I’ve been stepping up my low noise game, took a deep dive into decoupling (simulations), and am feeling a lot more confident about the performance of the boards I design. The only problem - I don’t have anything to verify this / measure PDN impedances above 5MHz or below 200mOhm.
The pros seem to use €30k VNAs and with €5k ground loop isolators to run S21 shunt through measurements. Is there a more affordable alternative out there? I’d be happy to get up to 200MHz / down to 5mOhm.
Would the cheaper VNAs suffice for this? Or do they not have the performance for it? Do I really need the ground loop isolators? Could I spin my own with a GHz differential amplifier?
I feel this is slightly out general hobby territory (but also feel people have to stop putting 100nFs everywhere like it’s the 80s, route those with way to long / thin traces compared to the capacitors ESL, and call it a day)
I’d love some advice - thanks!
I would say this is the realm of devices like the
Bode 500.
While not 30k€, it' still "comfortably" into the 5 digits. It's more specialized to the task than a general purpose VNA, though, on the hard and software side.
A LiteVNA64 is a lot cheaper, can be run from battery and joeqsmiths "
" software might enable the kind of low impedance measurement you'd need ("shunt-through measurement") - I'm not sure of the latter, though. If you're comfortable with software you should be able to implement the needed post processing yourself (on a quick glance "shunt-through" is based on S21 so a LiteVNA should get you there)
There are various low cost network analysers. Problem is though that many don't work well below -say- 20MHz while PDN analysis is most interesting int the range from several kHz to 200MHz.
That software looks like it makes those cheap VNAs into quite powerful tools!
I just bought a NanoVNA-H4, with a dynamic range of 70dB from 50kHz-300MHz, specified with a minimum frequency of 10khz. Cheap enough to play around with as a start.
Now I use my scope with its generator, a voltage + current channel, and a math channel, but it doesn't have the resolution / current capacity to measure low impedances, and doesn't have the bandwidth for the high frequencies. I might be able to wing a combination of the two to still measure DC to 50kHz it the impedance is high enough to measure on the scope.
Thanks for the advice! I'll post an update on how it goes
Do you already have a modern oscilloscope? A signal generator and some amplification will get you into milliohms.
I’ve been stepping up my low noise game, took a deep dive into decoupling (simulations), and am feeling a lot more confident about the performance of the boards I design. The only problem - I don’t have anything to verify this / measure PDN impedances above 5MHz or below 200mOhm.
...
Would the cheaper VNAs suffice for this? Or do they not have the performance for it? Do I really need the ground loop isolators? Could I spin my own with a GHz differential amplifier?
....
I’d love some advice - thanks!
Showing the $50 VNA making PDN measurements and explaining the limitations. You can find some of these details in the software manual as well.
***
Looking back at my testing, looks like 1mohm was about the limit with the original NanoVNA. The LiteVNA's I have have worse performance at these lower ranges. They recently released a newer version now of the Lite hardware, but I still doubt it would out perform the original NanoVNA.
I just bought a NanoVNA-H4
It seems that the one I procured did have slightly better noise that my NanoVNA. The firmware wasn't stable enough to use it and it sat on the shelf for about a year before I found a version that would pass my regression tests. I don't believe I tried to measure the standards I made for the PDN experiments. I also stopped developing software for the older protocol that the H4 and original NanoVNA supported. If you wanted to try it, you would need to use that last version I released, or roll your own.
Do you already have a modern oscilloscope? A signal generator and some amplification will get you into milliohms.
Our messages just crossed! As mentioned above I've indeed been using a scope for the purpose. How would you go about implementing 'some amplification'? Purchase a fancy premade preamp? Breadboarding something with a random op amp? Something in between?
Look up DG8SAQ VNWA 3.
That seems to have the perfect range for PDN applications, affordable too. Do you have direct experience with them? How's the software / user interface side of things?
I just bought a NanoVNA-H4
It seems that the one I procured did have slightly better noise that my NanoVNA. The firmware wasn't stable enough to use it and it sat on the shelf for about a year before I found a version that would pass my regression tests. I don't believe I tried to measure the standards I made for the PDN experiments. I also stopped developing software for the older protocol that the H4 and original NanoVNA supported. If you wanted to try it, you would need to use that last version I released, or roll your own.
Thanks for the video link (and the software in general!) I followed along with the NanoVNA-H4 on your older software. I'm not entirely sure what, but something was quite buggy, requiring lots of restarts of both the device and the software to keep things running. While it's running it does work like a charm.
Attached is a sweep of two 100nF and 10uF ceramic capacitors together to make a little resonant peak, on both the NanoVNA-H4 and an Analog Discovery 2 (which doesn't do high frequencies at all). Despite the AD2 circuit getting inductive from 200kHz onwards they agree on the sample PDN having a ~1.1Ω impedance at 10kHz and ~0.11Ω at 100kHz - bingo! That's quite usable already.
What kind of probes do people typically use to probe PDNs with VNAs? Some kind of T-junction cut open coax?
