Author Topic: Rigol's "long memory"  (Read 24062 times)

0 Members and 1 Guest are viewing this topic.

Offline cyberfish

  • Regular Contributor
  • *
  • Posts: 239
  • Country: gb
Re: Rigol's "long memory"
« Reply #25 on: September 12, 2010, 08:12:15 pm »
Yeah I know Rigol OEM low end Agilent scopes. Wonder how that works. Maybe they have higher quality control for those? Which makes sense because they have higher profit on those.

By testing ADCs I mean by Rigol, to make sure they run at higher frequency. I know ADI tests them, but only to rated frequency of course.

I don't think ADCs will fail. Probably just higher noise, which is hard to measure.
 

Offline cyberfish

  • Regular Contributor
  • *
  • Posts: 239
  • Country: gb
Re: Rigol's "long memory"
« Reply #26 on: September 12, 2010, 08:59:01 pm »
I was thinking about getting both the 40MHz and 100MHz versions and test them out to settle this issue once and for all, but no place has stock and AD doesn't have sampling for these chips.

EDIT: a few places have stock for the 40MHz one, but no place has 100MHz. AD lead time is till November for the 100MHz. On an interesting side note, it seems like Maxim's alternative is only 40MHz.
http://www.maxim-ic.com/alternatives.cfm/part/AD9288/pk/28

So it's not like anyone can make them.

EDIT2: Maxim does have 100Msps ADCs. Pricing is similar (high premium compared to slower parts). Maybe yield for those parts still aren't that high?
« Last Edit: September 12, 2010, 09:06:07 pm by cyberfish »
 

Offline Zad

  • Super Contributor
  • ***
  • Posts: 1013
  • Country: gb
    • Digital Wizardry, Analogue Alchemy, Software Sorcery
Re: Rigol's "long memory"
« Reply #27 on: September 12, 2010, 10:52:29 pm »
I was thinking about getting both the 40MHz and 100MHz versions and test them out to settle this issue once and for all, but no place has stock and AD doesn't have sampling for these chips.

EDIT: a few places have stock for the 40MHz one, but no place has 100MHz. AD lead time is till November for the 100MHz.

I'm getting a bit worried about AD. Here is an edited part of my current shopping basket with them:

AD8342ACPZ-REEL7                                           
Product Status: Production     
Availability: In 16 weeks

AD8343ARUZ                                
DC-to-2.5 GHz High IP3 Active Mixer       
Product Status: Production     
Availability: In 12 weeks   

AD8351ARMZ                                
Low Distortion Fully Differential RF / IF Amplifier       
Product Status: Production     
Availability: In 10 weeks   

AD8620ARZ                                
Precision, Low Input Bias Current, Wide BW JFET Op Amp (Dual)       
Product Status: Production     
Availability: Not Currently Available at this Quantity

I have the AD9910 DDS on my shopping list too, nowhere has them in stock.

For fast ADCs, try Texas Instruments, they have a huge range.


Offline rf-loop

  • Super Contributor
  • ***
  • Posts: 3453
  • Country: cn
  • Born with DLL21 in hand
Re: Rigol's "long memory"
« Reply #28 on: September 13, 2010, 05:28:56 am »


I don't think ADCs will fail. Probably just higher noise, which is hard to measure.

Yes, this kind of things there are and maybe there is also more differencies between individual scopes if compare to case if they use ADC as they are selected by AD. But we need remember, they sell scopes with scope specs, not with ADC specs by AD..

AD have not set max for ADC clock. ;) They tell only minimum quaranteed max conversion specs where all given data are qualid.
If you test 1000  -40 chips you may be find that in your temp area and you voltage and your more loose limits they do it example 100MSPS (conversion quality is maybe not same as datasheet but still acceptable for scope with scope itself specs.

One important thing for noise is ADC clock signal quality. And I quess that part of modified (and not modified) Rigol noise come there. Specially it can see with modified 1052. Maybe some individual scope kore and some less.

