PG506 repair question.
I'm working my way through my PG506 (SN B031555), trying to restore full
functionality before I try and calibrate it.
I am working on the high amplitude mode right now, and have a few questions
for the group.
I have uploaded accompanying images (including a page of schematics) here:
1. The trailing edge of the output pulse is distorted (
https://groups.io/g/TekScopes/photo/265034/3242704) This goes away at
smaller periods, or lower amplitudes. I believe I know why this is
happening (see below), and would appreciate confirmation/rebuttal whether
this happens to other PG506s of a similar vintage.
It just doesn't seem like something Tek engineers would have allowed to
happen on their watch, even if the rising edge was the important one.
I believe this is happening because the way the design works is to change
the -72V rail, so that the voltage drop across the output transistors
(Q745/755) remains roughly constant, so at low output (Amplitude pot CCW)
the rail is at -15V (the output signal is around 3Vp-p.) As you increase
the desired amplitude, the rail voltage is adjusted to stay several volts
more -ve than the lowest output value. The -ve rail is internally generated
by an on-board 25KHz inverter, and regulated by adjusting the voltage
regulator that supplies the low-voltage side drivers of the inverter
transformer (via the output of U840.)
The circuit description states: "The net result is that the voltage drop
across the output transistors remains relatively constant." This, combined
with the description that C800 ("memory capacitor") tracks the -ve peaks of
the output waveform makes the scope trace make some sense. When the pulse
goes high, C800 starts discharging, and as a result the control loop is
raising the -ve rail. When the falling edge arrives, it is fine until it
meets -ve rail, then the -ve output swing is limited by the ability of the
on-board inverter to slew.
Either this is just the way the design works, which would be nice to know
from other people's experience with a similar PG506; or something is borked
in mine. The time constant of C800 seems reasonable, given it is such a
small value (10nF.)
2) However, I do seem to have other problems in this section - the max -ve
output voltage I can get is about -50V (into 1MOhm load) and the "-72V
rail" seems to max out around -60V. This when the Pulse Amplitude control
is at about 3 o'clock. Between 3 o'clock and 5 o'clock (max), no change in
either the output amplitude or the -72V rail. Note that in other operating
modes, the 72V winding does generate 72V. I have replaced all the tantalums
and also replaced the bulk caps on the -72V rail (C145/150/732) - no
change. Note that the inputs to U840 stop changing, so it is not the
inability of the low-voltage side of the inverter to respond.
I am curious if anyone has seen similar behaviour, or has ideas to try and
triangulate on the issue. I suspect there may be some connection with the
behaviour mentioned in (1) above.
3) I am having some trouble wrapping my head around the subtleties of the
way this particular circuit operates (despite multiple readings of the
theory of operation.) I am still struggling to understand exactly how
R785A changes the output swing - it seems like that whole section of the
circuit is kept several volts more negative than the desired output
amplitude, and so it is able to keep asking for "more", and the control
loop tracks that and stays ahead. I puzzled over Q758 for a while - which
is not mentioned at all in the manual - until I figured out it is not
present in my serial number build (nor is W700.)
If anyone has worked through the operation of this circuit ahead of me, and
can share any insights, it would be much appreciated. Hoping that someone
has seen (and repaired) the exact same behaviour seems way too much to hope
On a side note, I do enjoy the process of getting inside the designer's
heads over time. I find I have to study the schematics for a while, and
re-read the theory of operation multiple times - combined with poking
around with a scope and meter, slowly enlightenment can happen - but I'm
coming up empty this time. Sadly, the manual doesn't have sample voltages
or traces to help debug.
I have two PG506 units. Neither one shows a trace like this as far as I can remember.
Thanks for the confirmation - as I said, I struggle to believe the Tektoggle quoted messageShow quoted text
engineers would have permitted that behaviour.
However, I also struggle to think how the design avoids it, without some
very different time constants in the bulk decoupling on the -72V line,
and/or the "memory capacitor" (C800).
On Fri, Jun 11, 2021 at 6:02 PM Mlynch001 <email@example.com> wrote:
I have two PG506 units. Neither one shows a trace like this as far as I
On Fri, Jun 11, 2021 at 01:52 PM, Andy Warner wrote:
3) I am having some trouble wrapping my head around the subtleties of theQ758 or Q755 is missing in your unit?
is not mentioned at all in the manual - until I figured out it is not--
I don’t have this equipment but here is what I can tell from the schematic you uploaded: It is simpler if you assume the ground of the circuit is at the anode of CR734 (or bottom of R784). R785A slider voltage changes ~ 0.3V to 3V (referred to anode of CR734) depending on amplitude setting. Q784 and Q780 form a unity gain buffer, Vbe of Q736 and Vbe of Q745/Q755 cancel each other. At the end the voltage across R745 and R748 is about same as the voltage at the amplitude control pot. Q760 provides ~ 6.7mA of current. When output is active (Q730 is off) Q745/Q755 sink current in the range of 0.3V/23.5ohm to 3V/23.5ohm total depending on amplitude setting. This current minus 6.7mA ends up on the load (R805//load at BNC). If terminated to 600-ohms at BNC, the output range is -300ohm*(0.3V/23.5ohm-6.7mA)= ~ -1.82V to -300ohm*(3V/23.5-6.7mA)=36.3V depending on amplitude control and R790 trim. Minus sign is because Q745/Q755 sink current from the load.
