TSP

Рет қаралды 22,619

Күн бұрын

In this episode Shahriar takes a look at the design and performance of the open-source ALK-PT1 passive probe. This probe uses is based on a classical transmission line feed structure and has gone through many iterations to improve its performance. The design tradeoffs of various passive probes are explored and the performance of the ALK-PT1 is measured both in S-parameters domain and time domain using NRZ serial data at various data rates.
You can get your own AKL-PT1 Passive Probe from Antikernel Labs:
www.antikernel.net/products.html
www.TheSignalPath.com
/ thesignalpath
www.Patreon.com/TheSignalPath

Пікірлер: 71

@AndrewZonenberg 3 жыл бұрын

Regarding the peaking: If you look carefully at the PCB around 3:30, you'll see that a short length of the coplanar waveguide just aft of the resistors is fatter than the rest. This is actually a distributed element notch filter tuned to the frequency of the peaking (which I believe is a resonance in the tip itself, since the tip forms a stub between the DUT and the high impedance of the resistors) and intended to produce an overall flat response. I designed v1.3 based on older characterization data that didn't correctly de-embed the loss of my test fixture at higher frequencies; as a result I underestimated the intensity of the peaking and didn't make the filter strong enough. Making the notch a few dB deeper (by widening the trace in the filter area) should flatten it out and is already planned for the v1.4 PCB revision, I just need to do more experiments to figure out the optimum filter response. Regarding the tips: These tips are made by PMK for their Tetris and Sonic active probe line (rebranded by Teledyne LeCroy as the ZSxxx probe family), part number 890-800-000. You can buy them on Digi-Key in a 4-pack for $15 by searching part number PK-ZS-001-ND. They also make a pogo variant, PMK part 890-800-001. I only have one of these and they're quite expensive ($54 for *one* tip) but if you'd like to try, you can get them on Digi-Key by searching PK-ZS-013-ND. A pair of these would more than double the BOM so I didn't include these in the standard accessory kit, but users can buy them separately if they want. I'll try to buy another one or two and do some characterization and see if they alter the response. For updates on development of this probe, and my other test equipment projects, check out my Twitter @azonenberg.

@pentachronic 3 жыл бұрын

One would assume that the step between the pins and the barrel of the pins is causing an impedance change and you are resonating it out with your Distributed Element filter/impedance transformer ?

@pentachronic 3 жыл бұрын

If it were possible to get pins that are thinner and have no barrel, wouldn't that alleviate the impedance change since they would be kept parallel ? SMA is rated up to about 6GHz before it starts dropping off FWIR.

@AndrewZonenberg 3 жыл бұрын

@@pentachronic Yeah, basically. My best simulation to date suggests that the tip is somewhere around 200 ohms impedance but unless I could get the tip up to ~500 ohms there would still be a mismatch when you hit the resistors. Higher is actually better vs being an exact 50, I think. These pins are socketed in order to be replaceable. Solder-on pins exist but most of the ones I've looked at are quite long (bad choice for high freq) or have other issues that make them not a good choice for this application. So far these are the best I've found. And good SMAs are good a lot higher than 6 GHz. The one I'm using from Amphenol RF is designed to transition smoothly to coplanar waveguide and is specced for operation out to 26.5 GHz.

@graybeardmicrowave3074 3 жыл бұрын

@@AndrewZonenberg maybe notch out the PCB between the pins and cover the end with black, anti static foam ensuring it fills the notch. That may de-Q the resonance along the length of the pins. Depending on the via inductance, may also need to bridge the top of the CPWG grounds with a zero ohm resistor to help equalize the ground currents. Unequal ground currents may build up a common mode resonance along the length of the probe.

@AndrewZonenberg 3 жыл бұрын

@@graybeardmicrowave3074 Field solver sim of the probe body matches experimental results of test coupons on the same CPWG geometry pretty well and I don't think there's any significant resonance there. The tip is trickier to model because it's 3D geometry and Sonnet is only 2.5D, so I rely more on experimental results for the tip. The foam is an interesting idea, I may have to play with it. I'm not sure how much is coupling between the sockets (possible to fix) vs the needles themselves (which I can't do much about since they need to remain separated to hit the DUT).

@ghwizz 3 жыл бұрын

Kudos to the engineer releasing their work open source - praise from Shahriar is rare - you should feel proud!

@SK-dv6yc 3 жыл бұрын

I just want to add, contrary to what someone else wrote, that these videos is what makes this channel awesome. Theory, teardowns and explanations of ridiculous high speed circuits and designs is so fascinating,even though i will probably never touch this stuff myself, at work or as a hobby. But thats what makes a video of a probe like this so interesting. No one wants to look at another 3 MHz amateur radio transmitter just because it's something 'everyone' can work with.

