Is 100Mhz enough? (calculations + charts included)

[pepelevamp]

TLDR: Is the 100MHz fast enough for my project? Will I get sloppiness?

Hello friends.
You may know me from such threads as https://www.eevblog.com/forum/testgear/noob-scope-like-the-tds-2024b where I'm deciding on a scope. Well, I'm stuck between two scopes.

One scope is 200MHz 2ch, the other scope is 100MHz 4ch. I want the 4 channels, but I am concerned if 100MHz is enough for my project.

I'll give you the details of what I'm trying to view in the scope and the actual scope so you can see if it'll work.
 
I wish to sample a 20,000Hz sine wave made up of 256 segments. An 8-bit DAC, unsmoothed.
I need to see each of the 256 digital segments as clearly as I can. Eg, the more squarey the better.

20,000Hz * 256 = 5,120,000Hz.

Since I'm wanting to view square waves, I am aware that its a game of 'how many upper harmonics of my fundamental can the scope see? enough to make it squarey?'.
I am aware that 100,000,000Hz is in the neighborhood of 19x my 5.1MHz fundamental.

But I know that scopes roll off near their upper bandwidth limit and I am concerned I may be in for a shit sandwich.


Attached is a chart showing a square wave (perfect,red) at the granularity I want for my project. I basically need to see square waves that sharp in the scope. My teensy will be generating them. That part is fine.

I couldn't low-pass using desmos, but I thought I would construct the square wave out of harmonics available up to 25MHz, with the idea that near 50MHz the scope will be rolling off.
The blue line is sorta what I'd accept as passable. I just need to see it in steps, discretely. I dont want a sine. I know in real-life it wouldn't look like that - it'd be a blobby wave, but its just how I'm trying to estimate precision.

The scopes I am eyeballing are:
Siglent SDS1202X-E 200 Mhz, 2 Channels
Siglent SDS1104X-U 100 Mhz,  4 Channels.


Please excuse the crudity of the chart - I didn't have time to make it to scale or paint it.



SDS1202X-E
https://siglentna.com/product/sds1202x-e/
Key Features

* 200 MHz Bandwidth
* 1 GSa/s Real-Time Sampling Rate
* 14 MPts Waveform Memory
* 500 uV/div Sensitivity
* Maximum waveoform capture rate: 100,000 frames / second (Standard mode), 400,000 frames / sec (Sequence mode)
* Serial bus decode & trigger: standard
* Digital trigger provides low jitter
* Intensity-Graded display / Color-Temperature display mode
* Video triggering / HDTV
* Cursors * Measurement functions
* USB and LAN interfaces
* 7″ (diagonal) display
* Sequence trigger mode
* History function
* Secure erase feature


-----------



SDS1104X-U
https://siglentna.com/product/sds1104x-u/
 100 MHz bandwidth
 Real-time sampling rate up to 1 GSa/s
 Record length up to 14 Mpts
 Serial bus triggering and decoder supports protocols I2C, SPI, UART, CAN, LIN
 Advanced measurements on full memory (14 Mpts)
 Large 7-inch TFT-LCD display with 800 * 480 resolution

[gf]

I wish to sample a 20,000Hz sine wave made up of 256 segments. An 8-bit DAC, unsmoothed.
I need to see each of the 256 digital segments as clearly as I can. Eg, the more squarey the better.

If you want to see each step, then I would worry more about the vertical resolution, than about 100 MHz bandwidth.
What is the reason why you think that you need to see each individual step? What kind of problem are you trying to investigate?

[bdunham7]

The 8-bit DAC will give you 256 vertical levels, but what is the sample rate of the DAC?  Also what is the analog bandwidth of the output circuitry of the DAC?  You may have trouble seeing the 'squarey' signal as you might imagine it should be, but it won't be because of the bandwidth limitation of any of those scopes.

[gf]

The 8-bit DAC will give you 256 vertical levels, but what is the sample rate of the DAC?  Also what is the analog bandwidth of the output circuitry of the DAC?  You may have trouble seeing the 'squarey' signal as you might imagine it should be, but it won't be because of the bandwidth limitation of any of those scopes.

In the OP, the sample rate was said to be  5.12 MSa/s, and the DAC output was said to be "unsmoothed". So 100MHz is likey well sufficient. But I wonder how to resolve 256 step w/o missing codes, when the scope has only an 8-bit ADC? This could only work if the levels of the DAC and scope ADC were exactly aligned.

