DSO2250 AC-DC coupling

[legg]

Trying to work up a Hantek DSO2250 into a useful tool, I notice that
the frequency compensation changes abnormally, as it is changed from
AC to DC coupling. By abnormally, I mean that AC coupling introduces
what looks like HF roll-off. Calibrated in the AC-coupled mode, a 10:1
probe overshoots when switched to DC coupling. (Using a 1KHz square
wave.)

I'm thinking that this can't be done by hardware, that it's got to be
some kind of signal processing bodge - even though you hear a
mechanical relay moving to physically short out the coupling
capacitor, as the mode change occurs.

How can things be screwed up in this manner, if not in software? Relay
shunt capacity in the wrong place?

It seems to have unequal effects on the seperate channels and even on
the rising or falling edge, if I'm not fooling myself.

The self-calibration feature doesn't seem to involve an AC signal.
Compensated probes can't be swapped from one channel to the other,
without re-adjustment

I'll have to run a frequency sweep to see what it looks like, though
I'm not sure if the display can be believed. Will I get differing
frequency response, depending on simple horizontal display settings,
for example....

RL

 

[Joerg]

Trying to work up a Hantek DSO2250 into a useful tool, I notice that
the frequency compensation changes abnormally, as it is changed from
AC to DC coupling. By abnormally, I mean that AC coupling introduces
what looks like HF roll-off. Calibrated in the AC-coupled mode, a 10:1
probe overshoots when switched to DC coupling. (Using a 1KHz square
wave.)

I'm thinking that this can't be done by hardware, that it's got to be
some kind of signal processing bodge - even though you hear a
mechanical relay moving to physically short out the coupling
capacitor, as the mode change occurs.

How can things be screwed up in this manner, if not in software? Relay
shunt capacity in the wrong place?

It seems to have unequal effects on the seperate channels and even on
the rising or falling edge, if I'm not fooling myself.

That looks like a layout screw-up. Traces of unequal length between
AC/DC settings and even between channels :-(

The self-calibration feature doesn't seem to involve an AC signal.
Compensated probes can't be swapped from one channel to the other,
without re-adjustment

I'll have to run a frequency sweep to see what it looks like, though
I'm not sure if the display can be believed. Will I get differing
frequency response, depending on simple horizontal display settings,
for example....

If it is a trace issue then some nifty kludges might work. The old trick
to produce a very small "variable capacitor" is to solder on some enamel
twisted pair, then snip - snipety - snip ... until just right. IOW you'd
have to determine the setting with the worst roll-off and then bring the
capacitance of all the other traces up until the same roll-off occurs there.

 

[legg]

That looks like a layout screw-up. Traces of unequal length between
AC/DC settings and even between channels :-(



If it is a trace issue then some nifty kludges might work. The old trick
to produce a very small "variable capacitor" is to solder on some enamel
twisted pair, then snip - snipety - snip ... until just right. IOW you'd
have to determine the setting with the worst roll-off and then bring the
capacitance of all the other traces up until the same roll-off occurs there.

Doesn't seem to have anything to do with the analog measurement
hardware function.

The same square wavform becomes decompensated as it is relocated from
the lower to the upper portions of the display screen.

A square wave that is properly compensated midscreen, shows overshoot
when shifted to the bottom, undershoot when shifted to the top.

The leading/trailing egde difference is in the order of 7LSB.

At the same time, the on-board ppk measurement increases
proportionally solely due to screen positioning. The overshot waveform
measures less ppk. Delta 70mV in a 2V measurement. This in the
opposite sense to the presented distortion.

How do they do it?

Why do they do it?

RL

 

[Joerg]

Doesn't seem to have anything to do with the analog measurement
hardware function.

But you wrote that when the relay clicks to the other position the
compensation gets messed up. That sure looks like a bug in the analog
hardware.

The same square wavform becomes decompensated as it is relocated from
the lower to the upper portions of the display screen.

Here we'd need the schematic. To see whether this offset function is
still analog. Not sure whether they send a USB command to the box and
some DAC or PWM adds an offset. In that case we could see slight changes
in a junction capacitance somewhere.

