Nice, older DSO advice

[DaJMasta]

Given the size of the posts on this page, I'd say we need a peak detect thread, then again, I'm a bit terrified at the prospects for such a thread 

Maybe a poll to see who uses it and finds it useful.  I don't think I've used the feature yet, though that goes for some of the other more exotic triggering schemes.




....anyone have a new scope model to look into? 

[nctnico]

As to Wurstunhund, yes I actually believe what he says is correct. Not only because he has tremendous knowledge but also because he can explain the reasons for his arguments so I can easily follow his reasoning.

His reasoning is usually in favour of Lecroy... working around the limitations of Lecroy's products in ways which are not necessarily the most optimal way of doing things in general... and if Lecroy says...  Also I think he is more involved in the sales side of oscilloscopes than in actually using them to make a living.

[Mechatrommer]

Given the size of the posts on this page, I'd say we need a peak detect thread, then again, I'm a bit terrified at the prospects for such a thread 

better the title is "Nice older DSO with peak detect advice highest BW/GSps at $5K or less" then we should clearly see which one is important, the peak detect or the money.

[Mr Nutts]

One is simply to detect aliasing.  If the peak detected signal is not consistent with the sampled signal, then aliasing has occurred.

Another place I have specifically seen it be advantageous is tracking the pulse per second output of a GPS under certain conditions (say after XOR or race condition edge detection) where even the longest record length is not sufficient because the capture time exceeds 1 second.

1s capture time? 

Segmented memory could be even more useful here, however many DSOs have peak detection without supporting segmented memory.  A dual delayed timebase may also be able to make this measurement.

How about a modern scope with just enough memory?

I had a look at the PPS signal of my China GPSDO with my Agilent DSO8064 set at 64M

Here's how it looks like with a capture time of 2 seconds:




This is a capture time of 5 seconds:




Or how about 50 seconds of capture?




Or how about 200 seconds of capture?




All without peak detect 

And here's the bonus: the signal timing is still correct, i.e. pulse width and the pause times are still correct, something that's not necessarily the case with peak detect






So where was the advantage of peak detect for seeing a 1PPS signal again?

[rsjsouza]

I fail to see your reasons to fight against something other people find useful. What is your reason for it?
Academic purity of measurement process? Fact that Wuerstchenhund says so so it must be true?
Fact that you do only limited scope of work with oscilloscopes and in your dealings you never had need for it?

The simple reason I ask is that, litterally, every time the discussion is about older higher BW scopes, it's stated that these and those scopes are no good because they lack peak detect

That seems to be considered as a statement of fact without every questioning if it's still true

I don't think it is a matter of true/false but instead a preference. When talking about old oscilloscopes the more limited memory certainly adds weight to the case for peak detect. Modern oscilloscopes and their larger memory reduce the advantages.

A similar case can be said about protocol decoding; although convenient to have it featured on an oscilloscope, the current offers for cheap and easily configurable PC-based logic analyzers greatly reduce the need to have it built into the scope - it certainly is not as worth as the option prices of the A-brands, where a few of the options can easily go beyond the price of a full featured Saleae or a ZeroPlus.

A while ago I interjected a discussion where both Mike and Nico explained to me their preference for having a peak detect oscilloscope. Despite the explanation made sense to me and backed by two very experienced users, I still haven't seen the need and use it sparsely. I have a 140Mpts DS4014, though.

Given the size of the posts on this page, I'd say we need a peak detect thread, then again, I'm a bit terrified at the prospects for such a thread 

Many threads already tore this subject to shreds. Don't even go that route... 

https://www.eevblog.com/forum/testgear/lets-talk-about-lecroy-scopes-aka-the-wuerstchenhund-holds-court-thread/msg1026148/#msg1026148
https://www.eevblog.com/forum/testgear/a-possibly-dumb-q-on-digital-oscilloscopes/msg1257270/#msg1257270
https://www.eevblog.com/forum/testgear/want-faster-and-newer-oscilloscope/msg1884350/#msg1884350

...and many more!