[dmulligan]

If you can stretch your budget a little more there is the SDS1104X-E which can be hacked to 200MHz.

[bdunham7]

In the OP, the sample rate was said to be  5.12 MSa/s, and the DAC output was said to be "unsmoothed". So 100MHz is likey well sufficient. But I wonder how to resolve 256 step w/o missing codes, when the scope has only an 8-bit ADC? This could only work if the levels of the DAC and scope ADC were exactly aligned.

The way he arrived at the conclusion regarding sample rates is questionable because he seems to imply that this is the natural result of using an 8-bit DAC.  As for the rest, let's just say I don't expect that he is going to see what he apparently expects to see in any case.

[Fungus]

I say "it's enough".  19x your fundamental is going to be very square, 200Mhz will make very little difference.

[tautech]

Well, I'm stuck between two scopes.

One scope is 200MHz 2ch, the other scope is 100MHz 4ch. I want the 4 channels, but I am concerned if 100MHz is enough for my project.

SDS1202X-E, SDS1104X-U
https://www.trademe.co.nz/Browse/Listing.aspx?id=3439619695
https://www.trademe.co.nz/Browse/Listing.aspx?id=3439619698

Yes those are the only 2 models we list on Trademe and at prices somewhat above what we can supply direct due to their exorbitant sales success fees. 
We have 4ch X-E and X+ DSO's too.
Drop by and try before you buy.

[kcbrown]

TLDR: Is the 100MHz fast enough for my project? Will I get sloppiness?

Hello friends.
You may know me from such threads as https://www.eevblog.com/forum/testgear/noob-scope-like-the-tds-2024b where I'm deciding on a scope. Well, I'm stuck between two scopes.

One scope is 200MHz 2ch, the other scope is 100MHz 4ch. I want the 4 channels, but I am concerned if 100MHz is enough for my project.

I'll give you the details of what I'm trying to view in the scope and the actual scope so you can see if it'll work.
 
I wish to sample a 20,000Hz sine wave made up of 256 segments. An 8-bit DAC, unsmoothed.
I need to see each of the 256 digital segments as clearly as I can. Eg, the more squarey the better.

20,000Hz * 256 = 5,120,000Hz.

Since I'm wanting to view square waves, I am aware that its a game of 'how many upper harmonics of my fundamental can the scope see? enough to make it squarey?'.
I am aware that 100,000,000Hz is in the neighborhood of 19x my 5.1MHz fundamental.

But I know that scopes roll off near their upper bandwidth limit and I am concerned I may be in for a shit sandwich.


Attached is a chart showing a square wave (perfect,red) at the granularity I want for my project. I basically need to see square waves that sharp in the scope. My teensy will be generating them. That part is fine.

What you're presuming is that the Teensy's slew rate is high enough to produce square wave harmonics that would exceed the 100 MHz scope's ability to show them.   That is highly questionable, at the very least.  At most it will depend on which Teensy you have. 

For the Teensy 3.2, which uses a Freescale MK20DX256VLH7, the slew rate is mentioned directly in the datasheet: at most 1.7V/us in high power mode.  This is as fast as it gets for that part, and for a 3.3 V P-P signal would result in at most a (roughly) 250 kHz signal.  The -3dB bandwidth of the DAC is 550 kHz at most, per the datasheet.

The Teensy 2.0 uses an Atmel ATMEGA32U4 microcontroller.  The datasheet doesn't specify slew rates or DAC bandwidths, but seeing how the speed of the MK20DX256VLH7 is at least an order of magnitude higher, it's highly unlikely that the ATMEGA32U4's DAC's performance will be any better, and it's likely worse.

The Teensy 4.0 (and 4.1) uses an NXP MIMXRT1062-DVL6A, which is another order of magnitude faster than the MK20DX256VLH7.  Its datasheet doesn't specify the slew rate or bandwidth of its DAC, so I can't say what to expect with it, but even if we assume an order of magnitude improvement in the slew rate, that's still easily within the 100 MHz scope's ability to see the edge clearly.


Put another way, the bandwidth of the output of your Teensy is almost certainly easily within the 100 MHz scope's ability to faithfully show anything you generate with it.

[Fungus]

The Teensy 2.0 uses an Atmel ATMEGA32U4 microcontroller.  The datasheet doesn't specify slew rates or ADC bandwidths, but seeing how the speed of the MK20DX256VLH7 is at least an order of magnitude higher, it's highly unlikely that the ATMEGA32U4's ADC's performance will be any better, and it's likely worse.

