EEVblog Electronics Community Forum
Products => Test Equipment => Topic started by: iraquois on November 01, 2020, 05:09:03 pm
-
Hi. I am currently workin on a project to have fast rise time about 1ns/250V. related to this question here:
https://www.eevblog.com/forum/testgear/very-fast-rise-time-5v-square-wave-generator/msg3258952/#msg3258952 (https://www.eevblog.com/forum/testgear/very-fast-rise-time-5v-square-wave-generator/msg3258952/#msg3258952)
But right now I decided to use relay instead of avalanche transistors. And got some relatively good results as 5-6ns/250V. I used a reed relay to switch charged capacitors to have this rise time. Circuit is like in the attachment.
But obviously I need sharper edge. I will try different relays like mercury relay. But I think there is another problem with measuring this fast signals. I have 1Ghz oscilloscope but my probe is only 400Mhz and its cable length is like 60-70cm. I think that might effect results. What kind of oscilloscope probe should I use to observe it clearly? Since probes are expensive is it possible to modify one of my probes or even build one?
-
Does changing the charging resistor (3 megohm in your drawing) have any effect on the transition times?
Can you build a 50-ohm matched attenuator to go from your 50 ohm output into the 50 ohm input on the oscilloscope through 50 ohm coax to test the rise and fall times? The voltage rating on the resistors may be important, but the power dissipation in the resistors for your waveform should be within reason
-
Hi. I am currently workin on a project to have fast rise time about 1ns/250V. related to this question here:
https://www.eevblog.com/forum/testgear/very-fast-rise-time-5v-square-wave-generator/msg3258952/#msg3258952 (https://www.eevblog.com/forum/testgear/very-fast-rise-time-5v-square-wave-generator/msg3258952/#msg3258952)
But right now I decided to use relay instead of avalanche transistors. And got some relatively good results as 5-6ns/250V. I used a reed relay to switch charged capacitors to have this rise time. Circuit is like in the attachment.
But obviously I need sharper edge. I will try different relays like mercury relay. But I think there is another problem with measuring this fast signals. I have 1Ghz oscilloscope but my probe is only 400Mhz and its cable length is like 60-70cm. I think that might effect results. What kind of oscilloscope probe should I use to observe it clearly? Since probes are expensive is it possible to modify one of my probes or even build one?
In the old times (Tektronix) they used pressurized mercury relays for below 1ns risetimes. Problem is ghastly jitter.
-
Tektronix managed rise times better than 250 ps back in the day before semiconductors that could manage this: http://w140.com/tekwiki/wiki/109 (http://w140.com/tekwiki/wiki/109)
But I think there is another problem with measuring this fast signals. I have 1Ghz oscilloscope but my probe is only 400Mhz and its cable length is like 60-70cm. I think that might effect results. What kind of oscilloscope probe should I use to observe it clearly? Since probes are expensive is it possible to modify one of my probes or even build one?
Since DC loading is not much an issue here, I would use a resistive divider probe, as discussed here (https://www.eevblog.com/forum/projects/lo-z-probe/). For example a resistive (at ~1 GHz) 47k resistor in series with a decent 50 Ohm coax cable into a 50 Ohm scope input would give roughly a 1000x attenuation. The parasitic capacitance / inductance from the physical layout will likely affects its performance. See the link to the Doug Smith design for one possible way of handling that.
Edit: if you output impedance is already a well-defined 50 Ohms, then building or buying a 50 Ohm attenuator capable of handling this power at 100s of MHz, as TimFox suggested, should perform better.
-
Tektronix managed rise times better than 250 ps back in the day before semiconductors that could manage this: http://w140.com/tekwiki/wiki/109 (http://w140.com/tekwiki/wiki/109)
But I think there is another problem with measuring this fast signals. I have 1Ghz oscilloscope but my probe is only 400Mhz and its cable length is like 60-70cm. I think that might effect results. What kind of oscilloscope probe should I use to observe it clearly? Since probes are expensive is it possible to modify one of my probes or even build one?
Since DC loading is not much an issue here, I would use a resistive divider probe, as discussed here (https://www.eevblog.com/forum/projects/lo-z-probe/). For example a resistive (at ~1 GHz) 47k resistor in series with a decent 50 Ohm coax cable into a 50 Ohm scope input would give roughly a 1000x attenuation. The parasitic capacitance / inductance from the physical layout will likely affects its performance. See the link to the Doug Smith design for one possible way of handling that.
