EEVblog Electronics Community Forum
Products => Crowd Funded Projects => Topic started by: rthorntn on December 28, 2013, 04:58:15 am
-
For $300 USD:
http://www.kickstarter.com/projects/1855991221/10-ghz-usb-oscilloscope (http://www.kickstarter.com/projects/1855991221/10-ghz-usb-oscilloscope)
Manual for his previous version: http://www.fastsampling.com/Products/DS800/DS800Manual14.pdf (http://www.fastsampling.com/Products/DS800/DS800Manual14.pdf)
To a beginner this appears to be heaps more bandwidth than my 100Mhz Rigol?
Thoughts?
-
I hold some strong reservations on that, mainly due to the plastic enclosure and the switch-mode supply being so close to the ADC,
-
$300 per scope.
They need $80k.
Ain't gonna happen.
It's a 10 GHz sampling scope with 1V max inputs. Highly specialized for a hobbyist. Are there really 270 people with $300 to burn on Kickstarter, who have a use for such a thing, who don't already have access to one? This is a 10 GHz sampling scope, not a µCurrent...
-
Both the µCurrent and this device are niche instruments. The µCurrent might be slightly more useful to the average user, but I expect the µCurrent to bring in much more money mainly because a) the people behind the sampling scope lack Dave's audience and reputation and b) because the µCurrent is much less risky. The µCurrent gold is an upgrade to an existing upgrade that's (unless Dave really screws up) guaranteed to be at least as good as the normal µCurrent, the sampling scope is a fairly ambitious product. I agree that it seems unlikely to meet its target unless they can convince a large audience that they need this product (similar to the µCurrent :P).
-
No probes, well, not like he can include a 10GHz probe without adding a few more zeros to the price.
And why do I get the impression that he's getting his spec through equivalent time sampling?
-
switch-mode supply being so close to the ADC,
what ADC? :wtf:
-
And why do I get the impression that he's getting his spec through equivalent time sampling?
rtfm or at least the description of that project / that SAMPLING scope :rant:
-
And why do I get the impression that he's getting his spec through equivalent time sampling?
rtfm or at least the description of that project / that SAMPLING scope :rant:
Which is another way of saying that "yes it's an equivalent time sampling scope" and "yes, you can read all about it in this here fine manual (http://www.fastsampling.com/Products/DS800/DS800Manual.pdf)".
I couldn't find the sampling rate however. He mentions 12-bit ADC and using a microcontroller, and no mention of a discrete ADC. :-//
-
For $300 USD:
Thoughts?
here a closer look on the DS800, there was clearly room for improvement (from a PCB size point of view),
which he finally did with DS100.
(http://www.mikrocontroller.net/attachment/105163/DS800.jpg)
his hardware works, the software is not modern or complex, but it does works as well.
I was thinking to buy one, however the price was bit high. These 300$ for DS100 are ok.
I couldn't find the sampling rate however. He mentions 12-bit ADC and using a microcontroller,
and no mention of a discrete ADC. :-//
The "sampling head" on the analog input is made from an comparator (ADCMP582), the trigger input is using ADCMP567. The max. sample rate is defined by the delay line resolution (on DS800 that 1ps) and the 12bit vertical resolution. Therefore is the maximum sample rate, for DS800, ~ 83GS/s
How sampling scope works has been described e.g. here: http://www.tek.com/dl/85W_23777_0_0.pdf (http://www.tek.com/dl/85W_23777_0_0.pdf)
-
I expect the µCurrent to bring in much more money mainly because a) the people behind the sampling scope lack Dave's audience and reputation and b) because the µCurrent is much less risky. The µCurrent gold is an upgrade to an existing upgrade that's (unless Dave really screws up) guaranteed to be at least as good as the normal µCurrent, the sampling scope is a fairly ambitious product.
the DS800 sampling scope exists since 4 years or so, there is no risk at all with the DS100. The developer seems to be however not good in marketing questions (the kickstarer page didn't looks very good, the 1GHz prototype picture is misplaced in my opinion, one ca think he is trying to build something where in principle he need some cash to produce enought units to lower costs).
