Best entry level oscilloscope for computer engineering?

[nbritton]

One of the points in favor of the USB LA was that you can view everything on a computer monitor. However, since the Rigol's have USB/Ethernet can't you also view these on a computer monitor too?

[Fungus]

I guess I am misunderstanding something perhaps, I only have the first model Logic 16, not the pro, but their website says 50MS/s for analog and 500MS/s for digital.  Is there some problem with the 500MS/s since everyone keeps saying the ADC is slow?

The ADC is for analog.

I don't really know why they bother adding the analog stuff. Maybe it's useful to be able to see the analog signals alongside the digital for timing issues. The bandwidth is awful though.

[Fungus]

One of the points in favor of the USB LA was that you can view everything on a computer monitor. However, since the Rigol's have USB/Ethernet can't you also view these on a computer monitor too?

Yes but with the same number of pixels as the scope's screen. And with slower update rate.

[alank2]

Hi,

The ADC is for analog.
I don't really know why they bother adding the analog stuff. Maybe it's useful to be able to see the analog signals alongside the digital for timing issues. The bandwidth is awful though.

I am with you, but you can choose analog or digital for each pin, and choose both.  If you choose "digital" then you get 500MSa/s, right?  Then it is using a comparator and not the ADC, right?

[D3f1ant]

In my opinion an Oscilloscope and a Logic Analyser are very different tools with different usage.
I use the oscilloscope primarily for ensuring signal integrity, does the waveform look right, is there noise etc. Once I know the signal looks good I might use it to ensure a few packets of data decode to what I expect and so its useful to have decoding on the scope.
I tend to prefer to use the PC based Logic Analyzer for decoding and examining large blocks of data, its more about the content of data streams than the shape of the signals. Its cumbersome working with tables of decoded data on an mixed signal oscilloscope, even if your scope supports plugging in a mouse and keyboard like the Agilent/Keysight scopes do. Yes you can copy data to PC from the scope, but those extra steps are not productive when your trying to fix a software problem.

You could consider a 4 ch Picoscope. They have very good decoding and free software updates. I doubt there is a stand alone scope that can decode a many different data types as the picoscope software does (and decode software is free). A USB scope is cumbersome in the opposite way the scope based LA is, but is compact and portable and they can have HUGE memory buffers (hundreds of MB) so you can drill into very long captures. You get all the benefits of a PC based LA, but not the number of channels, you really do need 4 or more. USB scopes take getting used to if your used to a scope. I forced myself for a while, but eventually the proper oscilloscope reappeared on my bench. Its very useful as a portable tool too, they are small and compact when your dragging a laptop or tablet around as well.

It comes down to what your predominately using it for, for maximum productivity you need the right tool for the job.

[D3f1ant]

Yeah I rarely use the analog channels on the Saleae Pro 16, but have done. Its sometimes useful to check the analog levels going into an ADC on a microcontroller. Its not a oscilloscope replacement, but it saves turning the scope (or even a multimeter) on to just verify something when you already have the thing out and hooked up to a board. Its also extremely portable.

[alank2]

In my opinion an Oscilloscope and a Logic Analyser are very different tools with different usage.

I completely agree with this.  I use my Saleae more than my scope and that is saying a lot because my scope sits on a table right next to me. 

Their pro 8 or pro 16 at 500MSa/s would have 10 samples per clock for 50 MHz SPI and 4 samples per clock for 125 MHz SPI.

[nbritton]

I find the hacked Rigol 1054Z very compelling. However, decoding SPI on this scope is limited to 25 MHz.

I don't know about the analogue channels, but the MSO1000Z LA channels will decode 50MHz SPI, down to a 250MHz sample rate.

Edit: I just tried it on the analogue channels, and while the signals ain't pretty, it does decode fine on a 50MHz SPI stream.

That's interesting because in theory the 1000Z is only rated for 25 MHz at 250 Sa/s when using 3 or 4 analog channels since it's all going through a single chip. I suppose even with attenuation the signal is enough to trigger the decoder, is the 1000Z decoding trigger adjustable? Someone stated that the 1000Z can do 150 MHz before it hits -3db attenuation:

https://www.eevblog.com/forum/testgear/new-rigol-ds1054z-oscilloscope/msg524797/#msg524797

1000Z Bandwidth:
Frequency      Amplitude
  10 MHz          0.0 dBm
100 MHz         -1.5 dBm
150 MHz         -3.0 dBm
393 MHz        -10.0 dBm
447 MHz        -20.0 dBm

[nbritton]

Their pro 8 or pro 16 at 500MSa/s would have 10 samples per clock for 50 MHz SPI and 4 samples per clock for 125 MHz SPI.