Edit: if you output impedance is already a well-defined 50 Ohms, then building or buying a 50 Ohm attenuator capable of handling this power at 100s of MHz, as TimFox suggested, should perform better.
Wont work. Just think of a the equivalent circuit of an 47K resistor at 1GHz. Only coaxial attenuators / power dividers all in 50ohms technology would work. You would also need a probe with risetime a lot smaller than 350ps to see a 1ns transition correctly.
-
Charge line pulse generators with mercury relays can do 250ps rise times at very high voltages. As pointed out earlier, jitter is terrible.
I pulled one of these out of an obscure piece of physics equipment used with a 1960s electron microscope. The whole piece of equipment was a glorified box holding the relay so I pulled it out to save space. The trigger out originally used a UHF connector which I promptly replaced with BNC. The pickup uses a two turn loop with half a ferrite toroid to couple to the magnetic field of the center conductor. Really not the best choice when you consider the high peaks from differentiating a rectangular waveform, but it "works". Some clamping zeners are a good idea when using the full 3kV on the output. The large coaxial connector at the top is an HN connector, a high voltage variant of the N connector.
(http://i.imgur.com/PsM8QJim.jpg) (https://imgur.com/PsM8QJi) (http://i.imgur.com/EpjuU90m.jpg) (https://imgur.com/EpjuU90)
(http://i.imgur.com/HJ9A0N1m.jpg) (https://imgur.com/HJ9A0N1) (http://i.imgur.com/pXQKOFFm.jpg) (https://imgur.com/pXQKOFF)
Though it doesn't help with the high voltage aspect, you may be able to achieve 1ns rise times by using a negative resistance oscillator. HP has a good write up on how to properly use the N-curve of such devices to achieve very fast rise times. The classic part is a tunnel diode, but you can use two j-fets to get something similar and possibly other discrete transistors as well. It's been awhile since I last made one of these to play around with and I lost all my old photos and data on this since my old hard drive decided to commit suicide.
A reed switch might also provide acceptable performance if you don't care about its lifespan.
-
So use coaxial attenuators. Multiple in series if necessary. The first in the chain should be rated well above the voltage you expect at 1 GHz.
With a 1 GHz scope, which will have a risetime in the order of 350 ps, a probe with a rise time much less than 350 ps won't help much to improve the system rise time.
A 350 ps probe with a 350 ps scope will give you about 500 ps. A 175 ps (2 GHz) probe with a 350 ps scope will give about 400 ps. Not that much improvement for doubling the probe bandwidth, which will likely take much more than twice the effort.
I agree that neither is great for studying a 1 ns edge, but what do you do with a 1 GHz scope?
-
It's not the document I remember reading, but the HP 213B manual covers some of what I was talking about with using the N shaped characteristic curve on page 4.1 (theory of operation).
https://bama.edebris.com/download/hp/213b/HP%20213B%20Pulse%20Generator%20-%20Operating%20and%20Service%20Manual.pdf (https://bama.edebris.com/download/hp/213b/HP%20213B%20Pulse%20Generator%20-%20Operating%20and%20Service%20Manual.pdf)
High impedance passive probes are a bad idea. A 50ohm divider might work okay if you use proper layout, but I'd go with coaxial attenuators as already talked about. It's much simpler and I've used them with out incidence. Just make sure you know the peak power rating of each part and that it's not exceeded and you will be alright. Barth Electronics makes attenuators specifically for this type of situation, though they are hard to find and use GR874 connectors that are a bit expensive.
250Vpk into 50ohms is 1250Wpk. The pulse width determines the total energy of the pulse. But let's assume its narrow, <=100ns, and has low duty cycle say 1% or less. So heating is not going to be an issue. Most high wattage attenuators can handle significant peak power from a short pulse. Realistically, you would need at most three attenuators in series, but two would probably work okay. E.g., a 25W or 50W 20-40dB attenuator would drop you down to 25 to 2.5 Vpk and then a much lower wattage attenuator can drop that to somewhere the scope's input would be comfortable with. Attenuator bandwidth might limit performance, but even with crappy attenuators this is the easiest/fastest way to test the pulse generator and can be upgraded as you go. Cabling also needs to be considered carefully. The faster the rise time, the higher the attenuation of the edges as the pulse travels along the transmission line. I wouldn't use RG-58 for this type of testing. If you can, connect everything as directly as possible and avoid using any extra cabling or adapters.
