New Rigol DS1054Z oscilloscope

[Fungus]

Nope. 5 * f(clk) is only the second harmonic.

f(clk) is the base frequency

3*f(clk) is the first harmonic

5*f(clk) is the second harmonic

etc.

So you're saying the Agilent doc is wrong?

I only count two harmonics in the list above. How many do you see?

[LokiChaos]

5 times the fundamental is the 5th harmonic, or 4th overtone.  However, it is the 3rd /component/ of the square wave.

Harmonics are just the integer multiples of the fundamental (the first harmonic being 1*f), the even ones just happen to have a zero amplitude in a square wave.

[Fungus]

5 times the fundamental is the 5th harmonic, or 4th overtone.  However, it is the 3rd /component/ of the square wave.

Harmonics are just the integer multiples of the fundamental (the first harmonic being 1*f), the even ones just happen to have a zero amplitude in a square wave.

OK.

Whatever naming scheme we use it's a fact that:
a) A 50MHz 'scope shows less than the first three harmonics of a 10MHz square wave.
b) That's not very square (see image above).

Also: One of the ideas of having more bandwidth is to know whether the slow rise times you're seeing are due to the 'scope or due to the wires. To do that you need a 'scope that's faster than the wires, not the same speed.

(in theory you can never have enough bandwidth)

So ... 50MHz bandwidth to look at a 10MHz signal? It's too close to the limit for my liking.

[heatbreak]

OK.

Whatever naming scheme we use it's a fact that:
a) A 50MHz 'scope shows less than the first three harmonics of a 10MHz square wave.
b) That's not very square (see image above).

Also: One of the ideas of having more bandwidth is to know whether the slow rise times you're seeing are due to the 'scope or due to the wires. To do that you need a 'scope that's faster than the wires, not the same speed.

(in theory you can never have enough bandwidth)

So ... 50MHz bandwidth to look at a 10MHz signal? It's too close to the limit for my liking.

I have to disagree.  If you look at the Agilent screen shots here http://cp.literature.agilent.com/litweb/pdf/5989-5733EN.pdf of a 100 Mhz signal captured using a 500 Mhz and a 1 GHz scope, the two square waves look almost identical.

[Assafl]

* I have other Irks - like the slow saving of screenshots to USB memory (I love to annotate debugging or problem resolution sessions, and screenshot a lot - and that drives me nuts sometimes).

Why don't you connect the scope directly to your pc instead of using a usb-pen?
That works much faster. Screenshots are saved on the pc almost instantly.
No need to hassle with a usb-pen.

I was working in the living room so not next to my computer - I could've used the laptop but was looking for an intermittent thump that took hours to capture - so the USB is very appealing to the lazy... Anyway - very appealing but performance could be better.
(the intermittent thump ended up being a 7824 that loosened, would thermally shutdown and send the analog sections to -24V only...).

Were I to use it for work - depending on the type of work - of course - it may not be up to the job. But as a home lab having the extra features at an affordable price makes sense.

I can't imagine a professional who uses a scope in this priceclass at work.
Imho, this is a scope for hobbyists for use at home.
Nothing wrong with that.

Not sure about that. The Pass/Fail functionality doesn't sound like a "home" type of use case. Perhaps as a cheap manufacturing scope for QA?
Also - given the education section on their website - I can definitely see my professors having a ball screwing with students for setting up the scope incorrectly....

In my experience technicians are way more picky than engineers... I can definitely imagine the aerospace technicians I worked with not liking the scope (very "hi end" brand oriented folk).

None of the design engineers I ever worked with really seemed to care much about what scope or device it was. Perhaps in the old days we were old school? (I even remember University days having to design a mixer - loved Mini-Circuits and their awesome design book back then - Just so we could measure the performance of the antenna with the Anritsu).

[Assafl]

If you wish to continue to snip and ignore points about signal integrity, then that's your choice.

Quoting a digital signal speed in MHz "Max SPI speed is 4 MHz" is revealing, and indicative.