According to the datasheet the Saleae Pro 8/16 are only capable of 500 MS/s on 4 channels, when using more than 4 channels it drops down to 100 MS/s. That's only 1 sample per clock at 100 MHz.

http://downloads.saleae.com/specs/Logic+Pro+16+Data+Sheet.pdf

[nbritton]

USBee went out of business yesterday: http://www.usbee.com/company.htm

[nbritton]

Can the DS4014 be hacked to 500 MHz? I was reading in the scope sizing guides that the analog bandwidth should be around 10 times the frequency you want to measure if you want an accurate waveform due to the rise times of square waves and such. It's becoming apparent to me that a 100 MHz scope is nothing but a toy when it comes to computer engineering applications, what do I need to really get into the game? I'd prefer not getting nickeled and dimed buying toy equipment that is ultimately useless for the task at hand.

My quest is to build a i486 based system from scratch. Linux dropped support for the i386, so I see the i486 as the oldest representation of the modern x86 architecture.

[Mark_O]

I don't like the USB logic analysers because either the memory is short or the sample rate low.

Nico, you left out:  or the software sucks. 

On the first 2 counts, I find Hantek's 4032L more than adequate (64M sample depth, 32 channels wide; at up to 400M samples/sec).  It also has a good front-end for handling logic levels.  It's on the 3rd count that it falls short.

[alank2]

nbritton the answer to your question is yes.

[Mark_O]

Their pro 8 or pro 16 at 500MSa/s would have 10 samples per clock for 50 MHz SPI and 4 samples per clock for 125 MHz SPI.

According to the datasheet the Saleae Pro 8/16 are only capable of 500 MS/s on 4 channels, when using more than 4 channels it drops down to 100 MS/s. That's only 1 sample per clock at 100 MHz.

Only 4 channels are needed for SPI.  And only 4 samples per clock will definitely decode SPI, though it won't give you very precise bit transitions.

[Mark_O]

It's becoming apparent to me that a 100 MHz scope is nothing but a toy when it comes to computer engineering applications...

That's certainly not a true statement, in general.

[nbritton]

I don't like the USB logic analysers because either the memory is short or the sample rate low.

Nico, you left out:  or the software sucks. 

I should note that I don't have a Windows PC. As a matter of principle I avoid Microsoft products because I make my living off of Linux and UNIX. So that leaves Mac or Linux as the host operating system for a USB LA.

[nbritton]

It's becoming apparent to me that a 100 MHz scope is nothing but a toy when it comes to computer engineering applications...

That's certainly not a true statement, in general.

How so, can you articulate on that? Pretty much everything I see in the computer space is running in the gigahertz range. Even the Arduino is running at 15 MHz, the scope sizing guidelines recommend 10x the bandwidth, so you would need a 150 MHz scope for this.

[nctnico]

Since you don't need to measure anything that is connected to the CPU directly (like in the old days) you only deal with peripherals and these run much slower. In many cases you can slow the peripherals (I2C, SPI, etc) down too to make signals measurable if necessary.

[Howardlong]

It's becoming apparent to me that a 100 MHz scope is nothing but a toy when it comes to computer engineering applications...

That's certainly not a true statement, in general.

How so, can you articulate on that? Pretty much everything I see in the computer space is running in the gigahertz range. Even the Arduino is running at 15 MHz, the scope sizing guidelines recommend 10x the bandwidth, so you would need a 150 MHz scope for this.

It's a case of using a rule of thumb, but rejecting that experience and analytical skills can work in your favour.

I'd have no problem in using a 1Gsa/s 500MHz bw scope to figure out if a 200MHz clock signal exists. However the signal won't look great even if I use reasonable probing techniques (i.e. probes that cost more than a DS1054Z, or use a homebrew 10:1 50 ohm resistive)

If you want to probe at GHz speeds on the board you will still need to know about the effects your probes will have, even though you may have spent $10k or $20k just on the probe, before you even bought the scope.

Designing, prototyping and testing PC motherboards at bus speed level is not the domain of the DS1054Z, or, in fact, the domain of 99%+ of the oscilloscopes on the planet, so most of the work is done in software before the first board is even fabricated.

As nctnico says, in embedded systems by far most of what you need to know and that's exposed isn't GHz. When it is, the scope's the least of your problems, it's figuring out how to probe it.

[nbritton]

Since you don't need to measure anything that is connected to the CPU directly (like in the old days) you only deal with peripherals and these run much slower. In many cases you can slow the peripherals (I2C, SPI, etc) down too to make signals measurable if necessary.

I understand that. However, one of the things I would like to do is build a computer (perferably x86 architecture) from scratch, i.e. selecting and laying out the components onto a PCB. Lets use a 486DX4 100 MHz as an example, the front side bus on this runs at 33.3 MHz. According to scope sizing guides I would need a 333 MHz scope to accurately represent a 33.3 MHz square wave due to waveform rise times. Furthermore, the PCI bus is 32-bit and also the 72-pin SIMM is 32-bit, so wouldn't this mean I need a 32 channel logic analyzer?

Am I being realistic here with my goals and requirements?