Why/how Rigol select what scope they give name DS1102 and what is then 1052 ;)
But I have compare (without good data out from test) original DS1102E and DS1052E modified to DS1102E.
By the eyes only without good test I have one "opinion". DS1102E/mod have more noise. It looks like sampling noise is more high. But this I can not proof becouse I do tests only from scope input.
If practice and theory is not equal it tells that used application of theory is wrong or the theory itself is wrong.
-
Harmony OS
 

Offline cyberfish

  • Regular Contributor
  • *
  • Posts: 239
  • Country: gb
Re: Rigol's "long memory"
« Reply #29 on: September 13, 2010, 06:04:04 am »
Quote
Yes, this kind of things there are and maybe there is also more differencies between individual scopes if compare to case if they use ADC as they are selected by AD. But we need remember, they sell scopes with scope specs, not with ADC specs by AD..
That's true. But you can also say bad soldering and fly wires are fine, too, as long as the scope works.

It's just that seeing those things lowers my confidence in them.

Quote
AD have not set max for ADC clock.  They tell only minimum quaranteed max conversion specs where all given data are qualid.
If you test 1000  -40 chips you may be find that in your temp area and you voltage and your more loose limits they do it example 100MSPS (conversion quality is maybe not same as datasheet but still acceptable for scope with scope itself specs.
That's possible, but they will have to test every single ADC, which I'm guessing they (Rigol) don't. For the cost of testing them they could just get 100MSPS parts.

It really depends on how good ADI's manufacturing process is. If most chips end up in the 40MHz bin, with very few ending up 100MHz, you probably won't find any that can do 100MHz in the 40MHz bin, because they would be in the 100MHz bin.

I'm guessing they just took 40MHz parts, and accept the higher noise, because most people probably won't notice. Or enough won't for the rest to not matter.
 

Offline scrat

  • Frequent Contributor
  • **
  • Posts: 608
  • Country: it
Re: Rigol's "long memory"
« Reply #30 on: September 15, 2010, 08:15:48 am »
I'm still in the middle between Rigol and Instek. The first has lower price, but seems to have also lower performances and builing quality.

Apart from the building quality (soldering, wires, etc) in my opinion there are a few things which must be taken into account.

If we were direct Rigol customers, so that we ordered a scope with some specs and followed its development, we could object to them for their choices, if we think quality level is insufficient with respect to requirements. But since we go on the market and take what is specified by someone other, unless it doesn't comply with specs (the whole product, not a single part), we can't say much.

One other thing is on ADC. Rigol scope versions management (which consists of only programming a different ID on 50 and 100 MHz machines) teaches that successful companies simply look at market demand, and try to satisfy it. It is also well known that many semiconductor companies sell different versions of their chips, but real difference is only a code in a non-volatile memory. All of the digital and some of the analog ICs have testing modes, which can be accessed via a peculiar sequence and allow for test after production, since the accepted failure rate on the market is usually low (1/1000, almost never it is 0%, also because it's impossible or too expensive to test for all of the working conditions). In some cases these modes also can be written on calibration data or "version" data, enabling or disabling some features (sometimes competitors have discovered this by looking at chips' microphotos which looked the same for different versions). In some other cases they simply mark differently their ICs to cover as many market segments as possible, like someone already posted here. Rigol uses those 40MHz parts at 100MHz, and it could be that ADCs are only marked differently, or that they do not perform the same way, but if the scope stays inside the specs, for what can we complain them? Of course, there can be failures, but which company on this world is able to produce only perfect products (staying in the low-cost market, too)? Companies guarantee for their products, which means there is a low probability of failure or no compliance, and they will pay for it. I will otherwise complain Rigol if for a defective product they do not obey their guarantee.

Are there some photos or reviews of the Instek? Is it really far better quality?
If possible, it would be nice to test the scope (instead of the ADCs) to see if it complies the specs...
One machine can do the work of fifty ordinary men. No machine can do the work of one extraordinary man. - Elbert Hubbard
 

Offline cyberfish

  • Regular Contributor
  • *
  • Posts: 239
  • Country: gb
Re: Rigol's "long memory"
« Reply #31 on: September 15, 2010, 08:01:04 pm »
There are pictures of the internals of both scopes a few posts up.

