Basic Digital Oscilloscope Spec Question

[bostonman]

I have a question about sampling rate of scopes.

The Tektronix TBS1052C two channel scope is listed as 50MHz bandwidth and a sample rate of 1GS/s "per channel".

The Agilent 54831 scope I own is listed as 600MHz (I've done the "upgrade" and it's now 1GHz) and up to 4GS/s sample rate.

In the case of the Tek, does the 1GS/s per channel mean if both are used, it's 1GS/s or does it decrease to 500MS/s?

In the case of of my Agilent, is it the same, 4GSa/s if only one channel is used and 1GSa/s per channel if all four are used?

Unless I'm wrong, performing the 1GHz upgrade means the sampling rate isn't increased, but also, the Agilent scope is listed as 600MHz, but that would be at 1GSa/s, I would only get (rounding up) two samples which wouldn't display a waveform well. This would imply my 600MHz scope is really 100MHz since it typically needs ten samples to make a somewhat accurate waveform.

[ataradov]

Tek datasheet : 1 GS/s sample rate on all channels
Agilent datasheet: 2 Channels Interleaved - 4 GSa/s, Each Channel - 2 GSa/s.

To reconstruct a sine wave you need two samples per period. More complex signals require more, of course.

[bostonman]

To reconstruct a sine wave you need two samples per period. More complex signals require more, of course.

Ooops, you're correct. I was thinking ten samples were needed to make a sine wave.

[SiliconWizard]

Nope, ten samples minimum per period are recommended for "digital signals" to get reasonably analyzable pulses. Otherwise Nyquist applies.

The benefit of having an analog bandwidth > Fs/2 is when using a "sampling" acquisition mode (which acquires several periods, each with a slightly different delay), rather than the "real-time" acqusition mode. It gets you a higher equivalent sampling rate, but only works for periodic signals.

[ArdWar]

To reconstruct a sine wave you need two samples per period. More complex signals require more, of course.

Manufacturers sometimes get a bit too optimistic with that

[Ice-Tea]

To reconstruct a sine wave you need two samples per period. More complex signals require more, of course.

Manufacturers sometimes get a bit too optimistic with that
(Attachment Link)

Fine for repeating signals. This was fairly common practice years ago where you would have a "100MHZ scope" with a 1MHz one-shot performance...

[radiolistener]

To reconstruct a sine wave you need two samples per period.

If we are being scrupulous, this is not entirely correct, as it depends on the signal phase.

For example, let's say we have a 1 GHz sample rate and two samples (see picture below): 0, 0 (sin(0)=0, sin(pi)=0)
It's obvious that you cannot distinguish whether it is a sine wave or DC.
And if we know that this is a sine wave, its amplitude still unknown.

So we have two samples per period and know that the signal is within the Nyquist bandwidth (Fs/2), but we are unable to reconstruct its waveform.

But if we shift the signal phase by 90 degrees, we get the samples +1, -1, which allows us to reconstruct the original sine waveform.

Therefore, it would be more accurate to say that two points are not enough to reconstruct the shape of a sine wave, but it is enough to reconstruct the shape of a cosine wave.



Regarding to oscilloscope requirements, as a rule of thumb, the oscilloscope's sample rate should be at least 10-20 times higher than its analog bandwidth. This allows the anti-aliasing filter slope to remain within the Nyquist bandwidth and ensures good flatness within the oscilloscope's analog bandwidth.

[2N3055]

You need more than 2 samples per sinewave period to reconstruct that sinewave. That is what math says.
To reconstruct complex waveform we need to have more than 2 samples per period of highest frequency component of said signal.
In practice, factor of 2.5 samples per period of highest frequency works fine.

[tggzzz]

Nope, ten samples minimum per period are recommended for "digital signals" to get reasonably analyzable pulses. Otherwise Nyquist applies.

The benefit of having an analog bandwidth > Fs/2 is when using a "sampling" acquisition mode (which acquires several periods, each with a slightly different delay), rather than the "real-time" acqusition mode. It gets you a higher equivalent sampling rate, but only works for periodic signals.

Period is irrelevant. All that matters is the risetime.

