500uV/Div Oscilloscopes

[plesa]

I have disassembled one. Needed to change the fan on it. I can show you few photos now, more will come soon. I need to do proper hi-res photos of the boards. Note the hidden button inside

Thanks a lot for sharing. It is more complex than I expected.Hi res photos will be helpful.
It is nice piece of equipment, I'm impressed what has been designed 20 years back ;-)

[Jon_Paul_Clarke]

Good point about the sensitivity boost before the front end via probe or black box contraption. I overlooked that because I thought that it was "cheating". I guess there's no such thing in electronics as long as it's soldered/crimped properly. Eh, Dave?

Many oscilloscopes essentially do this internally by changing the gain of the input preamplifier which also has the effect of lowering the bandwidth.  On the often discussed 2225, this is done by lowering the load (increasing the resistance) at the output of a transconductance stage to multiply the gain by 10 times.  Other oscilloscopes like the 2235 series do this for their highest sensitivity setting or settings by lowering the emitter resistance in a transconductance stage by a factor of 2.5 (5mV/div to 2mV/div) for a minor bandwidth hit and this is common in both old and modern DSOs which often have a different bandwidth for each of their higher sensitivity settings.

I had a funny feeling that BW was somehow inversely proportional to voltage sensitivity. I'm guessing that there is a rough correlation to do with the Q of the cct.

[David Hess]

Exactly what happens depends on the circuit configuration.

With the 2225, the output impedance of the transconductance stage is inversely proportional to the bandwidth and they actually added a little bit of capacitance to further lower it so the x10 switch was intended to do dual duty by selecting magnification and bandwidth limiting which make sense because otherwise excessive noise would render high sensitivity useless anyway.  The 7A15A works in a completely different way to achieve the same result.  DSOs do the same thing but on the digital side with their high resolution mode.

With other designs, the loss of bandwidth is minor and just a side effect of getting all of the available gain out of a given number of stages.  It is just more economical that way.  Would you rather have a maximum sensitivity of 5mV/div and constant bandwidth or a maximum sensitivity of 2mV/div or even 1mV/div with a minor hit to the bandwidth and very little extra design cost?  Higher performance (more expensive) vertical amplifiers are less likely to suffer from this issue and "constant bandwidth" may be an advertised feature.

If the input referred noise is high enough, then extra sensitivity is wasted unless the bandwidth can be limited.  The only fast amplifier, 100 MHz in this case, with a sensitivity better than 5mV/div in the Tektronix 7000 series is the 7A13 and it includes a switchable 5 MHz bandwidth limit.  Otherwise it is a constant bandwidth vertical amplifier and its noise at 1mV/div and full bandwidth is readily apparent.  I actually measured it at almost 0.5 divisions RMS (almost 500uV RMS and its specification is 400uV RMS or better) which neatly shows why oscilloscope sensitivities usually top out somewhere between 1mV/div and 5mV/div.

[Jay_Diddy_B]

Hi,
If you look at input sections of the classic Tektronix scopes 465 etc and the 7A18 and 7A26 plugins for the 7000 series. The signal path is very similar.

The input attenuator is arranged in 1, 2, 5 sequence. 10x attenuators are added to the signal path for 10, 20, 50x attenuation.

The basic sensitivity, with the input attenuator set to zero attenuation, in most of the Tektronix scopes,  is 10mV per division. For 5mV/div 2x amplification is added to the signal chain.

To obtain 500uV /div a 20x amplifier would be needed.

To maintain a 100 MHz bandwidth, the amplifier needs a GBW product of 2GHz.

Regards,

Jay_Diddy_B

[Jon_Paul_Clarke]

Thanks David and Jay,
           That information makes a lot of sense. That's true about making specs to a price, otherwise every CRO that came out would be no holds barred and be absolutely awesome in features and design. I remember Dave(EEVBlog Dave) mentioning something about how Tektronix scopes (DSO's anyway) arrived at a certain way of displaying information that was different to other products from other companies. Tektronix replied back saying something to Dave about how that they have always designed their 'scopes. So my point being is that I guess, some manufacturers can have a design philosophy that lasts decades (Years not Bandwidth!) and may take a lot of effort to problem solve for newer models for adding those extra features.
          Also having a large Gain BW for the Cct in the order of 2Ghz just to maintain an operating BW of 100Mhz sounds about right. I guess it's a bit like the gain on ye olde Op-amps: Infinite Gain!! - But it will cost you in practice.

Thanks again for the Info.

Oh, By the way: I think I'll be getting a Hitachi v-1050f. The 2225 is nice and small but the Hitachi seems to tick all of the boxes for me. Now to save...