I'm probably getting the Instek soon, so we can potentially test it and compare it with Rigol (that many people have), then. But I don't really know how to test oscilloscopes, and I'm not taking it apart :).
 

Offline rf-loop

  • Super Contributor
  • ***
  • Posts: 3453
  • Country: cn
  • Born with DLL21 in hand
Re: Rigol's "long memory"
« Reply #32 on: September 19, 2010, 09:03:39 am »
CPU overclocking is totally different case. It crash if some inside databus data is wrong becouse parallel digital data is not qualid exactly this time when it need be. (I do not explain all things, it is extremely complex to explain with bad english. There is also many other problems as temperature. Hot spots inside chip and also many wear and tear things on the chip.)

This AD's ADC is "100 MHz" chip, if it is labeled -40, -80 or -100. This chip digital parts work very easy with 100MHz without any digital data qualid problems inside chip. Indirectly this can read also datasheet. Also this kind of company as Rigol (no 1. or 2. in world for manufacturing oscilloscopes. They make also OEM). They can do all specifications tests for chips just same quality (or more) than chip manufacturer. (yes I wait time when Rigol co to they own chip designs or even manufacturing as HP or Tektronix before)
Also they can order standard chips with they own label. If they do this nobody tell any overclocking becouse nobody can not see any overclocking problems)

This "overclocking" what is not really overclocking in normal "CPU or Memory" meaning. And also not as PCB overclocking. Yes there is many times problems becouse PCB RF design is not good or not good for overclocking and bus timing may also fail there. 100ps is long time and travel distance on the PCB trip may be 20mm (trivial example) and more complex it go with strip lines becouse there is resistance capacitance and inductance. Also transmitting part have these and receiver chip.

CPU itself or whole system overclocking is totally different case and it may affect component lifetime and system reliability.

But what happend if put -40 ADC (what is really same chip as -100 chip). It works fine and without problem in most case in digital side. No corrupted data no timing problem what can conflict with data reading chips.
Problem is in ADConversion. (in most case) And if problem is other type these cips come not out from AD. If there still are chips what fails with 100M they can test out before go to scope PCB. After scope PCB is ready and if there is "bad" chips they fails and do not go to customers. (just as all do..., Nokia, Tektronix, Agilent etc)

Interesting part of this case is Analog Digital Conversion (ADC) itself quality.
-40 Chip do not reach same ADC quality as -100chip with 100MSPS. -100 chip ADC quality you can read directly from datasheet and this AD quarantee. (example around 7bit total accuracy or littlebit worse. Dynamic perfomance part of datasheet is meaningful and I "quess" there these -40 chips used as -100 chips do not reach same quality at all. SINAD/SNR and harmonic distortion may differ lot of.

But now it is important to understand that we do not look ADC specs. They are not promised Rigol scope specs.
Rigol do NOT promise even nearly same for oscilloscope.
Inside scope specs "window" +-4% or +-3% is lot of room for bad and poor ADC. Only they tell that still ADC resolution is 8bit. Yes it is, it can product to databus 8 bit bytes. But real accuracy is nearly as 4 or 5 bits. +-4% window have lot of room for also these things what come from ADC. (encode pulse timing is too short for this -40 labeled chip if we think what conversion quality is promised if it is used with more slow encode signal.)

This is quessing. Most reasons for -40 label on originally -100 designed and producted chip is inside anlog channel and S&H complex ("bad silicon" ... but maybe there can find also more selective and accurate explanation. Strip lines and components quality in analog part and capacitor and switch (sampling) components. More fast run more lost in analog and conversion accuracy there. And Rigol do NOT promise accuracy what can do with -100 chips. (but also Instek do not promise ;)

But what do Instek.
They overclock. Chip is 100M chip (also this -40 chip) and digital parts (databus and timing there) is quaranteed with chip design for this speed. Instek go over this! Rigol do not go over this. Going over this limit is more danger (but Instek go only littlebit (125/100) and maybe this do not make also any problem aspecially if (data etc) timing outside of chip is good (also need remember that specs data are made for 3.0Vdd but chip normal use area go up to 3.6V (not as abs max).

