High frequency, low current measurement
Alex Yuan, Fri Mar 17 2017, 08:12PMTrying to help a friend figure out how to measure the supply current of an SRAM chip. He'd like to see how much current the chip draws while writing a bit, which takes ~10nS.
Is a small resistor and a high bandwidth op amp the way to go?
Re: High frequency, low current measurement
Nik, Fri Mar 17 2017, 10:42PM
Could you supply it from a source with a nice big capacitor and measure the current before that cap? It would take out all the ripple and leave you with the average current draw.
Nik, Fri Mar 17 2017, 10:42PM
Could you supply it from a source with a nice big capacitor and measure the current before that cap? It would take out all the ripple and leave you with the average current draw.
Re: High frequency, low current measurement
Alex Yuan, Sat Mar 18 2017, 07:51PM
The problem is, he wants to see the transient current as well as the average current.
Alex Yuan, Sat Mar 18 2017, 07:51PM
The problem is, he wants to see the transient current as well as the average current.
Re: High frequency, low current measurement
Mattski, Mon Mar 20 2017, 09:39PM
Oscilloscope makers have grappled with this exact issue. Keysight has a new line of products they're calling "Device Current Waveform Analyzers" specifically for measuring small currents at high speeds.
Otherwise you'll need to go the more traditional route of a series resistor. There are some challenges to measure very small curents very quickly. In general you want a larger resistor the smaller your current is to increase the voltage signal, and you can also get away with that because the voltage drop to the chip will be low if the current is low. The preamp will need to have a low enough offset voltage relative to the signal you are measuring, suitable common mode range, and a small enough input capacitance to rise fast enough with the value of resistor you're using.
Mattski, Mon Mar 20 2017, 09:39PM
Oscilloscope makers have grappled with this exact issue. Keysight has a new line of products they're calling "Device Current Waveform Analyzers" specifically for measuring small currents at high speeds.
Otherwise you'll need to go the more traditional route of a series resistor. There are some challenges to measure very small curents very quickly. In general you want a larger resistor the smaller your current is to increase the voltage signal, and you can also get away with that because the voltage drop to the chip will be low if the current is low. The preamp will need to have a low enough offset voltage relative to the signal you are measuring, suitable common mode range, and a small enough input capacitance to rise fast enough with the value of resistor you're using.
Re: High frequency, low current measurement
Dr. Slack, Tue Mar 21 2017, 06:22AM
Two approaches spring to mind.
One is to use a series resistor and a high speed voltage gain amplifier. The other is to use a transimpedance amplifier. The difference between these is whether there is a significant or insignificant change in rail voltage respectively.
The choice starts by understanding whether the SRAM can tolerate a rail bounce and still work properly, which means not only store valid data, but also draw the current it normally draws. Although (say) 50mV should be plenty to read on a scoope and is well within its static rail tolerance, are there dynamic bias things going on inside an SRAM that mean that rail just has to be nailed still when it wants to draw write current? Those SRAM chips require rail decoupling for a reason.
One way to check this would be to measure the current with increasing values of resistor. Does the measurement show a systematic trend. At what resistance does the SRAM fail to write properly?
There are lots of app notes for the use of transimpedance amplifiers. I put 'high speed photodiode amplifier' into the usual search engine, and on the first page were pdfs for 1 1Mbit/s amp, Texas OPA380, and a 10Gbit/s amp MAX3970. OK, the second one is only available in dice, but you get the general conclusion that there's a wide range of devices out there. Add 'application note' to the search, and there's plenty of detailed help to get low noise and wide bandwidth out of the things.
Dr. Slack, Tue Mar 21 2017, 06:22AM
Two approaches spring to mind.
One is to use a series resistor and a high speed voltage gain amplifier. The other is to use a transimpedance amplifier. The difference between these is whether there is a significant or insignificant change in rail voltage respectively.
The choice starts by understanding whether the SRAM can tolerate a rail bounce and still work properly, which means not only store valid data, but also draw the current it normally draws. Although (say) 50mV should be plenty to read on a scoope and is well within its static rail tolerance, are there dynamic bias things going on inside an SRAM that mean that rail just has to be nailed still when it wants to draw write current? Those SRAM chips require rail decoupling for a reason.
One way to check this would be to measure the current with increasing values of resistor. Does the measurement show a systematic trend. At what resistance does the SRAM fail to write properly?
There are lots of app notes for the use of transimpedance amplifiers. I put 'high speed photodiode amplifier' into the usual search engine, and on the first page were pdfs for 1 1Mbit/s amp, Texas OPA380, and a 10Gbit/s amp MAX3970. OK, the second one is only available in dice, but you get the general conclusion that there's a wide range of devices out there. Add 'application note' to the search, and there's plenty of detailed help to get low noise and wide bandwidth out of the things.
Print this page