The developer tried to make some deals with other distributors, e.g. http://www.ichaus.de/product/iC227 (http://www.ichaus.de/product/iC227)
-
The "sampling head" on the analog input is made from an comparator (ADCMP582)
That's simplified to the point of uselessness. A comparator doesn't make a sampling gate, it could be part of a diode bridge sampling gate but making one with 10 GHz bandwidth seems highly non trivial.
-
Well, there is almost no point in building an expensive USB oscilloscope, because LCDs, buttons and rotary knobs are cheap these days.
-
Well, there is almost no point in building an expensive USB oscilloscope, because LCDs, buttons and rotary knobs are cheap these days.
Then it would cost a lot more.
-
That's simplified to the point of uselessness. A comparator doesn't make a sampling gate
:bullshit:
-
Well, there is almost no point in building an expensive USB oscilloscope, because LCDs,
buttons and rotary knobs are cheap these days.
the PC software is not complex, one can add an extra ARM board (with LCD), code simple UI and add some knobs,
then simply send the data via USB to the DS100 PCB - ready. Or even chepaer write an control app for smartphone.
-
From the manual "DS800 will work only with repetitive signals, since it requires multiple signal repetitions to
complete conversion"
That's what happens people with no high-speed interface knowledge try to make sophisticated circuits
The osciloscopes are all about transient events.
-
That's simplified to the point of uselessness. A comparator doesn't make a sampling gate
bullshit
I have my doubts about whether this is really a sampling scope in the same way as the Tek PDF you linked ... it seems like it compares the signal and it's digital approximation to increment/decrement the approximation (the only other implementation of this concept I know of called it a stroboscopic converter).
-
That's what happens people with no high-speed interface knowledge try to make sophisticated circuits
Sampling scopes have been around for decades. Mostly because they can achieve bandwidths beyond what's feasible for real-time scopes. Just look at the line-up of the big scope manufacturers: Agilent, Lecroy and Tektronix. They all have sampling scopes for the high bandwidth applications. The fastest digital real-time scope for $300 will probably have about 200 MHz bandwidth at best. That's almost 2 orders of magnitude worse than what this sampling scope offers.
The osciloscopes are all about transient events.
So analog scopes are useless?
-
Was his guy was on the forum a few years ago and everyone shot him down (rightly or wrongly im not capable of answering) as I'm sure Ive seen the prototype before on this forum.
-
Was his guy was on the forum a few years ago and everyone shot him down (rightly or wrongly im not capable of answering) as I'm sure Ive seen the prototype before on this forum.
no, definitely not on this forum, and honestly i haven't seens any DS800 prototype pictures before his current KS project.
-
An affordable, 10 GHz Bandwidth, USB Sampling Oscilloscope
This project is about production funding of a version of already existing product. Here is the link to the predesesor manual. http://www.fastsampling.com/Products/DS800/DS800Manual14.pdf (http://www.fastsampling.com/Products/DS800/DS800Manual14.pdf)
Stopped reading right there.
-
And why do I get the impression that he's getting his spec through equivalent time sampling?
rtfm or at least the description of that project / that SAMPLING scope :rant:
My bad, but still, that limits the market a great deal. If the thing could be used to look at say SATA or USB3 data there would be a lot more interest. As it is it's only good for checking signal quality. Granted you'll need LA's for those purposes but $300 just to check signal integrity is very limiting for a KS crowd.
-
As it is it's only good for checking signal quality. Granted you'll need LA's for those purposes but $300 just to check signal integrity is very limiting for a KS crowd.
It compares well with http://www.picotech.com/picoscope9200.html (http://www.picotech.com/picoscope9200.html) which at £5995 is described as "at a price you can afford".
Maybe the KS crowd would find the base price of an Agilent DSA90804A 8GHz real time scope at $104,745 a bit limiting as well.
The campaign isn't very well presented but I would say he/they look competent to deliver. He/they is trying to raise $80k to fund a worthwhile size production batch. I hope he makes it.
-
Was his guy was on the forum a few years ago and everyone shot him down (rightly or wrongly im not capable of answering) as I'm sure Ive seen the prototype before on this forum.
IIRC that was another one, looked much more advanced than this, and was real time.
