Tek p5205 HV differential probe teardown. BTW, what are the red and brown wires?

[onlooker]

Well, there is not much to teardown other than showing 2 internal pictures, then ask a few questions and ideas.

As it stands now, a tek p5205 differential probe (100MHz, 1000V, 50X-500X) in working order may cost several hundred dollars even in the second hand market. The 1103 probe power supply will cost a additional several hundred dollars.

The alternative for someone with low budget is to get the ones that occasionally pops up on eBay that had all cables cut and without any accessories. But, it can go about $30~$50 a piece, or may be even lower.

I opened one of such p5205. It looks like it used all off-the-shelf parts. The design also looked simple in contrast to the high original cost. I guess, the cost went into the professional design and the professional specs. 

To reconnect the necessary wires, I tried to find  the schematics, but, I failed on it. It will be nice if someone can share the schematics. Otherwise, I will continue to understand the wiring from the PCB, do some testing and hopefully not to damage anything.



Edit: bottom picture added.



The back side of the PCB is not very interesting. The 1st picture is the upper side of the PCB. The HV wiring (left side) and the coax wring on the LV (right) side are simple to redo.

The thin Black, Blue, Red and Brown wires are my concerns.  I have identified that the Black and Blue wires are the -+15V main power supplies.  When the -+15V are on, I can hear a brief beep and LED flash.

More importantly, when one of the HV wire was  touched, the LV coax showed the usual distorted mains waveform. This says that the device should be still functional.

Currently I am trying to understand the roll of the Red and Brown wires. According to the instruction manual, it is likely related to input of the offset adjustment voltage. But, why two wires. Anyone know the answer?

The official power supply 1103 also supplies +-5V. But they are not likely used in p5205 since p5205 have circuitry to split/convert the +-15V supplies to a pair of +-5V supplies.

I am also thinking to reduce the attenuation from  50x-500x to 10x-100x since I am really just interested in the differential capability, not the HV capability. Any suggestions?

As to the +-15V power supply, is switch mode supply good enough?

The 2nd picture is with the HV side wired with the wires taken from a old CRT monitor.

 

[MarkL]

One of the wires is probably the input to the scope which tells the scope what attenuation factor to use for the channel, x50 or x500.

Look for a fixed change in resistance to ground when you push the attenuator selector button.

[MarkL]

Ok, I popped the lid on a P5205 "compensation box".  First off, that's in quotes because there's no compensation in it.  The coax goes right into the scope.

Brown goes to the Tekprobe "Data" pin, which is the attenuation sense input (aka  "readout" pin).  It flips between 690 and 1140 ohms for x500 and x50 settings, but the probe needs to be under power for it to work.

Red goes to the center wiper of a 1-turn 10k pot, and the ends of the pot go to +15 and -15.  This is the offset adjustment accessible through the small hole in the compensation box

You are correct on the blue +15V and black -15v.

(Edit: Fix typo.)

[onlooker]

Excellent and clear.  And thanks MarkL.  The next time google will be friend again for this info.

For my power supply, I can now forgo the brown wire. It needs only 4 wires: one for GND, two for the +/-15V rails, and  plus a wire for -1V~+1V offset adjustment. A piece of phone cord should work; A coiled cord can even wrap around the coax.
 
I think there is also a pot on the p5205 PCB that does the offset adjustment during calibration. if I am willing to deface the device to expose the pot,  I can further omit the offset adjustment wire to the power supply.

To answer my own question about using switch mode supply, yes, I googled, in its TekVPI to TekProbe adapter, Tek was already using switching power to convert a 12V single rail to +/-15V dual rail. It is only a metter of if I can find one that works noise-wise. It should be no suprise since switching powers are used in powering scopes for some time.

[128er]

Hi,
bought four of the P5205 probes a few weeks ago for 50€. Now I'm planning a little powersupply like the Tek 1103.

Almost all Informations are given in this thread. Nevertheless I add my hand drawn schematic to this thread.

IIRC, the probe draws about 100 mA per rail (+15V/-15V).
   

[onlooker]

You certain got better deal. 

I did further test on mine. It passed the "Functional Check" (probing mains as described in the manual). And yes, it is a good news that the power supply is not demanding. Mine draws 71mA steady.

Thanks for sharing the drawing.

[128er]

Hi,

would that be an adequate way to detect the attenuation over the data pin? Or is it to complicated, or did i miss something?

Edit: VDD is +15V

Edit²: Could I use an LM324 for the window comparator and the CC source instead of real comparators and an extra opamp? Would reduce the number of packages

[David Hess]

The old Tektronix designs essentially did use a window comparator but with a lot more windows; the signals themselves used 15 volts of bias to generate 0 to 1 milliamps with 0.1 milliamp resolution.

