Hi guys,
I'd like to qualify and tune my decoupling networks for low impedance over a wide range of frequencies.
Is this something which is easily done with a Vector Network Analyzer? I don't have a budget for a very advanced tool. What I do have is a spectrum analyzer with a tracking output. It goes to 40MHz which will at least give me an initial picture of what's going on.
My first idea was to see how the tracking output is divided down by a reference resistor and the analyzer input, and then repeat without the reference resistor but with a coax soldered down on the board where the decoupling is present but the load is not. This should give me DUT impedance / reference impedance for a range of frequencies.
This will probably not be very accurate, since both reference resistor and DUT resistor are significantly lower than the 50Ohm impedance of the analyzer.
Is this where a Reflection Bridge VSWR box could come into play?
Alternatively, I've seen this: VNA 1MHz to 3GHz at $380. No idea if this could do the trick for low impedances, though. VNA 1M 3GHz Vector Network Analyzer miniVNA Tiny+ VHF/UHF/NFC/RFID RF Antenna Analyzer Signal Generator SWR/S Parameter/Smith on Aliexpress.com | Alibaba Group
Thanks,
Børge
I'd like to qualify and tune my decoupling networks for low impedance over a wide range of frequencies.
Is this something which is easily done with a Vector Network Analyzer? I don't have a budget for a very advanced tool. What I do have is a spectrum analyzer with a tracking output. It goes to 40MHz which will at least give me an initial picture of what's going on.
My first idea was to see how the tracking output is divided down by a reference resistor and the analyzer input, and then repeat without the reference resistor but with a coax soldered down on the board where the decoupling is present but the load is not. This should give me DUT impedance / reference impedance for a range of frequencies.
This will probably not be very accurate, since both reference resistor and DUT resistor are significantly lower than the 50Ohm impedance of the analyzer.
Is this where a Reflection Bridge VSWR box could come into play?
Alternatively, I've seen this: VNA 1MHz to 3GHz at $380. No idea if this could do the trick for low impedances, though. VNA 1M 3GHz Vector Network Analyzer miniVNA Tiny+ VHF/UHF/NFC/RFID RF Antenna Analyzer Signal Generator SWR/S Parameter/Smith on Aliexpress.com | Alibaba Group
Thanks,
Børge
A VNA looks at both level and phase. The return bridge samples the reflection which is a function of impedance. You can get an idea of the impedance magnitude from the bridge but its resistive/reactive aspects need the phase info. A spectrum analyzer with a tracking generator is a start but it has only one receiver unlike the 2-4 on a VNA so it can't see the phase differences.
Maybe you can better describe the decoupling networks so we have a better understanding of what you are looking for. There may be an easier way to confirm what you want to know.
Maybe you can better describe the decoupling networks so we have a better understanding of what you are looking for. There may be an easier way to confirm what you want to know.
Hi Demain - long time no see!
It's basically about designing ordinary capacitive decoupling networks at points of load. My idea is to populate the decoupling (and possibly a turned-on LDO) but not the load IC itself.
Where the IC sits I will then put in an SMA plug and attach a coax which leads to the instrument. The instrument will compare the impedance of the decoupling network to that of a reference resistor, say 0.1Ohm.
All capacitors become inductive after some self-resonant frequency. My goal is to be able to design networks of capacitors where there is a consistently low and flat impedance curve as the frequency increases. Not just at audio frequencies, but at the switching frequencies (and their harmonics) of delta-sigma converters.
My HP 3585A (with tracking gen and possibly 1MOhm input) goes to 40Mhz. I also have a 8590A which goes to 1.5GHz with 50Ohm input. I did a quick check, and it doesn't seem to have a meaningful output.
Seeing only magnitude shouldn't be such a big deal, as I guess it is obvious what is the inductive and what is the capacitive side of each impedance peak. But for the finer tunign, seeing the phase would certainly help. And the $380 VNA going to 3GHz is much further than my analyzers go anyway.
Børge
It's basically about designing ordinary capacitive decoupling networks at points of load. My idea is to populate the decoupling (and possibly a turned-on LDO) but not the load IC itself.
Where the IC sits I will then put in an SMA plug and attach a coax which leads to the instrument. The instrument will compare the impedance of the decoupling network to that of a reference resistor, say 0.1Ohm.
All capacitors become inductive after some self-resonant frequency. My goal is to be able to design networks of capacitors where there is a consistently low and flat impedance curve as the frequency increases. Not just at audio frequencies, but at the switching frequencies (and their harmonics) of delta-sigma converters.
My HP 3585A (with tracking gen and possibly 1MOhm input) goes to 40Mhz. I also have a 8590A which goes to 1.5GHz with 50Ohm input. I did a quick check, and it doesn't seem to have a meaningful output.
