Analog..digital.. grr (oscilloscopes)

[TriodeTiger]

I found a beautiful analogue Tek 2215 on sale in my price range (<200$) used of course, but am incredibly confused by what to expect.

Can one set the seconds per division to 5, and see the levels plotted out, or is that just for a DSO? For example, setting it to 5s/div and watching a capacitor discharge with nice curves shown all on one screen. Something I'd like to play with, not all too important.

If not, then how does it display more than one wave form at a time? Does it just repeat, and the goal is to get one cycle centred to be viewed well?

I intend to work with AM radio for hobby (just in-house low powered @ 1MHz), would it matter at all to have an FFT function in this case to view harmonics? Harmonics as in bleeding in to other channels, as my square wave oscillator (to be replaced) currently does, or does making it (or making) a sine wave remove the need to have FFT (and DSO) any longer?

I just cannot find anything anywhere about it. I am ruining my health by stressing myself so much.

[amspire]

Analog scopes are designed for repetative waveforms, so to see a capacitor discharge curve, the ideal is to repetitively charge and discharge it using perhaps a function generator to control the charge/discharge cycles.

At the slower scan speeds and with the brightness all the way up, the screen persist is long enough so you can sort of see the trace of a single shot event, but it is really ignoring trying to remember a waveform that is only visible for a fraction of a second.

At 5s/div, you will basically see a bright dot moving slowly across the screen showing the current voltage across the capacitor - you need  faster scan rate to actually see a proper line. I do not know what the slowest horizontal scan rate of the 2225 is but below 100mS/div, it is not much fun to use.

There is no form of capture to computer - that is DSOs only.

It can show two traces on the same screen. It does this either in Alternate mode where it draws the top trace fully and then the next scan it draws the bottom trace. The alternate method is Chop that is better for lower scan rates. It draws a bit of the top trace, then a bit of the lower trace,a bit of the top trace, and so on, so you still see it as two traces.

In general, an analog scope is good for AM radio work, but you can only trigger on either the rf AM frequency or the AM modulation at one time. A DSO can capture both in long memory and you can zoom in and out. The analog scopes do not have FFT, but the FFT in scopes it not great for RF. If you stick with RF, you will probably want to get a proper spectrum analyzer one day. I am not sure what you mean by "harmonics bleeding into other channels". The first harmonic of a 700kHz AM signal will be at 1400kHz. Is that what you mean?  If you mean bleeding into adjacent channels, then that is a filter problem, not a harmonic problem.

Overall though, I think the 2225 is an excellent scope and definitely useful for the type of work you want to do. An analog scope is a good thing to have as they will probably be harder to get as everyone starts buying digital scopes.

Richard.

[vk6zgo]

Before you go too far along this track,check your local licencing laws re this sort of stuff.
Canada may have considerably less favourable views on "low powered" MF AM transmitters than the USA,where a lot of information on this comes from.

If you are going to have a signal which won't produce multiple harmonics,you need to start with a sinewave oscillator.
Your square wave is useful.but not for Broadcasting!
I think 5 seconds/ div may be a bit slow for an Analog "scope,but if you can externally trigger it,maybe/

Must go now see you later!

[free_electron]

At 5 seconds per div your only option is to hook up a polaroid camera, complete a full sweep, pull out exposed film and develop.... That is howw it was done in the dinosaur age. Jim williams ,rip, was a master at that. Heck i sometimes do it for very special apps on machines that do not have modern equivalent ( like curvetracers. They are sometimes handier than building a setup with a sourcemeter and writing a sweep script. )

Today... Get a scope with some decent memory and fast sampling and off you go. There's plenty ones out there at a fair price.
You are better of saving up some money and going after a good used one . And beware of ebay. Lots of floggers think they are sitting on gold.

As for fft ...the way it is implemented in cheap scopes it is just a mathematical gimmick. Nowhere neer enough point are fast enough acquisituin to give any kind of meaningfull view. Good luck finding a spirious sideband signal if you only have 300 points and your rescan rate is once a second.

Here is a problem with digital scopes at the bottom end : they are grossly underpowered when it comes to processing. Sure theycmay have a fast sampler, but, then they have to process the samples, scale them and translate them into pixels on the screen. Screen refresh is slow, even in lcds... And thenthey have to some calculation to align cursors and show you some numbers. During this processing time there is no sampling taking place. The entire aquisitiun system sleeps. It can be so bad that the scope is blind for 99% of the time... Anything that happens there you will never see.