You could probably find out if you'd open the box and see if a DC level
in there changes if you move the offset via the GUI.

A square wave that is properly compensated midscreen, shows overshoot
when shifted to the bottom, undershoot when shifted to the top.

The leading/trailing egde difference is in the order of 7LSB.

Out of 8 bits? That's quite a lot.

At the same time, the on-board ppk measurement increases
proportionally solely due to screen positioning. The overshot waveform
measures less ppk. Delta 70mV in a 2V measurement. This in the
opposite sense to the presented distortion.

How do they do it?

Sounds more and more like they do the offset before the ADC and the
analog circuitry was done by a rookie

Why do they do it?

It is probably not intentional and they either didn't notice or ignored it.

 

[legg]

But you wrote that when the relay clicks to the other position the
compensation gets messed up. That sure looks like a bug in the analog
hardware.

The apparent change was due to the shift of the waveform on the
display. It was a positive pulse waveform that shifted down-screen
when AC coupled.

All center-screen compensated waveforms can be caused to overshoot by
shifting baseline down or to round-off by shifting baseline up. This
with unchanged input or coupling method.

Here we'd need the schematic. To see whether this offset function is
still analog. Not sure whether they send a USB command to the box and
some DAC or PWM adds an offset. In that case we could see slight changes
in a junction capacitance somewhere.

You could probably find out if you'd open the box and see if a DC level
in there changes if you move the offset via the GUI.

Out of 8 bits? That's quite a lot.



Sounds more and more like they do the offset before the ADC and the
analog circuitry was done by a rookie



It is probably not intentional and they either didn't notice or ignored it.

I suppose that screen offsets have to be physically enforced as a DC
component introduced before the ADC, as the coding expires at the
display limits ( this evidenced by manipulating math functions to
overdiven signals )

The unipolarity of effect suggests that the ADC is seeing a unipolar
input - perhaps a protection network's capacitance is being modulated
by this shift.

I would have thought, however that internal measurements of ppk would
be generated from ADC output, and this demonstratedly cannot be the
case, if the GUI display is unaltered from the ADC output.

Input protection is specified at 35V levels, and +/-20V levels are
expected. The behavior is visible using the unipolar 2Vppk square wave
signal provided as calibration, by the device.

Anyone else see this effect on other models of similar devices?

I've only been working with this one for two days.

RL

 

[Joerg]

The apparent change was due to the shift of the waveform on the
display. It was a positive pulse waveform that shifted down-screen
when AC coupled.

Ok, then I misunderstood. Now it sounds less like layout and more like
sub-optimal analog circuitry.

[...]

I suppose that screen offsets have to be physically enforced as a DC
component introduced before the ADC, as the coding expires at the
display limits ( this evidenced by manipulating math functions to
overdiven signals )

The unipolarity of effect suggests that the ADC is seeing a unipolar
input - perhaps a protection network's capacitance is being modulated
by this shift.

Now it really sounds like an analog circuit issue in there.

I would have thought, however that internal measurements of ppk would
be generated from ADC output, and this demonstratedly cannot be the
case, if the GUI display is unaltered from the ADC output.

I would be very surprised if they didn't take the cheap route and do it
off of the ADC output. Who knows what the software does.

Input protection is specified at 35V levels, and +/-20V levels are
expected. The behavior is visible using the unipolar 2Vppk square wave
signal provided as calibration, by the device.

Anyone else see this effect on other models of similar devices?

I've only been working with this one for two days.

It's time to take a 2nd analog scope and probe around in there. I
suspect the analog DC offset changes some junction capacitance and they
do not correct for that in SW. Or worse, haven't noticed ...

If this is indeed the case then probing around with an analog scope
should bring out the truth rather quickly.

 

[legg]

Doesn't seem to have anything to do with the analog measurement
hardware function.

The same square wavform becomes decompensated as it is relocated from
the lower to the upper portions of the display screen.

A square wave that is properly compensated midscreen, shows overshoot
when shifted to the bottom, undershoot when shifted to the top.

The leading/trailing egde difference is in the order of 7LSB.