The front end bandwidth is what matters; sampling rate is little more than marketing wank. Extreme example: my Philips scope has a 200ps risetime (1.7GHz bandwidth) but manages to use transistors with f_T=300MHz in the signal path.

In digitising scopes a higher sampling rate is used so some of the internal systems are easier to implement. As a user that is barely relevant.

[Ice-Tea]

Period is irrelevant. All that matters is the risetime.

Highlighted for emphasis. Important for scopes, signal integrity, EMI, EMC, layout, ....

[David Hess]

In the case of the Tek, does the 1GS/s per channel mean if both are used, it's 1GS/s or does it decrease to 500MS/s?

In the case of of my Agilent, is it the same, 4GSa/s if only one channel is used and 1GSa/s per channel if all four are used?

You got it right.  The Tektronix DSO has a separate digitizer and acquisition record for each channel, so sample rate and record length does not change with the number of used channels.  The Agilent shares digitizers and acquisition records between channels.

The front end bandwidth is what matters; sampling rate is little more than marketing wank. Extreme example: my Philips scope has a 200ps risetime (1.7GHz bandwidth) but manages to use transistors with f_T=300MHz in the signal path

How did they manage that?  Usually the transistor f_T needs to be several times the bandwidth even in a cascode.

[tggzzz]

The front end bandwidth is what matters; sampling rate is little more than marketing wank. Extreme example: my Philips scope has a 200ps risetime (1.7GHz bandwidth) but manages to use transistors with f_T=300MHz in the signal path

How did they manage that?  Usually the transistor f_T needs to be several times the bandwidth even in a cascode.

The Old Skool way: a sampling diode. After that it is BC107s and similar Admittedly the trigger circuit has three BFW92s (f_T 1.3-2.4GHz) as a diff-pair with cascode output driving the inevitable tunnel diodes.

The Philips PM3410 is a weird and wonderful beast. Definitely not a general purpose scope, but I like it. The slowest timebase is "20µs/cm". It has GR874 connectors and a couple of very thick 30ns delay lines. It is about the same size as an HP1740, but lighter.

[David Hess]

The front end bandwidth is what matters; sampling rate is little more than marketing wank. Extreme example: my Philips scope has a 200ps risetime (1.7GHz bandwidth) but manages to use transistors with f_T=300MHz in the signal path

How did they manage that?  Usually the transistor f_T needs to be several times the bandwidth even in a cascode.

The Old Skool way: a sampling diode. After that it is BC107s and similar Admittedly the trigger circuit has three BFW92s (f_T 1.3-2.4GHz) as a diff-pair with cascode output driving the inevitable tunnel diodes.

The Philips PM3410 is a weird and wonderful beast. Definitely not a general purpose scope, but I like it. The slowest timebase is "20µs/cm". It has GR874 connectors and a couple of very thick 30ns delay lines. It is about the same size as an HP1740, but lighter.

Oh, sampling oscilloscope, got it.

Tektronix made 7000 series sampling plug-in which is like that, with delay lines so it can show the leading edge, but the delay lines limit bandwidth to 1 GHz.

[bostonman]

Tek datasheet : 1 GS/s sample rate on all channels
Agilent datasheet: 2 Channels Interleaved - 4 GSa/s, Each Channel - 2 GSa/s.

Still not entirely certain I understand the specs. For the Tek, it's 1GS/s on each channel regardless if one channel is used or both?

My Agilent is interleaved, so it takes BW from other channels to improve the other used channels? Would using a single channel mean I get 4GSa/s or 2GSa/s? Using all four channels would use 1GSa/s then?

[David Hess]

Tek datasheet : 1 GS/s sample rate on all channels
Agilent datasheet: 2 Channels Interleaved - 4 GSa/s, Each Channel - 2 GSa/s.

Still not entirely certain I understand the specs. For the Tek, it's 1GS/s on each channel regardless if one channel is used or both?

Yes, that is correct.

My Agilent is interleaved, so it takes BW from other channels to improve the other used channels? Would using a single channel mean I get 4GSa/s or 2GSa/s? Using all four channels would use 1GSa/s then?

The sample rate is divided between channels.  The bandwidth does not change.