-
My bad, but still, that limits the market a great deal. If the thing could be used to look at say SATA or USB3 data there would be a lot more interest. As it is it's only good for checking signal quality. Granted you'll need LA's for those purposes but $300 just to check signal integrity is very limiting for a KS crowd.
The issue is that it's quite limited unless you have expensive high bandwidth high impedance probes to go with it.
-
It compares well with http://www.picotech.com/picoscope9200.html (http://www.picotech.com/picoscope9200.html) which at £5995 is described as "at a price you can afford".
here we go, the 11GHz version of DS800 can be bought for $2,602.
http://shop.ichausamerica.com/ProductDetails.asp?ProductCode=iC227+Dual+11GHz++Oscilloscope (http://shop.ichausamerica.com/ProductDetails.asp?ProductCode=iC227+Dual+11GHz++Oscilloscope)
-
I couldn't find the sampling rate however. He mentions 12-bit ADC and using a microcontroller,
and no mention of a discrete ADC. :-//
The "sampling head" on the analog input is made from an comparator (ADCMP582), the trigger input is using ADCMP567. The max. sample rate is defined by the delay line resolution (on DS800 that 1ps), and the vertical resolution (12bit 500kps/s ADC in PIC24HJ64). Therefore is the maximum sample rate, for DS800, ~ 83GS/s
I'm aware of how equivalent time sampling works. But you did answer my question, it is as I feared... use the PIC for sampling. I don't quite understand why he would go to the trouble of making a nice frontend etc, and then use such a limited actual, non-equivalent, quite real sampling rate of 500 ksps. High speed at 12-bit isn't all that expensive.
If you have a signal that has reasonable but not great long term stability, then you can make do with equivalent time sampling as long as you don't take eons to gather all your samples. However I can see the use case for this where 500 ksps is sufficient (checking SI on your pcbs etc). It's only too bad IMO, because even a lowly 40 MSPS 12-bit ADC would be a nice improvement. That said, even with the 500 ksps limitation this looks quite interesting.
-
I don't quite understand why he would go to the trouble of making a nice frontend etc, and then use such a limited actual, non-equivalent, quite real sampling rate of 500 ksps. High speed at 12-bit isn't all that expensive.
If you have a signal that has reasonable but not great long term stability, then you can make do with equivalent time sampling as long as you don't take eons to gather all your samples. However I can see the use case for this where 500 ksps is sufficient (checking SI on your pcbs etc). It's only too bad IMO, because even a lowly 40 MSPS 12-bit ADC would be a nice improvement.
my guess: because then he would have to split the analog signal, add second path with attn/opamps and care much more about the higher speed ADC. Sure, the problem with splitter in RF path can be easy solved with second input, but then this would be not good from usability point of view (when i have to swap cables, well then i swap to another decent RT DSO). The higher speed ADC is not necessary for the sampling scope, 500ks/s is fast enought to get all the updates captured. With higher sampling there would be no improvement, but only additional noise source (and cost factor). The only improvement i see might be 14bit ADC to get real 12bit resolution, decent enclosure and more modern-looking software. He said, he is not using the ADC at all.
But that for DS800, no idea what he will use as ADC on DS100.
-
I wonder how he gets reasonable slew rates at reasonable sampling rates. As I said, it's not a proper sampling scope ... if you look the input is directly connected to the inverting input of the comparator (well, through a coupling capacitor in the picture). There is no room for a sampling gate on the PCB.
Does he try to predict the slew rate of the signal and increase the approximation accordingly rather than single stepping it?
-
However I can see the use case for this where 500 ksps is sufficient (checking SI on your pcbs etc). It's only too bad IMO, because even a lowly 40 MSPS 12-bit ADC would be a nice improvement.
If the front end is a high speed comparator I don't see how it is especially using the processor ADC at all.
My guess is it is a successive approximation ADC using the front end comparator with the 'wrinkle' that the comparator is sampled after a precise time delay from the trigger input. It needs a DAC of some kind to set comparator reference voltage and each point on the captured waveform will require n trigger events to achieve n bit vertical resolution.
Knowing that adjacent samples are likely to have similar values there would be scope for optimising the successive approximation. It might be quicker to implement some kind of tracking converter.