Rigol go over analog specs. Instek go over chip designed digital speed. What happend if data is not qualid. It can be random 8 bits. In Rigol case there is more like only analog conversion itself quality problem if compare to datasheet quality promises. (and these Rigol do not promise.) Rigol promise DC aacuracy +-3  to +-4%. (these limits includes also ADC conversion itself error. Btw 8% from 255 is 20,4.  4bit is 16. ;) So there is huge room also for ADC origin error. Wery easy can drop one bit out and after it we can accept -40 chip used as -100 chip but specs read not now signal to noise ratio 44dB or differential nonlinearity typical +-0.5LSB)

If there happend some other "mystified" things as they do "overclocking" I am very interested to find any mystified problem origin from this "overclocking".

Yes I know lot of stability problem and crashed computers. But please do not compare these if you really do not know AND understand  theoretical and practical difference to this case.

Rigol, Instek and others they sell oscilloscopes and they have published specs for they product. I buy oscilloscope not ADC chips. If Oscilloscope specs meet my needs and I believe quality is ok and price...  yes of course I sometimes want buy also name.
One but small scope manufacturer in world is Tektronix. But in the "golden years" it have good name specially with scopes. Yes there are maybe also some reasons for this thinking about "good" with this name but... in the truth.. they have made many many bad designs and bad production but I do not have pictures about extremely bad "hop wires" and bad solderings and bad designs and specially bad component selections.

More difficult is "overclocking" but you can not know it becouse chips are Tektronix design and tektronic manufacturing.
How peoples can buy oscilloscopes what are made with components "no datasheet". ;)
What is designed durability. 10000hr? (what you think about 500MHz vertical amplifier hybrid chipa (this case mixed digital and analog) where case temperature in normal use is more than 90 celsius in normal room temperature.
Yes you can buy these Tektronix good brand oscilloscopes in ebay. Very typical is that seller put picture where scope is power off. "we have no knowledge or possible to test"...hehe)

I have never understand this big mouth talking about "overclocking".
There is only clever engineers who find this solution. There is also lot of power for testing.

But where are these units what have real problems. I have use lot of time for surfing internet and find anything but only peoples self made fails. Where are these reliability problems. Scope have sell lot of and long time. Where are these bad solderings errors and ADC fails becouse "overclocking" ... I have seen lot of Intel and AMD crashed with overclocking...  burned CPUs burned display drivers etc and etc. But I have not seen or heard any crashed ADC or burned ADC about Rigol. Mybe there are fails... of course there are normal fails but if this is any kind of problem there is full of internet talking it.  How many they have sell total DS1052E and 1102E?

I hope some specialist (who also really know) write with better english also this kind of things what are in theoretical and practical things related to this kind of using AD9288 out of its specification sheet.

But my tight opinion with my all knowledge is that comparing to CPU's and this kind of overclocking is not relevant.
« Last Edit: September 19, 2010, 09:19:50 am by rf-loop »
If practice and theory is not equal it tells that used application of theory is wrong or the theory itself is wrong.
-
Harmony OS
 

Offline cyberfish

  • Regular Contributor
  • *
  • Posts: 239
  • Country: gb
Re: Rigol's "long memory"
« Reply #33 on: September 19, 2010, 05:35:18 pm »
The problem is, no one here knows how to properly test oscilloscopes, so we have to rely on indirect observations, like build quality, soldering quality, component selection, etc. And Rigol is not looking good in that department.

Perhaps we can do some simple tests after I got mine? But of course, such test won't be too accurate because of many factors (different signal sources, different places/temperature/humidity, etc).
 

alm

  • Guest
Re: Rigol's "long memory"
« Reply #34 on: September 19, 2010, 06:01:10 pm »
Perhaps we can do some simple tests after I got mine? But of course, such test won't be too accurate because of many factors (different signal sources, different places/temperature/humidity, etc).