There is a lot of difference between something capable of basic visualisation of UART/SPI/GPIO levels (a 10MHz scope would suffice), and a 50/100NHz scope.

I will mostly use off the shelf modules, ie Arduinio and its shields etc, so signal integrity should not be an issue.  But even a 50 MHz scope is capable of troubleshooting "signal integrity" for a 10 MHz signal no?  I don't know how a 100 Mhz one is much better than a 50 MHz one.  Slightly better, but you'll still leave wanting more.

Ok.  So you have all convince me to maybe spend more $ to get a 100 or even a 200 MHz one.

Actually - the same guy that instilled the doubts about the Rigol's UI shows that given a single channel the BW (due to the 1GS/s sample rate) goes rather nicely to 200MHz - https://www.youtube.com/embed/aiIr3j_EyLY?list=PL5zLSmFo0HELyDnrxbC2tEjSeAl0KaYM0 and in another video shows a 400MHz (substantially attenuated) sine using the DS1054Z.

So assuming he is correct - a square wave of a few MHz to perhaps 50MHz measured on one channel should look pretty good (albeit a bit rounded at the top end)... 

[JPortici]

But they are not! The edges are different. Read the paragraph

It doesn't really matter if the frequency is 1GHz or 1 Hz, the key parameter is the rise/fall time and that is the source of all our problems.
Take any arm processor and program it to endlessly switch on and of an output and probe the edges. they won't look like squares at all, there will be reflections unless the ground lead is placed correctly and the trace is not properly terminated.
Now increase the frequency to the point that the reflections will interfere and your squarish trace will generate false conditions where it goes to

Now, switch on the bandwidth limit of the scope channel. No reflections there.. so they not exists.. mmm..
Also, notice that on the scope parameter sheet you have bandwidth and rise time.

this is the reason why you have control over the GPIO speed. It is a limit on the rise/fall time.

[Assafl]

But they are not! The edges are different. Read the paragraph

It doesn't really matter if the frequency is 1GHz or 1 Hz, the key parameter is the rise/fall time and that is the source of all our problems.

Is it critical races you are trying to find? I can't say I love a good race condition error - but to say that that is "the source of all our problems" which I assume to mean "the only use for a scope" - I strongly disagree. I like using a scope on analog circuits as well, on sensors, as well as for power supplies, and sometimes even to see if a silly circuit is doing anything or stone dead...

However, if races are your thing - and especially with high speed digital circuits - the DS1054 is probably not your scope.

[rstofer]

OK.

Whatever naming scheme we use it's a fact that:
a) A 50MHz 'scope shows less than the first three harmonics of a 10MHz square wave.
b) That's not very square (see image above).

Also: One of the ideas of having more bandwidth is to know whether the slow rise times you're seeing are due to the 'scope or due to the wires. To do that you need a 'scope that's faster than the wires, not the same speed.

(in theory you can never have enough bandwidth)

So ... 50MHz bandwidth to look at a 10MHz signal? It's too close to the limit for my liking.

I have to disagree.  If you look at the Agilent screen shots here http://cp.literature.agilent.com/litweb/pdf/5989-5733EN.pdf of a 100 Mhz signal captured using a 500 Mhz and a 1 GHz scope, the two square waves look almost identical.

Not even close to identical...

On the 100 MHz scope the rise time is 2.986 ns
On the 500 MHz scope the rise time is 812 ps
On the 1 GHz scope the rise time is 615 ps
And, finally, on the 2 GHz scope the rise time is about right at 520 ps.

The difference between the 100 MHz and 500 MHz scope is huge.  At 5x, things start to look better but it's not until 20x that the waveform is nearly right.  Pick 10x...  This is consistent with most other measurement 'truisms'.  If you want to measure to 0.x V, make sure the meter can read to 0.0x V.

[Fungus]

I have to disagree.  If you look at the Agilent screen shots here http://cp.literature.agilent.com/litweb/pdf/5989-5733EN.pdf of a 100 Mhz signal captured using a 500 Mhz and a 1 GHz scope, the two square waves look almost identical.