That is why the real scopes use multiport memory and dedicat processing hardware. Those can crunch the dat as it is being sampled and they dump the result in a memory that can simultaneously be accessed by the display engine. This is not something you can do in an fpga based scope like the little toasterboxes. You need some heavy iron asic to do that. One of my colleagues works in the division that performs the backend design of these asics... Not easy stuff and extremely expensive silicon, out of reach.

So, don't be mislead by a pretty mathematical trick in. A 200$ scope... Its pretty mich useless for any kind of serious work... Looking at a carrier. Ok. Looking at a harmonic , ok .. Looking at sidelobes of a modulated signal. Tricky as your processing is too slow... Looking for spurious stuff.. Forget it.

[TriodeTiger]

I am glad I could figure out the difference finally, analogue looks neat but I probably would not find it as useful without storage. I am tempted to buy it just to have a Tek, seems a little shallow, who cares if I don't?

As for fft ...the way it is implemented in cheap scopes it is just a mathematical gimmick.

For some contest (555? Renases?).., someone amplified signals received from an antenna and used FFT (built in, or made themselves) showing all sorts of signals being broadcast, shortwave, etc. like a DIY spectrum analyser. Just got me interested, I love viewing such things. Just a dream.

The first harmonic of a 700kHz AM signal will be at 1400kHz. Is that what you mean?  If you mean bleeding into adjacent channels, then that is a filter problem, not a harmonic problem.

That makes complete sense. The only AM broadcast thing I could find for my Arduino (on hand,without amplification/tuning/inductors) was a 1MHz square wave done with interrupts or something crazy. I had no clue how to measure or know what it was doing, but it broadcast in nearly every single frequency on the AM dial (and below FM), major noise, but it transmitted morse code 2 feet.

I am purchasing one of those XM-something function generators to create a carrier and do some more simple things. I thought I really needed a scope to see why it was transmitting on every station (even with simple RC filtering), but I am using unknown code that I do not understand anyway. I'll start with a clean waveform and go from there.

Overall though, I think the 2225 is an excellent scope and definitely useful for the type of work you want to do. An analog scope is a good thing to have as they will probably be harder to get as everyone starts buying digital scopes.

It's a 2215, I am not sure how that differs, but it appears to be 60MHz rather than the 2225 (50MHz, which Dave reviewed when I looked it up, I remember his video now.)

I'll find the scope very useful to diagnose wave forms I wish to create and all that, no need to go above 60 megs or get a silly digital scope that is just so darn expensive for the noise, shoddy firmware and construction quality.

You guys have helped me process a lot. Thank you, cannot wait to review and use it for real things.

[elliott]

If you have fairly simple digital needs, then maybe an older Tek hybrid scope would suit you. I have a 2221A and it can do a lot of what you are asking for. If you out grow its abilities, then get a modern digital scope, press the Store/Non-Store button on the Tek and you have a great analog scope for when you need it.

[TriodeTiger]

The history of these things sparks my interest a lot -

At 5 seconds per div your only option is to hook up a polaroid camera, complete a full sweep, pull out exposed film and develop

Just reading that gave me a warm fuzzy feeling

Learning to work with it, and how it works, will teach me a lot along the way.

Alexander.

[vk6zgo]

First thing to do is to wrap the Arduino up in Aluminium foil & put it in the back of the cupboard.
You can use a squarewave & lowpass filter,but at this point it is a lot harder than just starting out with a sinewave
oscillator.

amspire & free_electron have pretty much covered what I was going to say,but I'd like to make the following points:

Your 2215 seems to have delayed timebase,so you can look at the RF signal at both the modulation  frequency rate,& the carrier rate,but it is a bit fiddly,& you will be playing with triggering to get a useable result.
Looking at video signals at both field & line rate simultaneously is very much easier,because they have a fixed phase relationship,whereas your mod & RF frequencies will not.
w2aew has a number of useful links to Oscilloscope tutorials in another thread.

Storage is not a big deal with AM modulated RF signals---most of your work will be realtime.
You will mostly be looking at the modulated RF signal to see if you are modulating correctly.

"Overmodulation" is the big bugbear,----if you modulate over 100%,you will "cut carrier" & produce,not only numerous harmonics,but also extra sidebands spaced at multiples of the modulating frequency.

