How good is your scope? (problem with capturing offset signals)

Hi

So I wanted to check the VCE of a transistor in a small converter
running at 100kHz, expected to see about 100mV above ground as the
saturation voltage of the NPN, increasing linearely since magnetising
current increasing.

I saw nothing like that, debugging and found the scope (Tek DPO5000)
somehow saturates with high voltage inputs. The signal is a square
wave of 100kHz, 25V peak to peak, offset so low voltage is close to
GND (measured by a 10:1 probe)

See plot (green trace):

www.electronicsdesign.dk/tmp/2.jpg

Low value is about 400mV

Ok, changed the scale div from 3V/div to 2V/div then this:

www.electronicsdesign.dk/tmp/1.jpg

Now low value is about 2.4V !

I checked it was not my setup and all is fine.

So I then hooked up the HP33120A function generator, with a BNC
directly to the scope, now the problem still exists, but just at
another volt/div setting (probably since the probe P500 has internal
scaling etc). The signal, square wave, 9Vpp, 4.5V offset, 100kHz) (a
levelshifted 9Vpp signal above GND)

My suspicion is that the input stage of the scope is saturated since
the top level of the signal is way out of bounds. If I reduce the
frequency to 1kHz, then the signal starts to look good again, which
tells me the input stage of the scope has settled and recovered from
the overdrive.

Then for a sanity check I hooked up a Tek TDS3032B and it showed the
correct signal all the way down to 20mV/Div setting

Then I took a Tek DPO4000 scope, and it had the same fault, allthough
at a different volt/div setting

Tried with a TDS220 (low end scope), it has the fault beginning at
below 2V/div

Tried a PicoScope 3000 series, had the fault at below 2V/div

Tried a Agilent DSO 4000 series, had the fault at AFAIR 5V/div

Then dusted off my first scope, the Hameg 205-3, it worked all the way
down to 100mV/div :-)

To the question, do any of you guys know the front-ends of scopes and
if so is my assumption that the input stage is saturated correct (so
the amp takes time to recover)?

Seems I should throw out the 10k USD scope and bring my 1k USD Hameg
to work for some real measurements....

Regards

Klaus

Klaus Kragelund wrote:
> Hi
> 
> So I wanted to check the VCE of a transistor in a small converter
> running at 100kHz, expected to see about 100mV above ground as the
> saturation voltage of the NPN, increasing linearely since magnetising
> current increasing.
> 
> I saw nothing like that, debugging and found the scope (Tek DPO5000)
> somehow saturates with high voltage inputs. The signal is a square
> wave of 100kHz, 25V peak to peak, offset so low voltage is close to
> GND (measured by a 10:1 probe)
> 
> See plot (green trace):
> 
> www.electronicsdesign.dk/tmp/2.jpg
> 
> Low value is about 400mV
> 
> Ok, changed the scale div from 3V/div to 2V/div then this:
> 
> www.electronicsdesign.dk/tmp/1.jpg
> 

I believe you set it from 3V/div to 600mV/div for channel 4.


> Now low value is about 2.4V !
> 

Can't follow that but I think it's 400mV. Which is weird, assuming you
haven't changed the offset (or the scope didn't do that because it
thought you might have wanted to ...).


> I checked it was not my setup and all is fine.
> 
> So I then hooked up the HP33120A function generator, with a BNC
> directly to the scope, now the problem still exists, but just at
> another volt/div setting (probably since the probe P500 has internal
> scaling etc). The signal, square wave, 9Vpp, 4.5V offset, 100kHz) (a
> levelshifted 9Vpp signal above GND)
> 
> My suspicion is that the input stage of the scope is saturated since
> the top level of the signal is way out of bounds. If I reduce the
> frequency to 1kHz, then the signal starts to look good again, which
> tells me the input stage of the scope has settled and recovered from
> the overdrive.
> 
> Then for a sanity check I hooked up a Tek TDS3032B and it showed the
> correct signal all the way down to 20mV/Div setting
> 
> Then I took a Tek DPO4000 scope, and it had the same fault, allthough
> at a different volt/div setting
> 

That is rather surprising.