[2N3055]

@Mr Nutts:

So basically you reason for this discussion is that you like riling people up and you like to be right at all cost.
We gave you plenty of reasons why peak detect is not evil and can be useful  sometimes, but you chose to ignore it all, and basically concluded that we use peak detect because we are backwards, luddites and generally not smart enough to know that there is a better way, your way....

Now I'm gonna go to a dark place, sit in a corner, crying softly and reflect upon my life, deeply touched by this great revelation..

NOT.

 

[jjoonathan]

Aliasing sucks.



Even if the Nyquist criteria is met, sampling without an AA filter sucks.


Gambling that you aren't aliasing sucks. You'll win 9 times out of 10 but you'll have enough fun on that 10th time to make up for everything you missed out on the other 9 times.

Doing due diligence to make sure you aren't aliasing sucks. You have to ensure that mF+nFs for all expected input frequencies F and integer m,n is >> 1/Tfeature for all relevant features you want to see. Sure, you can do it, if you know which input frequencies to expect.  If you're an RF guy you might even be good at this calculation, but is it really what you want to spend your brain cycles on? Peak-hold lets you invest those brain cycles in something more productive, like daydreaming, or thinking about the problem you are hunting.

> (paraphrasing) If your scope has enough memory it's not a problem

See that scope above? It has 800Mpt of memory and can draw 1 million Wfm/s. By default, it has a 10Mpt acquisition depth limit enabled in its settings. If you zoom to longer timebases, it will decimate to keep depth at 10Mpt (or whatever you configure the limit to). Why? Because one typically prefers fast updates and deep history [1] to full resolution long acquisitions. See that third knob in the timebase section, in addition to the usual scroll and zoom knobs? The third knob controls acquisition depth, and if you want to sacrifice your history and Wfm/s to get very deep acquisitions, it's easy, you just turn that knob to the right. I prefer not to turn the knob to the right.

I prefer fast updates and deep history and I also prefer not to calculate beat frequencies vs feature frequencies. Peak detect lets me have my cake and eat it too. Thank you, peak detect!

[1] It's segmented unless you tell it not to be and they call it "history." If you hit stop you can browse to previous acquisitions. It really makes more sense than throwing away previous acquisitions by default.

[2N3055]

Aliasing sucks.

Thank you for explaining what I couldn't. That is one fierce scope..
Best regards
Sinisa

[Kosmic]

I think the problem is not if people like peak-detect or not. Or if it's useful some time to some people.

The problem is more that some people freely declare to everybody "you can't buy an oscilloscope without peak-detect". Which I think is not entirely true. Really, depend of your usage and personal taste.

[Mr Nutts]

I don't think it is a matter of true/false but instead a preference. When talking about old oscilloscopes the more limited memory certainly adds weight to the case for peak detect. Modern oscilloscopes and their larger memory reduce the advantages.

If it was preference I had no problem with it. But it's not that. You can go into any discussion about slightly older scopes as in this thread and there is at least one person who constantly bring up peak detect as a necessity - not as a personal preference.

A similar case can be said about protocol decoding; although convenient to have it featured on an oscilloscope, the current offers for cheap and easily configurable PC-based logic analyzers greatly reduce the need to have it built into the scope - it certainly is not as worth as the option prices of the A-brands, where a few of the options can easily go beyond the price of a full featured Saleae or a ZeroPlus.

And still in those threads no-one complains that older scopes don't have protocol decoding or dismisses certain scopes because they don't have protocol decoding

A while ago I interjected a discussion where both Mike and Nico explained to me their preference for having a peak detect oscilloscope. Despite the explanation made sense to me and backed by two very experienced users,

I did read that, too, but so so far I haven't read anything beyond "I use it because I'm used to use it". I'm not saying it doesn't work for these users, but there's nothing in there showing how peak detect in a modern deep memory scope solves a problem that without it would be unsolvable or even just more difficult to solve.