-
But right now I decided to use relay instead of avalanche transistors. And got some relatively good results as 5-6ns/250V. I used a reed relay to switch charged capacitors to have this rise time. Circuit is like in the attachment.
What are you using for your "capacitors"? To get fast ristimes you need to use coaxial cable charged to twice the desired pulse amplitude. So in your case, 500V. The pulse width is twice the electrical length of the chargeline.
Also, how are you connecting the reed switch? Some pictures of your setup would help when giving advice on how to improve it.
-
Since the relay needs to be in a transmission line environment for such fast rise times, why use an oscilloscope probe at all? Connect the 50 ohm oscilloscope input directly to the 50 ohm output from the relay.
-
Since the relay needs to be in a transmission line environment for such fast rise times, why use an oscilloscope probe at all? Connect the 50 ohm oscilloscope input directly to the 50 ohm output from the relay.
I don't think his scope would like a 250V pulse at the input :-BROKE ;D. But I agree, the output of the generator should be connected as directly as possible to the input of the scope. Otherwise the measurements are meaningless. The nice thing about charge line pulse generators is that the rise time is independent of amplitude, so you can simply charge it with a lower voltage while measuring directly and be confident that you will get similar results when you apply the full voltage.
-
Since the relay needs to be in a transmission line environment for such fast rise times, why use an oscilloscope probe at all? Connect the 50 ohm oscilloscope input directly to the 50 ohm output from the relay.
I don't think his scope would like a 250V pulse at the input :-BROKE ;D. But I agree, the output of the generator should be connected as directly as possible to the input of the scope. Otherwise the measurements are meaningless. The nice thing about charge line pulse generators is that the rise time is independent of amplitude, so you can simply charge it with a lower voltage while measuring directly and be confident that you will get similar results when you apply the full voltage.
Does anybody specify their 50 ohm attenuators for peak power? I checked the ones I have and Mini-circuits does not say.
-
Since the relay needs to be in a transmission line environment for such fast rise times, why use an oscilloscope probe at all? Connect the 50 ohm oscilloscope input directly to the 50 ohm output from the relay.
I don't think his scope would like a 250V pulse at the input :-BROKE ;D. But I agree, the output of the generator should be connected as directly as possible to the input of the scope. Otherwise the measurements are meaningless. The nice thing about charge line pulse generators is that the rise time is independent of amplitude, so you can simply charge it with a lower voltage while measuring directly and be confident that you will get similar results when you apply the full voltage.
Does anybody specify their 50 ohm attenuators for peak power? I checked the ones I have and Mini-circuits does not say.
AFAIK Anritsu has some (N-Connector, expensive, ...)
-
Does anybody specify their 50 ohm attenuators for peak power? I checked the ones I have and Mini-circuits does not say.
Off the top my head HP/Agilent/Keysight do with their 849x step attenuators. They specify 1W average or 100W peak if <= 10us pulse width, which is pretty impressive for precision parts.
If anyone is interested, I just put together a homebrew charge line pulser using a normal reed switch to compare with the one made by iraquois. I should have time to test later on tonight. Hopefully I can get 1ns switching times, but who knows. I tried to make it fully coaxial by inserting the reed switch in the middle of an 8-inch piece of RG-58 and used a small 1in brass tube that barely fit over the dielectric. It looks like a monstrosity but it's good enough to test the idea.
-
(http://i.imgur.com/bvyk0kWl.jpg) (https://imgur.com/bvyk0kW)
Well... it works. Measured around 500 to 600 ps in realtime mode on my 500MHz, 2GS/s scope. So a normal reed switch can do fast pulses just fine. Even with terrible construction, like my unit, you can easily get <1ns rise times. I bet you could do better with microstrip PCB construction and edge mounted SMA connectors.
The charge line uses about 10 or 12 feet of RG-400 with a 1Mohm charging resistor. The current setup uses a magnet to switch the relay on. Since that is a hassle to do each time, I plan on winding a coil around the brass tube so I can trigger the switch automatically.