Thank you for the analysis - I appreciate it, and it is very
helpful. The perspective that this design is all about current, and the
voltages just follow (as best they can) is a great help. You can make the
case that this is true of all circuits, but I fall into the trap of
thinking about designs as voltage-driven too easily - appreciate the fresh
My unit has Q745 & Q755, but does not have Q758, W700, C755 and some other
I spent some more time with the ‘scope looking at the actual behaviour, and
I believe that Q745 is dead (no voltage swing at it’s emitter) - and looked
at through the prism of a current-driven design, that might explain the low
Naturally Q745/755 are long obsolete GE D40E7 parts, which are now
unobtainium; I’ve done my best to pick some suitable modern replacements to
try (such as BD237G, in a TO-225 package), there will be fun because of the
differing pinouts and packages, but that is to be expected. I’ll upload a
photo of how tightly packed in this area of the PCB is, and how the tabs on
the GE transistors are folded over.
Onwards… I appreciate the help.
On Sat, Jun 12, 2021 at 03:02 Ozan <firstname.lastname@example.org> wrote:
On Fri, Jun 11, 2021 at 01:52 PM, Andy Warner wrote:Andy3) I am having some trouble wrapping my head around the subtleties of theQ758 or Q755 is missing in your unit?
I had similar problems, waveform and low output, with my PG506 in high amplitude mode, and spent many hours....well, OK, weeks, trying to understand what was going on in that circuitry...mostly due to my own ignorance. With a lot of help from friends, we ultimately concluded that the -72 V rail was not able to keep up with what was being asked of it. That, coupled with a pretty hot switching transformer got me down a wonderful rabbit hole of figuring out what was going on inside that switcher, including learning about magnetics and flux walking, which appeared to be happening on this switcher. In the end, my problem turned out to be the diodes in the -72V rail supply that had been replaced with standard slow recovery rectifier diodes, killing the performance of that supply, and really wreaking havoc with the switching circuits on the primary side. Fixing that took care of the low output, but I'm seeming to remember some of the distorted trace at very low frequency settings remained. A bit later today, I'll get it stuffed back in a TM and confirm.
I wouldn't expect that there would be two PGs out there that have had those diodes replaced with the wrong type, but maybe don't discount the -72 supply or even the switcher as a potential source of problems.
Thanks for the tip. In my unit, CR145-148 look original, these are listed
as WG1308/NDP566/PG2312 parts (again, seemingly mostly unobtainium.)
They test out OK, with a Vf of ~0.5V. What did you replace yours with
(assuming they were modern, available parts) ?
If you are willing to see if the falling edge of your high amplitude output
has a similar glitch at low frequencies, that would be awesome. The manual
talks about the rising edge being the “reference edge”, I’m still unclear
if that is supposed to imply that it’s the only edge worth caring about, or
just to show it’s relationship with the trigger output.
On Sat, Jun 12, 2021 at 10:20 Tom Norman <email@example.com> wrote:
With regard to your photo https://groups.io/g/TekScopes/photo/265034/3242704 I can confirm that my PG506 with S/N B032228 (very close to yours) shows exactly the same behavior. I know I have repaired something in the high amplitude circuits but don't remember exactly what the problem was. I can't guarantee the my unit is 100% in order.
The schematic changes are annoying. IIRC schematic 5 in the B04000-up manual is useful even for our lower S/N.
When you do internal measurements in High Ampl mode then don't forget that the supply is floating (also IIRC).
(continued) I'm glad that you showed a schematic for B039999_below from 1980, more recent than my paper manual from 1977. I downloaded the scan by Håkan. Now I found CR374 and indeed there also is a sawtooth-like voltage at "-72V".
Thank you for checking. If I understand you correctly, you see similar
artifacts on the trailing edge of high amplitude, low frequency output
signals. But neither of us are sure if this is an artifact of the design,
or just a common failure mode we both have.
Note that one of the additional late changes was the insertion of yet
another zener diode (VR792) that would have added another 3 volts of
headroom between the -ve rail and the sampled low output amplitude.
As an engineer, the proliferation of component and circuit changes,
followed by the abrupt redesign of the product suggests to me that there
are some mix of performance, reliability, maintainability issues
surrounding these first generation units. A colleague of mine used to call
these “dizzy boards” because of the number of board re-spins needed.