-
Sampling scopes have been around for decades. Mostly because they can achieve bandwidths beyond what's feasible for real-time scopes. Just look at the line-up of the big scope manufacturers: Agilent, Lecroy and Tektronix. They all have sampling scopes for the high bandwidth applications. The fastest digital real-time scope for $300 will probably have about 200 MHz bandwidth at best. That's almost 2 orders of magnitude worse than what this sampling scope offers.
Ok. Let's examine this case. Assume I have a scope that sampling rate is much lower than the signal and uses equivalent-time sampling thing. When I dedice to make some RF thing such as 1GHz oscilator, amplfier etc.., that scope is completly useless. Why ? because the signal must repeat identically each time however there could be huge phase nosie in my hypothetical circuit or a mismatch or intermodulation etc... and I can not able to troubleshoot with that type of osciloscope (probably I need a spectrum analyer for that :P)
My point is that I need an osiloscope to evaluate undeterminted responses of my circuits. Those signals may or may not repeat them selfs. That's why I'm using an osciloscope to find out.
-
My point is that I need an osiloscope to evaluate undeterminted responses of my circuits. Those signals may or may not repeat them selfs. That's why I'm using an osciloscope to find out.
Then go spend $104,000 on an real time 8GHz Agilent and stop moaning here.
-
Then go spend $104,000 on an real time 8GHz Agilent and stop moaning here.
Thank you for your enlightening and insightfull response :palm:
-
The issue is that it's quite limited unless you have expensive high bandwidth high impedance probes to go with it.
That's not the case. These high speed sampling head systems are used for design verifcation rather than debugging arbitrary circuits, and suitable monitoring circuitry and test connectors will have been designed into the system under test.
-
This is a really nice device at an excellent price if he can get the backing. There are scores of people involved in picosecond electronics who have to rely on old boat anchor sampling systems from TeK and HP (which are getting harder and harder to maintain) which would buy into this if they are aware of it. I hope the fellow partners up with someone with some marketing flair.
-
The issue is that it's quite limited unless you have expensive high bandwidth high impedance probes to go with it.
That's not the case. These high speed sampling head systems are used for design verifcation rather than debugging arbitrary circuits, and suitable monitoring circuitry and test connectors will have been designed into the system under test.
That and for testing some asynchronous fpga abuse in timing applications for which it would be just the ticket. The $300 price level is a no-brainer. The only reason I haven't backed it already is that I cannot really miss $300 on a gamble right now. If I was sure I'd plonk down $300 for that sort of functionality since right now I have nothing in that segment of gear.
-
However I can see the use case for this where 500 ksps is sufficient (checking SI on your pcbs etc). It's only too bad IMO, because even a lowly 40 MSPS 12-bit ADC would be a nice improvement.
If the front end is a high speed comparator I don't see how it is especially using the processor ADC at all.
My guess is it is a successive approximation ADC using the front end comparator with the 'wrinkle' that the comparator is sampled after a precise time delay from the trigger input. It needs a DAC of some kind to set comparator reference voltage and each point on the captured waveform will require n trigger events to achieve n bit vertical resolution.
and it is exact like that, the misleading information was the 12bit resolution and 12bit ADC inside PIC. In the reality he isn't using the ADC to measure anything.
-
Ok. Let's examine this case. Assume I have a scope that sampling rate is much lower than the signal and uses equivalent-time sampling thing. When I dedice to make some RF thing such as 1GHz oscilator, amplfier etc.., that scope is completly useless. Why ? because the signal must repeat identically each time however there could be huge phase nosie in my hypothetical circuit or a mismatch or intermodulation etc... and I can not able to troubleshoot with that type of osciloscope (probably I need a spectrum analyer for that :P)
I'm getting the impression you've never used a sampling scope before. Sampling scopes require a trigger clock input, including the scope we're discussing here, which can be generated by splitting the input waveform into two -3 dB signals. The scope triggers off this clock and not from another clock (such as an internal 10 MHz) that has no relation to the input signal. If there is phase noise in the clock, you will still see the clock waveform on the scope, and then you will be able to see that the edges of the waveform are fuzzy from the phase noise or jitter. This is actually useful information for debugging your circuit and will inform you that you have a phase noise problem. If you then put the input signal through an amplifier, you will fully be able to see how the amplifier has affected the signal. If the amplifier is oscillating or having a drastic problem, then you will see noise instead of a clean, amplified signal. This will also tell you that there is something wrong with your circuit, and when it comes out as a clean, amplified signal, then you know your circuit is working.