Now that we know it can switch fast enough, we still have some unanswered questions: Can it do HV switching? And how much do attenuators affect the rise time? Well, you'll have to tune in tomorrow. A man’s got to get his beauty sleep.
But before I go, here's some pictures detailing the construction of the pulser. The reed switch is from an old alarm system. I wrapped some teflon tape around so it would fit snugly in the center of the brass tube. The coax is RG-58 from the 1960s and the BNC connectors are random junk I had floating around. I have some nice semi-rigid coax I could have used but I decided not to sacrifice it to the god of idle curiosity.
(http://i.imgur.com/b1CovSwm.jpg) (https://imgur.com/b1CovSw) (http://i.imgur.com/kO0sc6qm.jpg) (https://imgur.com/kO0sc6q)
(http://i.imgur.com/GUArsWdm.jpg) (https://imgur.com/GUArsWd) (http://i.imgur.com/UeX4lY2m.jpg) (https://imgur.com/UeX4lY2)
(http://i.imgur.com/FrYpryQm.jpg) (https://imgur.com/FrYpryQ) (http://i.imgur.com/7SNRH11m.jpg) (https://imgur.com/7SNRH11)
Here's what the whole setup looks like.
(http://i.imgur.com/jqq4h1Rl.jpg) (https://imgur.com/jqq4h1R)
And here's with the timebase zoomed out a touch.
(http://i.imgur.com/ngNvuIol.jpg) (https://imgur.com/ngNvuIo)
-
Hello garrettm,
I don't understand how you get such small puls with a magnet?
Best regards
egonotto
-
I don't understand how you get such small puls with a magnet?
Hi egonotto, the magnet is used to trigger the reed switch, not actually induce the pulse. Reed switches are made from a ferrous metal whose contacts will "touch" (i.e. close the circuit) when there is a large enough magnetic field present. Ideally the leads of the reed switch should complete the magnetic circuit, but in practice you can place the magnet in many different directions to get the contacts to touch. The field that a soleniod coil makes is the ideal choice for operating the reed switch: the reed completes the magnetic circuit (with field lines parallel to the contacts) and the coil allows for easy control.
-
Egonotto, the pulse duration is just determined by the langth of the "storage cap coax", i.e. the stub "wire" that's on the charging end of the reed switch. The whole discharge takes place within the first episode of contact bounce (in case a dry reed switch is used). It doesn't really matter how long the contacts are closed, this just determines when the next pulse can be produced. The pulse duration is twice the wave propagation time of the stub end coax. The wave is generated by closing the contacts, travels to the open end of the stub, is reflected without phase reversal (since the wave resistance of the charging resistor placed there is for all practical reasons infinite) and travels back to the contacts and the output connector.
-
If anyone is interested, I just put together a homebrew charge line pulser using a normal reed switch to compare with the one made by iraquois. I should have time to test later on tonight. Hopefully I can get 1ns switching times, but who knows. I tried to make it fully coaxial by inserting the reed switch in the middle of an 8-inch piece of RG-58 and used a small 1in brass tube that barely fit over the dielectric. It looks like a monstrosity but it's good enough to test the idea.
Check out how Tektronix used reed relays in GHz+ designs:
http://w140.com/tekwiki/images/3/3c/Tek-7t11-right.jpg (http://w140.com/tekwiki/images/3/3c/Tek-7t11-right.jpg)
http://w140.com/tekwiki/images/4/43/7t11-trigger.JPG (http://w140.com/tekwiki/images/4/43/7t11-trigger.JPG)
-
Check out how Tektronix used reed relays in GHz+ designs:
http://w140.com/tekwiki/images/3/3c/Tek-7t11-right.jpg (http://w140.com/tekwiki/images/3/3c/Tek-7t11-right.jpg)
http://w140.com/tekwiki/images/4/43/7t11-trigger.JPG (http://w140.com/tekwiki/images/4/43/7t11-trigger.JPG)
That's some incredible workmanship and engineering! Though I think they were just using those as signal level switches. The structure of a reed switch and its ability to work coaxially really lends itself to RF transmission lines and microstrip PCB layout. I love the nice smooth traces that Tektronix used in their older equipment. I have some HP gear made by their branch in Germany that uses what I'd call hideous right-angle-only PCB traces that confuses me to no end. My older HP gear used nice smooth traces as shown in the Tek pictures. So why did HP Germany choose to route their traces the way they did?