Based on the specifications, I think the Agilent has 2 digitizers for each pair of channels, so 2 channels can operate at 4 GS/s each, and 4 channels can operate at 2 GS/s each.

Agilent 54831 offers 4 channels, 4 GSa/s sampling rate in 2-channel mode, 2 GSa/s sampling rate in 4-channel mode, 600 MHz bandwidth.

[ataradov]

Still not entirely certain I understand the specs.

Tek - 1 GSa/s for all channels regardless of how many channels are used.
Agilent - 2 GSa/s if all channels are used, 4 GSa/s if only one or two channels are used. You may need to use channels 1 and 3, I don't known the exact details on which channels are interleaved. It is easy enough to find out - just try to enable different combinations and see what sample rate you get.

My Agilent is interleaved, so it takes BW from other channels to improve the other used channels?

Bandwidth remains the same, only sample rate changes.

[bostonman]

just try to enable different combinations and see what sample rate you get.

How would I know the sample rate?

Bandwidth remains the same, only sample rate changes.

Yes, I accidentally interchanged BW and sample rate.

Unless I'm wrong, my Agilent 54831 is (still) a great scope. I thought it was a great scope when I first got it due to four channels, color, Windows based, etc... but seems the BW and sample rate are still great specs.

[ataradov]

How would I know the sample rate?

It would be written somewhere on the  screen. Looking at some pictures of the UI, it is in the top left corner.

Unless I'm wrong, my Agilent 54831 is (still) a great scope.

Yes, it is not a bad scope.

[bostonman]

Funny, I haven't used the scope that much, but I find myself searching the screen for various pieces of information each time I use it because I forget where everything is.

I have a sample print from a few years ago that shows it had automatic sampling and was 250MSa/s.

I'll try running more tests sometime this week.

[ataradov]

Yes, it normally would adjust the sample rate based on the time scale. You can dig though the menus and set a fixed rate. In most cases, automatic sampling is fine. There are a few cases where you want to set a fixed value.

[bostonman]

I set the sample rate to manual and the highest it went is 4GSa/s.

From what I saw, it was as follows:

Channel 1 or 2 only: 4GSa/s

Channel 1 and 2: 2GSa/s

Channel 1 and 3 or 4: 4GSa/s

Channel 2 and 3 or 4: 4GSa/s

Channel 3 or 4 only: 4GSa/s

Channel 3 and 4: 2GSa/s

[wasedadoc]

Ch1 and Ch2 share one ADC.  Ch3 and Ch4 share a second ADC.

[bostonman]

I forgot to include when all four channels are on, it showed 2GSa/s.

Does this mean if I use channels 1 and 2, or 3 and 4, or all four, I get 2GSa/s on each channel?

Do any disadvantages exist using a scope that interleaves two channels versus a scope that doesn't?

[radiolistener]

Do any disadvantages exist using a scope that interleaves two channels versus a scope that doesn't?

Yes, there is some little disadvantage.

For signle channel the disadvantage is that interleaved ADC have some characteristics difference between two ADC and it may affect waveform. But for oscilloscope it should be almost not noticeable because this difference is very small and can be noticed only when you process sampled data very precisely in matlab.

For two/more channels the disadvantage is that interleaved ADC may have some small parasite interference between channels. But it will be very small, almost not noticeable and can be noticed only when you process sampled data very precisely in matlab.

[bostonman]

Makes sense that it would have such disadvantages.

I'm so use to old slow scopes and then begin questioning if something is wrong with my scope when I see anomalies on this model.

The reason I asked in a previous post about whether it's still a good scope. I'm seeing these tiny Tek models for >$500 and my Agilent model is selling used for around $1k. Seems this is still a great scope and well worth the $1k.

I like it enough that I purchased a second one months ago. It was listed as making loud noises but boots and had a feeling it could be fixed. We settled on $225, it arrived, and the only "loud noise" was the beeping from the motherboard.

Turned out the micro didn't have enough thermo compound. I removed the heat sink, cleaned it, added thermo grease, and it was fixed. After I did some upgrades (including the 1GHz BW upgrade) and now I have two of these model scopes.