Given the size of the posts on this page, I'd say we need a peak detect thread, then again, I'm a bit terrified at the prospects for such a thread 

Many threads already tore this subject to shreds. Don't even go that route... 

https://www.eevblog.com/forum/testgear/lets-talk-about-lecroy-scopes-aka-the-wuerstchenhund-holds-court-thread/msg1026148/#msg1026148
https://www.eevblog.com/forum/testgear/a-possibly-dumb-q-on-digital-oscilloscopes/msg1257270/#msg1257270
https://www.eevblog.com/forum/testgear/want-faster-and-newer-oscilloscope/msg1884350/#msg1884350

...and many more!

Na, I'm not going there 

Aliasing sucks.

I'm sorry to say this but so does your way of attaching pictures because I can't see them.

Someone recommended to attach images to the post and then use the image link in the text to make sure other people can see them

Gambling that you aren't aliasing sucks. You'll win 9 times out of 10 but you'll have enough fun on that 10th time to make up for everything you missed out on the other 9 times.

Doing due diligence to make sure you aren't aliasing sucks. You have to ensure that mF+nFs for all expected input frequencies F and integer m,n is >> 1/Tfeature for all relevant features you want to see. Sure, you can do it, if you know which input frequencies to expect.  If you're an RF guy you might even be good at this calculation, but is it really what you want to spend your brain cycles on? Peak-hold lets you invest those brain cycles in something more productive, like daydreaming, or thinking about the problem you are hunting.

Not sure what you want to say here. Of course there is aliasing in my screenshots. So what? Peak detect is even less true to the original signal, which is why proponents claim to use it only to see if there are pulses or if there are glitches

But as shown in my screenshots I don't need to select peak detect to see the 1PPS pulses, and the fact that the signal has some aliasing doesn't change the fact that the basic timing parameters of the waveform are correct, something that would not be the case with peak detect

See that scope above?

No, unless it looks like two blue blobs

It would have caused no harm had you at least stated brand and model

It has 800Mpt of memory and can draw 1 million Wfm/s. By default, it has a 10Mpt acquisition depth limit enabled in its settings. If you zoom to longer timebases, it will decimate to keep depth at 10Mpt (or whatever you configure the limit to). Why? Because one typically prefers fast updates and deep history [1] to full resolution long acquisitions.

That sounds like one of the new R&S scopes WH has told me about. I think they can only use a small part of their memory in normal mode and can only use the large memory for I believe history and other special modes. Is that right?

1 million waveforms per seconds sounds great but to me that's just marketing wank (others have done it before, too). Update rates are way overrated

I prefer fast updates and deep history and I also prefer not to calculate beat frequencies vs feature frequencies. Peak detect lets me have my cake and eat it too. Thank you, peak detect!

If you prefer that, who am I to tell you it's not OK?

I have no experience with R&S scopes but frankly if you need to use a lot of brain power just to use extended memory or do frequency calculations (what fo, btw?) then I would say the scope isn't very good

On the Agilent Infinum I created the screenshots with, using Peak Detect would need slighty more effort while giving me inferior results

good luck (or retry) finding 10MHz glitch (100ns pulse width) with that setup. (glitch = something we didnt program nor expect but circuit went haywired)

I would't do that with that setup, that's what advanced triggers are for

@Mr Nutts:

So basically you reason for this discussion is that you like riling people up and you like to be right at all cost.
We gave you plenty of reasons why peak detect is not evil and can be useful  sometimes, but you chose to ignore it all, and basically concluded that we use peak detect because we are backwards, luddites and generally not smart enough to know that there is a better way, your way....

Now I'm gonna go to a dark place, sit in a corner, crying softly and reflect upon my life, deeply touched by this great revelation..

NOT.