Thanks for the video link (and the software in general!) I followed along with the NanoVNA-H4 on your older software. I'm not entirely sure what, but something was quite buggy, requiring lots of restarts of both the device and the software to keep things running. While it's running it does work like a charm.
Attached is a sweep of two 100nF and 10uF ceramic capacitors together to make a little resonant peak, on both the NanoVNA-H4 and an Analog Discovery 2 (which doesn't do high frequencies at all). Despite the AD2 circuit getting inductive from 200kHz onwards they agree on the sample PDN having a ~1.1Ω impedance at 10kHz and ~0.11Ω at 100kHz - bingo! That's quite usable already.
What kind of probes do people typically use to probe PDNs with VNAs? Some kind of T-junction cut open coax?
After wasting so much time trying to find stable firmware, mine now just sits on shelf. Good to see you at least got it to somewhat work. It's really too bad that the LiteVNA and/or V2Plus4 didn't perform as well as the original NanoVNA in all aspects. IMO, that was a big miss. Then again, sells more product.
Probe wise, the video and manual show my use of coax, normally soldered directly to the PCB. If you are looking for more information, there is a lot of free videos on this subject. Search YT. If you are looking for a book, try "RIGHT THE FIRST TIME, A PRACTICAL HANDBOOK ON HIGH SPEED PCB AND SYSTEM DESIGN"
As I described in that video and manual, the problem is breaking the ground loop. Assuming you are doing this for a company, I recommend just buying a commercial transformer. If you want to roll your own, I'm sure the data Brian collected for me is still available and the materials I used are documented in the manual/video.
Look up DG8SAQ VNWA 3.
That seems to have the perfect range for PDN applications, affordable too. Do you have direct experience with them? How's the software / user interface side of things?
I have it. Software is very powerful but acquired taste and needs some time to learn.
There is great community support and it has decent dynamic range in it's limited BW.
Software and manual can be downloaded. Take a peek.
Look up DG8SAQ VNWA 3.
That seems to have the perfect range for PDN applications, affordable too. Do you have direct experience with them? How's the software / user interface side of things?
I have it. Software is very powerful but acquired taste and needs some time to learn.
There is great community support and it has decent dynamic range in it's limited BW.
Software and manual can be downloaded. Take a peek.
I’d be happy to get up to 200MHz / down to 5mOhm
Good enough to measure the 0.005 ohms? The original NanoVNA, with homemade transformer and blocks could measure down that low but I was really pushing its limits.
Do they support modes specific for PDN?
Do you already have a modern oscilloscope? A signal generator and some amplification will get you into milliohms.
Our messages just crossed! As mentioned above I've indeed been using a scope for the purpose. How would you go about implementing 'some amplification'? Purchase a fancy premade preamp? Breadboarding something with a random op amp? Something in between?
No problem.
Now I use my scope with its generator, a voltage + current channel, and a math channel, but it doesn't have the resolution / current capacity to measure low impedances, and doesn't have the bandwidth for the high frequencies. I might be able to wing a combination of the two to still measure DC to 50kHz it the impedance is high enough to measure on the scope.
Amplification can be on either (or both) sides of the DUT depends which is easier/cheaper for you or what you already have. You might only need a higher output signal generator or a simple LNA depending on how accurate/fast/noisy you need the measurements to be.
Some other things which can help you:
Use a transformer from the output of the signal generator, it breaks the ground loops and can be selected/wound with an appropriate turns ratio to better match the impedances (50ohm generator vs 0.1ohm PDN).
Measure the current being fed into the PDN either with a floating/isolated shunt or a clip on current probe.
@nctnico I've seen your DC to daylight differential amplifier on here - would it be possible to somehow spin that circuit to have a 50 ohm input impedance, and make the ground loop breaker that way?
@joeqsmith I've seen the coaxes on the breadboard - how'd you do it for an actual circuit board? Solder two together, run the through calibration, and then solder both to a pin pair you'd like to measure? How'd you do the other calibrations?
@Someone amplification on the current side could be done with something like this?
https://www.ti.com/lit/ds/symlink/buf634a.pdfDo you have any suggestions for parts on the other side?
@joeqsmith I've seen the coaxes on the breadboard - how'd you do it for an actual circuit board? Solder two together, run the through calibration, and then solder both to a pin pair you'd like to measure? How'd you do the other calibrations?
The trick I've learned is to have a box of (semi-rigid) coaxes, terminated to e.g. SMA on one side, cut to precisely the same length on the other.
Then you can e.g. solder two together as a cal rig, solder the other two to the circuit, have a few spare for other stuff - but still share the cal rig...
@nctnico I've seen your DC to daylight differential amplifier on here - would it be possible to somehow spin that circuit to have a 50 ohm input impedance, and make the ground loop breaker that way?
For PDN measurement purposes, the 26dB dampening of my DIP1400 probe is not what you want. Using a transformer is a better option.
@nctnico I've seen your DC to daylight differential amplifier on here - would it be possible to somehow spin that circuit to have a 50 ohm input impedance, and make the ground loop breaker that way?
For PDN measurement purposes, the 26dB dampening of my DIP1400 probe is not what you want. Using a transformer is a better option.