-
Hat this thing does is trigger at a certain point of the incoming signal. A variable delay line then shifts the sample point back and forward. So they adjust the delay between where they trigger and where they actually measure. That is how they reconstruct the signal. So with a few k samples a second you can still trap a very fast signal. If you lcd display of the scope os 800 pixels horizontal you can sample 800 samples a second a still have a refreshed image every second.
The key thing is having a very fast sample gate.
To find the first point : set delay to zero. Sweep a simple dac to find the point where the comparator toggles. The dac value is the incoming level at time 0.
Increment the delay between trigger and sampler. Sweep your dac again. That gives you the value at trigger + 1 delay step.
Increment delay again. And so on...
The only important thing is the sample gate and the resolution of the delay generator.
This guy is using timing equalisers coming from the fiber optics world. Companies like Vitesse and Ti and Synertec have special chips with programmable delay generators in steps of a few picoseconds.
Most likely that is what he is using. His frontend is a laser diode pickup amplifer.
The pic is a slowpoke , doesnt matter. He uses two optocouplers to isolate the thing and sends it over a silabs usb uart.
The only really fast elements are the trigger comparator , the delay generator chip and the sampler gate. The rest is kilohertz domain.
-
I'm getting the impression you've never used a sampling scope before. Sampling scopes require a trigger clock input, including the scope we're discussing here, which can be generated by splitting the input waveform into two -3 dB signals. The scope triggers off this clock and not from another clock (such as an internal 10 MHz) that has no relation to the input signal. If there is phase noise in the clock, you will still see the clock waveform on the scope, and then you will be able to see that the edges of the waveform are fuzzy from the phase noise or jitter. This is actually useful information for debugging your circuit and will inform you that you have a phase noise problem. If you then put the input signal through an amplifier, you will fully be able to see how the amplifier has affected the signal. If the amplifier is oscillating or having a drastic problem, then you will see noise instead of a clean, amplified signal. This will also tell you that there is something wrong with your circuit, and when it comes out as a clean, amplified signal, then you know your circuit is working.
I understand your point it's a convincing one and You are right I've never use this type of scope before. If I may say so, It seems more usefull as a production test scope.
-
thse machines are only useful for things like eye diagrams and jitter detection.
signal must be repetitive based on the trigger.
-
Then go spend $104,000 on an real time 8GHz Agilent and stop moaning here.
A couple 100 point digitizer could be made for far far less ... hmm, might make an interesting kickstarter.
-
Then go spend $104,000 on an real time 8GHz Agilent and stop moaning here.
A couple 100 point digitizer could be made for far far less ... hmm, might make an interesting kickstarter.
Indeed. I recently read an interesting kickstarter for a nice sampler that is far far far less than $104,000 (http://www.kickstarter.com/projects/1855991221/10-ghz-usb-oscilloscope).
-
Single shot digitizer.
-
Single shot digitizer.
Memory depth is not the #1 cost driver of those fast real time samplers. Nor #2. Or #3. Maybe #4, who knows. :-//
But I like being wrong, because in this case being wrong is advantageous. Lets go with couple hundred being 512 sample points. Lets go with old school 8 bit, so none of that newfangled 12-bit stuff. The bleeping expensive $100k+ scope under discussion ==> 8 GHz bandwidth signal is to be sampled in real time, 8 bit samples, 512 points. Define far far far less expensive, and rough idea of proposed solution.
How do you propose a real time sampler that can capture that 8 GHz BW signal for 512 points under lets say 1k?
And just to clarify, sampling real time 8 GHz BW signal != 8 GSPS ...
-
Memory depth isn't an issue with an ADC, but I'm suggesting a massively parallel sampler.