@BarriosGroupie 3 жыл бұрын

Interesting idea and good luck to the company selling these. As I'm sure you know, the famous Techtronix probes book mentions transmission line probes where the series resistance is integrated into the transmission line using a resistance wire if my memory serves me right to get the required astonishing bandwidth back in the 1960s. "My Early Tektronix Days By John Kobbe" has more on this.

@BlakeJohnson1 3 жыл бұрын

The crossover episode we’ve all been waiting for!

@artiem5262 3 жыл бұрын

what fantastic work! congratulations!

@k7iq 3 жыл бұрын

Nice job Andrew !

@BrianB14471 3 жыл бұрын

We solder the probe right to the traces when measuring PCIe with an active diff probe.

@AndrewZonenberg 3 жыл бұрын

The AKL-PT2, planned to launch in April, is a solder in passive probe based on this same architecture. Limited to about 4 GHz by losses in the fairly long flex PCB but a shorter version, or one made on a lower loss substrate, would perform better. I also have a 4 GHz differential probe (AKL-AD1 amplifier plus AKL-PD1 tip) under development, as well as some faster stuff, planned but no ETA on those yet.

@navadeep.ganesh 3 жыл бұрын

Great testing! Feels like some magical hooo-haah is happening when eye diagrams comes up and fades away :)

@fataliity101 3 жыл бұрын

Pretty sure those eyes are the wavelength of the data being carried

@rtsp 3 жыл бұрын

Great product.

@LightningHelix101 3 жыл бұрын

This test equipment is crazy good. No product needs the level of performance or flexibility that we expect from this stuff

@miroslawkaras7710 3 жыл бұрын

... Caraze good... E5071 is just average one - " low cost". I usualy tray avoid unsing that one if I have somting serious to measure. PNAs series are world class.

@LutzSchafer 3 жыл бұрын

Nice work. Would be interesting how the better one compares to the commercial high impedance one although it's not a fair comparison. Greetings from Canada :)

@pocoapoco2 3 жыл бұрын

Maybe just having one of the pins sprung and the other rigid would help minimize any negative effects.

@BrendaEM 3 жыл бұрын

Is the discontinuity in the center conductor to somehow compensate for the large vertical tab? Or should that not matter because of the via stitching? Could the center conductor transition be smoother, perhaps drawn with a Bézier, or even a gradual taper? Sadly, the legs on SMA connectors usually used for afixing them to boards has been found to be an issue at high frequencies, something about circular modes a problem. I believe that Eric Bogatin's associates characterized them.

@miroslawkaras7710 3 жыл бұрын

Does limiting the thhickness of the metalization to skin deapht at the highest frequency will flatten the responce? In such case characteristic inductance over frequency might become constant. It might be simpler to cotrol interface capacitance

@d.jeffdionne 3 жыл бұрын

Now that is just fantastic. Where is the order page, and do you have a github for the design? :)

@AndrewZonenberg 3 жыл бұрын

The design files are at github.com/azonenberg/starshipraider/tree/master/boards/probes. Right now it's called "handheld-resistive-probe" which was the early development name, it will be renamed to "akl-pt1" soon as part of some cleanup. They're not for sale until I've made a version 1.4 PCB fixing the 5 GHz peaking. There's no official waiting list and I'm not taking preorders at the moment, but if you email sales at antikernel.net I'll let you know when assembled units are available for purchase.

@berndk1456 3 жыл бұрын

Looks good. Is there a Download for the Gerber-Files?

@AndrewZonenberg 3 жыл бұрын

You can find the KiCAD design files at github.com/azonenberg/starshipraider/tree/master/boards/probes (directory is currently named "handheld-resistive-probe" for legacy reasons, but will be renamed to "akl-pt1" soon). There's no published Gerber release because at least one more PCB revision is pending before I consider the project "done" but if you want to try your hand at building a v1.3 board you can grab a copy of KiCAD and generate gerbers from the CAD design now.

@alifeleparanj3688 3 жыл бұрын

Nice Video

@rfengr00 3 жыл бұрын

Neat probe. You can make a similar one by carefully trimming a piece 0.086” semirigid coax at a 45 degree diagonal and soldering on a 1k 0603 resistor. Of course you don’t have the nice pogo pins to make contact.

@KX36 3 жыл бұрын

that sounds pretty cool, where'd you get that from?

@miroslawkaras7710 3 жыл бұрын

@Kevin A. LOL

@rfengr00 3 жыл бұрын

@@KX36 nowhere, just made one up. Can also snip tungsten needle probes in half and install a SMT resistor with silver epoxy. Makes a high impedance DC probe for probing RF thin film circuits and ICs.

@andymouse 3 жыл бұрын

Love your channel, don't understand any of it ! (maybe some bits) by 'landing' your talking about probe contact no ? what does a bad landing look like ? do you take the probe off and do it again ? how do you know a landing is bad and not just a rubbish piece of kit your testing ?...cheers.