At the same time, the on-board ppk measurement increases
proportionally solely due to screen positioning. The overshot waveform
measures less ppk. Delta 70mV in a 2V measurement. This in the
opposite sense to the presented distortion.

I've tried posting a few images to a.b.s.e., but my own server doesn't
seem to pick them up.

RL

 

[legg]

I use giganews... your posts are there.

...Jim Thompson

Thanks.
RL

 

[Joerg]

Thanks.

Problem is, I bet that >95% of participants in this NG have no access to
a.b.s.e. any longer. Better to find some free picture-hosting server,
preferably one without nasty scripting or excessive ads but one with
mild ads.

 

[Joerg]

Why not? Too CHEAP for $5/month ?

What good does it do to sign up if almost nobody else can access binary
groups?

My newsserver costs around $15 a year but the nice thing is that they
seem to do a stellar job of hosing off spam. Lately I took the
googlemail domain filter out and ... no more flooding with ads for fake
watches and handbags.

Unfortunately, after Cuomo's tirades, binary groups are history for
engineers. You and I and a lot of others aren't liking it but it's a
fact :-(

 

[legg]

The apparent change was due to the shift of the waveform on the
display. It was a positive pulse waveform that shifted down-screen
when AC coupled.

Ok, then I misunderstood. Now it sounds less like layout and more like
sub-optimal analog circuitry.

[...]

I suppose that screen offsets have to be physically enforced as a DC
component introduced before the ADC, as the coding expires at the
display limits ( this evidenced by manipulating math functions to
overdiven signals )

The unipolarity of effect suggests that the ADC is seeing a unipolar
input - perhaps a protection network's capacitance is being modulated
by this shift.

Now it really sounds like an analog circuit issue in there.

I would have thought, however that internal measurements of ppk would
be generated from ADC output, and this demonstratedly cannot be the
case, if the GUI display is unaltered from the ADC output.

I would be very surprised if they didn't take the cheap route and do it
off of the ADC output. Who knows what the software does.

Input protection is specified at 35V levels, and +/-20V levels are
expected. The behavior is visible using the unipolar 2Vppk square wave
signal provided as calibration, by the device.

Anyone else see this effect on other models of similar devices?

I've only been working with this one for two days.

It's time to take a 2nd analog scope and probe around in there. I
suspect the analog DC offset changes some junction capacitance and they
do not correct for that in SW. Or worse, haven't noticed ...

If this is indeed the case then probing around with an analog scope
should bring out the truth rather quickly.

I've thrown away the warranty, for what it's worth, and given the
circuit a quick look.

After the input attenuator, a DC shift is introduced before the DAC,
as expected, to provide a unipolar ADC input and to control the screen
centering position. There's no sign of alterations in rise-time or
overshoo on either side of the buffer feeding the ADC, as the DC shift
is modulated.

At the same time, these errors are visible on the PC display.

The protection network is composed of the CB junction of a vhf bipolar
transistor to the +3V5 rail (~1pF), and a DO214(SMA) diode body to the
common return. This diode is soldered manually and is marked with the
signature (RD)found on all of
-a 54V transil,(~120pF)
-an 18V transil (~470pF)
-a conventional 200V rectifier (10-7pF)
-others?

One thing that wasn't noticed before, is that the effect is most
pronounced on channel 2 ( the channel used to demonstrate in the
images posted on a.b.s.e.). The effect on channel 1 is probably small
enough to have been ignored by a bored operator.

The ADCs are identity-ground-off. There's a nasty void in the one used
for channel 2, probably caused by metallic grit accidentally included
in the grinding operation. Some of this metal is imbedded in the void.
Image on a.b.s.e..

RL

 

[legg]

The apparent change was due to the shift of the waveform on the
display. It was a positive pulse waveform that shifted down-screen
when AC coupled.

Ok, then I misunderstood. Now it sounds less like layout and more like
sub-optimal analog circuitry.

[...]

At the same time, the on-board ppk measurement increases
proportionally solely due to screen positioning. The overshot waveform
measures less ppk. Delta 70mV in a 2V measurement. This in the
opposite sense to the presented distortion.