I wish I had a faster oscilloscope. The pulser is clearly much faster than 600 ps, which itself is a little suspect. A 500 MHz scope using the 0.35 / BW[GHz] = tr[ns] formula would suggest 700 ps is the minimum measureable risetime. With single shot real-time mode, the scope reports anywhere from 550 to 650 ps. I'm unable do equivalent-time measurements with averaging since I haven't wound the trigger coil yet. But I think it's safe to say the scope has a little extra breathing room than the 500MHz printed on the case.
I think a little microstrip board with one of those Tektronix reed switches would yield even better results. I wouldn't dare use HV on it though, gold contacts aren't good for that type of abuse. Maybe 20V max (for a 10V output) would be more than enough for most applications.
The nice thing about these type of pulsers is that they are incredibly cheap to make and even random junk parts can achieve impressive results.
-
I wish I had a faster oscilloscope. The pulser is clearly much faster than 600 ps, which itself is a little suspect. A 500 MHz scope using the 0.35 / BW[GHz] = tr[ns] formula would suggest 700 ps is the minimum measureable risetime. With single shot real-time mode, the scope reports anywhere from 550 to 650 ps. I'm unable do equivalent-time measurements with averaging since I haven't wound the trigger coil yet. But I think it's safe to say the scope has a little extra breathing room than the 500MHz printed on the case.
Would there be a possibility to use frequency domain analysis instead ?
The faster the pulse, the highest the peak frequency, right ?
Just thinking out loud
-
I love the nice smooth traces that Tektronix used in their older equipment.
The traces are smooth because the layout was done by hand on transparent film or paper vellum using tape, although when I did analog and RF board layouts in AutoCAD, I also made smooth traces.
But I think it's safe to say the scope has a little extra breathing room than the 500MHz printed on the case.
The 500 MHz bandwidth of your DSO severly limits the measurement accuracy and no amount of fiddling will improve it much. Sometimes there is no substitute for massive bandwidth. This is the kind of application that in the past would use a sampling oscilloscopes which presents difficulties at such low repetition rates but measurement is still possible; see the www.amplifier.cd (http://www.amplifier.cd) link below.
The Tektronix Type 109 pulser used a relay to achieve about 300 picoseconds:
https://www.amplifier.cd/Test_Equipment/Tektronix/Tektronix_other/109.html (https://www.amplifier.cd/Test_Equipment/Tektronix/Tektronix_other/109.html)
https://w140.com/tekwiki/wiki/109 (https://w140.com/tekwiki/wiki/109)
-
(https://i.imgur.com/bvyk0kW.jpg)
I think this the used principle:
(https://w140.com/tekwiki/images/0/06/Charge_line_animation.gif)
Do you trigger a single shot by using a permanent magnet?
You chosed Rs = 1 MOhms for 50 Ohms coax cable and Zo of the DUT (scope)?
Your coax cable lenght is 10 feet, and the part between reed contact and scope input is about 1/3 feet?
For my understanding the waveform at Zo (scope input) should look like this:
(https://www.eevblog.com/forum/testgear/shortest-rise-time-possible-with-a-relay/?action=dlattach;attach=1115106;image)
But it looks like this:
(https://i.imgur.com/bvyk0kW.jpg)
Where is the peak/plateau ratio 2:1?
Am I wrong?
-
I finally wound a trigger coil around the reed switch.
To do this, I cut out two circular bobin ends from some rigid 0.5mm plastic with a slit going to the center hole. Then I bent the plastic so they would fit onto the cable near the brass tube. Next I used some hot glue to hold them in place and slid the start of the winding through one of the bobin slits to hold it in place while winding.
After each layer I put a layer of transformer tape (3M type 74, 0.02mm polyester film) on to keep the windings from loosening. Once the coil looked like it had enough turns I pulled the other end through the bobin slit on the oposite side and used more hot glue to fix them in place. Finally I cut the bobin down to the size of the coil.
Overall, I think it looks decent enough--and to be honest, over kill for such a prototype.
(http://i.imgur.com/WqzULNal.jpg) (https://imgur.com/WqzULNa)
Using a current source, I found that the coil would switch in at about 44 mA and drop out at 34 mA, to guarantee it would always trigger I settled on 50 mA. With a coil resistance of 5 Ω and an inductance of 1.5 mH, the signal generator needs to be set for about 1.4 Vpp with 0.7 Vdc offset (into 50 Ω or double open circuit). One could play with the offset and amplitude to get the reed to just barely move between states, but the above seemed to work fine during testing.