Later more resistance values were added to handle attenuation factors other than x10, x100, and x1000.

[MarkL]

Hi,

would that be an adequate way to detect the attenuation over the data pin? Or is it to complicated, or did i miss something?

Edit: VDD is +15V

Edit²: Could I use an LM324 for the window comparator and the CC source instead of real comparators and an extra opamp? Would reduce the number of packages

I didn't analyze your circuit in detail, but I think your approach would work.   Since you're only trying to differentiate between two resistance values, I think it could be simplified.  And it could be really trivial if you use a comparator with a push/pull output, such as the TLC3702.  See below (caveat: NOT built/tested).

The real utility of the attenuation sense pin is to allow the attached scope to automatically adjust the V/div so that cursors and other measurements appear correctly.  Lighting an LED can certainly be done instead, but isn't that a lot of extra trouble when you can glance over at the x50/x500 button?

[128er]

 And it could be really trivial if you use a comparator with a push/pull output, such as the TLC3702.  See below (caveat: NOT built/tested).

Thanks a lot for your input! The push pull solution is nice and would save me a comparator. But the TLC3702 and the LM393 comes with two comparators either way. So it's not that important to reduce the circuit to one comparator. Maybe just for kicks or minimalism.

The real utility of the attenuation sense pin is to allow the attached scope to automatically adjust the V/div so that cursors and other measurements appear correctly.  Lighting an LED can certainly be done instead, but isn't that a lot of extra trouble when you can glance over at the x50/x500 button?

I had the same thought while I planed the circuit. In the end my thought was: If I have the possibility to read out the data pin, then why not. 
It's not a lot of extra trouble and cost. Furthermore, the probes have long cables and can lay anywhere on the bench or in a switching cabinet. It's nice to have LED's in the power supply housing as a central point for fast indication.

One more thing...

After I saw your schematic with the voltage divider for the reference voltage. That was my first idea too. But the reference voltage depents on the supply voltage and it could be unstable. With my circuit i try to sense +- ~250 mV.
Is the solution with the zener regulator as a reference overkill, seen from a practical point? The 7815 should regulate well enough, that it won't have any effect on the circuit.

Probably I put too many thoughts at such a small thing.

[Mechatrommer]

I am also thinking to reduce the attenuation from  50x-500x to 10x-100x since I am really just interested in the differential capability, not the HV capability. Any suggestions?

if you are well versed in "dividers matching" and "frequency compensation" then its possible. all you have to do is replacing the dividers in HV left section with "quality" components of the same sizes. otherwise dont do it, messing with attenuation is messing with differential capability/performance.

As to the +-15V power supply, is switch mode supply good enough?

if it doesnt screw the noise and CMRR performance.

[David Hess]

I am also thinking to reduce the attenuation from  50x-500x to 10x-100x since I am really just interested in the differential capability, not the HV capability. Any suggestions?

if you are well versed in "dividers matching" and "frequency compensation" then its possible. all you have to do is replacing the dividers in HV left section with "quality" components of the same sizes. otherwise dont do it, messing with attenuation is messing with differential capability/performance.

Especially do not do it unless you are prepared to recalibrate the AC and DC common mode rejection.  Of course you may have to do this anyway if you change the probe leads.

You may be able to adjust the attenuation on the output side which is low impedance but I would sure want a schematic before trying it which seems like a lot of work for a minor convenience.  Why not just set the oscilloscope vertical input to a higher sensitivity?

[MarkL]

One more thing...

After I saw your schematic with the voltage divider for the reference voltage. That was my first idea too. But the reference voltage depents on the supply voltage and it could be unstable. With my circuit i try to sense +- ~250 mV.
Is the solution with the zener regulator as a reference overkill, seen from a practical point? The 7815 should regulate well enough, that it won't have any effect on the circuit.

Perhaps it might not be obvious as drawn, but it's really just a classic measurement bridge.  It's not dependent on supply voltage variation.

For example, if the +5V supply goes up 10% and throws off the 2.5V reference voltage by +10% on the - input, the measured node into the + input is also up by +10%.  They track each other.  The only things critical to this design is the tolerance of the resistors, and the offset voltage of the comparator (TLC3702 = 6.5mV max).

You can work out the exact requirements for the tolerances, but we're trying to detect an almost +/- 25% change in resistance on the brown-to-GND leg of the bridge.  There's a lot of leeway.