The LMH3401 by itself has a positive gain right? Wouldn't an alternative input resistor scheme allow for 50 ohm input impedance while still offering some gain at the end?
@nctnico I've seen your DC to daylight differential amplifier on here - would it be possible to somehow spin that circuit to have a 50 ohm input impedance, and make the ground loop breaker that way?
For PDN measurement purposes, the 26dB dampening of my DIP1400 probe is not what you want. Using a transformer is a better option.
The LMH3401 by itself has a positive gain right? Wouldn't an alternative input resistor scheme allow for 50 ohm input impedance while still offering some gain at the end?
If the resistors are changed, then you could turn it into a buffer or even a pre-amplifier for sure but the bandwidth will be much lower ofcourse. I'm not sure what the effect of the added noise will be.
If the resistors are changed, then you could turn it into a buffer or even a pre-amplifier for sure but the bandwidth will be much lower ofcourse.
Enough for my desired 200MHz range? Do you happen to have a simulation I could play with?
If you use the LMH3401 model, you should be able to get a decent idea of how it would perform. I don't expect surprises at 200MHz.
@joeqsmith I've seen the coaxes on the breadboard - how'd you do it for an actual circuit board? Solder two together, run the through calibration, and then solder both to a pin pair you'd like to measure? How'd you do the other calibrations?
The trick I've learned is to have a box of (semi-rigid) coaxes, terminated to e.g. SMA on one side, cut to precisely the same length on the other.
Then you can e.g. solder two together as a cal rig, solder the other two to the circuit, have a few spare for other stuff - but still share the cal rig...
Add a cal standard to the board, maybe on a rail
Solder them together and then to the board
Insert a thru adapter and call it good
In the case the breadboard, left the board unpopulated and called that good
I am not sure what "other calibrations" would be.
Solving the ground loop could certainly be done without a transformer using a diff amp. You could easily out perform my transformers at lower frequencies going with solid state. However, the one transformer I use has 25db@10k and 37db@100k. The 3dB point is around 100Hz. Consider that these low cost VNAs can't run down this low, it may be good enough for your needs. Yes, I know they have all sorts of lower limits. I think I had seen one allowing 1Khz operation. That doesn't suggest you can measure anything useful that low.
Another thing with the transformer, it's very robust. You may want to consider investing in some very low capacitance TVSs and attach them to the VNA.
***
Another option would be to purchase the parts you need. Maybe have a look at what Pico offers.
https://www.picotest.com/solutions/pdn-measurement-forpower-integrity/
I've been playing with- and modifying my Analog Discovery 2 today, messing with the input attenuation and gain circuitry of the scope inputs to give me in total 26dB extra gain, giving me a great improvement in usable impedance measurement range. Here's a stack of five 10uF 0805 capacitors without any averaging:
Problem however with this new milliohm range extension is... inductance. Even at 1MHz a little 10mm back-and-forth of wire has an AC impedance of around 200mΩ, severely limiting my usable frequency range now I can measure down to the milliohms. For the image above I soldered the capacitors on the AD2 PCB itself. Can't really solder an entire PDN on there...
So I believe it's back to the NanoVNA-H4!
No. Your cabling is 50 Ohm so there is no wire inductance. What you can do is add some attenuation (like 6dB) at where you inject and measure the signal to improve impedance matching a bit but it should affect the measurements much. See the source's output resistor as a current source. When calculating the impedance of the board, the current from the source is used together with the measured voltage level to calculate the impedance (R = V / I).
@nctnico you're right! Turns out I was accidentally measuring the inductance / impedance of a shared ground path. I messed up the wiring of one of the differential inputs. Measurements are starting to make a lot more sense now I've got that sorted!
Here's the impedance plot of a 3.3V LDO, both when on and off:
I'm getting 60mΩ at 1MHz there - 10nH - I'd say that makes sense based on how I wired it.
Here's one soldered directly over an 0805, that also has 200uF of bulk capacitance on the same net:
Giving 30mΩ at 10MHz - 500pH - sounds quite appropriate for an 0805.
Glad I got that sorted, thanks for the push nctnico. I can't really use it as a scope anymore with all the modifications, but am pretty happy with the little device it has become.
I'm still getting back to the NanoVNA-H4, to dive more into the higher frequencies, but as an addition instead of a replacement now.
I ended up designing a little adapter board for the Analog Discovery series!
Here's the board measuring a low impedance capacitor array:
Here's a bunch of verification resistors (with varying inductances):
I realized I didn't need to measure higher than ~25MHz - once I see the inductance slope at the end I'll known enough about what happens afterwards. The board has a 200mA output current buffer, and two fully differential opamps. This gives me both more current and enough resolution for all the low stuff. The bottom limit is about 100uΩ.
The flat tinned bars are spaceholders for tantalum capacitors, to be able to measure with a DC Bias as well. Works great so far.
Thanks for all the tips along the way!
I ended up designing a little adapter board for the Analog Discovery series!
Here's the board measuring a low impedance capacitor array:
Hi,
Could you please share the schematic?