You mean this?


Pure marketing crap.

Obviously you don't believe a word I say, so... I'll throw it open: Who here can measure a square wave at 20% of their scopes rated bandwidth and get a beautiful image like that?

Back here in the real world we have images like this:



That's a 10MHz signal on a 25MHz 'scope, and guess what? It agrees perfectly with the theory of harmonics I posted above.

( it's the best image I could find in ten seconds of googling...but it demonstrates it beautifully)

If you're happy with square waves that look like that then carry on reading Siglent sales brochures. 

[Assafl]

I have to disagree.  If you look at the Agilent screen shots here http://cp.literature.agilent.com/litweb/pdf/5989-5733EN.pdf of a 100 Mhz signal captured using a 500 Mhz and a 1 GHz scope, the two square waves look almost identical.

You mean this?


Pure marketing crap.

Obviously you don't believe a word I say, so... I'll throw it open: Who here can measure a square wave at 20% of their scopes rated bandwidth and get a beautiful image like that?

Back here in the real world we have images like this:



That's a 10MHz signal on a 25MHz 'scope, and guess what? It agrees perfectly with the theory of harmonics I posted above.

( it's the best image I could find in ten seconds of googling...but it demonstrates it beautifully)

If you're happy with square waves that look like that then carry on reading Siglent sales brochures. 

I think they are talking trying to resolve critical race conditions for high speed circuits - in which case the justification for going as high BW as one can is clear. I just don't think it is the only use case for a scope...

[JPortici]

But they are not! The edges are different. Read the paragraph

It doesn't really matter if the frequency is 1GHz or 1 Hz, the key parameter is the rise/fall time and that is the source of all our problems.

Is it critical races you are trying to find? I can't say I love a good race condition error - but to say that that is "the source of all our problems" which I assume to mean "the only use for a scope" - I strongly disagree. I like using a scope on analog circuits as well, on sensors, as well as for power supplies, and sometimes even to see if a silly circuit is doing anything or stone dead...

However, if races are your thing - and especially with high speed digital circuits - the DS1054 is probably not your scope.

I meant in the scope of this argument why 50 MHz is no good when it should be.. and as an example i brought an arm mcu which is used by a lot of hobbyist

from how i see it the problem is the rise time which has nothing to do with the signal frequency, which can't be overlooked but it will be if the risetime of your scope is much slower and the risetime is heavily influenced by the channel bandwith (in my example, the reflections will vanish when i insert the bandwidth limit but it doesn't mean they aren't there)

[heatbreak]

I have to disagree.  If you look at the Agilent screen shots here http://cp.literature.agilent.com/litweb/pdf/5989-5733EN.pdf of a 100 Mhz signal captured using a 500 Mhz and a 1 GHz scope, the two square waves look almost identical.

You mean this?


Pure marketing crap.

Obviously you don't believe a word I say, so... I'll throw it open: Who here can measure a square wave at 20% of their scopes rated bandwidth and get a beautiful image like that?

Back here in the real world we have images like this:



That's a 10MHz signal on a 25MHz 'scope, and guess what? It agrees perfectly with the theory of harmonics I posted above.

( it's the best image I could find in ten seconds of googling...but it demonstrates it beautifully)

If you're happy with square waves that look like that then carry on reading Siglent sales brochures. 

25 MHz doesn't even give the 3rd harmonic.  You need to cover at least the 3rd harmonic (a 30 MHz scope) to get some kind of square wave.  A 50 MHz scope would give you a nice square wave.

[heatbreak]

Not even close to identical...

On the 100 MHz scope the rise time is 2.986 ns
On the 500 MHz scope the rise time is 812 ps
On the 1 GHz scope the rise time is 615 ps
And, finally, on the 2 GHz scope the rise time is about right at 520 ps.

The difference between the 100 MHz and 500 MHz scope is huge.  At 5x, things start to look better but it's not until 20x that the waveform is nearly right.  Pick 10x...  This is consistent with most other measurement 'truisms'.  If you want to measure to 0.x V, make sure the meter can read to 0.0x V.