 

So in short, instead of answering my question you decide to go all drama queen on me (David Hess has at least given a real example I could replicate, kudos for that). Fair enough, although considering your clear and obvious misconceptions about peak detect and sampling theory it appears you're pretty butt hurt

I think I slowly start to understand what Wurstunhund meant when he said "you'll find out yourself" when I asked him about the kick-out (he didn't tell me except that I'd find out myself if I hang around here). Clearly one shalt not ask too many questions if it touches established mantra it seems

So I'll grant you your safe space and bow out of this thread

[Mechatrommer]

All without peak detect 
And here's the bonus: the signal timing is still correct, i.e. pulse width and the pause times are still correct, something that's not necessarily the case with peak detect

good luck (or retry) finding 10MHz glitch (100ns pulse width) with that setup. (glitch = something we didnt program nor expect but circuit went haywired)

[rsjsouza]

I think the problem is not if people like peak-detect or not. Or if it's useful some time to some people.

The problem is more that some people freely declare to everybody "you can't buy an oscilloscope without peak-detect". Which I think is not entirely true. Really, depend of your usage and personal taste.

If it was preference I had no problem with it. But it's not that. You can go into any discussion about slightly older scopes as in this thread and there is at least one person who constantly bring up peak detect as a necessity - not as a personal preference.

Yes, from time to time we get some heated discussions around here. However, in all fairness, while searching for the threads above I found Nico softened his speech over the years - from "must have" or "useless" to "I prefer" or "I would never buy". From what I recall, he is the one that was the most vocal for PD.

And still in those threads no-one complains that older scopes don't have protocol decoding or dismisses certain scopes because they don't have protocol decoding

Yes, but decoders are quite optional and a much more recent feature of oscilloscopes. 

[2N3055]

So in short, instead of answering my question you decide to go all drama queen on me (David Hess has at least given a real example I could replicate, kudos for that). Fair enough, although considering your clear and obvious misconceptions about peak detect and sampling theory it appears you're pretty butt hurt

I think I slowly start to understand what Wurstunhund meant when he said "you'll find out yourself" when I asked him about the kick-out (he didn't tell me except that I'd find out myself if I hang around here). Clearly one shalt not ask too many questions if it touches established mantra it seems

So I'll grant you your safe space and bow out of this thread

You are the one that refuses to understand despite being explained many times. It tells more about you than anything else.
Also first you call me stupid and backwards and now you're implying you buttfucked me. Strike two.

And as Wuerstchenhund was, you are abrasive, rude and disrespectful. He wasn't banned because of his knowledge that was wast.
The man is a walking encyclopedia of scopes. Shame he's gone, really.
He was banned because he was getting in constant fights with members, and because he was asshole about it.

And it seems he's your mentor in more than scopes.

You are not a judge, I don't have to explain myself to you. I explained enough. It seems it was understandable enough to everyone else except you.

So let me repeat once more: peak detect is useful. Even if you have lots of memory. It helps with undersampling, and when using it, you can get good visual representation of signal. Yes, many people are using scopes visually, not only as a offline data analysis tools.
Actually that is how scopes are mostly used, sorry.

I don't think peak detect is most important feature of the scope. I disagree with that.
I also strongly disagree this crusade against it that you are doing. I disagree with that too, together with your manners.

I hope you will agree we stop here.  I don't like personal attacks and insults. And certainly I'm not the one to turn the other cheek.
I feel we are not going the right way.
Also it is really off topic. If we can't keep it nice, it should not be at expense of spoiling it for other people.

Sinisa

[Kosmic]

Easy now, you're the one insulting others.

This is probably a translation problem. Ther's a big difference between "butt hurt" and like you said "buttfucked"   

"butt·hurt"
adjective: butthurt; adjective: butt-hurt
overly or unjustifiably offended or resentful.
"they're all butthurt that she released the album online first"

[nctnico]

And as Wuerstchenhund was, you are abrasive, rude and disrespectful. He wasn't banned because of his knowledge that was wast.
The man is a walking encyclopedia of scopes. Shame he's gone, really.
He was banned because he was getting in constant fights with members, and because he was asshole about it.