> Tried with a TDS220 (low end scope), it has the fault beginning at
> below 2V/div
> 

I've never trusted those things.


> Tried a PicoScope 3000 series, had the fault at below 2V/div
> 
> Tried a Agilent DSO 4000 series, had the fault at AFAIR 5V/div
> 
> Then dusted off my first scope, the Hameg 205-3, it worked all the way
> down to 100mV/div :-)
> 

Well, good old German quality :-)


> To the question, do any of you guys know the front-ends of scopes and
> if so is my assumption that the input stage is saturated correct (so
> the amp takes time to recover)?
> 
> Seems I should throw out the 10k USD scope and bring my 1k USD Hameg
> to work for some real measurements....
> 

In a nutshell, yes. Maybe not throw out the $10k scope though but I'd
send this info to Tektronix support. If they can't answer I'd ask the
sales rep who sold it to you to come out there.

I can't count the times where I've had to ask clients "Do you have a
real scope?" and out came a big old Tektronix boat anchor where people
took a step back when I pulled the power switch handle ... bzzzt ...
TUNGGGG ... blueish green traces showed up and ... tada ... fuzz could
be seen that the DSO simply was unable to show.

Here I have a Tektronix 7704, a Telequipment D66 and I also still have
my old Hameg that I had as a highschool student (but that doesn't get
used much anymore). I would not want to be without a real scope. They
don't make them like that anymore. So if your workplace does not have
something like a Tek 7000 series or a Tek 2465 I'd ask them to go on
Ebay and get one :-)

-- 
Regards, Joerg

http://www.analogconsultants.com/

On Mon, 26 Sep 2011 16:28:18 -0700 (PDT), Klaus Kragelund
<klauskvik@hotmail.com> wrote:

><snip>
>Seems I should throw out the 10k USD scope and bring my 1k USD Hameg
>to work for some real measurements....

Somehow this reminds me of the old Sorcerer in Aladdin crying
out, "New lamps for old.  New lamps for old," trying to
recover that special, magic lamp.

(Replace lamp with oscilloscope....)

:)

Jon

On 27 Sep., 08:25, Jon Kirwan <j...@infinitefactors.org> wrote:
> On Mon, 26 Sep 2011 16:28:18 -0700 (PDT), Klaus Kragelund
>
> <klausk...@hotmail.com> wrote:
> ><snip>
> >Seems I should throw out the 10k USD scope and bring my 1k USD Hameg
> >to work for some real measurements....
>
> Somehow this reminds me of the old Sorcerer in Aladdin crying
> out, "New lamps for old. =A0New lamps for old," trying to
> recover that special, magic lamp.
>
> (Replace lamp with oscilloscope....)
>
> :)
>
> Jon
>
>

Hmm, just found an app note from Tektronix on this:

http://cp.literature.agilent.com/litweb/pdf/5989-0068EN.pdf

http://www.kemt.fei.tuke.sk/predmety/KEMT434_RM/_materialy/Doplnkove%20zdro=
je/UnderstandingOscilloscopeBandwidth.pdf

Overdrive seems to be the problem

However it is typically specified as less than 20ns, i am seeing a lot
more. But perhaps thats due to excessive overdrive. Funny enough the
DPO5034 does not mention anything about that in the datasheet

I have been in contact with Tek, and they keep telling me to update
the SW of the scope :-)

Regards

Klaus

Klaus Kragelund <klauskvik@hotmail.com> writes:

> Hi
>
> So I wanted to check the VCE of a transistor in a small converter
> running at 100kHz, expected to see about 100mV above ground as the
> saturation voltage of the NPN, increasing linearely since magnetising
> current increasing.
>
> I saw nothing like that, debugging and found the scope (Tek DPO5000)
> somehow saturates with high voltage inputs. The signal is a square
> wave of 100kHz, 25V peak to peak, offset so low voltage is close to
> GND (measured by a 10:1 probe)
>
> See plot (green trace):
>
> www.electronicsdesign.dk/tmp/2.jpg
>
> Low value is about 400mV
>
> Ok, changed the scale div from 3V/div to 2V/div then this:
>
> www.electronicsdesign.dk/tmp/1.jpg
>
> Now low value is about 2.4V !
>
> I checked it was not my setup and all is fine.
>
> So I then hooked up the HP33120A function generator, with a BNC
> directly to the scope, now the problem still exists, but just at
> another volt/div setting (probably since the probe P500 has internal
> scaling etc). The signal, square wave, 9Vpp, 4.5V offset, 100kHz) (a
> levelshifted 9Vpp signal above GND)
>
> My suspicion is that the input stage of the scope is saturated since
> the top level of the signal is way out of bounds. If I reduce the
> frequency to 1kHz, then the signal starts to look good again, which
> tells me the input stage of the scope has settled and recovered from
> the overdrive.
>
> Then for a sanity check I hooked up a Tek TDS3032B and it showed the
> correct signal all the way down to 20mV/Div setting
>
> Then I took a Tek DPO4000 scope, and it had the same fault, allthough
> at a different volt/div setting
>
> Tried with a TDS220 (low end scope), it has the fault beginning at
> below 2V/div
>
> Tried a PicoScope 3000 series, had the fault at below 2V/div
>
> Tried a Agilent DSO 4000 series, had the fault at AFAIR 5V/div
>
> Then dusted off my first scope, the Hameg 205-3, it worked all the way
> down to 100mV/div :-)
>
> To the question, do any of you guys know the front-ends of scopes and
> if so is my assumption that the input stage is saturated correct (so
> the amp takes time to recover)?
>
> Seems I should throw out the 10k USD scope and bring my 1k USD Hameg
> to work for some real measurements....


Interesting - it sounds like the overdrive issue Jim Williams has
addressed several times:

http://cds.linear.com/docs/Application%20Note/an74f.pdf
http://cds.linear.com/docs/Application%20Note/an79.pdf
http://cds.linear.com/docs/Application%20Note/an120f.pdf

But I would not have expected it to show up so badly for such a modest
overdrive. 

-- 

John Devereux

Klaus Kragelund <klauskvik@hotmail.com> writes:

> On 27 Sep., 08:25, Jon Kirwan <j...@infinitefactors.org> wrote:
>> On Mon, 26 Sep 2011 16:28:18 -0700 (PDT), Klaus Kragelund
>>
>> <klausk...@hotmail.com> wrote:
>> ><snip>
>> >Seems I should throw out the 10k USD scope and bring my 1k USD Hameg
>> >to work for some real measurements....
>>
>> Somehow this reminds me of the old Sorcerer in Aladdin crying
>> out, "New lamps for old. &nbsp;New lamps for old," trying to
>> recover that special, magic lamp.
>>
>> (Replace lamp with oscilloscope....)
>>
>> :)
>>
>> Jon
>>
>>
>
> Hmm, just found an app note from Tektronix on this:
>
> http://cp.literature.agilent.com/litweb/pdf/5989-0068EN.pdf
>
> http://www.kemt.fei.tuke.sk/predmety/KEMT434_RM/_materialy/Doplnkove%20zdroje/UnderstandingOscilloscopeBandwidth.pdf
>
> Overdrive seems to be the problem
>
> However it is typically specified as less than 20ns, i am seeing a lot
> more. But perhaps thats due to excessive overdrive. Funny enough the
> DPO5034 does not mention anything about that in the datasheet
>
> I have been in contact with Tek, and they keep telling me to update
> the SW of the scope :-)

Did you run a virus scanner? :)