Initially checking modulation will be  done using a fixed tone--1kHz is common.
The modulation envelope should follow a true sinewave shape,again making sure that you don't overmodulate .
Any deviation from a true sinewave shape indicates distortion.

An oscilloscope display is not a very sensitive indicator of distortion,so in Commercial Broadcasting,a device called a Noise & Distortion Test Set (N&D set)  is used.

In use,a high grade receiver,or a Modulation Monitor is coupled to the Transmitter  output,& the demodulated audio fed to the N&D set.
The Transmitter is then fed with a fixed tone,as before.
440Hz (middle "C" on a piano) is often used.*
*See comment by baljemmett
A tuneable filter is used to reject the original tone,&
what is left is distortion +noise.
A soundcard type audio spectrum analyser fed from the same audio output could probably be used instead of the N&D set.

If you are worried about RF harmonics,etc being generated,you could look for them with a Shortwave Radio Receiver.

[baljemmett]

440Hz (middle "C" on a piano)is often used.

Sorry to nitpick, but 440Hz is the A above middle C.  At least, it is in modern standard concert pitch; historically the accepted standard frequency for that note has varied a little, but middle C is significantly lower (around 260Hz given A440 tuning).

[vk6zgo]

Yep you're right, it's a long time since I've tried to play the piano,but I shouldn't have been caught by that one!
I had a little niggle of doubt at the time,& nearly deleted it!

[Kilroy]

At 5 seconds per div your only option is to hook up a polaroid camera, complete a full sweep, pull out exposed film and develop.... That is howw it was done in the dinosaur age.

Um...what about analog variable persistence storage scopes? I'm still getting lots of real time love from an HP1741A and an HP1201A. Just beautiful scopes for any kind of slow speed work.

HP sure took it all off when it came to designing variable persistence storage CRTs. They were very good at that stuff.

[free_electron]

Um...what about analog variable persistence storage scopes?

yeah. Tek and Gould also had storage scopes using a floodgun in their tubes. But those too have gone the way of the dinosaur ....
Good luck finding a scope today that does NOT use an lcd display ...

[tekfan]

At 5 seconds per div your only option is to hook up a polaroid camera, complete a full sweep, pull out exposed film and develop.... That is howw it was done in the dinosaur age.

You can still easily do it today with a digital camera that has variable shutter time. Although you have to do the photos in either complete darkness with just a spot barely visible on the CRT or with a hood/lightshield thing between the CRT and camera to elliminate stray light from entering. You can get very good results using this method.

[free_electron]

You can still easily do it today with a digital camera that has variable shutter time. Although you have to do the photos in either complete darkness with just a spot barely visible on the CRT or with a hood/lightshield thing between the CRT and camera to elliminate stray light from entering. You can get very good results using this method.

You could do that. But why go to the trouble. Just hit the quicksave button and the machine dumps a png file straight onto the file server.

[slateraptor]

That is why the real scopes use multiport memory and dedicat processing hardware. Those can crunch the dat as it is being sampled and they dump the result in a memory that can simultaneously be accessed by the display engine. This is not something you can do in an fpga based scope like the little toasterboxes. You need some heavy iron asic to do that. One of my colleagues works in the division that performs the backend design of these asics... Not easy stuff and extremely expensive silicon, out of reach.

I'm curious to hear your justification for the highlighted comment above. If FPGAs are good at anything, it's custom pipeline architectures, which would be the natural sol'n to the repetitive nature of the problem you've described.

[free_electron]

Let me rephrase it ... : This is not something you can do in 'toasterbox scopes based on a cheap 10$ fpga'

Sure you can do it in an fpga ... but it aint gonna be a 10 $ fpga like in the toasterbox scopes. Its going to be a 1000$ fpga ...
The little guys have no where near enough memory on board ( you need multiported ram... pipelining is not the answer as the clockspeed inside the fpga would still go nuts. if data flies in at 200Mhz the entire fpga machine has to be able to sustain that speed.... very tricky, and expensive)
The technology is not complicated, you just need high clockpseeds and lots of dualported memory ... only found in large fpga's.
When you go flat out and need memory depts in the megasamples range your fpga solution is unworkable because there are no such large memories. ( neither in fpga , or external )

