vce here is about 9V and current is about
750mA... The clipping seems to come from base current being limited at a
certain level and I tried then to lower the value of Q1486 (to about 3K)
and, since it doesn't make any difference, I conclude that the base current
is being limited by the action of the feedback itself.
Since this transistor apparently have a larger beta, lowering the current
from R1483 was apparently enough to put the circuit in "closed-loop", but
since the (AC) gain is still high, the regulation occurs at a point at
which base current is limited by the control circuit.
I attempted to insert (up to) 220R resistor in series with the base and no
change in behavior... Then I tried to make it a miller integrator, to add a
dominant pole to this amplifier... and added a capacitor of as much as 1nF
from the collector to the base. As per my calculations, 220R and 1nF should
have been enough to eliminate the 520KHz ringing, but there was no change
to that as well.
Although I`m already sold-out in my theories, I conclude that this ringing
is just resonation from the transformer due to the abrupt current
"clipping", and not being amplified by the transistor.
Well, you have stumped me so I am going to throw some ideas out. The
transistor should only be conducting during the part of the cycle
where the Vce is low and any spurious oscillation should only occur
during that time.

I checked the ST datasheet for the 2N3772 and they are indeed slow so
I have little idea where the 520 kHz ringing is coming from or why the
slow 2N3772 would produce it but the also slow MJ15015 does not. But
I do have some ideas

Is it possible that the 2N3772 is counterfeit?

The ON MJ15015 datasheet says that it is effectively a 0.8 MHz
super-premo-deluxe 2N3055 with tighter specifications so it would be a
good choice for a replacement anyway. I did not actually know they
made such a part. I will have to add it to my list. Mouser says it
is a 6 MHz part which is misleading (see next paragraph); sometimes
they list the minimum and sometimes they list the maximum Ft making it
difficult to compare parts just based on their selection guide and
that is why I missed it.

The variation of Ft from part to part is of course wide just like the
variation in hfe. The ON MJ15015 datasheet lists a range of 0.8 to
6.0 MHz and I have no reason to think the 2N3772 does not vary just as
much. This of course just point to why Tektronix may be been
selecting parts.

Another thing that occurs to me is if the 2N3772 shows a spurious
oscillation during the whole cycle, maybe it is actually conducting
during the whole cycle. Try adding a 0.8V / 2.5mA = roughly 320 ohm
parallel resistor between the base and emitter to draw off any leakage
and stored charge. If that makes any difference, then find the value
where it just barely stops the problem and then halve it.

If leakage is what is causing the problem, then this may be what
Tektronix was actually selecting parts for; leakage is an easy test to
do.

Since this transistor works apparently well (at least in comparison to the
next one), I appreciate any suggestions on:
A. If this operation regime is already acceptable although with some
ringing (and probably more EMI generation)
B. Suggestions on other ways to modify the circuit in order to compensate
for this transistor's higher gain...
I do not consider the spurious oscillation acceptable. Even if the
circuit works, it indicates marginal operation.

See above for my suggestions about circuit operation and
modifications.

4. There are 2 pictures of the 2nd attempt to replace Q1486, by a 2N3772. A
picture of the transistor itself and a dual trace picture of the collector
voltage and emitter current (sensed with a 0.22R resistor). This time,
there's a remarkable spurious oscillation at approx. 520KHz).
Curiously enough, the 2N3772 should have had a lower fT and therefore less
tendency to oscillate... but in this case, I tend to believe that this
transistor is a fake.
The transistor is an ST part and, although the labeling is weird, I could
see on the net that this labeling over a silver background is normal to
those ST transistors.
Still I think this one might be a fake.
Ya, I wondered about whether the 2N3772 is fake also.

As I point out above, there is a lot of variation in Ft, hfe, and
leakage from transistor to transistor so it is quite possible that
Tektronix was selecting parts so they would work correctly in this
circuit. Personally I would have worked to make the circuit more
robust but maybe that proved to be impracticable although I outlined
in an earlier post changes Tektronix made in this design which seem to
indicate they did manage to eventually do this.