I agree. Sometimes you just have to walk away. You can lead a thirsty horse to water but the horse has to decide to drink or not.

[2N3055]

Easy now, you're the one insulting others.

This is probably a translation problem. Ther's a big difference between "butt hurt" and like you said "buttfucked"   

"butt·hurt"
adjective: butthurt; adjective: butt-hurt
overly or unjustifiably offended or resentful.
"they're all butthurt that she released the album online first"

Thank you for explanation and concern.. My English is not much, surely, I appreciate any help that I can get.
Just a simple explanation. Being butthurt is something very derogatory. Ethimology of that word is that somebody anally raped you and now you're hurting. With time, it was used in context " stop whining, it's not like you're buthurt" implying that your whining is not proportional to what you experienced.  And while carrying distasteful symbolism, that would have the meaning you said.  Not nice but, hey, OK.

What he said, in English means what I understood, especially in the otherwise derogatory tone. I don't care if there is some local slang that takes that and interprets it differently. Citizens of the world cannot know all of that. You take sentence by its word meaning, not some arbitrary phrased meaning you don't even know exist.

Again, thank you for caring and trying to help. Much appreciated.
I don't want a fight with anybody. Especially about something not very important.


So if I went overboard and overreacted I apologize to everybody, including Mr. Nutts.
I'm not gonna go back and delete or edit anything. I will leave it as it is so all is clear. If I make a mistake I own it.
Maybe someone else will learn how easy is to go wrong direction.

Thanks Kosmic.

Regards,

[Mechatrommer]

someone who live on a desert will say the world is yellow, someone else who live in the rain forest will say its green, someone on the ocean will say its blue, and then everyone of them...

and then the admin is the one who has the kill switch and left with fewer members, who win? nobody... some wiser people around here from whom i learnt are already left/inactive, sad.

[David Hess]

Large acquisition memories come with a cost besides economics.  They proportionally increase processing time and I have watched many of these DSOs grind to a halt when a large acquisition memory is selected.

Sounds like the Tek scopes at college

Sounds like every long record length DSO I have tested.

But yes, at a given sample rate it takes longer to fill a larger memory than a smaller one. There's no way around that

It does not take any longer when the extra record length is filled at a faster sample rate.
 

Otherwise why allow selecting a shorter record length at all?  Why don't they always use the longest record length unless there is some advantage to be had by using a smaller one?

Because a smaller memory comes with a higher screen update rate, which is important if you want to use the scope visually
 
On my Lecroy scopes I usually select a size of 100k if I want to get fast screen updates. If I need it for more than just staring at waveforms then I select the full memory
 
That is why better scopes allow manual memory management

Wait, so which is it?  Does it slow down or not?
 

But that method is unreliable as just because your scope doesn't show a glitch doesn't mean there is none, simply because your scope is still blind most of the time (approx 9x as likely to miss it than to find it)

DPO operation solves this issue and can be considered a superset of peak detection.

No, it does not
 
Even in DPO mode a Tek scope is still suffering from huge blind time, as does every other scope in persistence mode
 
With these methods you have no idea if your signal is really free of anomalies because most of the time you'd just miss it

What are you talking about?  DPO mode reduces blind time.

The original DPO oscilloscopes manged 50 thousand acquisitions per second which is comparable to an analog oscilloscope but later ones were 100s of thousands.
 

The point of DPO operation is to find every unknown by capturing every signal characteristic.  It is useful when I do not know specifically or even exclusively what I am looking for.  If you measure something 'funny,' record the amount of 'funny'.  (2)

But it doesn't. It only shows glitches that appear outside the blind time, the latter which is roughly 9x as long as the time the scope actually "sees" something
 
The chance to miss a glitch is approx 9x as high as the chance that the scope captures it

More recent DPOs acheive blind times of less than 5% even at high sample rates.