The ADC in those ATMegas (ie. Arduinos) can only handle about 10kHz. 

You can clock the ADC higher but there's no point, the internal sample-and-hold capacitor won't charge any faster, all you'll get is a low-pass filtered version of the signal.

If you switch between channels the capacitor won't fully charge and you'll get crosstalk between channels.

[NaxFM]

To me the 1104-U is really not worth the money.
The 1104-E, for 100 bucks more gives you much, MUCH more.
Easily hackable for 200MHz bandwidth, dual ADC (double the sampling frequency), mixed signal capabilities and usable with arbitrary function generation. The 200MHz bandwidth alone is worth the money.
If you're on a budget, just wait and collect those 100 bucks before buying the 1104-U, IMHO

Now, regerding seeing the steps. It's a bit hard with an 8 bit oscilloscope, you can but as PeDre demonstrated you have to apply a massive vertical zoom to see anything.

If I absolutely had to see the steps as clear as possible and have no budget limit, I'd go for the siglent SDS 2104 which has a 10 bits ADC up to 100 MHz, but that's well over 1000bucks... (Hackable to 500MHz tho, well worth the money)

[kcbrown]

The Teensy 2.0 uses an Atmel ATMEGA32U4 microcontroller.  The datasheet doesn't specify slew rates or ADC bandwidths, but seeing how the speed of the MK20DX256VLH7 is at least an order of magnitude higher, it's highly unlikely that the ATMEGA32U4's ADC's performance will be any better, and it's likely worse.

The ADC in those ATMegas (ie. Arduinos) can only handle about 10kHz. 

Sigh.  I typed "ADC" when I meant "DAC" (I've since corrected this).  But what I was looking up was the DAC specs.

Anyway, yeah, I'd expect the DAC performance in the ATmega to be around the 10 kHz range.

[kcbrown]

If I absolutely had to see the steps as clear as possible and have no budget limit, I'd go for the siglent SDS 2104 which has a 10 bits ADC up to 100 MHz, but that's well over 1000bucks... (Hackable to 500MHz tho, well worth the money)

The 1104X-E (and, I presume, the 1104X-U) has an "eres" acquisition mode where you can increase the effective resolution through oversampling.

[nctnico]

If I absolutely had to see the steps as clear as possible and have no budget limit, I'd go for the siglent SDS 2104 which has a 10 bits ADC up to 100 MHz, but that's well over 1000bucks... (Hackable to 500MHz tho, well worth the money)

No, it has an 8 bit ADC. There is a 10 bit mode, but that is still coming from an 8 bit ADC. If you want true 10 bit (or more), then you'll need to look somewhere else (R&S RTB2004 for example).

Then again, the OPs project goals make next to no sense at all so there is little use in suggesting anything. An oscilloscope is not a high precision instrument for measuring amplitudes anyway. If the OP wants to measure DAC steps accurately, then he/she needs to set the DAC to a specific output code (DC level) and use a >4.5 digit DMM to measure the DAC's output value.

[pepelevamp]

Hello friends. I'm absolutely staggered at the effort you've gone to with your replies. While I read through things here (which will take some time), I'll answer some questions about the project.

TLDR: I need to discretely and explicitly control the phase & timing of two channels. One analog, one digital. It sounds all ass about face, but I gotta do it this way for 'reasons'.

Voltage range & voltage-step-size isn't too much of an issue. I know the linearity of the 'DAC' and I'm more concerned about the timing - the stops/starts of these steps. I know they're in time - I just want a scope that will let me see it without too much smoothing of my edges.

The 'DAC' and microcontroller generating this is a solved problem: I'm using parallel-IO into a resistor-ladder. This is all working just fine on my 84MHz SAM3X8E (Arduino due).
My objective is to have two channels tightly in sync, one is a sine wave, and the other is digital logic. This is easy since I can push 32 bits in parallel. I'm using about 9 i/o pins at the moment. 8 of them go to the "DAC".

This is all just fine, as measured by the Tektronix TDS 2024B scope (200mhz) that I've borrowed. It all works perfectly fine. Problem is, I've returned the scope.

I'm just upgrading to the Teensy 4.0 (600MHz) because I need to do more floating-point math and conditionals inside my 5.1MHz interrupt routine than the SAM3X8E can handle. I know about not doing stuff inside ISRs - its necessary here but it'd take too long to explain why.