That's whay the agilent machines use an asic that has the ram built in. They use some clever tricks like sliding window peak detectors that scan the data as it is streaming. they look at how many samples are in your 'viewport' ( your timebase and sampling speed ), know the resolution of the viewport ( how many pixels horizontally and vertically make up the viewport )  and set the peak detector window to the pixelsize. the detectors run during the acquisition ( not after it ) so even if there is a minute glitch it will show up on the screen.
if you stop acquisition ( put scope in stop or single ) all they need to do is re-cycle the captured data through this engine. Since that memory is always going flat-out you have instantaneous response. The entire acquisition mode memory is a cyclic buffer ( endless fifo principle ). the output of the 'viewport scrubber' ( the peak detectors and other clever bits to reconstruct the image ) is always running over this cyclic buffer and spits its output into a multiport ram.

there are multiple of these multiport rams , each forming one graphics plane . There are typically up to 16 graphics and two text planes.
1 graphics plane per channel or waveform store
1 graphics plane used for 'static images' like the logo , graticules
1 graphics plane for menu's
1 graphics plane for cursors.

And so on. Another block of logix combines all these planes and hands it off to the lcd refresh logic.
That's why scopes like the 2000 and 3000 and 7000 respond so fast to panel operations. the little cpu in there does barely nothing... its all hardware.

[slateraptor]

It's pretty much a given that strictly relying on fabric resources is not cost effective; there's no question that external memory is needed. That being said, dual-port is very tempting, but I'm not convinced that it's a necessity for devices short of bleeding edge. If anything, it would only be really useful for waveform sample storage, so an implementation which interlaces multiple frames is irrelevant since the other frames are not critically dependent on real-time data.

Here's a hypothetical description running through my mind:
Lack of external dual-port RAM with insufficient capacity is replaced with single-port RAM that's at least twice as wide and twice as fast (which is not unrealistic). A dual-port controller is then synthesized in the FPGA with priority given to writes from the sampling sub-system; sure it's not true dual-port, but in this model, it doesn't have to be either. Read data is then streamed through a pipeline architecture for processing and stalled as necessary, with post-process data stored in a different, much slower RAM for display purposes; any intermediate processing like peak detection can be directly incorporated into the pipeline.


EDIT:

( you need multiported ram... pipelining is not the answer as the clockspeed inside the fpga would still go nuts. if data flies in at 200Mhz the entire fpga machine has to be able to sustain that speed.... very tricky, and expensive

Pipelining pretty much is the only reasonable answer.

P.S. A $10 FPGA may not be suitable, but surely a $20 FPGA can make some significant headway...especially when purchased in bulk.

[krenzo]

Pipelining pretty much is the only reasonable answer.

P.S. A $10 FPGA may not be suitable, but surely a $20 FPGA can make some significant headway...especially when purchased in bulk.

Yeah, Spartan 6s at speed grade 3 can access DDR memory at 600 MHz (300 MHz but on the rise and fall of the clock) these days.  I don't know why you would try to store everything on the FPGA's RAM.  Xilinx's next generation of FPGAs (that cost ~$1k right now because supply is so low) is going to have tens of Megabytes of onboard RAM.  Once those get cheap enough, you won't even need external RAM.

[ejeffrey]

As for fft ...the way it is implemented in cheap scopes it is just a mathematical gimmick. Nowhere neer enough point are fast enough acquisituin to give any kind of meaningfull view. Good luck finding a spirious sideband signal if you only have 300 points and your rescan rate is once a second.

FFT on almost all scopes is a joke.  Maybe there are a few that get it right, but I have never used one.  High end scopes may do slightly better than the el-cheapo units, but not by much, and certainly not enough to be worthy of their high-end status.

Which is really a shame, really.  There is no affordable bench instrument that can satisfy the need for high resolution spectral analysis in the frequency range 1-200 MHz.  Dedicated FFT analyzers for the audio domain (like the SR760) top out at 100 kHz to 1 MHz, and aren't much better than a PC sound card.  Low cost Chinese RF spectrum analyzers have pretty terrible performance, and name-brands will cost you a kidney largely because they start at 3 or 6 GHz.

Right now, the only way to get good spectral analysis for low RF frequency at an affordable price is to buy a PCIe digitlzer/scope card with a ton of memory and do the FFT on the host processor.  That works, but it is much less convenient than if scopes in the same price range were worth a damn.