Jon

[David Hess]

If you look at input sections of the classic Tektronix scopes 465 etc and the 7A18 and 7A26 plugins for the 7000 series. The signal path is very similar.

You are in a maze of twisty little vertical amplifier building blocks, all alike.

The 7A26 and later amplifiers benefited from integration which allowed functions like paraphase amplifiers for switchable and variable gain blocks.

The 2225 series has fewer custom ICs than the 2235 series.  That and other differences make me suspect it was intended as a cost reduced version.

The input attenuator is arranged in 1, 2, 5 sequence. 10x attenuators are added to the signal path for 10, 20, 50x attenuation.

The 2225 series did not do it but the 2235 series and later oscilloscopes include a set of low impedance attenuators after the input amplifier buffer which is a common feature in later oscilloscopes.  This is something the older 7A13 does as well.

The basic sensitivity, with the input attenuator set to zero attenuation, in most of the Tektronix scopes,  is 10mV per division. For 5mV/div 2x amplification is added to the signal chain.

When I do calibration, I usually go through the schematics to find out what the basic sensitivity is because a couple of the Tektronix service manuals have incorrect calibration procedures.  Usually this is the second to the highest sensitivity.

To obtain 500uV /div a 20x amplifier would be needed.

I am not sure where you get 20x for the 2225 unless you are treating the differential amplifier as two single ended amplifiers.  Its basic sensitivity is 5mV/div so the switchable differential amplifier is x10.  The 2235 is the same but has a switchable x2.5 amplifier for 2mV/div.

To maintain a 100 MHz bandwidth, the amplifier needs a GBW product of 2GHz.

Are you counting from the input to the CRT deflection plates?

[David Hess]

Oh, By the way: I think I'll be getting a Hitachi v-1050f. The 2225 is nice and small but the Hitachi seems to tick all of the boxes for me. Now to save...

The Hitachi looks nice and easy to maintain.  I think they or another Japanese brand was suppose to have the sharpest CRTs after the Tektronix 547 era.  I wonder how much noise is shown on the CRT at 500uV/div where it apparently operates at full bandwidth.

[Jon_Paul_Clarke]

Well Dave,

When I get one I'll let you know! I'll do some experiments to find out.

[ivaylo]

So why sensitivity better than 500uV is so difficult in an oscilloscope? Theoretically it should only be a matter of amplification, no?

[KJDS]

So why sensitivity better than 500uV is so difficult in an oscilloscope? Theoretically it should only be a matter of amplification, no?

I may have got my maths wrong, but a quick calculation shows that with a 1meg input resistance and 100MHz of bandwidth, that's 1.3mV of rms noise.

[ivaylo]

a 1meg input resistance and 100MHz of bandwidth, that's 1.3mV of rms noise

What would be the formula for this if you don't mind?

And if input impedance is the obstacle here, there are instruments with even higher impedance measuring way bellow .5mV (voltmeters, spectrum analyzers, etc.), no?

[David Hess]

So why sensitivity better than 500uV is so difficult in an oscilloscope? Theoretically it should only be a matter of amplification, no?

I may have got my maths wrong, but a quick calculation shows that with a 1meg input resistance and 100MHz of bandwidth, that's 1.3mV of rms noise.

It is not that bad because the input resistance is shunted by the input capacitance.  20pF would make the effective noise bandwidth only 8kHz and the noise do to the input resistance 15 microvolts RMS.  But that does nothing for the noise of the input buffer which will be considerable because it is essentially a small gate (low capacitance) RF JFET or MOSFET.  You can verify this by shorting out the input to the oscilloscope and observing that the noise level barely changes if it changes at all.

It is interesting to note that the Tektronix 7A22 which uses low noise but slow JFETs in its input stage does indeed get close to the calculated input noise level for a 1 megohm input resistance shunted by 47 picofarads and has a maximum input sensitivity of 10 microvolts per division to match it but it still needs low pass filtering after the input amplifier to be useful.

[David Hess]

a 1meg input resistance and 100MHz of bandwidth, that's 1.3mV of rms noise

What would be the formula for this if you don't mind?

http://en.wikipedia.org/wiki/Johnson%E2%80%93Nyquist_noise

Integrated noise bandwidth will be about 1.6 times higher than the bandwidth of a single pole response.

And if input impedance is the obstacle here, there are instruments with even higher impedance measuring way bellow .5mV (voltmeters, spectrum analyzers, etc.), no?

These instruments make measurements over very limited bandwidths which are displayed over larger bandwidths.

