Is DAC ultrasonic/RF output important?

[Arpiben]

As always one should know the limits of one's measuring equipment. This is true for analog scopes as well...[/QUOTE]

Fully agreeing. Spectral analysis also may turn quite difficult with limited ADC front end analog stages.
This is true with Picoscope as well as some well known (and expensive) 110 /170 GHz USB spectrum analyzers. Recently trying one I encountered strong aliasing with only two carriers at first trial.

 

[chris719]

I second that with a Tek DSO we have in the company. I always use the old Tek 2465 to find signals in our NMR spectrometers. But that's just me. All my younger colleagues hate the 2465 and much prefer the DSO. I guess it's just what you're used to work with, and that in my case the manual of the DSO is not available on the spot.

You can always press the AUTO-SETUP knob on a DSO, it should show you something for a start.

Having my own private DSO at home now (Siglent SDS 1202X-E) I can say that it is actually a joy to use. Of course you have to read the manual (220 pages!) to be able to use its higher functions but it can do so much more than my old Tek465. And you can easily shoot yourself in your foot as I did, of course.

As always one should know the limits of one's measuring equipment. This is true for analog scopes as well. I recall debugging our control software with my boss for several hours because a 300 MHz RF pulse was missing until I realized that the Tek 465 we had at hand is a 100 MHz scope (I was used to a 485 which is 300 MHz) and just could not show 300 MHz. There was no bug in the software.

Certainly there are differences between analog and digital scopes. I find the software is a big differentiator among DSOs also. The Keysight DSO I use mostly at work is the better of the scopes I have access to, but I find the Teledyne LeCroy software to be more intuitive. If you are used to an analog scope, you'll probably like the higher end Agilent/Keysight DSOs because of the very high waveform capture rate - they are good at finding infrequent events and runts that other digital scopes miss.

 

[DonH56]

Certainly there are differences between analog and digital scopes. I find the software is a big differentiator among DSOs also. The Keysight DSO I use mostly at work is the better of the scopes I have access to, but I find the Teledyne LeCroy software to be more intuitive. If you are used to an analog scope, you'll probably like the higher end Agilent/Keysight DSOs because of the very high waveform capture rate - they are good at finding infrequent events and runts that other digital scopes miss.

Interesting... We use Tek for the same reason... But we have both in-house and I have not played with Keysight as much.

 

[Music1969]

This Picoscope is the low-end one, not enough memory for anything but 4k FFTs. Going to swap it for one with more memory (enough for 1M FFT size)

Which model did you end up getting @pkane ?

 

[pma]

This Picoscope is the low-end one, not enough memory for anything but 4k FFTs. Going to swap it for one with more memory (enough for 1M FFT size)

It is a good tool in case that one knows what he is doing. It needs knowledge, experience and understanding of the issue, to avoid mistakes, but then it is a good tool. Yes not enough to a plain beginner.

 

[pkane]

It is a good tool in case that one knows what he is doing. It needs knowledge, experience and understanding of the issue, to avoid mistakes, but then it is a good tool. Yes not enough to a plain beginner.

A 4k FFT is fine for simple things, but not for the distortion measurements I was looking for. It's a nice tool, but one with more memory is better for some things. Maybe you can tell me how to get a greater frequency resolution than about 23Hz/bin from a 4k FFT in Picoscope when measuring a 96k signal?

 

[DonH56]

Y'all know this already, but for non-EEs reading this, the main problem with low (short, small) record length is keeping the sampling rate high when you have broadband signals. If you are looking for low-frequency noise modulating a high-frequency signal (e.g. power supply noise or LF clocks on a Gb/s data streams), or looking for high-frequency spurs riding on low-frequency signals (e.g. SMPS noise or RFI on audio signals), you need a long enough record to capture the low frequencies whilst not aliasing (or missing) the high frequency content. That is the main reason the scopes I use at work have 64 to 128 million sample memories (record lengths) to go along with their 33~100 GHz bandwidth. Capturing 10 MHz clock noise on a 16 GHz signal sampled at 100+ GS/s takes a lot of memory.

The bandwidth is lower, but if you are looking across say 10 MHz to catch SMPS spurs sneaking into you audio signal, then you probably need to sample around 50 MS/s (most DSO companies suggest 5:1 oversampling), thus capturing a single 10 Hz cycle takes about 5 million samples (50 MS/s / 10 Hz). The sampling rate must be more than twice the highest signal frequency of interest (technically bandwidth, but assume we are going from DC on up), so you must sample very fast to catch the highest frequencies without aliasing. If you need to measure low frequencies at the same time, you need enough samples to capture a cycle (usually several cycles) of the low frequency. That means a long record length (memory depth) is needed.

Example:

• Goal is to capture noise up to 1 MHz with minimum signal frequency of 10 Hz (desired bottom of the audio band)
• 10 Hz period = 100 ms (0.1 second)
• 1 MHz period = 1 us (0.000001 second)
• Assuming 5x oversampling is needed by the DSO to accurately resolve the signal so it is sampling at 5 MHz, a period of 0.2 us
• Now 0.1 second / 0.2 us = 500,000 samples are needed to capture both the 10 Hz to 20 kHz audio band and noise spurs up to 1 MHz

You can filter and sample a set of limited frequency bands, assuming the 'scope's front end has that capability or you have external filters to do it, to reduce the record length requirements. That way you can look at just the audio band using a lower sampling rate to optimize the memory depth (sample memory, record length -- different manufacturers use different terms), and then lower (narrow) bandwidths centered at higher frequencies using a higher sampling rate, like a window around 1 MHz if you suspect noise there. Then stitch the bands together in post-processing if needed.