@jaro6985 3 жыл бұрын

Yes, landing probe poking onto the copper on the PCB. Bad is when he had pressed it down but either ground or signal is not making contact, so he'd see nothing useful on the oscilloscope.

@gustavlicht9620 3 жыл бұрын

Does anone know good open source EM simulation software?

@vonnikon 3 жыл бұрын

I would like to see a passive differential probe. Based on a transformer. Yes, it would be limited to only DC balanced differential signals. But the cost would be so much lower than active differential probes. Plus, it would be isolated! And by using something like a 10:1 transformer, the loading could be made really low while still providing a good strong 50ohm signal to the oscilloscope.

@AndrewZonenberg 3 жыл бұрын

Interesting idea, and I've thought about it. I'm actually working on an open hardware 4 GHz active differential probe as well. Right now it looks like around $500 for an assembled unit including profit and assembly costs is viable; the BOM cost for a DIY assembled version would be in the $150-200 range.

@tfinmoraes 3 жыл бұрын

@@AndrewZonenberg Great! Please let us know about any follow up on this, it is surely a valuable contribution to the open source community! And thank you @Shahriar for bringing his work to the audience!

@VE2XJS 3 жыл бұрын

could you calibrate for the probe?

@AndrewZonenberg 3 жыл бұрын

The pro edition will include individually serialized 2-port S-parameters for de-embedding. You can't do this on a sampling scope like Shahriar is using in this video because it relies on being able to see changes in the waveform over time to compensate for frequency dispersion etc, but with a sufficiently fast realtime scope it can significantly improve results.

@friedmule5403 3 жыл бұрын

Shahriar do we have some sort of misunderstanding? :-) A person did write the word "boring" as a comment to this video, and I asked him about what was boring. He replied that it was because it was about HF that not many cared about. I replied that he then would hate another channel about HF and PCB design. I may suspect that you perhaps are thinking it was me who wrote "boring", while it instead was me who asked why the commenter did write that. :-) If I for some reason have done anything to offend you, please understand that it in no way was my purpose and that I am sincerely and deeply apologizing!!

@Thesignalpath 3 жыл бұрын

Not at all. I appreciate that you wrote me. There are tens of thousands of people who come across these videos. Some are surely going to find the topic boring. It is impossible to please everyone. Just leaving a comment as “boring” is not constructive of course, since it conveys no useful information. Thanks again for jumping in. :)

@anatolyzapadinsky5995 3 жыл бұрын

What setup you need to probe a USB2 line? USB is bidirectional. Is it even possible to analyse a bidirectional lines like that using a sampling oscilloscope?

@AndrewZonenberg 3 жыл бұрын

USB2 is bidirectional, but the the TX and RX sides take turns sending. It's not like 1000base-T where both ends send simultaneously using echo cancellation. You could probably use an edge trigger on a sampling scope to get an eye pattern off it, but USB2 is slow enough that a halfway decent realtime scope can analyze it too. As a bonus, with a realtime scope you can do protocol decoding, and maybe even separate TX/RX eyes (possible in principle although I'm not sure if any software does this). The AKL-PT1 would be a poor choice for probing USB2 because of the 500 ohm loading; USB2 relies on pullup/pulldown resistors to detect device speed etc and also is DC coupled with fairly large swings. USB3 is AC coupled high speed serial with a smaller swing, so it would probably work decently for that.

@anatolyzapadinsky5995 3 жыл бұрын

@@AndrewZonenberg USB devices do not expose its internal state and it is hard to say when it is sending or receiving. It is just sending packets and retransmitting them until the checksum test is passed. In a DIY projects you just have two devices soldered to a board or connected through a custom cable. You may have an access to its clock generator but this is all. How can you troubleshoot this USB line and test its signal integrity? A realtime scope that is able to show an eye diagram of an old 480MHZ USB2 interface would already cost around $10k. This is the other story though. Hobbyists are struggle to test their rather old digital interfaces and we just desperately need a cheap 5Ghz sampling scope just to test the plain old USB or HDMI signals integrity. The industry on the other hand only produce new sampling scopes with a much much higher bandwidth than the current realtime scopes. So the only option you have is to buy an old used sampling scope. Because the new realtime 5Ghz scope would cost you $10k and the new sampling scopes bandwidths starts from 50Ghz and they are even more expensive.

@AndrewZonenberg 3 жыл бұрын

@@anatolyzapadinsky5995 I have a partially abandoned design for a 10 GHz sampling scope that I may revive at some point. I started work on it a while back when I knew a lot less about signal integrity than I do now, so a fair bit of it needs to be redone. It would have been somewhere in the $1K range probably.