Seeing only magnitude shouldn't be such a big deal, as I guess it is obvious what is the inductive and what is the capacitive side of each impedance peak. But for the finer tunign, seeing the phase would certainly help. And the $380 VNA going to 3GHz is much further than my analyzers go anyway.
Børge
Borge-
1) do you really need to see what happens at 2 GHz?
2) I think you will learn more by checking the through path from power in to the chip. Add a .1 uF DC block + the SMA connection at each end where the power originates and at the chips. You can then see what the losses are going each way as well as see the noise in real time. The VNA will require a tedious cal to de-imbed the cable and connectors before you can get any real indication of impedance and then you need to figure what you can do with the info. Measuring the loss each way will give a good indication of how effective the noise isolation is. The spectrum will also help you see what needs to be addressed. A deep insertion loss where there was no noise won't accomplish much.
1) do you really need to see what happens at 2 GHz?
2) I think you will learn more by checking the through path from power in to the chip. Add a .1 uF DC block + the SMA connection at each end where the power originates and at the chips. You can then see what the losses are going each way as well as see the noise in real time. The VNA will require a tedious cal to de-imbed the cable and connectors before you can get any real indication of impedance and then you need to figure what you can do with the info. Measuring the loss each way will give a good indication of how effective the noise isolation is. The spectrum will also help you see what needs to be addressed. A deep insertion loss where there was no noise won't accomplish much.
Borges, I'm using exactly your scenario very often with my trusty W&G SNA-2 Network Analyzer (180MHz upper limit, perfect for the task).
Measure a known good low-impedance reference resistor (0.1Ohm), preferably in 2-port shunt-through mode (that is, two cables to the DUT, not only one), adjust the readout baseline to that so you'll have dBOhms display, measure DUT impedance and see what's going on there.
https://www.edn.com/Pdf/ViewPdf?contentItemId=4458562
https://www.emctest.it/public/pages...Parameters Using Vector Network Analyzers.pdf
Protecting the input from DC *and* initial coupling capacitor glitch is the most important thing here when you plan to measure a live supply (not necessarily needed, most of the impedance effects you'll see readily in unpowered state).
For analyzing decoupling quality absolute precision and de-embedding parasitics etc isn't needed, IHMO.
Measure a known good low-impedance reference resistor (0.1Ohm), preferably in 2-port shunt-through mode (that is, two cables to the DUT, not only one), adjust the readout baseline to that so you'll have dBOhms display, measure DUT impedance and see what's going on there.
https://www.edn.com/Pdf/ViewPdf?contentItemId=4458562
https://www.emctest.it/public/pages...Parameters Using Vector Network Analyzers.pdf
Protecting the input from DC *and* initial coupling capacitor glitch is the most important thing here when you plan to measure a live supply (not necessarily needed, most of the impedance effects you'll see readily in unpowered state).
For analyzing decoupling quality absolute precision and de-embedding parasitics etc isn't needed, IHMO.
SNR limits the utility of 1-port measurements to Z>0.1R.
Keysight application note on ultra-low impedance measurement. http://literature.cdn.keysight.com/litweb/pdf/5989-5935EN.pdf
btw -- did you know that the 8590 can be programmed via the Hewlett Packard Interface Loop (HP-IL)?
Keysight application note on ultra-low impedance measurement. http://literature.cdn.keysight.com/litweb/pdf/5989-5935EN.pdf
btw -- did you know that the 8590 can be programmed via the Hewlett Packard Interface Loop (HP-IL)?
& KSTR, thanks! Interesting stuff.
I once have made this writeup: http://www.hoffmann-hochfrequenz.de/downloads/experiments_with_decoupling_capacitors.pdf
also with a W&G TSA-2, and also a 2port measurement. My TSA-2 is ill, it has noisy traces for narrow bandwidths and does no more calibrate for scalar network analysis.
I will have to repair it some day. The ZVB8 starts at 300 KHz, the 89441A cannot do log sweeps, the most ubiquitous hing that I have is DG8SAQ VNWA that I bought when it still was available as a kit. In the meantime, the price has tripled :-(
cheers, Gerhard
Last edited:
Thank you all for the hints and litterature links! I will read up a bit before asking more questions.
Cheers,
Børge
Cheers,
Børge
Based on the texts you've shown me here, a likely way to go initially seems to use the tracking generator as output and RF input as input. Measure on reference resistor and de-embed at will. Connect in shunt-through topology with a couple SMA plugs on the DUT.
In your opinion, would a Reflection Bridge VSWR make sense?
I have an instrument which lets me do that up to 40MHz, which is a good start. For my 1.5GHz spectrum analyzer there is no tracking output. In that case I'll need a new instrument.
Thanks,
Børge
In your opinion, would a Reflection Bridge VSWR make sense?
I have an instrument which lets me do that up to 40MHz, which is a good start. For my 1.5GHz spectrum analyzer there is no tracking output. In that case I'll need a new instrument.
Thanks,
Børge
- Status
- Not open for further replies.