[Fungus]

It's becoming apparent to me that a 100 MHz scope is nothing but a toy when it comes to computer engineering applications...

That's certainly not a true statement, in general.

Signals don''t vanish above the rated bandwidth, they just attenuate (get smaller). This affects harmonics and changes the shapes of things. If you know signal theory and practice a bit then you can compensate in your head (up to a point).

eg. A DS1054Z will show a 50MHz square wave with 'ringing' artifacts. Those artifacts aren't coming from your PCB, they're there because the third harmonic isn't coming through loud and clear.

Knowing how to probe? That's a whole other layer of learning. The overshoots on your rising edges? Probably the way you connected the probe. A more expensive oscilloscope won't save you from knowing that.

Bottom line: Even the most expensive oscilloscaope won't just show you the beautiful waveforms you might be imagining when you arrive from a purely digital world. It takes knowledge and practice.

[tautech]

It's becoming apparent to me that a 100 MHz scope is nothing but a toy when it comes to computer engineering applications...

That's certainly not a true statement, in general.

Signals don''t vanish above the rated bandwidth, they just attenuate (get smaller). This affects harmonics and changes the shapes of things. If you know signal theory and practice a bit then you can compensate in your head (up to a point).


Knowing how to probe? That's a whole other layer of learning. The overshoots on your rising edges? Probably the way you connected the probe. A more expensive oscilloscope won't save you from knowing that.

Bottom line: Even the most expensive oscilloscaope won't just show you the beautiful waveforms you might be imagining when you arrive from a purely digital world. It takes knowledge and practice.

As Howard points out: $$$$ probes

[D3f1ant]

You can probably design and build the whole thing without measuring any hardware at all. After its built (and if it's not working) you might need to measure some clock lines, to verify they are actually clocking, but I very much doubt you need to measure anything on a 32bit wide memory bus. Again you might probe a line or two to ensure it looks ok.  Because your hardware will be fairly expensive to build, you would be wise to invest time in setting up software simulations to ensure theoretical signal integrity of your board design. A basic 100mhz scope will be fast enough to indicate if a 33mhz clock line is working or not. I'd say it would be a fairly ambitious project if you don't have a heap of pcb layout experience.

[smgvbest]

Since you don't need to measure anything that is connected to the CPU directly (like in the old days) you only deal with peripherals and these run much slower. In many cases you can slow the peripherals (I2C, SPI, etc) down too to make signals measurable if necessary.

I understand that. However, one of the things I would like to do is build a computer (perferably x86 architecture) from scratch, i.e. selecting and laying out the components onto a PCB. Lets use a 486DX4 100 MHz as an example, the front side bus on this runs at 33.3 MHz. According to scope sizing guides I would need a 333 MHz scope to accurately represent a 33.3 MHz square wave due to waveform rise times. Furthermore, the PCI bus is 32-bit and also the 72-pin SIMM is 32-bit, so wouldn't this mean I need a 32 channel logic analyzer?

Am I being realistic here with my goals and requirements?

I think there's confusing of Bandwidth of the scope versus it's Sampling Rate.   a 100Mhz Bandwidth scope can show that square wave just fine regardless if its a analog or a dso type  and for an analog scope that's the end of the question, 

For a DSO you now have to bring in the sampling rate.   This is why the DS1000Z series have 1Ghz sampling for 100Mhz.   That is of course for 1 channel, at 2 ch it's 500Mhz sampling and all 4 is 250Mhz per channel for the sampling rate.
if you use all 4 channels for a SPI bus then you have 100Mhz Bandwidth but 250Mhz sampling rate per channel.   Can it do what you're talking about I say yes,   would a higher sample rate DSO do better sure.  But the DS1000z can do it.  Also keep in mind,  do you really need to look at CS, CLK, MISO and MOSI.  if you just look at CLK and MISO for example you have 500Mhz sample rate.  again is it idea, no but you can do it.

And as has been said,  for Bandwidth,  that is the specified -3db point not the end of life of the scope.  with care you can go past it.
I feed in a 500Mhz signal in the mso1000z I have and at 1vpp in i still saw it but at <100mV (dont' remember the actual value)  but it was there.
also,  my MSO1000z I measured the bandwidth at 148Mhz by measuring the -3db point

That my 2 cents FWIW.

[Howardlong]

I'd be pretty unfazed (see what I did there) about using a 100MHz BW 250Msa/s per ch scope on a 33MHz bus.

Remember that a lot of what you're seeing isn't 33MHz, maximum data change rate on worst case is half that.

For things like RAS and CAS, you're looking at relative timing and phase differences, and for set up and hold times etc, 250Msa/s may just about cut it, if not go up to 500Msa/s per ch by only using 2ch and work from there.

Back in the days of the 33MHz bus CPU, in the early 90s, there were pretty much no DSOs that would accomplish single shot 1Gsa/s, but I stand to be corrected of course.

Edit: Changed some GSa/s to MSa/s due to operator error.