P.S. Regarding startup, curiously enough... (and I can't explain how it is
like it is)... Instead of an initial H.V. overshoot (due to the slow time
constant of the feedback, as earlier mentioned by David), when we look at
the Q1486 collector's waveform while turning-on, we see the AC voltage rise
slowly (in about 1.5 seconds) to about 1/2 of the peak-to-peak voltage of
the normal regime... and it stays at this apmlitude for a about 1/2
second... only then, it increases (quicker this time), to the final
voltage, which is about 30Vpp, with the lower peak at +9V and the higher
peak at +39V (as can be seen on the pictures).
By looking at this, it seems that there's something else at play during the
startup, that actually limits the oscillation amplitude, before the control
circuit can kick-in.

Rgrds,

Fabio
If collector to base leakage is occurring like I suggested above, then
this might be starting the oscillator at a low level until the
regulator loop can catch up.

There is an excellent discussion of startup problems with these types
of inverters written by Jim Williams in Linear Technology application
note 55. Page 14 shows an example of the destructive overshoot I
mentioned:

http://www.linear.com/docs/4144

The examples on page 16 show the behavior you describe with the
inverter starting at a low level and then quickly changing to full
power.

Incidentally, one of the real uses for a DSO with peak detection like
a 2230 or 2232 is to capture the startup behavior of a power supply or
inverter. An analog storage oscilloscope can do it also but is not
quite as convenient.

Fabio Trevisan
 

Hello David,
Thanks for your always comprehensive and clarifying answers... I`m learning
a lot.
My comments next to yours...
Rgrds,
Fabio
2016-11-08 10:01 GMT-02:00 David @DWH [TekScopes] <
TekScopes@...>:

A shorted or partially shorted transformer might make the waveform
look cleaner than it should be.
I can understand that... but following that path invites me to believe the
"not so clean" signal as it is from the MJ15015 currently installed would be
correct, but from your comments further down, I don't think it's the case.

Thanks for posting photos of your results and photos of the
transistors you used. They make working on the problem much easier.
It took me a while to understand the Yahoo interface. It's so unintuitive!
I only wish I had measured emitter current back when the circuit was
original (before replacing both the transformer and Q1486)...
But back then I was still looking for reasons why the HV was collapsing
after warming, and none of the initial voltage waveforms (collector and
base)
abnormal so to make further measurements.

Well, you have stumped me so I am going to throw some ideas out. The
transistor should only be conducting during the part of the cycle
where the Vce is low and any spurious oscillation should only occur
during that time.
Well, I was stumped too... but I may have missed to clarify one thing.
When I added the miller integrator (220R in series with base and 1nF
from base to collector of Q1486), I did it only for the MJ15015.
The waveforms on the 2N3772 was so "off of the beaten track" that I
didn't insist on it besides the minimum necessary to take the pictures.
In complement to the story, only one thing made sense with the
2N3772, was that I needed to change R1483 back to 5K6 in order for it
to oscillate properly....

I checked the ST datasheet for the 2N3772 and they are indeed slow so
I have little idea where the 520 kHz ringing is coming from or why the
slow 2N3772 would produce it but the also slow MJ15015 does not. But
I do have some ideas

Is it possible that the 2N3772 is counterfeit?
When I saw those waveforms with the 2N3772, the first thing I did too was to
check ST's datasheet, just to find out that it's basically the same specs
as the
ON part. (so, it shouldn't explain those spikes).
I also think it's a counterfeit as it couldn't be switching that fast as
shown by the
waveforms if it would be 200KHz fT.
Brazil lose almost completely its native electronics industry, so there's
little
business for electronic parts resellers and the reputable ones don't serve
the
hobbyist market (they only sell wholesale, for the few remaining specialized
industries).
Even Farnell / Element 14, opened business here in the 2010s and closed
operations on 2015.

The ON MJ15015 datasheet says that it is effectively a 0.8 MHz
super-premo-deluxe 2N3055 with tighter specifications so it would be a
good choice for a replacement anyway. I did not actually know they
made such a part. I will have to add it to my list. Mouser says it
is a 6 MHz part which is misleading (see next paragraph); sometimes
they list the minimum and sometimes they list the maximum Ft making it
difficult to compare parts just based on their selection guide and
that is why I missed it.
I also wasn't aware of it... I knew the famous MJ15003/4 (also
counterfeited a lot).
I got to it searching for 2N3055 replacements on the net.
Interesting enough is that ON lists the 2N3055 itself on the same dataheet,
as
well as there's a datasheet of the 2N3055 alone.
The highlight of the MJ is that it's an 2N3055 with a better S.O.A.

The variation of Ft from part to part is of course wide just like the
variation in hfe. The ON MJ15015 datasheet lists a range of 0.8 to
6.0 MHz and I have no reason to think the 2N3772 does not vary just as
much. This of course just point to why Tektronix may be been
selecting parts.
In the case of this 2N3772 in particular, I can only think it's a
counterfeit
Or I was very lucky to get the part with 6MHz (the spikes at the emitter
current shows about a full turn on-off cycle of less than 2us).
If I would be shopping for 6MHz, I would probably find only the slow ones.

Another thing that occurs to me is if the 2N3772 shows a spurious
oscillation during the whole cycle, maybe it is actually conducting
during the whole cycle. Try adding a 0.8V / 2.5mA = roughly 320 ohm
parallel resistor between the base and emitter to draw off any leakage
and stored charge. If that makes any difference, then find the value
where it just barely stops the problem and then halve it.
Indeed, but the middle trace, which is emitter current, shows almost 0
current during most of the cycle...
I think that most of the ringing we see is from the transformer itself,
and since the transistor is not conducting to dampen it, it keeps ringing.
Still, I will try the base transistor (to ground)... not with the 2N3772 as
I already put the MJ15015 back.
And since also the MJ doesn't show any emitter current through most
of the cycle, it reinforces my thinking that the transistor is not
contributing
to the sustaining of the ringing... only to starting it.
I wonder if there's any snubber that I can add, that could kill that 520KHz
energy without compromising the oscillation at the fundamental...
Their frequencies are more than 10 times apart, so it should be possible.

If leakage is what is causing the problem, then this may be what
Tektronix was actually selecting parts for; leakage is an easy test to
do.

I do not consider the spurious oscillation acceptable. Even if the
circuit works, it indicates marginal operation.
See above for my suggestions about circuit operation and
modifications.
We shall see... I will try the resistor from base to ground and see what
comes out.

Ya, I wondered about whether the 2N3772 is fake also.
Regarding this 2N3772, in view of the Brazilian market... I`m almost
positive
about that.
I`ll try to get one from a different source, as a "second opinion"...
Preferably one
that is not from ST, as if those are fake, the market may be flooded with
that
particular fake.

As I point out above, there is a lot of variation in Ft, hfe, and
leakage from transistor to transistor so it is quite possible that
Tektronix was selecting parts so they would work correctly in this
circuit. Personally I would have worked to make the circuit more
robust but maybe that proved to be impracticable although I outlined
in an earlier post changes Tektronix made in this design which seem to
indicate they did manage to eventually do this.
I wonder if I can implement some of those improvements to this one.
What particular scope you know that this circuit seems less dependent
on the transistor characteristics, or in which Tek didn't use "selected"
parts?

If collector to base leakage is occurring like I suggested above, then
this might be starting the oscillator at a low level until the
regulator loop can catch up.
Let's see what the base resistor will do...

There is an excellent discussion of startup problems with these types
of inverters written by Jim Williams in Linear Technology application
note 55. Page 14 shows an example of the destructive overshoot I
mentioned:

http://www.linear.com/docs/4144

The examples on page 16 show the behavior you describe with the
inverter starting at a low level and then quickly changing to full
power.

Incidentally, one of the real uses for a DSO with peak detection like
a 2230 or 2232 is to capture the startup behavior of a power supply or
inverter. An analog storage oscilloscope can do it also but is not
quite as convenient.
I've read quite a few of Jim William's ANs, but I missed this one.
I`m going to take a look at it, to see if I can learn a little bit from
and contribute more to the discussion... :-)



bobh@joba.com
 

It’s probably not much help for the high frequency oscillation issue but the Power Supply Circuits book (one of the Tek Circuit Concepts books) on page 122 discusses a similar circuit. The .pdf version is on TekWiki.
Bob.
From: Fabio Trevisan fabio.tr3visan@... [TekScopes]
Sent: Tuesday, November 08, 2016 2:39 PM
To: TekScopes@...
Subject: Re: [TekScopes] Re: Hello from newcomer Fabio Trevisan - My first Tek Scope 464 + DM44


I do not consider the spurious oscillation acceptable. Even if the
circuit works, it indicates marginal operation.
See above for my suggestions about circuit operation and
modifications.
We shall see... I will try the resistor from base to ground and see what
comes out.


[Non-text portions of this message have been removed]

Fabio Trevisan
 

Hello Bob,
Thanks for the tip. I was familiar with the Concept series of books but
never had looked into this one (I don't know why but in my twisted mind, an
H.V. power supply doesn't fit in the "power supply" bill :-) and this one
never got my attention)
It was interesting to see the description of the role of the capacitor, and
how the voltage coming from the feedback circuit, controls the discharging
of this capacitor, as much as, by increasing the drive to the output's
transistor base, affects also how the capacitor is charged (indirectly, via
transformer's feedback winding).
Still, the circuit of the 464 is a little bit different in that area and -
at a first glance - I don't see the only capacitor there (C1483) playing
the same role... but I must be mistaken... I confess I'm sort of "blind"
following up negative charging currents and positive discharging currents.
It's like a short circuit to my mind.
It sounds silly, but sometimes I draw a "negative" a circuit upside down,
using everything in reverse (transistors, diodes, etc...), to see currents
flowing in the "normal" (to me) way, so that I can understand it.




2016-11-08 22:03 GMT-02:00 'bobh' bobh@... [TekScopes] <
TekScopes@...>:



It’s probably not much help for the high frequency oscillation issue but
the Power Supply Circuits book (one of the Tek Circuit Concepts books) on
page 122 discusses a similar circuit. The .pdf version is on TekWiki.
Bob.
From: Fabio Trevisan fabio.tr3visan@... [TekScopes]
Sent: Tuesday, November 08, 2016 2:39 PM
To: TekScopes@...
Subject: Re: [TekScopes] Re: Hello from newcomer Fabio Trevisan - My first
Tek Scope 464 + DM44

I do not consider the spurious oscillation acceptable. Even if the
circuit works, it indicates marginal operation.
See above for my suggestions about circuit operation and
modifications.
We shall see... I will try the resistor from base to ground and see what
comes out.
[Non-text portions of this message have been removed]



[Non-text portions of this message have been removed]

 

On Tue, 8 Nov 2016 20:39:51 -0200, you wrote:

2016-11-08 10:01 GMT-02:00 David @DWH [TekScopes] <
TekScopes@...>:

Well, you have stumped me so I am going to throw some ideas out. The
transistor should only be conducting during the part of the cycle
where the Vce is low and any spurious oscillation should only occur
during that time.
Well, I was stumped too... but I may have missed to clarify one thing.
When I added the miller integrator (220R in series with base and 1nF
from base to collector of Q1486), I did it only for the MJ15015.
The waveforms on the 2N3772 was so "off of the beaten track" that I
didn't insist on it besides the minimum necessary to take the pictures.
In complement to the story, only one thing made sense with the
2N3772, was that I needed to change R1483 back to 5K6 in order for it
to oscillate properly....
The network I would experiment with is a series connected capacitor
and resistor between the collector and base.

As I point out above, there is a lot of variation in Ft, hfe, and
leakage from transistor to transistor so it is quite possible that
Tektronix was selecting parts so they would work correctly in this
circuit. Personally I would have worked to make the circuit more
robust but maybe that proved to be impracticable although I outlined
in an earlier post changes Tektronix made in this design which seem to
indicate they did manage to eventually do this.
I wonder if I can implement some of those improvements to this one.
What particular scope you know that this circuit seems less dependent
on the transistor characteristics, or in which Tek didn't use "selected"
parts?
The 465B and 468 may have been the last oscilloscopes which used this
inverter design but the slightly earlier 455 and 465M appear to use
the most advanced version.

They use a high Ft transistor and 100 ohm base-emitter shunt. There
are other aspects of the circuit though which make me suspect that
overshoot protection was needed.