LLNL digitizer (http://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&cad=rja&ved=0CC8QFjAA&url=http%3A%2F%2Fwww.llnl.gov%2Fetr%2Fpdfs%2F04_94.1.pdf&ei=5ZPEUofiCseS0AXNsoG4Cw&usg=AFQjCNF9mUEDx3IB_9ccIjh2No2BR9rkBg&sig2=C1hOMTqGU1Eo8mg_9Uvc7A) ... the patents have expired I think.
-
Memory depth isn't an issue with an ADC, but I'm suggesting a massively parallel sampler.
LLNL digitizer (http://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&cad=rja&ved=0CC8QFjAA&url=http%3A%2F%2Fwww.llnl.gov%2Fetr%2Fpdfs%2F04_94.1.pdf&ei=5ZPEUofiCseS0AXNsoG4Cw&usg=AFQjCNF9mUEDx3IB_9ccIjh2No2BR9rkBg&sig2=C1hOMTqGU1Eo8mg_9Uvc7A) ... the patents have expired I think.
Given your low "a few hundred" I thought it might go into optimistic "a bunch of discrete points along a transmissions line" territory. So lets say 512 to keep it somewhat usable. So which diodes at 1000 units do you propose using? Yeeeeears ago I looked into something similar and it wasn't all that cheap at that point in time. But maybe you know the magic part at a nice price @1000 units at digikey, for use as S&H? In the meantime fpga's have progressed, so the generation of gate pulse etc has become simpler. Anyways, I don't see that becoming all that nice and cheap compared to the $300 equivalent time sampler from the OP.
-
It's optimistic to say I could do it ... but it's not optimistic as in not physically feasible, they did it all with commodity parts 20 years ago.
If you can avoid the truly low (<0.1pf) capacitance diodes and use one of the non exotic RF schottkys from a mainstream manufacturer you're only going to be out around 100$ for 1000 diodes.
-
How do you implement different timebase speeds?
-
You make a new PCB.
PS. unless you can find some really low cost way to make a 100 ps range pulse generator, so you can have one pulse generator per tap ... then you could do coarse delay in a FPGA, fine tuned with a varactor circuit (and on-line calibration).
-
I got the DS800E 11 Ghz version.
I did many test on pulses and RF and work very well.
Of course it is a sampling scope, only for use with repetitive signals.
I compared my test with a Tek TDS820 (6 Ghz) and Agilent MSO-X (1Ghz) and got similar results.
The software could be improved but I think this will be done.
Good value for the money.
eurofox
-
Can you post some pictures Eurofox. What's the build quality like? How was the delivery time?
-
Can you post some pictures Eurofox. What's the build quality like? How was the delivery time?
I intend to make a review, pictures, test with different sources (pulses in the ps second range, RF ....), I just order a 10 Ghz VCO.
I'm really surprise, this little scope perform very well.
Some work to do on the software to improve it but I will manage that.
eurofox
-
Im very much looking forward to your review eurofox :)
Might get myself one of these
-
I hope the price for ADC12D2000 will drop one day... Now it is $4,499 at Digi-Key. :-BROKE
-
The "sampling head" on the analog input is made from an comparator (ADCMP582), the trigger input is using ADCMP567. The max. sample rate is defined by the delay line resolution (on DS800 that 1ps) and the 12bit vertical resolution. Therefore is the maximum sample rate, for DS800, ~ 83GS/s
Based on photo that is not how it works and an ADCMP582 is not nearly fast enough to support that kind of input bandwidth. The two sampling inputs are at the bottom next to what I assume are a pair of RF mixers repurposed as sampling bridges. The four inputs at the top going to the comparator are the trigger inputs. This is a traditional sampling oscilloscope design.
-
I'm aware of how equivalent time sampling works. But you did answer my question, it is as I feared... use the PIC for sampling. I don't quite understand why he would go to the trouble of making a nice frontend etc, and then use such a limited actual, non-equivalent, quite real sampling rate of 500 ksps. High speed at 12-bit isn't all that expensive.
If you have a signal that has reasonable but not great long term stability, then you can make do with equivalent time sampling as long as you don't take eons to gather all your samples. However I can see the use case for this where 500 ksps is sufficient (checking SI on your pcbs etc). It's only too bad IMO, because even a lowly 40 MSPS 12-bit ADC would be a nice improvement. That said, even with the 500 ksps limitation this looks quite interesting.