[pascal_sweden]

Sounds like someone needs to build small box to boost signals in front of oscilloscope... I wonder how terrible demanding that would be considering bandwidth requirements and signal intergity...

This has been one of my to-do project for a long time.
Last I had a brief look at it, I think 20MHz (typical test gear noise measurement bandwidth) was quite doable for x10
That would give any scope at least 500uV/DIV, most 100uV/DIV

What is the current status of this project?

[AccountRemovedPerUsersRequest]

Since I'm not a RF designer myself in any standard, I'm also giving mental support for this project :-)

Axel.

[TimFox]

The input impedance is not the relevant parameter to calculate thermal noise.
The correct parameter is the source resistance (real part of the source impedance).
This could include any series resistance inside the CRO amplifier as part of the input protection circuit.
The amplifier input device (JFET) also has a noise voltage (quoted as a density per "root Hz") and a noise current (another density) that contributes a third noise voltage that increases with increasing source impedance.
Only if the input is open-circuited does the shunt resistor that determines the 1 megohm input resistance become the source resistance.

[tautech]

Stumbled on this thread while searching for other stuff.

It needs updating.
The Siglent SDS1000X series has a full BW 500uV range;
http://www.siglentamerica.com/pdxx.aspx?id=1369&T=2&tid=1

As described in this thread:
https://www.eevblog.com/forum/testgear/siglent-sds1000x-series-oscilloscopes/msg825532/#msg825532

New model in the series:
SDS1202X-E
http://www.siglentamerica.com/pdxx.aspx?id=5109&T=2&tid=1

[innkeeper]

An analog scope to add to the 500 uv/div list

Leader LBO-516 (also the lab-volt AA-798 - institutional re-badge version)

100mhz 3 channel delayed timebase scope.

[bson]

Sounds like someone needs to build small box to boost signals in front of oscilloscope... I wonder how terrible demanding that would be considering bandwidth requirements and signal intergity...

This has been one of my to-do project for a long time.
Last I had a brief look at it, I think 20MHz (typical test gear noise measurement bandwidth) was quite doable for x10
That would give any scope at least 500uV/DIV, most 100uV/DIV

x10, DC-500MHz wideband
Using a TI THS3202 CFB op amp, in two stages
Works like a charm even using cheapo 2L FR-4 (dirtypcbs).

At some point, when I cease being this lazy or pigs fly whichever comes first I'm planning to panel mount one of these in a small metal box with a tiny toroidal and a dual rail regulator...

[Marco]

I may have got my maths wrong, but a quick calculation shows that with a 1meg input resistance and 100MHz of bandwidth, that's 1.3mV of rms noise.

It's the shunt capacitance on the 9 Meg divider leg which determines the noise, should be in the range of 100 uV.

With active x10 probes you might be able to quarter that, with active x1 probes quarter it again. All theoretical limits obviously.

[Andreas]

The Siglent SDS1000X series has a full BW 500uV range;

Hello,

but for what?
a 1 Meg input impedance gives 1800uV (eff) or around 12 mVpp noise at 200 MHz.
so without bandwidth limiting this is not very useful.

With best regards

Andreas

[Marco]

a 1 Meg input impedance gives 1800uV (eff) or around 12 mVpp noise at 200 MHz.
so without bandwidth limiting this is not very useful.

The impedance is a lot lower for most of that bandwidth because of capacitance.

[PartialDischarge]

As mentioned the 7A22 plugin does 10uV/div which is very useful for measuring noise in SMPS. My recently acquired nicolet sigma scope, a rare one, does 100uV/div but is 12bit and only 5mhz, in certain situations it is usable and useful

[Andreas]

As mentioned the 7A22 plugin does 10uV/div which is very useful for measuring noise in SMPS.

Hello,

10uV/div is of course benchmark. Question how is the noise floor / which bandwidth do you have to set to take measurements?

My scope has only 2mV/div but has up to factor 50 of zoom (which is possible due to 16 bit ADC) resulting in 40uV/div.
At 100 kHz the noise floor is around 100uVpp.
Of course you can get lower than that by further reducing bandwidth.

with best regards

Andreas

[Neomys Sapiens]

x10, DC-500MHz wideband
Using a TI THS3202 CFB op amp, in two stages
Works like a charm even using cheapo 2L FR-4 (dirtypcbs).

At some point, when I cease being this lazy or pigs fly whichever comes first I'm planning to panel mount one of these in a small metal box with a tiny toroidal and a dual rail regulator...

Are those PCBs available somewhere or are they eval boards for this part? What did you want the toroidal coil for?