Shown below is operation with a 20 dB attenuator and the charge line at 20 V--giving a 1 V pulse on the scope. The attenuator greatly reduces the effects of reflections and doesn't appear to slow the signal down much, if at all.
The sharp peak on rising edge is actually from the scope's input capacitance, which needs to be adjusted. Tesing each channel of the scope shows little consistency regarding the amplitude of this spike. I need to pick up one of Leo Bodnar's pulsers to recalibrate the scope's rise time adjustment.
(http://i.imgur.com/ppXDFxLl.jpg) (https://imgur.com/ppXDFxL)
Here's some close ups of the pulse with the attenuator in place. I increased the amplitude at the PSU to better fill the display graticules.
(http://i.imgur.com/6K26utyl.jpg) (https://imgur.com/6K26uty)
(http://i.imgur.com/ETuLZmXl.jpg) (https://imgur.com/ETuLZmX)
And now for some tests with the attenuator removed and the scope set to 1 MΩ input termination.
When zoomed out we see the typical exponential decay of a charged capacitor into a resistive load, asympotically falling to 1/2 the applied voltage from the PSU since source and load resistance are equal (1 MΩ). When zooming back in, we start to see transmission line effects due to the missmatch between source, load and line (not contact bounce).
(http://i.imgur.com/7ye6Dvml.jpg) (https://imgur.com/7ye6Dvm)
(http://i.imgur.com/K5OKvInl.jpg) (https://imgur.com/K5OKvIn)
(http://i.imgur.com/t5spcNal.jpg) (https://imgur.com/t5spcNa)
-
Neat! And thanks to you I added more things to my lab shopping list... :-DD
I don't know if it's been mentioned, but Tektronix also used a reed relay in the calibration-step generator in the 519 oscilloscope; this is specified to approximately 100 ps, although contact bounce is an issue and you can't always eliminate extraneous traces because of it (I can verify this is true because my 519 has a working reed).
-
I think this the used principle:
(https://w140.com/tekwiki/images/0/06/Charge_line_animation.gif)
Correct. The accepted model is that two waves travel in opposite directions inside the charge line, the second half reflecting off of the essentially open end of the line used for charging and the other wave immediately terminates into the load resistor. Thus causing the amplitude to appear as 1/2 the charge voltage and the pulse width to be twice the transit time of the cable.
Do you trigger a single shot by using a permanent magnet?
Indeed, but as my previous post shows, I've since wound a trigger coil over the reed switch, making it much simpler to use.
You chosed Rs = 1 MOhms for 50 Ohms coax cable and Zo of the DUT (scope)?
Correct. The scope's internal 50 Ω termination was used to match the RG400 coax, which is also 50 Ω, and the charging resistor was 1 MΩ.
Your coax cable lenght is 10 feet, and the part between reed contact and scope input is about 1/3 feet?
That number was a rough guesstimate, as I didn't actually measure the length of the RG400, but the pulse width of the signal is roughly 24.50 ns.
Using a velocity factor of 69.5%, we have that
l_cable = VF/100 * c * t_pw/2
= 0.695 * 299792458[m/s] * 12.25[ns]
= 2.55 meters
The pulser itself is about 20cm in length, end-to-end. The reed switch is in the center of the pulser. Factoring in half the length of the pulser (10 cm), we have 2.45 meters or 8 feet, which is actually what I just measured!
A charge line pulser gives a simple way to measure cables too long to reliably measure with hand tools.
For my understanding the waveform at Zo (scope input) should look like this:
(https://www.eevblog.com/forum/testgear/shortest-rise-time-possible-with-a-relay/?action=dlattach;attach=1115106;image)
[...]
Where is the peak/plateau ratio 2:1?
That 2:1 spike looks like the RC exponential decay I showed in the previous post. I've never observed that happening with any of the charge line systems I've worked with.
Are you solving this via D'Alembert's method? I remember working this type of problem out in PDEs at university.
-
Are you solving this via D'Alembert's method?
No, I don't even know this method.
I just found the animated schematic above and thought about what waveform comes to the oscilloscope input.