The trouble with testing scopes is that you need signal sources that are better than the scope under test, to be able to test the limits of the scope. I've seen some tests that suggested the attenuator wasn't very linear over the frequency range, but this was with an unknown, uncalibrated pulse generator, so it's hard to tell what's going on. Running through the performance verification procedure (or calibration procedure without doing any adjustments) in the service manual is my usual method, except that Rigol doesn't publish any procedures (neither does Agilent with the re-branded versions). Plus I don't have a Rigol scope. Required equipment would probably include:
- Leveled sine wave source from <=1MHz or so up to well beyond 100MHz
- Fast pulse source with edge rate much faster than 3.5ns
- Accurate DC source (this one is easy, stable lab supply + basic DMM should be plenty accurate)
- For testing attenuators, you might also need to be able to generate fast pulses with a large amplitude, or large amplitude leveled sine waves

This is unlikely to be cheaper than to buy a good brand name scope in the first place. This is not something that can be done with your average function generator. Plus they either need to be calibrated or verified with a known good scope that's much faster than the Rigol. Ideally, these tests should be performed over the operational temperature and humidity range.

The avalanche pulser design by Jim Williams (from memory it was in LT AN-47) discussed before on this forum is the best you can make as DIY design for a few bucks, but you really need a much faster scope to determine the output parameters before using it to verify a scope (he used an ancient sampling scope).

I prefer to just buy from a manufacturer that I can trust, although I do verify as much as I can.
 

Offline cyberfish

  • Regular Contributor
  • *
  • Posts: 239
  • Country: gb
Re: Rigol's "long memory"
« Reply #35 on: September 23, 2010, 04:48:37 am »
Just got my GDS-1062A today. Have only been playing with it.

No major (or minor) complaints so far. I'm pretty happy with it, but then I don't have anything to compare it to (except the Tektronix in school lab).

Scrolling is smooth. Tried different time bases, etc. No noticeable lag any time.

I want to test it out for future potential buyers' benefit, but don't really know how. And I don't have a signal generator (I do have MCUs, passive components, diodes, transistors, common chips, etc). Any suggestions? I don't even have something that can generate short enough rise time to test the bandwidth!

Just a note before I forget - Yes, 1GS/s can be used with 2M long memory mode. 100us is the maximum timebase for 1GS/s (2ms recorded). It switches to 500MS/s at 250us.

For equivalent time sampling -
25ns -> 1GS/s
10ns -> 2.5GS/s
5ns -> 5GS/s
2.5ns -> 10GS/s
1ns (min) -> 25GS/s

So when the manual says the long memory cannot be used at <25ns, it just means it cannot be used with ETS. So yes, it is better than Rigol in this regard (1GS/s with 2M).

At 1x probe, DC coupling, averaging off, 1Gsps, probe shorted to ground clip, 2mV/div, 1ms/div, I am getting
Vavg = -117uV, Vp-p = 400uV
If that means anything (I'm curious. Anyone want to do the same on Rigol and see what you get?).

Interestingly, if I switch to ground coupling, Vavg becomes 89.1uV, and Vp-p = 0.
 

Offline joelby

  • Frequent Contributor
  • **
  • Posts: 634
Re: Rigol's "long memory"
« Reply #36 on: September 23, 2010, 05:02:08 am »
One thing I've seen done when trying to find a signal's noise floor is clipping the probe tips together and the grounds together and looking at the amount of stray noise.

It's not clear if it's the room I'm in or my scope, but my Rigol DS1102E registers about 7.5 mV RMS of noise, pretty much all 50 Hz. I compared this to a much more expensive Tektronik scope (both in the same room) and it was far, far better, though I don't remember what the signal was.

Quote
At 1x probe, DC coupling, averaging off, 1Gsps, probe shorted to ground clip, 2mV/div, 1ms/div, I am getting
Vavg = -117uV, Vp-p = 400uV

The DS1102E registers Vavg 360uV, Vpp 2.00mV, which does seem a bit noisier than yours.

Quote
Interestingly, if I switch to ground coupling, Vavg becomes 89.1uV, and Vp-p = 0.

Does it have a self-calibration routine? The DC offset on mine seems to drift a fair bit over time.
 

Offline cyberfish

  • Regular Contributor
  • *
  • Posts: 239
  • Country: gb
Re: Rigol's "long memory"
« Reply #37 on: September 23, 2010, 06:20:51 am »
Quote
One thing I've seen done when trying to find a signal's noise floor is clipping the probe tips together and the grounds together and looking at the amount of stray noise.
Is that to measure background noise?