I assume the slew rate of the Teensy blows the Arduino Due out of the water & will also be of no concern.

This will all eventually get pushed out to some amplification, which may low pass things, but at that stage of the pipeline I'm fine with the smoothing. I'll filter off the sine wave at something phase-friendly. I just need explicit control over every step of that wave form before its smoothed into the sewer.


Thank you all for your huge replies - I'm going to fully read through them now. You guys are heroes.
 

[pepelevamp]

If you want to see each step, then I would worry more about the vertical resolution, than about 100 MHz bandwidth.
What is the reason why you think that you need to see each individual step? What kind of problem are you trying to investigate?

I'm trying to have digital logic signals go high/low at the exact time of my choosing relative to the 256-step sine-wave. Ive slowed everything down & checked the vertical resolution & my generator is a-okay. I need to see every individual voltage step so I can be certain of the timing between my digital logic & steppy sine wave.

My question is ultimately - is the 100MHz scope fast enough to let me still see my signals before its own bandwidth limitation starts rounding everything off?

In the future it may not be a simple sine wave Its software generated.

[pepelevamp]

The 8-bit DAC will give you 256 vertical levels, but what is the sample rate of the DAC?  Also what is the analog bandwidth of the output circuitry of the DAC?  You may have trouble seeing the 'squarey' signal as you might imagine it should be, but it won't be because of the bandwidth limitation of any of those scopes.

Its all good mate, I made the DAC out of resistors. I've had it working already. I'm just giving back the scope I'd borrowed, and and I'm uncertain if the scope I wish to buy will be able to see the signals I'm making.

[Fungus]

The ADC in those ATMegas (ie. Arduinos) can only handle about 10kHz. 

Sigh.  I typed "ADC" when I meant "DAC" (I've since corrected this).  But what I was looking up was the DAC specs.

Anyway, yeah, I'd expect the DAC performance in the ATmega to be around the 10 kHz range.

They don't have a DAC. 

[pepelevamp]

If you can stretch your budget a little more there is the SDS1104X-E which can be hacked to 200MHz.

OooOooo. So the U is the wimpy version of the E, but kinda 'pretend' wimpy where its downgraded in the firmware aye. This is a rabbit hole worth looking into.

[pepelevamp]

To me the 1104-U is really not worth the money.
The 1104-E, for 100 bucks more gives you much, MUCH more.
Easily hackable for 200MHz bandwidth, dual ADC (double the sampling frequency), mixed signal capabilities and usable with arbitrary function generation. The 200MHz bandwidth alone is worth the money.
If you're on a budget, just wait and collect those 100 bucks before buying the 1104-U, IMHO

This is very good information. I hadn't noticed that the 'E' version of the 1104 was available. You're right - its a much smarter option. You may have actually touched the magic spot & convinced me. I'm going to pursue down this rabbit hole.

Now, regerding seeing the steps. It's a bit hard with an 8 bit oscilloscope, you can but as PeDre demonstrated you have to apply a massive vertical zoom to see anything.

If I absolutely had to see the steps as clear as possible and have no budget limit, I'd go for the siglent SDS 2104 which has a 10 bits ADC up to 100 MHz, but that's well over 1000bucks... (Hackable to 500MHz tho, well worth the money)

Damn. Yeah this would be where I'd go if my needs were simply multiplied up to 2x and I could spend 2x the money. Thanks mate, this helps because it sorta gives me an upper-bound & calibrates my expectations of features<>costs.

[pepelevamp]

Then again, the OPs project goals make next to no sense at all so there is little use in suggesting anything. An oscilloscope is not a high precision instrument for measuring amplitudes anyway. If the OP wants to measure DAC steps accurately, then he/she needs to set the DAC to a specific output code (DC level) and use a >4.5 digit DMM to measure the DAC's output value.

Saying project goals make no sense when the project goals weren't specified -that's a paddlin.
Saying theres no use in suggesting anything when the performance metric was specified - that's a paddlin.
Giving a course of action based on the absence of information, and when by your own admission you don't know of a project goal- that's a paddlin.

The world will be a better place without people such as yourself who lean into providing 'advice' yet belittle projects at the same time - all before you even understand what is happening.
Nobody needs that bullshit, and I won't tolerate it.

I'm after help about scope roll-off & bung frequency response and others here have been very helpful. 

