Memory Depth in Basic Oscilloscope

[dave_robinson_022]

Hi,

First of all I am a machine guy (mills, lathes, etc) who works with electronics sometimes.  As part of getting into electronics more I began building a bench power supply and realised that an oscilloscope would be useful; I also work with automation ie steppers and servos.

The thing is I have a lot of expensive toys and I can't really justify a large cost for something I may not use on a regular basis.  I have read a lot about the requirements of memory depth in a scope and understand why more is better.

What I haven't got to the bottom of is; what exactly would a specific memory size limit me to... i.e if I have a 100MHz 1GS/s scope; what is the tangible difference (not mathematical difference) in what i can use 25Kpt, vs 40Kpt vs 1Mpt?

Also, is it ever possible to upgrade memory on a scope or is this an impossibility with firmware etc?

[ataradov]

What I haven't got to the bottom of is; what exactly would a specific memory size limit me to... i.e if I have a 100MHz 1GS/s scope; what is the tangible difference (not mathematical difference) in what i can use 25Kpt, vs 40Kpt vs 1Mpt?

At 1GS/s you can save 40 us worth of data with 40 Kpts, but 1 ms at 1 Mpts.

In practice you chose how much time you want to capture, and the scope automatically selects appropriate sample rate based on the amount of the available memory. So the more memory you have, the higher actual sample rate would be for a given time span you want to capture.

Also, is it ever possible to upgrade memory on a scope or is this an impossibility with firmware etc?

Impossible with exception of a few cases where memory is actually installed, but sold separately as an option. Rigol 1054z is one of those devices. And the new ones come unlocked anyway.

[rstofer]

In recent times, the Rigol DS1054Z (when unlocked to 100 MHz, details elsewhere) is quite popular.  For me, the 4 channels was important.  I haven't had to use it on my CNC mill but I would still rather have 4 channels than just 2.

https://www.amazon.com/Rigol-DS1054Z-Digital-Oscilloscopes-Bandwidth/dp/B012938E76

For 2 channels, the newer Siglent SDS 1202X-E seems to be the latest and greatest for bandwidth (200 MHz) while the SDS1104X-E is a 100 Mhz 4 channel version.  The 200 MHz 4 channel is a little out of my price range but apparently the 1104X-E can be unlocked to 200 MHz, details elsewhere

https://siglentna.com/digital-oscilloscopes/sds1000x-e-series-super-phosphor-oscilloscopes/

These 3 are currently trending (to borrow an awful phrase) with the Siglent purported to have a much nicer user interface.  But it's a few years newer...

[dave_robinson_022]

So how do I determine how much time I actually need and therefore how much memory I.e I want to look at opamp output on a power supply, or I want to look at output from a stepper driver, or something on the SPI bus?

Is for example the Hantek Dso5102p with its 40kpts so bad? If so who buys them?

What could I not do with a scope like this? Or are things just more difficult and time consuming?

[ataradov]

That depends on what you want to do with that data. What are you looking for?

If you are talking about SPI bus and want to capture meaningful amounts of data, then  you want more memory.

Generally I would say it is not worth having a new general purpose digital scope with less than 1Mpts of memory.

[TK]

Having a scope that markets a specific sample memory does not mean it will use it all the time or when you set the value manually.  Siglent scopes are known to use only what is needed for a given sample rate, and what you are setting is the MAXIMUM limit. 

i.e. you set max sample memory to 20Mpoints, but it will not use it if your timebase only requires 20Kpoints, it will use 20Kpoints.

They also stop decoding serial protocols like SPI at 3000 packets, even when you are sampling over 10,000 or 15,000 packets.

[vk6zgo]

Very old DSOs, or new "very cheap" ones have a problem due to lack of memory depth.

This is very apparent when doing one of the standard tests with analog TV waveforms.
When you need to look at a video waveform at "field rate" (50 or 60 Hz) it is necessary to use a time/div setting of around 5ms/div.(2ms/div is a bit too fast).

The Oscilloscope assumes you are looking at a sine wave at that frequency, so reduces the sampling rate so as not to exceed the memory depth.

Unfortunately, a video signal is not a sine wave, as it contains frequency components up tpo 5MHz, which is well above the reduced sampling rate.

The resulting aliased display is totally unuseable!

The analog video example is a "worst case" scenario, but if you want to look at Mains rate "hum" on any pulse train with a substantial high frequency component, the same problem will occur.

Modern, relatively inexpensive DSOs like the Rigol DS1054Z have adequate memory depth, & do not exhibit this problem.

[David Hess]

What I haven't got to the bottom of is; what exactly would a specific memory size limit me to... i.e if I have a 100MHz 1GS/s scope; what is the tangible difference (not mathematical difference) in what i can use 25Kpt, vs 40Kpt vs 1Mpt?

A longer record length allows for a higher sample rate at slower time/div settings, assuming of course that the maximum sample rate has not already been reached.  In some cases however, a longer record length has the disadvantage of lowering the display update rate because the processor is not able to handle the data quickly enough.

How much is enough?  Unless you are capturing long runs of digital data to decode, I would not worry about it.

Also, is it ever possible to upgrade memory on a scope or is this an impossibility with firmware etc?

Optionally longer record length used to be common but now it is pretty rare.  This is because memory is less expensive than it used to be and more tightly integrated which makes replaceable memory more difficult.