I get 10mV Vpp, ~2mV Vavg if I do that. But it jumps around quite a bit if I just put my hand near it.

Quote
It's not clear if it's the room I'm in or my scope, but my Rigol DS1102E registers about 7.5 mV RMS of noise, pretty much all 50 Hz. I compared this to a much more expensive Tektronik scope (both in the same room) and it was far, far better, though I don't remember what the signal was.
I live in a country with 60Hz AC, but my noise is pretty much uniform on FFT. No spikes.

Quote
Does it have a self-calibration routine? The DC offset on mine seems to drift a fair bit over time.
It does, but I don't have a BNC-BNC cable, so I can't really use it.
 

alm

  • Guest
Re: Rigol's "long memory"
« Reply #38 on: September 23, 2010, 06:39:42 am »
Just got my GDS-1062A today. Have only been playing with it.
Congratulations!

I want to test it out for future potential buyers' benefit, but don't really know how. And I don't have a signal generator (I do have MCUs, passive components, diodes, transistors, common chips, etc). Any suggestions? I don't even have something that can generate short enough rise time to test the bandwidth!
The Jim Williams design in an appendix of LT AN-47 (other description here) is the closest thing you get to a fast pulse generator from jellybean parts. But the parts aren't terribly common, and you really need a fast scope (eg. sampling scope) to determine the real amplitude. The amplitude depends on the individual transistor behavior and strays, and is critical for calculating rise time (which is based on 10% to 90% of the original amplitude). Jim Williams has an improved design with a piece of hard coax as charge line, but it's more complex. This one generates a step instead of a pulse. Probably not worth the effort, I wouldn't worry about something as basic as bandwidth, they probably got that one right. Something like linearity of the attenuator over frequency range is more interesting (I suspect the Rigol is not great in that regard), but hard to do without expensive equipment.

I would check Dave's review of the Rigol and see how it compares on the good and bad points he mentions. Also subjective things like build quality (on the outside), user interface, responsiveness (especially with long records and things like measurements or math) and fan noise.

Some complaints I've heard about the Rigol are that triggering is not terribly stable (the wave always dances around slightly, even with a stable and noise-free signal) and that the measurements are based on the image on the screen, not the original signal. Things like bad accuracy compared to precision, and measurements that change depending on the vertical position on screen. Since I don't have a Rigol scope, I can't give you specific examples. It would be useful to know if the Instek scope has the same issues.

So when the manual says the long memory cannot be used at <25ns, it just means it cannot be used with ETS. So yes, it is better than Rigol in this regard (1GS/s with 2M).
Thanks for clearing that up! That's a clear pro over the Rigol. Nobody cares about ETS (at least I don't), so that doesn't matter.

At 1x probe, DC coupling, averaging off, 1Gsps, probe shorted to ground clip, 2mV/div, 1ms/div, I am getting
Vavg = -117uV, Vp-p = 400uV
Note that that makes a pretty good antenna. Doug Smith (EMC guru) actually uses it as a magnetic field detector. So it's probably not a great test. Look at what Dave did in one of his unexplained scope phenomenon blogs with the aluminum foil. Even better would be attaching a BNC short or terminator (50ohm is pretty close to a short for a 1Mohm input) directly to the female BNC connector.

Interestingly, if I switch to ground coupling, Vavg becomes 89.1uV, and Vp-p = 0.
Try letting it warm up for an hour or so, and then run signal path compensation (or however that's called on the Instek scope). That's designed to null temperature-dependent offsets, and usually recommended every time the ambient temp changes by more than a few K or so and you want to do accurate measurements.
 