FFTs impact sampling rate and record length because you need a long enough record in time to capture the lowest frequencies of interest, but at high enough rate to not only capture the highest frequencies but also provide narrow enough frequency bins, or good enough frequency resolution, to pick out individual signals across the spectrum. Too few points means too many frequencies land in the same "bin" of the FFT so you cannot tell them apart. For the previous post's example, sampling at 96,000 S/s with record length of 4000 points yields 96,000 / 4000 = 24 Hz frequency steps. If you have frequencies closer than 24 Hz, they will fall into the same FFT bin, and you won't be able to tell them apart. Sometimes that does not matter, depends upon what you are trying to measure. Or you can use other methods such as filtering, capturing multiple records (and aligning them in time, which can be tricky), or various interleaving techniques to provide an effectively larger record.

HTH - Don

 

[Music1969]

Does anyone know if any labs / suppliers sell any used Picoscope 4262 ?

North America or Europe ?

Nothing coming up in Google but I thought you experts might know contacts for used / refurbished

 

[DonH56]

There are a number of used test equipment companies in the USA, and I would guess in Europe as well, but their stock varies. I don't have a list but have managed to somehow get on several mailing lists at work. I think Test Equity is one, Liberty another, don't remember the other few. I am not sure I have seen a Picoscope listed but I do not look at them often. eBay is another potential source. When I did a quick search for "Picoscope 4262" a few used units came up in eBay and some used equipment sites so you might try that. My search skills are not that great...

I picked up a Digilent kit a few years ago but haven't taken it out of the box yet. It's an Analog Discovery 2 bundle that runs around $450 but was on sale for ~$300 or $350 a couple of years ago. Probably not as robust as the Picoscope units, but you get a lot for the money: https://digilent.com/shop/analog-di...ope-logic-analyzer-and-variable-power-supply/ -- that is just the measurement device, the bundle adds a bunch of accessories.

 

[mareclodaniel]

Are you sure you're not undersampling those high frequencies and getting aliasing? That's the only way I can produce anything resembling your result.

Speed Test​

 

[BDWoody]

I picked up a Digilent kit a few years ago but haven't taken it out of the box yet.

This thread made me want to dig out the QuantAsylum USB scope I picked up a couple years ago but also never opened.

I like the 'FYI' printed on top of the case. I wonder how many figure they can start poking probes into stuff before having enough of the basics down to be safe.

The more I learn, the more I'm amazed I haven't killed myself by now.

 

[DonH56]

The more I learn, the more I'm amazed I haven't killed myself by now.

I started as a TV/stereo repair tech after years of Radio Shack, Heathkit, HAM radio, and misc. DIY stuff and have the scars to prove it. Worked through college in the EE lab and tech jobs at a couple of local stereo and TV shops, plus worked as an electrician for a while (as well as odd jobs working on a ranch, running wheat harvest, radio station under 1st class supervision, etc.) High-power RF transmitters (HAM stations, radio stations) can give you a nasty burn before you realize what is going on! Working on ICs was a pleasant change, no HV stuff past the power supplies, no burns from brushing the anode of the CRT or flyback transformer, etc. I did have way more practical experience than most of my classmates in college; sometimes helped, sometimes not.

I dug out my Digilent box so at least it's on top of the pile, then spent another 12-hour day at work and put it aside again. Maybe when (if) I retire...

On the subject at hand, highly-oversampled (delta-sigma) DACs and switch-mode power supplies (SMPS) can generate spurs into the MHz region, and may excite much higher resonances. A few years ago a SMPS provided the launch energy into what turned out to be a resonant power plane that sprayed 300 MHz common-mode noise throughout a large server (hundreds of disk drives used in an enterprise storage array). Debugging that was painful. One of the clues was that drives at regular (physical) intervals failed, and that interval correlated with the resonance frequency.

The basic arguments to me are (1) equipment operating in the audio band should have sufficient EMI/RFI rejection to block such problems, and (2) audio equipment should be shielded and filtered so as not to cause such trash to begin with. For audio DACs, image suppression should be high, so clock and SMPS noise would be the main culprits unless there is some sort of resonance/oscillator elsewhere. One problem I have seen are that very wideband op-amps may be used for I-V converters to provide fast recovery (providing those very clean edges before the output filter), and if not well-shielded and decoupled they provide a very nice "sneak" path for RF radiation. For power amps, a common sneak path (again IME) is from RFI impinging on the speaker wires that manages to sneak through the feedback network to the input stage and create a nice RF oscillator. Those can be insidious; the circuit's bandwidth will often prevent them from reaching full scale, but you can't hear them, so a few watts dumped into the tweeter constantly for hours can kill it. Sometimes tweeters get replaced several times before someone (ok, me) thinks to put a 'scope or spectrum analyzer on the speaker terminals to see what is going on, especially since it may only happen in the customer's system and not on the bench into a load resistor. I had some big caps and inductors around to do some spot checks for oscillation after amp repairs.