Cycle period of 100 MHz = 10 nSecs.  The error between the 1GHz and the 500 Mhz scope is 0.2 nSecs.  Relative error between the two scopes compared to the signal frequency is 0.2/10 = 2%.  2% error should be ok for home use no?

[heatbreak]

I have to disagree.  If you look at the Agilent screen shots here http://cp.literature.agilent.com/litweb/pdf/5989-5733EN.pdf of a 100 Mhz signal captured using a 500 Mhz and a 1 GHz scope, the two square waves look almost identical.

You mean this?


Pure marketing crap.

Obviously you don't believe a word I say, so... I'll throw it open: Who here can measure a square wave at 20% of their scopes rated bandwidth and get a beautiful image like that?

Back here in the real world we have images like this:



That's a 10MHz signal on a 25MHz 'scope, and guess what? It agrees perfectly with the theory of harmonics I posted above.

( it's the best image I could find in ten seconds of googling...but it demonstrates it beautifully)

If you're happy with square waves that look like that then carry on reading Siglent sales brochures. 

25 MHz doesn't even give the 3rd harmonic.  You need to cover at least the 3rd harmonic (a 30 MHz scope) to get some kind of square wave.  A 50 MHz scope would give you a nice square wave.

After checking out the link you showed me, although the 4 MHz capture on a 25 MHz looks passable for home use, but I think you're right, I do want at least a 100 MHz scope.  Thanks for showing me the way!

[JPortici]

Not even close to identical...

On the 100 MHz scope the rise time is 2.986 ns
On the 500 MHz scope the rise time is 812 ps
On the 1 GHz scope the rise time is 615 ps
And, finally, on the 2 GHz scope the rise time is about right at 520 ps.

The difference between the 100 MHz and 500 MHz scope is huge.  At 5x, things start to look better but it's not until 20x that the waveform is nearly right.  Pick 10x...  This is consistent with most other measurement 'truisms'.  If you want to measure to 0.x V, make sure the meter can read to 0.0x V.

Cycle period of 100 MHz = 10 nSecs.  The error between the 1GHz and the 500 Mhz scope is 0.2 nSecs.  Relative error between the two scopes compared to the signal frequency is 0.2/10 = 2%.  2% error should be ok for home use no?

You are missing the point. the difference is in the edge because of the different risetime of the scope, which is heavily influenced by the channel bandwidth

[heatbreak]

You are missing the point. the difference is in the edge because of the different risetime of the scope, which is heavily influenced by the channel bandwidth

I guess I'm really missing the point.  I know the rise time is not gonna be as accurate as the one with the higher bandwidth (but the higher bandwidth is not exactly the same as the actual rise time anyway), but say a 50 MHz scope has 15 nSecs extra error over a 100 MHz one, if I write my code to add at least say 30 nSecs over any timing specs, then why does a 15 nSecs error matter?  And at 10 Mhz, I can comfortably add 30 nSecs to any timing spec.

[tggzzz]

You are missing the point. the difference is in the edge because of the different risetime of the scope, which is heavily influenced by the channel bandwidth

I guess I'm really missing the point.  I know the rise time is not gonna be as accurate as the one with the higher bandwidth (but the higher bandwidth is not exactly the same as the actual rise time anyway), but say a 50 MHz scope has 15 nSecs extra error over a 100 MHz one, if I write my code to add at least say 30 nSecs over any timing specs, then why does a 15 nSecs error matter?  And at 10 Mhz, I can comfortably add 30 nSecs to any timing spec.

Sigh. What on earth makes you think that in a real system you will have an ideal input signal, e.g. like a square wave from a signal generator

In the real world you have imperfect signal sources transmitted through imperfect interconnections to imperfect receivers.  Then you have to add in the effects of imperfect power and grounds, e.g. ground bounce and decoupling. Not to mention crosstalk and interference.