[pepelevamp]

What you're presuming is that the Teensy's slew rate is high enough to produce square wave harmonics that would exceed the 100 MHz scope's ability to show them.   That is highly questionable, at the very least.  At most it will depend on which Teensy you have. 

For the Teensy 3.2, which uses a Freescale MK20DX256VLH7, the slew rate is mentioned directly in the datasheet: at most 1.7V/us in high power mode.  This is as fast as it gets for that part, and for a 3.3 V P-P signal would result in at most a (roughly) 250 kHz signal.  The -3dB bandwidth of the DAC is 550 kHz at most, per the datasheet.

The Teensy 2.0 uses an Atmel ATMEGA32U4 microcontroller.  The datasheet doesn't specify slew rates or DAC bandwidths, but seeing how the speed of the MK20DX256VLH7 is at least an order of magnitude higher, it's highly unlikely that the ATMEGA32U4's DAC's performance will be any better, and it's likely worse.

The Teensy 4.0 (and 4.1) uses an NXP MIMXRT1062-DVL6A, which is another order of magnitude faster than the MK20DX256VLH7.  Its datasheet doesn't specify the slew rate or bandwidth of its DAC, so I can't say what to expect with it, but even if we assume an order of magnitude improvement in the slew rate, that's still easily within the 100 MHz scope's ability to see the edge clearly.


Put another way, the bandwidth of the output of your Teensy is almost certainly easily within the 100 MHz scope's ability to faithfully show anything you generate with it.

Thanks mate this is great advice. I appreciate your help going into detail with the slew rate.  Yup I got the Teensy 4.0. I'm really going hardout on the CPU.
I'm using parallel-IO into a resistor ladder instead of using the built-in DACs because its been too hard to sync the DAC to digital IO.

Its way, way, waaaaaaaaaaay faster - at least on my SAM3XE on my Arduino DUE. Waaaaaaaaaaaaay faster.

Slew rate on the DAC had indeed  been a problem - trying to sync things up when the slew rate is all sloppy woppy.

Thats why I'm using a resistor-ladder from parallel-IO. Super duper fast. Later on I'll be amping things up & bandwidth-limiting etc - but by that stage I know everything's in sync etc.

[pepelevamp]

I have recreated the question out of interest with the DS1104Z (8-bit, 100 MHz) and RTB2000 (10-bit, 300 MHz), perhaps the screenshots will be helpful.
The arbitrary signal comes from the Rigol DG4162, and is terminated with 50 ohms at the oscilloscope. The 2 Vpp, 20 kHz sine consists of 256 parts in the vertical.

Peter

Dude this is extreme. You really went to town for me here. Thank you heaps.
Its quite clear on the 100mhz 8-bit guy that when zoomed out the top of the wave gets pretty hairy. But zooming in lets you see the voltage steps quite well, even if those straight lines have a 5 o'clock shadow.

I really appreciate you going to the effort of creating these charts. As one might expect the global amplitude of the 256-step sine wave does affect the temporal precision because it requires a faster slew rate to accomplish. It does look like 100MHz will be enough to capture what I want to see.

I guess the next question is the other specs - the maximum precision that the scope can see. Mind you, for 2v peak to peak divided by 255 is 7.5mV. This scope I'm eyeballing can see 0.5mV. So I'm in good company. Thank you mate.

[pepelevamp]

The ADC in those ATMegas (ie. Arduinos) can only handle about 10kHz. 

Sigh.  I typed "ADC" when I meant "DAC" (I've since corrected this).  But what I was looking up was the DAC specs.

Anyway, yeah, I'd expect the DAC performance in the ATmega to be around the 10 kHz range.

They don't have a DAC. 

Where we're going - we don't need DACs.

[tautech]

I have recreated the question out of interest with the DS1104Z (8-bit, 100 MHz) and RTB2000 (10-bit, 300 MHz), perhaps the screenshots will be helpful.
The arbitrary signal comes from the Rigol DG4162, and is terminated with 50 ohms at the oscilloscope. The 2 Vpp, 20 kHz sine consists of 256 parts in the vertical.

Peter

Dude this is extreme. You really went to town for me here. Thank you heaps.
Its quite clear on the 100mhz 8-bit guy that when zoomed out the top of the wave gets pretty hairy. But zooming in lets you see the voltage steps quite well, even if those straight lines have a 5 o'clock shadow.

Yes, as expected when looking closely at a 14 bit AWG waveform, nothing too special in that.