-- 

John Devereux

On Sep 26, 7:28=A0pm, Klaus Kragelund <klausk...@hotmail.com> wrote:
> Hi
>
> So I wanted to check the VCE of a transistor in a small converter
> running at 100kHz, expected to see about 100mV above ground as the
> saturation voltage of the NPN, increasing linearely since magnetising
> current increasing.
>
> I saw nothing like that, debugging and found the scope (Tek DPO5000)
> somehow saturates with high voltage inputs. The signal is a square
> wave of 100kHz, 25V peak to peak, offset so low voltage is close to
> GND (measured by a 10:1 probe)
>
> See plot (green trace):
>
> www.electronicsdesign.dk/tmp/2.jpg
>
> Low value is about 400mV
>
> Ok, changed the scale div from 3V/div to 2V/div then this:
>
> www.electronicsdesign.dk/tmp/1.jpg
>
> Now low value is about 2.4V !
>
> I checked it was not my setup and all is fine.
>
> So I then hooked up the HP33120A function generator, with a BNC
> directly to the scope, now the problem still exists, but just at
> another volt/div setting (probably since the probe P500 has internal
> scaling etc). The signal, square wave, 9Vpp, 4.5V offset, 100kHz) (a
> levelshifted 9Vpp signal above GND)
>
> My suspicion is that the input stage of the scope is saturated since
> the top level of the signal is way out of bounds. If I reduce the
> frequency to 1kHz, then the signal starts to look good again, which
> tells me the input stage of the scope has settled and recovered from
> the overdrive.
>
> Then for a sanity check I hooked up a Tek TDS3032B and it showed the
> correct signal all the way down to 20mV/Div setting
>
> Then I took a Tek DPO4000 scope, and it had the same fault, allthough
> at a different volt/div setting
>
> Tried with a TDS220 (low end scope), it has the fault beginning at
> below 2V/div
>
> Tried a PicoScope 3000 series, had the fault at below 2V/div
>
> Tried a Agilent DSO 4000 series, had the fault at AFAIR 5V/div
>
> Then dusted off my first scope, the Hameg 205-3, it worked all the way
> down to 100mV/div :-)
>
> To the question, do any of you guys know the front-ends of scopes and
> if so is my assumption that the input stage is saturated correct (so
> the amp takes time to recover)?
>
> Seems I should throw out the 10k USD scope and bring my 1k USD Hameg
> to work for some real measurements....
>
> Regards
>
> Klaus

I believe I've seen similar things with DSO's.  Sometimes negative
'overdrive' is different than positive 'overdirve'.  Which can also
lead to you seeing things that aren't really there.  The older DSO's
in general seem worse... An old HP 54600B is note worthy.

George H.

Klaus Kragelund wrote:

> Hi
> 
> So I wanted to check the VCE of a transistor in a small converter
> running at 100kHz, expected to see about 100mV above ground as the
> saturation voltage of the NPN, increasing linearely since magnetising
> current increasing.
> 
> I saw nothing like that, debugging and found the scope (Tek DPO5000)
> somehow saturates with high voltage inputs. The signal is a square
> wave of 100kHz, 25V peak to peak, offset so low voltage is close to
> GND (measured by a 10:1 probe)
> 
> See plot (green trace):
> 
> www.electronicsdesign.dk/tmp/2.jpg
> 
> Low value is about 400mV
> 
> Ok, changed the scale div from 3V/div to 2V/div then this:
> 
> www.electronicsdesign.dk/tmp/1.jpg
> 
> Now low value is about 2.4V !
> 

Interesting. Just a question of understanding (I'm not a user of Tek 
scopes, hence apologies if I miss the obvious information): in the 
screen shots there are different numbers of acquisitions displayed (86 
vs. 226). Is the scope set into averaging mode? And what is the coupling 
you selected, AC or DC?