All of those real world effects are classified  "signal integrity", and lead to intermittent operation. If you want something to work reliably, then you have to measure the analogue effects that will affect your "digital" signals.  That means measuring transients that are within your logic's response times - and those are independent of your circuit's clock rate.

But don't worry; many expensive consultants make a good living sorting out problems cause by that kind of lack of understanding.

[heatbreak]

Sigh. What on earth makes you think that in a real system you will have an ideal input signal, e.g. like a square wave from a signal generator

In the real world you have imperfect signal sources transmitted through imperfect interconnections to imperfect receivers.  Then you have to add in the effects of imperfect power and grounds, e.g. ground bounce and decoupling. Not to mention crosstalk and interference.

All of those real world effects are classified  "signal integrity", and lead to intermittent operation. If you want something to work reliably, then you have to measure the analogue effects that will affect your "digital" signals.  That means measuring transients that are within your logic's response times - and those are independent of your circuit's clock rate.

But don't worry; many expensive consultants make a good living sorting out problems cause by that kind of lack of understanding.

We're switching from rise/fall time to "signal integrity" again?  I mean come on, I'm driving a SPI/IC2 signal from the uC to some chip/module at home.  How imperfect could it get? Unless your home is a power substation, then you shouldn't have to worry about "signal integrity".  I mean if you really want to catch all possible signal integrity problems, then mine as well skip the 100 MHz ('cause it ain't much different than a 50 Mhz if you looking at a 10 MHz signal) get a 1GHz and get it over with.

[tggzzz]

Sigh. What on earth makes you think that in a real system you will have an ideal input signal, e.g. like a square wave from a signal generator

In the real world you have imperfect signal sources transmitted through imperfect interconnections to imperfect receivers.  Then you have to add in the effects of imperfect power and grounds, e.g. ground bounce and decoupling. Not to mention crosstalk and interference.

All of those real world effects are classified  "signal integrity", and lead to intermittent operation. If you want something to work reliably, then you have to measure the analogue effects that will affect your "digital" signals.  That means measuring transients that are within your logic's response times - and those are independent of your circuit's clock rate.

But don't worry; many expensive consultants make a good living sorting out problems cause by that kind of lack of understanding.

We're switching from rise/fall time to "signal integrity" again?  I mean come on, I'm driving a SPI/IC2 signal from the uC to some chip/module at home.  How imperfect could it get? Unless your home is a power substation, then you shouldn't have to worry about "signal integrity".  I mean if you really want to catch all possible signal integrity problems, then mine as well skip the 100 MHz ('cause it ain't much different than a 50 Mhz if you looking at a 10 MHz signal) get a 1GHz and get it over with.

Transition time and signal integrity are intimately intertwined. You very deeply fail to understand, and have some reading to do.

For a taster, have a look at http://www.edn.com/collections/4435129/Bogatin-s-Rules-of-Thumb
I suggest an order of 0, 8, 7, 19, then the others.

If you don't understand why inductance and capacitance are important, then please don't design/integrate any hardware for a system that has to be reliable.

[heatbreak]

I have to disagree.  If you look at the Agilent screen shots here http://cp.literature.agilent.com/litweb/pdf/5989-5733EN.pdf of a 100 Mhz signal captured using a 500 Mhz and a 1 GHz scope, the two square waves look almost identical.

You mean this?


Pure marketing crap.

Obviously you don't believe a word I say, so... I'll throw it open: Who here can measure a square wave at 20% of their scopes rated bandwidth and get a beautiful image like that?

Back here in the real world we have images like this:



That's a 10MHz signal on a 25MHz 'scope, and guess what? It agrees perfectly with the theory of harmonics I posted above.

( it's the best image I could find in ten seconds of googling...but it demonstrates it beautifully)

If you're happy with square waves that look like that then carry on reading Siglent sales brochures. 

25 MHz doesn't even give the 3rd harmonic.  You need to cover at least the 3rd harmonic (a 30 MHz scope) to get some kind of square wave.  A 50 MHz scope would give you a nice square wave.