One thing not obvious is that this circuit is operating off +5V and not +15V (so, yes, there's regulator not shown).  5V operating voltage is on purpose because from testing we know the resistance on the brown wire is not valid unless the probe is powered up.  That tells you instantly there's some active electronics also on that brown wire and not just a switch with a couple of resistors.  On a TDS3054, the brown pin was measured as +5V, so that's the only thing I'm comfortable with, without knowing what else is on the brown wire.  Your current source has a compliance voltage > +5V on that pin.  I don't know if that's safe or what effect it might have on the probe internals.

In fact, that contact on the TekProbe interface, along with another, make up the connections for 5V I2C communication with some of the more expensive probes.  That's why the connection is called "Data" (and the other "Clock").  The pin is multiplexed by the scope between the two functions of I2C and attenuation sense.

And one additional thing, as you point out the TLC3702 is a dual comparator.  If you're not using the other half you should connect one input high and the other low.   It has extremely high input impedance and it's probably pretty easy for it to oscillate with open inputs.  It's generally good practice to tie levels on unused gates on any chip anyway.

Probably I put too many thoughts at such a small thing.

Sometimes the fun is in the designing and building, more than the final outcome!

[128er]

On a TDS3054, the brown pin was measured as +5V, so that's the only thing I'm comfortable with, without knowing what else is on the brown wire.  Your current source has a compliance voltage > +5V on that pin.  I don't know if that's safe or what effect it might have on the probe internals.

That's a huge no go for +15V. Didn't thought of that, or missed it in the thread. Thanks again. One more power rail required. One thing leads to another . . .

[NiHaoMike]

Put in a zener to prevent that pin from going above 5V.

[onlooker]

Thanks for all the suggestions and comments. I think I will do the easy part first, that is, to  do minimum rewiring  to make it usable again.

The first thing is the coax. I have just received some eBay version of RG174. Its small diameter is the reason for the choice. I need to have both the coax and the power wires go through the existing outlet.

From an online coax loss calculator, for a run of 5ft of generic RG174, the loss is about 0.1dB@10Mhz and 0.4dB@100MHz (comparing to RG58 with <0.1dB@10Mhz and 0.2dB@100MHz). 

Is a generic RG174 good enough or I need to have a low loss coax? Additionally, is this 0.1dB vs 0.4dB difference enough of a concern?

The other main part is the power. I did more tests. For a 50Mhz signal, the current consumption of the device is already more than 200mA, which is significantly more than that for a low frequency signal that I checked earlier.

Lastly,  two pictures are attached for the entertainment. The pictures are just quick checks of 10MHz aquare and triangle waveformss before (yellow) and after (red) P5205. They are not with proper termnation or best possible cabling; the square one looks particularly ugly.

[Mechatrommer]

From an online coax loss calculator, for a run of 5ft of generic RG174, the loss is about 0.1dB@10Mhz and 0.4dB@100MHz (comparing to RG58 with <0.1dB@10Mhz and 0.2dB@100MHz). 

no. by the right mind, you should concern more on the impracticality of the bigger RG58, which you cant do pretty much anything about. the major artistic side of it is "to make it work" by hook or by crook, 58 or 174. i think you are fine with 174, just recompensate if the result is not good, thats the biggest part anyway.... knowing your stuffs

edit: btw thanks for the pcb picture of the probe. if you can post the back side of it i will appreciate it. from the signal capture... triangle wave got roll off at peak and bottom, thats normal due to BW limitation, at 10MHz, triangle still got so many harmonics which your probe cant show. second picture... too much ringing on both red and yellow signal. i'm suspecting reflection (or black magic) on yellow stuff. not sure how you test it.

wait a minute, the yellow is original triangle signal and red is probe output? you have overcompesated probe redo again with proper termination.

[onlooker]

Ok, a picture of the bottom side is just added to the 1st post. The PCB has visible middle layer(s).

Yes, I will try RG174 1st, though the little leftover of the original cable shows it was more close to RG58 in size.

And yes, the P5205 output is closer to the true waveform shape; the yellow traces were taken before P5205 with an x10 probe having its tip touching one of the input terminals of P5205.

And yes, I found the problem, I did not use the probe GND connection since I  mistakenly thought the GND connection was already there.  In fact, it was indeed there, but it was in the form of a big ground loop going from the FG GND to the scope GND. This loop dumped the yellow traces and caused the delay of the yellows. 

After connecting the probe GND to another one of the input of P5205, the views are much better. Again the yellows are at the input of P2505; and the reds are at the output of P2505. Now the yellows are ahead of the reds.





Anyway, I think, if nothing else, these pictures showed that 1). the waveforms are centered well enough without needing any additional offset compensation (though for small signals, the offset is more pronounced); and 2). the amplitudes are roughly right.