Sampling rate is limited either by the maximum strobe rate like if the strobe is generated by an avalanche pulse generator or the delay time generator or counter which has to reset between samples. Higher sample rate designs grab multiple samples along a transmission line with one strobe as with the linked LLNL design.
My sampling oscilloscope only operates at about 50 ksps and is completely usable providing a real time display into the 10+ GHz range.
-
It's optimistic to say I could do it ... but it's not optimistic as in not physically feasible, they did it all with commodity parts 20 years ago.
My sampling oscilloscope is old enough to drink. :)
If you can avoid the truly low (<0.1pf) capacitance diodes and use one of the non exotic RF schottkys from a mainstream manufacturer you're only going to be out around 100$ for 1000 diodes.
Avago sells the needed diodes to get into at least the 4 to 6 GHz range and they are cheap. The strobe generator and timing circuits are much more difficult.
-
PS. unless you can find some really low cost way to make a 100 ps range pulse generator, so you can have one pulse generator per tap ... then you could do coarse delay in a FPGA, fine tuned with a varactor circuit (and on-line calibration).
An FPGA would need to be dedicated if used in the strobe or trigger critical path because of pattern dependent jitter and FPGA time to digital converters have pretty low resolution compared to the alternatives although they are mighty fast.
Time to voltage conversion could do better than 10ps over 50ns 20+ years ago so I would be disappointed in a modern design which could not interpolate within 1ps while using a 50 or 100 MHz counter. ECL logic was used back then and I would expect it to be used now in the critical paths.
-
I'm getting the impression you've never used a sampling scope before. Sampling scopes require a trigger clock input, including the scope we're discussing here, which can be generated by splitting the input waveform into two -3 dB signals. The scope triggers off this clock and not from another clock (such as an internal 10 MHz) that has no relation to the input signal. If there is phase noise in the clock, you will still see the clock waveform on the scope, and then you will be able to see that the edges of the waveform are fuzzy from the phase noise or jitter. This is actually useful information for debugging your circuit and will inform you that you have a phase noise problem. If you then put the input signal through an amplifier, you will fully be able to see how the amplifier has affected the signal. If the amplifier is oscillating or having a drastic problem, then you will see noise instead of a clean, amplified signal. This will also tell you that there is something wrong with your circuit, and when it comes out as a clean, amplified signal, then you know your circuit is working.
I understand your point it's a convincing one and You are right I've never use this type of scope before. If I may say so, It seems more usefull as a production test scope.
Just to give an example of what can be done, here is an oscillograph from my sampling oscilloscope loafing at 200 ps/div while measuring pattern dependent jitter in a TTL delay circuit that I was testing. Sample rate is about 50 kSamples/second. A standard CRT display was completely adequate for mark 1 eyeball use despite the low sample rate but in this case a CRT storage display was used to make a better photograph. Measurement jitter is insignificant on this scale at about 10ps.
-
The "sampling head" on the analog input is made from an comparator (ADCMP582), the trigger input is using ADCMP567. The max. sample rate is defined by the delay line resolution (on DS800 that 1ps) and the 12bit vertical resolution. Therefore is the maximum sample rate, for DS800, ~ 83GS/s
Based on photo that is not how it works and an ADCMP582 is not nearly fast enough to support that kind of input bandwidth.
He didn't say anything about the bandwidth, he said the maximum (equivalent) sample rate is 83 GS/s.
The two sampling inputs are at the bottom next to what I assume are a pair of RF mixers repurposed as sampling bridges.
The package is a match for the ADCMP582 and pin 3 on the ADCMP582 is the negative input ... and I would not expect mixer diodes in such a package.
-
It looks like you are right. I had to check the datasheet but it shows that the ADCMP582 has an input bandwidth of 8 GHz. That explains why that sampling design has such a slow sampling rate but on the other hand does not need a strobe driver or sampling bridge. Discrete versions of the later can certainly achieve higher sampling rates and bandwidth in a surface mount design but the strobe is is not trivial.
There are some RF mixers in similar packages which might work as sampling bridges.