After checking out the link you showed me, although the 4 MHz capture on a 25 MHz looks passable for home use, but I think you're right, I do want at least a 100 MHz scope.  Thanks for showing me the way!

After thinking about it more, I take back what I said about a 50 MHz scope couldn't produce a nice square wave.  It could indeed produce a very nice square wave because isn't the scope front end is just a low pass filter?  If so, the wave form you showed in your picture captured from the OWON Scope makes no sense.  There must be something very wrong with the OWON.  I have simulated a low-pass filter with cut-off at 50 MHz and fed in a 10 MHz square wave using LTSpice and as you can see from the picture below (green input, red output), the output from the filter matches what Agilent showed in their document.  So Agilent is right, you can choose the bandwidth of your scope using the formula

f(bw) = 5 * f (clk)

So a 50 MHz scope can produce a very nice square wave of 10 Mhz except the rise time of the pulse will be slightly off from a 100 MHz scope.

[heatbreak]

Transition time and signal integrity are intimately intertwined. You very deeply fail to understand, and have some reading to do.

For a taster, have a look at http://www.edn.com/collections/4435129/Bogatin-s-Rules-of-Thumb
I suggest an order of 0, 8, 7, 19, then the others.

If you don't understand why inductance and capacitance are important, then please don't design/integrate any hardware for a system that has to be reliable.

You should only be worry about those if designing a new system from scratch.  Unlikely to occur if you're using off the shelf modules, but ok fine I'll bite.  How does a 100 MHz scope be much better than a 50 MHz scope in catching the problems of "Signal Integrity"?  And how can a 100 MHz scope capture a "signal integrity" problem that only a 200 MHz scope can catch?

[jjoonathan]

A 1-pole AFE filter would only make sense if you had plenty of sample rate to burn. Most scopes are more aggressive.

[heatbreak]

A 1-pole AFE filter would only make sense if you had plenty of sample rate to burn. Most scopes are more aggressive.

I found a video showing a 10 MHz and 20 MHz on a 100 MHz Rigol scope.  Of course the 10 Mhz looks better, but to me the 20 MHz is doable for home use.  The peak voltage measured are the same and so are the rise/fall times.  4:21 and 4:31 time mark.

[heatbreak]

Dropping CS' is a separate line of C that occurs before spi.write() because the datafield can be arbitrarily long.  Same with raising CS'.  There's not much chance I can violate the setup and hold times using C but it's good to measure anyway.  It sometimes happens that the SPI gadget is still shifting when CS' is raised.  Since I didn't write the mbed SPI library, I don't know whether they waited for transmission of the last byte to be complete or returned after the last byte was stuffed in the shift register.

I didn't use the cursors to check the SCK MOSI or MISO, I just shifted the traces up over the SCK so I could check the edges and verify I was in mode 0.

Another gotcha with C is using bitfields.  They aren't guaranteed to be in any particular alignment within an unsigned char and that can make a HUGE difference (easily discovered by decoding the SPI string) when writing the 3rd byte of the W5500 command.  I had to scrap that idea.  But I WANT bit fields!  So, I'm looking at coding the thing in Ada.

But, you're right, for your needs the more responsive controls are the deciding factor.  I simply don't care about the HMI.  I don't spend nearly as much time twiddling the knobs as I do analyzing the screen display.  And I didn't want to spend much money on a digital scope...

If you don't know what the spi.write() function does, you shouldn't be doing any code development at all.  I mean that's pretty important to know whether that function returns after last bit left the wire or after the shift register is written.  You need to find out how to determine when the MOSI and CLK are at their final resting places before de-asserting CS.  Also in some processors like Arm with the Matrix, you need to also make sure that the CS assert has left the wire before calling the 1st spi.write(..)

Some compilers allow you to define the order of the bit-fields, but I don't remember how the alignment definition extends to multi-bit bit-fields.  But if you really want to write portable code, then you could just write a function that converts the bit-field into an unsigned char suitable for spi.write(..).