How do they do it?

Sounds more and more like they do the offset before the ADC and the
analog circuitry was done by a rookie


Why do they do it?

It is probably not intentional and they either didn't notice or ignored it.

I suppose that screen offsets have to be physically enforced as a DC
component introduced before the ADC, as the coding expires at the
display limits ( this evidenced by manipulating math functions to
overdiven signals )

The unipolarity of effect suggests that the ADC is seeing a unipolar
input - perhaps a protection network's capacitance is being modulated
by this shift.

Now it really sounds like an analog circuit issue in there.

I would have thought, however that internal measurements of ppk would
be generated from ADC output, and this demonstratedly cannot be the
case, if the GUI display is unaltered from the ADC output.

I would be very surprised if they didn't take the cheap route and do it
off of the ADC output. Who knows what the software does.

Input protection is specified at 35V levels, and +/-20V levels are
expected. The behavior is visible using the unipolar 2Vppk square wave
signal provided as calibration, by the device.

Anyone else see this effect on other models of similar devices?

I've only been working with this one for two days.

It's time to take a 2nd analog scope and probe around in there. I
suspect the analog DC offset changes some junction capacitance and they
do not correct for that in SW. Or worse, haven't noticed ...

If this is indeed the case then probing around with an analog scope
should bring out the truth rather quickly.

I've thrown away the warranty, for what it's worth, and given the
circuit a quick look.

After the input attenuator, a DC shift is introduced before the DAC,
as expected, to provide a unipolar ADC input and to control the screen
centering position. There's no sign of alterations in rise-time or
overshoo on either side of the buffer feeding the ADC, as the DC shift
is modulated.

At the same time, these errors are visible on the PC display.

The protection network is composed of the CB junction of a vhf bipolar
transistor to the +3V5 rail (~1pF), and a DO214(SMA) diode body to the
common return. This diode is soldered manually and is marked with the
signature (RD)found on all of
-a 54V transil,(~120pF)
-an 18V transil (~470pF)
-a conventional 200V rectifier (10-7pF)
-others?

One thing that wasn't noticed before, is that the effect is most
pronounced on channel 2 ( the channel used to demonstrate in the
images posted on a.b.s.e.). The effect on channel 1 is probably small
enough to have been ignored by a bored operator.

The ADCs are identity-ground-off. There's a nasty void in the one used
for channel 2, probably caused by metallic grit accidentally included
in the grinding operation. Some of this metal is imbedded in the void.
Image on a.b.s.e..

Removing both transient protection components from their position in
channel 2 makes no difference. The frequency compensation is still
dependent on display screen position, without visible alteration at
the transient limiting node.

I also note that the actual frequency compensation waveform displayed
seems to bear little resemblance to the actual waveform before the
ADC, regardless of position, on either channel. In an analog circuit,
I might not be overly concerned, as there is the possibility of
intentional peaking or tweaking farther downstream that need not be
DC-dependent.

The intention of probe calibration (in the x10 setting) is simply to
tweak high frequency response of the probe, to adapt to scope input
capacitance.

A linear input signal should give a linear display, within the limits
of the circuitry involved. Signals that are rolled off dependent on
screen position would be difficult to get your mind around, no matter
what their actual content.

I'm guessing that there's some kind of ADC error that is time and
dc-voltage dependent, in the 1/4 millisecond range. Possibly a S/H
section that is not properly 'flushed' during reset.

Perhaps there's an attempt to correct for this in SW, if so, there may
be firmware corrections in the pipeline. I somehow doubt that it could
be effectively applied at the full bandwidth of the instrument, or to
two channels with non-identical characteristics.

I will make inquiries, but I can only expect an offer to accept a
returned unit, for servicing or replacement (under warranty...).

RL

Excuse duplicate, if present. Server issues again.

 

[Joerg]

The apparent change was due to the shift of the waveform on the
display. It was a positive pulse waveform that shifted down-screen
when AC coupled.

Ok, then I misunderstood. Now it sounds less like layout and more like
sub-optimal analog circuitry.