Besides set I only can tell you that I never ran into this problem with 
my LeCroy scope and I did that kind of overdrive measurements on an 
almost daily basis. So, if this is problem is for real, then I dare 
conclude that not all scopes are affected by it :-)

Klaus

Klaus Kragelund wrote:
> On 27 Sep., 08:25, Jon Kirwan <j...@infinitefactors.org> wrote:
>> On Mon, 26 Sep 2011 16:28:18 -0700 (PDT), Klaus Kragelund
>>
>> <klausk...@hotmail.com> wrote:
>>> <snip>
>>> Seems I should throw out the 10k USD scope and bring my 1k USD Hameg
>>> to work for some real measurements....
>> Somehow this reminds me of the old Sorcerer in Aladdin crying
>> out, "New lamps for old.  New lamps for old," trying to
>> recover that special, magic lamp.
>>
>> (Replace lamp with oscilloscope....)
>>
>> :)
>>
>> Jon
>>
>>
> 
> Hmm, just found an app note from Tektronix on this:
> 
> http://cp.literature.agilent.com/litweb/pdf/5989-0068EN.pdf
> 

Quote "There is a misconception that some oscilloscopes can tolerate
overdrive conditions, whereas others cannot."

Tell Tektronix that this is not a misconception but it is true. Or
better yet, tell them to visit some nursing homes around Beaverton
during lunch time and ask anyone who was an engineer at Tektronix to
raise their hand. Then they should ask them how it's done right :-)

I just tried it out. I have to send my Instek GDS-2204 (digital scope)
into 10x overdrive until it distorts. It shouldn't even distort there
but, oh well, this is a <$2k digital scope. And as expected, the Tek
7704 has none of those problems at all, remains true all the way to
5mV/div. This is why I was saying all along, Tektronix should operate a
nursing home for very old engineers and their wifes on their campus.
Kind of like the Japanese way, where older generations teach the younger.


> http://www.kemt.fei.tuke.sk/predmety/KEMT434_RM/_materialy/Doplnkove%20zdroje/UnderstandingOscilloscopeBandwidth.pdf
> 
> Overdrive seems to be the problem
> 

It shouldn't be. There are risetime limits, of course, but it should not
suddenly and drasticaly alter the waveform or lose the offset. If I had
designed ultrasound frontends that way they'd have fired me. Those are
expected to remain true even in a +40dB overdrive situation.


> However it is typically specified as less than 20ns, i am seeing a lot
> more. But perhaps thats due to excessive overdrive. Funny enough the
> DPO5034 does not mention anything about that in the datasheet
> 
> I have been in contact with Tek, and they keep telling me to update
> the SW of the scope :-)
> 

That's the standard answer I guess. Like when you call your telco
because the broadband went out. "Sir, please re-install the software".
Yeah, right :-)

-- 
Regards, Joerg

http://www.analogconsultants.com/

Klaus Bahner wrote:
> Klaus Kragelund wrote:
> 
>> Hi
>>
>> So I wanted to check the VCE of a transistor in a small converter
>> running at 100kHz, expected to see about 100mV above ground as the
>> saturation voltage of the NPN, increasing linearely since magnetising
>> current increasing.
>>
>> I saw nothing like that, debugging and found the scope (Tek DPO5000)
>> somehow saturates with high voltage inputs. The signal is a square
>> wave of 100kHz, 25V peak to peak, offset so low voltage is close to
>> GND (measured by a 10:1 probe)
>>
>> See plot (green trace):
>>
>> www.electronicsdesign.dk/tmp/2.jpg
>>
>> Low value is about 400mV
>>
>> Ok, changed the scale div from 3V/div to 2V/div then this:
>>
>> www.electronicsdesign.dk/tmp/1.jpg
>>
>> Now low value is about 2.4V !
>>
> 
> Interesting. Just a question of understanding (I'm not a user of Tek
> scopes, hence apologies if I miss the obvious information): in the
> screen shots there are different numbers of acquisitions displayed (86
> vs. 226). Is the scope set into averaging mode? And what is the coupling
> you selected, AC or DC?
> 
> Besides set I only can tell you that I never ran into this problem with
> my LeCroy scope and I did that kind of overdrive measurements on an
> almost daily basis. So, if this is problem is for real, then I dare
> conclude that not all scopes are affected by it :-)
> 

Overdrive it past 10x and see. If it can do 100x without flinching then
they did the amp design about right.

-- 
Regards, Joerg

http://www.analogconsultants.com/