Offline cyberfish

  • Regular Contributor
  • *
  • Posts: 239
  • Country: gb
Re: Rigol's "long memory"
« Reply #39 on: September 24, 2010, 12:16:34 am »
Quote
The Jim Williams design in an appendix of LT AN-47 (other description here) is the closest thing you get to a fast pulse generator from jellybean parts. But the parts aren't terribly common, and you really need a fast scope (eg. sampling scope) to determine the real amplitude. The amplitude depends on the individual transistor behavior and strays, and is critical for calculating rise time (which is based on 10% to 90% of the original amplitude). Jim Williams has an improved design with a piece of hard coax as charge line, but it's more complex. This one generates a step instead of a pulse. Probably not worth the effort, I wouldn't worry about something as basic as bandwidth, they probably got that one right. Something like linearity of the attenuator over frequency range is more interesting (I suspect the Rigol is not great in that regard), but hard to do without expensive equipment.
The circuit looks pretty simple, but i don't have a boost converter to get 90V DC. Actually, wouldn't it be easier to just rectify and filter my 110V AC from the wall? I don't have enough training to play with mains voltage, though. So I will leave that alone.

I was hoping to be able to find out the "actual" bandwidth.

Perhaps a high speed FPGA's output would be fast enough? Sadly I don't have one around me either.

As for the subjective things -
Build quality is good. Nothing filmsy. User interface is well laid out. Didn't notice any peculiar sequences needed to access any function.
Responsiveness is good when scrolling without math. No noticeable lag any time. With FFT it starts lagging noticeably at high time bases. Gets to about 2 frames per second.
Fan noise I don't know. I am in a room with 5 computers and probably 20 fans =) It sounds like jet engine.

I can try the other things you mentioned (triggering, measurements, accuracy), but do you have more detail on how to test?

Quote
Note that that makes a pretty good antenna. Doug Smith (EMC guru) actually uses it as a magnetic field detector. So it's probably not a great test. Look at what Dave did in one of his unexplained scope phenomenon blogs with the aluminum foil. Even better would be attaching a BNC short or terminator (50ohm is pretty close to a short for a 1Mohm input) directly to the female BNC connector.
That makes sense. Will try that tonight.

Quote
Try letting it warm up for an hour or so, and then run signal path compensation (or however that's called on the Instek scope). That's designed to null temperature-dependent offsets, and usually recommended every time the ambient temp changes by more than a few K or so and you want to do accurate measurements.
I don't have a BNC-BNC cable, so I can't really do that. For some reason just using the 2 probes with the tips and ground clips connected doesn't work. Impedance too high probably.
 

Offline Bored@Work

  • Super Contributor
  • ***
  • Posts: 3932
  • Country: 00
Re: Rigol's "long memory"
« Reply #40 on: September 24, 2010, 04:23:59 am »
The circuit looks pretty simple, but i don't have a boost converter to get 90V DC

The thing doesn't require any specific DC/DC converter, just a low amp, approx. 90V one. Jim Williams has published a few variants of the circuit over time in other application notes, sporting other DC/DC converters and triggerable versions. Some have DC/DC converters without the diode voltage doubler, and use converters going straight to 90V.

Years ago I build one such avalanche generator. Not having the particular IC from AN-47 at hand, and not having a HV DC/DC converter IC at hand, I designed my own DC/DC converter. I used some bog-standard DC/DC IC, dimensioned it as a 9V -> 30V boost converter, followed by a diode voltage-tripler and closing the feedback loop over the tripler and the IC with a 90V:1.25V resistive divider. Pretty much what is in AN-47, just with a tripler instead of a doubler.

Efficiency of my converter wasn't great. I overestimated the power requirements and ended up with the converter working in non-continuous mode. Using a 9V battery as power source also wasn't too clever (they have the worst energy density of batteries). Still it worked good enough.

I would say, except for the 90V it is a fun weekend project. But remember, 90V are already in the dangerous voltage territory, and if you mess up the converter you can easily get higher voltages before the magic smoke is released from the circuit.

I delete PMs unread. If you have something to say, say it in public.
For all else: Profile->[Modify Profile]Buddies/Ignore List->Edit Ignore List
 

Offline cyberfish

  • Regular Contributor
  • *
  • Posts: 239
  • Country: gb
Re: Rigol's "long memory"
« Reply #41 on: September 24, 2010, 05:36:35 am »
1. Measurements are done on the actual data. I can still get the 1kHz frequency reading from the compensation output even if I zoom out to 100ms/div (becomes big yellow band).
2. I have no way of measuring accuracy, but vertical offset (tried from -30V to 30V) does not change the readings appreciably (on the compensation output).
3. Used soundcard to generate 8kHz, 2.3Vpp sine wave. Appears to trigger reliably. I put it on continuous trigger mode, and moved the trigger from peak to peak. Waveform display is always very stable, and trace is thin. Tried manual triggering a few times, too. All images look identical.