On Sun, Jun 13, 2021 at 07:40 Albert Otten <firstname.lastname@example.org> wrote:
(continued) I'm glad that you showed a schematic for B039999_below fromAndy
Thank you for checking. If I understand you correctly, you see similarYes.
But neither of us are sure if this is an artifact of the design,In my opinion it's all part of the game. The time constant R832xC800 is 1 ms, by far insufficient to maintain the "-72V" at 100 Hz output. The "reservoir" capacitors are 3x2 uF. As I understand it the Q760 and Q790 bias currents, together 6 mA, flow in the "-72V" anyway, plus the current via R154, < 1 mA. This alone gives a fall of slightly over 5 V during a half period of 5 ms. I may be overlooking loads. Your and my measurements show some 10 V in 5 ms.
However, under the same conditions (40 V pp, 10 ms period) the effect on the output is much less than in yours, hardly visible. The effect increases fast when I rotate VAR to larger period. I guess I get your situation at 20 ms period.
Note that one of the additional late changes was the insertion of yetIt might well be that mine has more what you call "headroom".
Despite the S/Ns are so close only my PG506 has that VR792 (according to the parts list)!
I am in a holding pattern until I get some new transistors for Q745/755 and
figure out how to squeeze them in, I will report back. I posted some photos
of how crammed in this part of the circuit is (and they kept updating the
design to add yet more stuff), and how the tabs of the original Q745/755
are bent over.
As always, I appreciate the thoughtful conversation and analysis from all
over the globe.
On Sun, Jun 13, 2021 at 12:34 Albert Otten <email@example.com> wrote:
Andy,AndyThank you for checking. If I understand you correctly, you see similarYes.
My PG506 is s/n B033xxx, so, unfortunately, the high amplitude circuit is not the same
as yours, though I think it is the very next revision (the change was at s/n B031980).
On mine, Q755 is gone, and Q745 is a Motorola 2N3866, with a finned heat sink.
Q758 shows up at the same time, attached to a rather substantial heat sink.
Having said that, here are some similarities: I can also reproduce your 10V negative-
going ramp (HIGH AMPL set to -40V), but I have to increase the period to 23 ms.
With a 10 ms period, this "ramp" is barely noticeable and less than 1V.
The HIGH AMPL maxes out at -62.0V, but this occurs with PULSE AMPLITUDE close
to the 3 o'clock position, much like yours. Minimum amplitude is 0.1V.
I think the positive edge is really the only one specified: under Electrical
Characteristics, High Amplitude Output section, Polarity is explicitly stated as positive.
Technically, even the signal you are seeing meets spec. What doesn't meet spec, of
course, is the limited signal amplitude, which should be at least 60V into 1 MOhm.
The only time I can remember using frequencies lower than 1 kHz with HIGH AMPL is to
drive the Overdrive Recovery Calibration Fixture (067-0608-00), and even that is with
frequencies higher than 200 Hz. Does your HIGH AMPL square wave have a better shape
at 1 kHz? Are you able to reach 60V then?
Since the PG506 is a calibration generator, not a general purpose square wave
generator, it seems reasonable that Tek might not have been too concerned with
the shape of an unspecified edge at uncommonly low frequencies. For what it's
worth, though, the high amplitude output of my Type 106 (which the PG506 was
designed to replace), has a nicely shaped negative edge, all the way down to
Before you removed the original transistors, did you measure the risetime of the
HIGH AMPL signal into 50 Ohm, and note the total aberrations of the
positive edge step?
I wanted to follow through on my earlier response. My PG is serial B035037, so similar to Dan's I think. It also shows the same negative going edge distortion you show in your posted picture at 100 Hz, and pretty well cleans up around 250 Hz. Output voltage into 1 Meg is 64V P-P, at maximum amplitude setting, across the frequency range of the instrument.
I too saw the reference to the rising edge being the "reference" edge, so after repairing the PS issue, I just accepted that they (Tek designers) didn't care too much about what was happening to the negative going edge at longer periods. It's good to know that at least two of you are seeing a similar thing. My ignorance is keeping me from understanding why the designers would have chosen to regulate the entire switching supply rather than just locally regulating a fixed -72 rail.
And on the -72 supply diode replacement, I ended up just trying some 1N914 diodes, which seem to have worked well. Also used those to replace CR80 and CR125, which had taken an absolute beating as a result of the strange switching transients that resulted from the "backflow" due to the slow rectifier recovery times. 1N4148 seems like it would work too.
Thanks for the additional verification, and advice on diodes.toggle quoted messageShow quoted text
Given the data points from the group, I plan to ignore the falling edge
artifact, and focus on getting the amplitude correct once I have
replaced D745/755 with modern parts.
On Mon, Jun 14, 2021 at 8:07 PM Tom Norman <firstname.lastname@example.org> wrote:
I have another off list confirmation that the falling edge of the hightoggle quoted messageShow quoted text
amplitude waveform, at least on first gen PG506's, routinely has an
aberration as shown in: https://groups.io/g/TekScopes/photo/265034/3242704
The exact period & amplitude it kicks in at varies between units, but for
the next person that trips over this: it seems like it is just part of the
design, and the rising edge is the one that matters for this particular
unit's role in life.
On Mon, Jun 14, 2021 at 9:14 PM Andy Warner <email@example.com> wrote:
Thanks for the additional verification, and advice on diodes.
The first round of candidates for Q745/755 in the PG506 high amplitudetoggle quoted messageShow quoted text
output section arrived today.
I have uploaded screenshots using the BD179G parts.
https://groups.io/g/TekScopes/photo/265034/3246294 is 10us period,
amplitude max, terminated into 1MOhm.
https://groups.io/g/TekScopes/photo/265034/3246295 is the same, but
terminated into 50Ohm.
The good news is that the full amplitude output is now restored, I think
Ozan's current-based analysis was right on the money here, and without the
ability to dump the current with both transistors, the full output voltage
swing was not possible.
Note the rise and fall times, which preclude the 1us period being any use
at all - the amplitude of the output signal is limited by the rise and fall
This is clearly shown in https://groups.io/g/TekScopes/photo/265034/3246296
I think I clearly need to look for faster parts than the BD179G. That has
an F(t) of 3MHz, while the venerable old D40E7 has F(t) of 230MHz.
I think I have one more chance at reworking the Q745/755 pads before I
start lifting traces.
As an aside, the aberration on the falling edge at low frequencies (which I
now accept as a fact of life) goes away when terminated in 50Ohm.
On Tue, Jun 15, 2021 at 11:44 AM Andy Warner <firstname.lastname@example.org> wrote:
I have another off list confirmation that the falling edge of the high
Final update from me (I hope.)
Looking for modern replacements for the GE D40E7 used in Q745/755 did not
result in a direct modern cross that I could identify from non-exotic
= 80V, >= 2A, F(t) > 200MHz, C(ob) ~9pF does not seem to be the sweet spotfor modern BJTs, nor is the TO-202 case, and more importantly pinout.
I found a way to make modern TO-225A packages work physically, but
eventually settled on the CEN-U07 from Central Semi (available from Mouser
This is a TO-202 package, so they drop right in and I even get to bend the
tabs over like Tek did originally (pro-tip, support the tab at the package
body with needle nose pliers when bending it, to avoid breaking off parts
of the case.)
The F(t) of the CEN-U07 is only 50MHz, and C(ob) is 30pF, so rise and fall
times are not stellar. At the 1us period, the output does indeed swing the
full amplitude, but the signal is not a great square wave (
If anyone has alternate recommendations for a better modern part to use,
that is not made of unobtainium, please let me know, otherwise I recommend
the CEN-U07 as the next best thing, if like me one of your output
transistors is fried (symptom was low max output voltage.)
Additional photos and scope traces are shown in the album:
Next up - aligning and tuning up the unit.
Thanks again for the help and encouragement, both on and off list. I hope
this might be useful to someone who needs to follow a similar path.
On Tue, Jun 15, 2021 at 10:21 PM Andy Warner <email@example.com> wrote:
The first round of candidates for Q745/755 in the PG506 high amplitude--
Yes, transistors with those specs definitely fall into a black hole. The MPS-U06 would be a very good cross, but it, too, is pretty much unobtainium these days. I see that Mouser still sells them but the minimum quantity is 500 pieces (and they quote a very long lead time).toggle quoted messageShow quoted text
In this circuit, you might be able to get by with a 2N2219 selected for higher than normal breakdown voltage. Both Q745 and Q755 are driven from a fairly low impedance, so the relevant breakdown voltage spec is BVCBO. Finding 2N2219s with >80V BVCBO is not unheard of, so if you aren't happy with the CEN-U07/MPS-U07, you may have other options.
Just a thought.
Prof. Thomas H. Lee
Allen Ctr., Rm. 205
350 Jane Stanford Way
Stanford, CA 94305-4070
On 6/17/2021 20:38, Andy Warner wrote:
Final update from me (I hope.)
Michael W. Lynch
On Fri, Jun 18, 2021 at 12:20 AM, Tom Lee wrote:
I know how much everyone here seems to rail against E-Bay; but there are several lots of these transistors listed on that platform. There are a couple of lots that appear to be NOS MPSU06 Motorola and the price for 2 pieces is not outrageous. To my mind, better to use the best available part than to try to find some substitute part that will be marginal at best. But that is just my opinion. What am I missing?