[...]


At the same time, the on-board ppk measurement increases
proportionally solely due to screen positioning. The overshot waveform
measures less ppk. Delta 70mV in a 2V measurement. This in the
opposite sense to the presented distortion.

How do they do it?

Sounds more and more like they do the offset before the ADC and the
analog circuitry was done by a rookie


Why do they do it?

It is probably not intentional and they either didn't notice or ignored it.
I suppose that screen offsets have to be physically enforced as a DC
component introduced before the ADC, as the coding expires at the
display limits ( this evidenced by manipulating math functions to
overdiven signals )

The unipolarity of effect suggests that the ADC is seeing a unipolar
input - perhaps a protection network's capacitance is being modulated
by this shift.

Now it really sounds like an analog circuit issue in there.


I would have thought, however that internal measurements of ppk would
be generated from ADC output, and this demonstratedly cannot be the
case, if the GUI display is unaltered from the ADC output.

I would be very surprised if they didn't take the cheap route and do it
off of the ADC output. Who knows what the software does.


Input protection is specified at 35V levels, and +/-20V levels are
expected. The behavior is visible using the unipolar 2Vppk square wave
signal provided as calibration, by the device.

Anyone else see this effect on other models of similar devices?

I've only been working with this one for two days.

It's time to take a 2nd analog scope and probe around in there. I
suspect the analog DC offset changes some junction capacitance and they
do not correct for that in SW. Or worse, haven't noticed ...

If this is indeed the case then probing around with an analog scope
should bring out the truth rather quickly.

I've thrown away the warranty, for what it's worth, and given the
circuit a quick look.

After the input attenuator, a DC shift is introduced before the DAC,
as expected, to provide a unipolar ADC input and to control the screen
centering position. There's no sign of alterations in rise-time or
overshoo on either side of the buffer feeding the ADC, as the DC shift
is modulated.

At the same time, these errors are visible on the PC display.

The protection network is composed of the CB junction of a vhf bipolar
transistor to the +3V5 rail (~1pF), and a DO214(SMA) diode body to the
common return. This diode is soldered manually and is marked with the
signature (RD)found on all of
-a 54V transil,(~120pF)
-an 18V transil (~470pF)
-a conventional 200V rectifier (10-7pF)
-others?

One thing that wasn't noticed before, is that the effect is most
pronounced on channel 2 ( the channel used to demonstrate in the
images posted on a.b.s.e.). The effect on channel 1 is probably small
enough to have been ignored by a bored operator.

The ADCs are identity-ground-off. There's a nasty void in the one used
for channel 2, probably caused by metallic grit accidentally included
in the grinding operation. Some of this metal is imbedded in the void.
Image on a.b.s.e..

Removing both transient protection components from their position in
channel 2 makes no difference. The frequency compensation is still
dependent on display screen position, without visible alteration at
the transient limiting node.

I also note that the actual frequency compensation waveform displayed
seems to bear little resemblance to the actual waveform before the
ADC, regardless of position, on either channel. In an analog circuit,
I might not be overly concerned, as there is the possibility of
intentional peaking or tweaking farther downstream that need not be
DC-dependent.

There should be not intentional peaking, at least not right at the ADC.
That could potentially lead to a signal becoming "different" with higher
amplitude and would make the scope rather useless as a tool.

The intention of probe calibration (in the x10 setting) is simply to
tweak high frequency response of the probe, to adapt to scope input
capacitance.

Does the waveform at the input to the ADC change when moving the offset?
It should not ...

A linear input signal should give a linear display, within the limits
of the circuitry involved. Signals that are rolled off dependent on
screen position would be difficult to get your mind around, no matter
what their actual content.

I'm guessing that there's some kind of ADC error that is time and
dc-voltage dependent, in the 1/4 millisecond range. Possibly a S/H
section that is not properly 'flushed' during reset.

I hope not. That would be a major design screw-up on their part.

Perhaps there's an attempt to correct for this in SW, if so, there may
be firmware corrections in the pipeline. I somehow doubt that it could
be effectively applied at the full bandwidth of the instrument, or to
two channels with non-identical characteristics.

I will make inquiries, but I can only expect an offer to accept a
returned unit, for servicing or replacement (under warranty...).

RL

Excuse duplicate, if present. Server issues again.

No problem. If the mfg can't fix this strange compensation error on the
DSO2250 I'd insist on getting my money back.

 

[legg]

Trying to work up a Hantek DSO2250 into a useful tool, I notice that
the frequency compensation changes abnormally, as it is changed from
AC to DC coupling. By abnormally, I mean that AC coupling introduces
what looks like HF roll-off. Calibrated in the AC-coupled mode, a 10:1
probe overshoots when switched to DC coupling. (Using a 1KHz square
wave.)

I'm not sure if the display can be believed. Will I get differing
frequency response, depending on simple horizontal display settings,
for example....

Contact at Hantek states that this is normal, and within specified
behaviour.

Is there a quick way of estimating the difference in frequency
response, demonstrated by a 100mV difference in rising edge amplitude
of a 2Vppk 1KHz square wave? I know the rise-time requirement rule of
thumb, but never stopped to consider the other, though probe
compensation is one of the first things you run into, with normal
scope use. Just never stopped to consider....

RL

 

[legg]

Simple spice calculations show that the simple compensation scheme's
of 200MHz probes cannot produce overshoot with the time frame
illustrated in the pictured 1KHz calibration signal - it should
normally be all over and done with in the first 100uSec of risetime.

RL

Every post is a frigging test.....

 

[Joerg]

Simple spice calculations show that the simple compensation scheme's
of 200MHz probes cannot produce overshoot with the time frame
illustrated in the pictured 1KHz calibration signal - it should
normally be all over and done with in the first 100uSec of risetime.

I do not have access to a.b.s.e., so can't see the pic. 100usec is a
whole lotta time.

But it couldn't possibly be the probe anyhow because the probe doesn't
know whether an offset is entered or whether the scope is on AC. My DSO
does not do that, no matter where I move the waveform to it doesn't change.

RL

Every post is a frigging test.....

Our whole life is one big test

 

[legg]

In any case, I'll stick the screenshots up.

http://www.magma.ca/~legg/TVS/111126a_under.jpg
http://www.magma.ca/~legg/TVS/111126b_under.jpg

The DSO2250 is one of the USB scope interfaces being offered in the
last few years. I would expect the input structure to be similar to
any of the lighter-weight 8-bit scopes, with LCD or phosphor displays,
made since Y2K.

RL

 

[legg]

The ADCs are identity-ground-off. There's a nasty void in the one used
for channel 2, probably caused by metallic grit accidentally included
in the grinding operation. Some of this metal is imbedded in the void.
Image on a.b.s.e..

The part is actually a DPDT video switch, not an ADC.

RL

 

[legg]

Geesh, can you send that POS back?
George H.

Not after voiding the warranty, I shouldn't think. With the box being
previously opened to temporarily remove sppressors (no effect), I'm
left with the alternative. Putting together a rough schematic to see
if some simple adjustment to part type, supply biasing or other might
correct things sufficiently to serve.

There appear to be jfets used as buffers/limiters, feeding a DPDT
video switch - so not as conventional as hoped. Both could be
misapplied to introduce nonlinearity. Patience and research needed
here.

RL

 

[Joerg]

In any case, I'll stick the screenshots up.

http://www.magma.ca/~legg/TVS/111126a_under.jpg
http://www.magma.ca/~legg/TVS/111126b_under.jpg

The DSO2250 is one of the USB scope interfaces being offered in the
last few years. I would expect the input structure to be similar to
any of the lighter-weight 8-bit scopes, with LCD or phosphor displays,
made since Y2K.

If this happens just by scooting the offset then it looks like a
screwed-up design :-(

Since you wrote that the warranty is possibly toast I guess you can only
go in there with another scope to find out where they messed up and
correct that, or live with it.

It's not always hopeless. I once found and fixed a design bug in a
shortwave transceiver that was know to "always have this oscillation on
one band". It ended up needing a rather minor change after which it
became a very nice piece of equipment.