@anatolyzapadinsky5995 3 жыл бұрын

@@AndrewZonenberg just curious, did you also base your scope design on fast comparators?

@AndrewZonenberg 3 жыл бұрын

@@anatolyzapadinsky5995 FREESAMPLE? Yeah it's based on a latching comparator, PLL+delay line, and slow DAC. But I also had a hardware CDR and some other nice features.

@TechMechanism 3 жыл бұрын

wow.

@stevec5000 3 жыл бұрын

Looks like vaporware! At antikernel all the probes say "coming soon" and their distributor Elegant Invention seems to be closed!

@stevec5000 3 жыл бұрын

Now they have available a Passive Solder-In Probe which is just an expensive piece of transmission line, the others are still vaporware!

@andrewnambudripad762 3 жыл бұрын

I'm in software and took basically only philosophy classes, so forgive me if none of this makes any sense...I learnt most of what I know (whcih, again, is very little) from the ARRL ham radio books....Can you test that out with pre-emphasis? (sorry I don't know the Agilent layout so I have no idea if you've already enabled it)? What was he using to model that out, ADS?ANSYS? And/or a semi-rigid connections and/or on a x/y stage ? It didnt' look like you SOLT calibrated, on a 3 port system can you just skip that part (like does the probe have the cal co-efficients on the afe?) You can get some cheap gains moving off of FR4, 100% sure of that. Also maybe experiment with the copper alloys/weights? And is the board house controlling the oxide built up on the other layers? Probably not, but maybe he can find a board house that'll be flexible haha. www.fr4-pcb.com/provide-99166-rogers-pcb These guys seem to offer a pretty flexible assortment of Rogers materials. Like it looks like you can end up requesting a few hundred different types of board permutations. "Bubble bag" is offered, so maybe they even offer production in nitrogen or some inert environment... Placement of..everything is important, I'd imagine. And the oils on your hand is probably not too good for something like this. Some dude was taking coax stock and building a triax setup, he was wearing gloves and all, and everyone says not to touch the pins on SMA connectors, so Ithat might be messing with the linearity too? AD has just a crazy number of microwave and mmwave stuff in different 3-5 processes since they got all the Hittite stuff, www.digikey.com/en/product-highlight/a/analog-devices/hmc8410-hmc8401-wideband-lna Just pick that eval board up and see how it fares. 650$ is a steal if it's really .1 to 10 ghz...

@PHANAMANA 3 жыл бұрын

The WD40 paint job is surely worth extra points.

@pentachronic 3 жыл бұрын

I believe Doug Smith may have a better solution. You put the resistors at the end of the coax and make a purely resistive probe launching straight into the coax. This way you don't get parasitics and surface issues from the PCB microstrip. emcesd.com/1ghzprob.htm It would be interesting to see these 2 designs side by side. I'm sure Doug Smith would love to see the results too. Nice idea and kudos for sharing.

@AndrewZonenberg 3 жыл бұрын

His probe needs extra compensating capacitance to flatten out the response, which increases loading. It's only a 1 GHz design and I don't see any S11 or input impedance curves anywhere on that page. It was one of several existing designs I looked at, as well as www.sigcon.com/Pubs/straight/probes.htm, to draw inspiration from. I ended up going with flip chip microwave resistors from the Vishay FC0402 series to get lowest parasitics. My current design uses a series string of 200+200+50 as the best tradeoff between the various issues (reflections between the resistive elements, high frequency peaking from shunt capacitance, available component values, etc).

@pentachronic 3 жыл бұрын

@@AndrewZonenberg Very cool. A fun project for sure. Thanks for sharing.

@SerenityMae11 3 жыл бұрын

First

@cogito451 3 жыл бұрын

I too have nothing to add.

@lasersbee 3 жыл бұрын

I can't stand sellers not showing prices until you jump through their Data Collection hoops.

@AndrewZonenberg 3 жыл бұрын

This was a pre-release beta unit and it's not available for sale yet. There's no "data collection hoops" to see prices. It will be available for immediate purchase, with published prices via my distribution partner, once I've addressed the peaking issues. No customer information is collected except what's needed for shipping and billing, and that's not retained after fulfillment is complete. Expected retail price (not including tax/shipping) will be: * $250 for standard edition (probe body and basic grounding accessories only), discounted to $150 to students / noncommercial users upon special request * $550 for pro edition. This will include an SMA cable, SMA-BNC adapter, and full characterization report including machine readable S-parameters for de-embedding probe response from your measurements. Based on this review, I may also include a set of pogo tips you can use instead of the rigid ones.

@lasersbee 3 жыл бұрын

@@AndrewZonenberg Thanks for taking the time to respond with some prices and explanations... When I'm required to register my identity information on a site to get info (pricing/etc.)... I call that "data collection hoops" and I will not continue on that site. nice new product.