I had a look at the PPS signal of my China GPSDO with my Agilent DSO8064 set at 64M

Now go back and do it again with the signal I suggested.

Your capture time of 2 seconds could miss a 40 nanosecondsecond pulse and it just gets worse from there.  Even my almost 30 year old Tektronix 2232 with 100MS/s can do better than that with a 1 kSample record length.

[jjoonathan]

Peak detect is even less true to the original signal... the fact that the signal has some aliasing doesn't change the fact that the basic timing parameters of the waveform are correct, something that would not be the case with peak detect

https://en.wikipedia.org/wiki/Aliasing

We are not talking about signals becoming blocky but otherwise remaining true to form as a result of lower sample rate. We are talking about the effect where the sample frequency and signal frequency have a relationship that "conspires" to turn a signal of one form into a signal of a completely different form, thoroughly corrupting its shape and timing parameters.

Consider the attached example. Aliasing causes a 100MHz burst to appear as a 10kHz burst. Note that peak detect mode is immune to this problem.

I have no experience with R&S scopes but frankly if you need to use a lot of brain power just to use extended memory or do frequency calculations (what fo, btw?)

It is possible to mathematically prove that the above effect will not corrupt your particular signal at your particular sample rate. If you don't use peak detect mode, you will either need to do this math or you will need to gamble on a favorable result. It sounds like you have been gambling and winning. That's fine, but it isn't for everyone.

1 million waveforms per seconds sounds great but to me that's just marketing wank (others have done it before, too). Update rates are way overrated

I have a literal knob that lets me choose between fast update rates and deep memory. This puts me in a good position to be impartial. I usually choose fast update rates.

[jjoonathan]

That sounds like one of the new R&S scopes WH has told me about. I think they can only use a small part of their memory in normal mode and can only use the large memory for I believe history and other special modes. Is that right?

No, it is not correct. RTO1000s with 800Mpt of memory can take single 10GS/s 800MPt captures no problem.

Someone unfamiliar with R&S scopes might get the wrong impression after playing with one for a few minutes, though. By default they are configured with a relatively low "Record length limit" of 10MPt or so. If you simply scale out the timebase, it will begin to decimate so that the record length does not exceed 10Mpt. This maintains a fast update rate (like a Keysight scope) and uses memory depth for history. If you want a single 800Mpt record, you must reconfigure the limit, but it will absolutely let you do so.

[Kosmic]

Ok jjoonathan, you have to stop posting pictures of your scope. You will have me starting shopping for a new one 

[james_s]

For $5k I'd be looking at new or high end used scopes, or maybe spend $3-4k on the scope and $1-2k on probes and other accessories.

Personally I've been really happy with my TDS784, 1GHz and nice familiar Tek interface for about 1/10th your budget however the newer Tek DPOs are nice too and a lot more compact.

[0culus]

Wow....this thread escalated quickly. 


...lots to read. I'm leaning towards finding something used + investing in probes as james says.

[Neomys Sapiens]

Oculus, keep straight please. LeCroy stinks badly, especially the user IF. I told about that trigger setting somewhere.
I have never personally encountered one of those R&S scopes, so I can't say.
I have the TDS644 at home and it is doing good service. Of course, I have a DPO70604 at work, which is beautiful.
But my suggestion would be the newer version of the DPO3054, with the larger screen and USB.

[0culus]

Oculus, keep straight please. LeCroy stinks badly, especially the user IF. I told about that trigger setting somewhere.
I have never personally encountered one of those R&S scopes, so I can't say.
I have the TDS644 at home and it is doing good service. Of course, I have a DPO70604 at work, which is beautiful.
But my suggestion would be the newer version of the DPO3054, with the larger screen and USB.

Tell us how you really feel!!! 


In all seriousness, I'd prefer to see why you think that (with evidence) rather than just stating that it stinks. UI preferences are highly subjective.