BTW, I did not like the forced mental math with the x50/x500 attenuation settings. This is also a good motivation to change it to x10/x100 later. 

[MarkL]

BTW, I did not like the forced mental math with the x50/x500 attenuation settings. This is also a good motivation to change it to x10/x100 later.

Does your scope not support manually entered attenuation factors?  Surely it must have x1 and x10, but not x50 and x500?  (Sorry, I don't recognize the make/model from your screen shots.)

[onlooker]

Right, it has only x1, x10, x100 and x1000. It is the now "infamous" Owon SDS7102. Ok, I can do  divide by 2 or multiply by 5.

[MarkL]

You could go in the other direction and make it x100 and x1000 by doing a simple divide by 2 (with compensation) on the coax side to the scope.

If your plan was to modify the input side of the probe, this alternative preserves the CMRR and other input characteristics.

[Mechatrommer]

agreed, making divider on the output is much less risk. looking at your picture, 50X is unterminated, simply adding proper 50 ohm termination at dso end you'll get 1/100X output (@ 50X setting) and 1/1000X (@500X setting), set your dso to suit, no need mental math

[MarkL]

Yes, this works to a first approximation and a simple 50 ohm pass-through terminator can be used.  However, be aware of this from the probe specs:

Output Type: Single-ended. Source Impedance of 50 ohms drives 1 M ohm oscilloscope input. Load impedance must be greater than 50 k ohm for stated accuracy.

So, a compensated divide by two attenuator with an impedance greater than 50k is needed if you want to stay accurate.  You could use a couple of 25k resistors as a divider, or you might even be able to get away with a single 1M and rely on the 1M inside the scope.  In either case you'll need an adjustable cap for compensation across the resistor, around 5 to 25pF.  If you put the arrangement in the probe box instead of at the scope you'll need even more pF to compensate for the cable.

There's no information I can find that says exactly *which* specs are invalidated if you go with 50 ohms (I suspect accuracy with max signal input), so in reality the above may be overkill for your purposes.  I would say try the 50 ohm termination with various signals that you typically measure to see if it works well enough.

(Edit: Ohms symbol didn't make through cut/paste from specs.  Changed  to "ohms".)

[David Hess]

You could go in the other direction and make it x100 and x1000 by doing a simple divide by 2 (with compensation) on the coax side to the scope.

So, a compensated divide by two attenuator with an impedance greater than 50k is needed if you want to stay accurate.  You could use a couple of 25k resistors as a divider, or you might even be able to get away with a single 1M and rely on the 1M inside the scope.  In either case you'll need an adjustable cap for compensation across the resistor, around 5 to 25pF.  If you put the arrangement in the probe box instead of at the scope you'll need even more pF to compensate for the cable.

What I find really odd about this discussion is that such a device used to exist!

Oscilloscope input "normalizers" are used to adjust the input capacitance of the vertical inputs of an oscilloscope as well as the compensation of the input attenuators but these normalizers also function as compensated precision divide by 2 attenuators.  I have a couple of them in my collection of calibration instruments.

There's no information I can find that says exactly *which* specs are invalidated if you go with 50 ohms (I suspect accuracy with max signal input), so in reality the above may be overkill for your purposes.  I would say try the 50 ohm termination with various signals that you typically measure to see if it works well enough.

The accuracy of the 50 ohm termination will degrade the calibration.  The high voltage differential probes I am familiar with specify two different attenuation factors depending on if their output is terminated into 50 ohms or not.

[MarkL]

What I find really odd about this discussion is that such a device used to exist!

Oscilloscope input "normalizers" are used to adjust the input capacitance of the vertical inputs of an oscilloscope as well as the compensation of the input attenuators but these normalizers also function as compensated precision divide by 2 attenuators.  I have a couple of them in my collection of calibration instruments.

I figured there had to be; this didn't seem like such an unusual situation.   But I couldn't find any to say, "Here, buy this!"  Thanks for the pointer.  Is there a compensation adjustment on a "normalizer"?

There's no information I can find that says exactly *which* specs are invalidated if you go with 50 ohms (I suspect accuracy with max signal input), so in reality the above may be overkill for your purposes.  I would say try the 50 ohm termination with various signals that you typically measure to see if it works well enough.

The accuracy of the 50 ohm termination will degrade the calibration.  The high voltage differential probes I am familiar with specify two different attenuation factors depending on if their output is terminated into 50 ohms or not.

Right.  I'm expecting the P5205 won't be able to swing to the peaks of its specified output with 50 ohms and the waveform will get squashed.