Quote
The thing doesn't require any specific DC/DC converter, just a low amp, approx. 90V one. Jim Williams has published a few variants of the circuit over time in other application notes, sporting other DC/DC converters and triggerable versions. Some have DC/DC converters without the diode voltage doubler, and use converters going straight to 90V.

Years ago I build one such avalanche generator. Not having the particular IC from AN-47 at hand, and not having a HV DC/DC converter IC at hand, I designed my own DC/DC converter. I used some bog-standard DC/DC IC, dimensioned it as a 9V -> 30V boost converter, followed by a diode voltage-tripler and closing the feedback loop over the tripler and the IC with a 90V:1.25V resistive divider. Pretty much what is in AN-47, just with a tripler instead of a doubler.

Efficiency of my converter wasn't great. I overestimated the power requirements and ended up with the converter working in non-continuous mode. Using a 9V battery as power source also wasn't too clever (they have the worst energy density of batteries). Still it worked good enough.

I would say, except for the 90V it is a fun weekend project. But remember, 90V are already in the dangerous voltage territory, and if you mess up the converter you can easily get higher voltages before the magic smoke is released from the circuit.
That sounds like fun!

I'm not ready for high voltage, yet, though. So I think I will pass on that for now.
 

alm

  • Guest
Re: Rigol's "long memory"
« Reply #42 on: September 24, 2010, 08:11:54 am »
The circuit looks pretty simple, but i don't have a boost converter to get 90V DC. Actually, wouldn't it be easier to just rectify and filter my 110V AC from the wall? I don't have enough training to play with mains voltage, though. So I will leave that alone.
I would recommend against that for safety reasons, you would have to treat the whole setup as dangerous mains voltage, and you would short neutral to ground by connecting it to the scope (don't even think about floating the scope/disconnecting the ground connection).

I was hoping to be able to find out the "actual" bandwidth.
You either need a very fast edge (significantly faster than the rise time of your scope, usually 3.5ns for 100MHz bandwidth) or a leveled sine wave from low frequency (<=1MHz or so) to beyond the bandwidth of your scope.

Perhaps a high speed FPGA's output would be fast enough? Sadly I don't have one around me either.
You need a rise time faster than 1ns or so, but a relatively long pulse length. I'd just stop worrying about it.

As for the subjective things -
[...]
Sounds good so far.

I can try the other things you mentioned (triggering, measurements, accuracy), but do you have more detail on how to test?
Someone with a Rigol scope would probably able to provide better reference values, my information about issues with the Rigol scope is purely based on what I heard.

I don't have a BNC-BNC cable, so I can't really do that. For some reason just using the 2 probes with the tips and ground clips connected doesn't work. Impedance too high probably.
Does it require BNC connections? Usually it just requires all cables to be removed, but I'm not familiar with the Instek scope.
 

Offline DJPhil

  • Frequent Contributor
  • **
  • Posts: 511
Re: Rigol's "long memory"
« Reply #43 on: September 24, 2010, 10:58:39 am »
Cyberfish: I'm loving all this information, would you consider starting a new thread with a summary of your observations?

I just feel like we're wandering off topic and I'm pretty sure we'll eventually have a lot of questions pile up soon.

Good idea?
 

Offline cyberfish

  • Regular Contributor
  • *
  • Posts: 239
  • Country: gb
Re: Rigol's "long memory"
« Reply #44 on: September 24, 2010, 06:11:59 pm »
Cyberfish: I'm loving all this information, would you consider starting a new thread with a summary of your observations?

I just feel like we're wandering off topic and I'm pretty sure we'll eventually have a lot of questions pile up soon.

Good idea?

Terribly good idea.

Let's continue the discussion here
https://www.eevblog.com/forum/index.php?topic=1358.0
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf