IoT consumption logging/profiling

[Jorge]

Hey everyone!

I'm searching for a device to measure power consumption of a custom IoT device that ranges from a few uA in sleep up to 2A peak during data tx.

Already have a ucurrent gold for small burst profiles, but this has a lot of limitations with the burden voltage and the lack of long-term logging.

The budget would be about 2kish, and for what I found my options are basically Joulescope or ZS1100A. Also found the power analyzer N6705C from Keysight, but it’s well over by budget. 

Any idea if there’s any like 5x cheaper N6705C alternative? 


Cheers.
Jorge

[thm_w]

Already have a ucurrent gold for small burst profiles, but this has a lot of limitations with the burden voltage and the lack of long-term logging.

You would use an appropriate external DMM for logging for that purpose.

https://www.eevblog.com/forum/testgear/how-to-measure-energy-consumption-on-a-iot-sensor/
https://www.eevblog.com/forum/testgear/test-equipment-for-integrated-average-current-down-at-the-ua-level/
https://www.eevblog.com/forum/testgear/looking-for-good-and-fast-ua-meter/
https://www.eevblog.com/forum/testgear/power-integrating-multimeter/
https://www.eevblog.com/forum/testgear/equipment-advice-battery-lifetime-testing/

[Cerebus]

Another route is powering from a measurement power supply, like a HP/Aglient/Keysight Model 66312A "Dynamic Measurement DC Source System" which is basically a poor man's SMU (Source Measure Unit), or a fully fledged SMU from someone like HPAK or Keithley.

I pick the 66312A just because I happen to have one. They turn up on the secondhand market from time to time at quite reasonable prices - usually around the £300 GBP region. It's a 0-20V, 0-2A bench power supply with more extensive, more precise, faster measurements than a typical bench PSU. It can measure at a resolution of 2.5uA on its low range (-20mA to +20mA) and 250uA on high range, and can switch measurement automatically between the two. Will store 4096 readings with a sampling rate range of 15.6us to 31,200s. Needs GPIB for fast readings, slow ones and the 4096 reading store can be read out on an RS232 port. It sounds like it might fit your requirements.

If you need faster or finer resolution you're probably into a fully fledged SMU. Again, you might get lucky and find a secondhand one within your budget, but the Keithley ones usually change hands for a bit more than 2k, HPAK ones are like hen's teeth and I'm poorly informed on other brands.

[hans]

Already have a ucurrent gold for small burst profiles, but this has a lot of limitations with the burden voltage and the lack of long-term logging.

You would use an appropriate external DMM for logging for that purpose.

https://www.eevblog.com/forum/testgear/how-to-measure-energy-consumption-on-a-iot-sensor/
https://www.eevblog.com/forum/testgear/test-equipment-for-integrated-average-current-down-at-the-ua-level/
https://www.eevblog.com/forum/testgear/looking-for-good-and-fast-ua-meter/
https://www.eevblog.com/forum/testgear/power-integrating-multimeter/
https://www.eevblog.com/forum/testgear/equipment-advice-battery-lifetime-testing/

DMM's have two issues.
uA to A is a dynamic range of 6 orders of magnitude (120dB, 20-bit). This is not out of the ordinary for 6,5 dig or more DMMs, but then the 2nd issue pops up: usually those devices only reach a high measurement resolution with multiple line cycles. For an IoT product the current profile can be rapidly switching, therefore also requiring ksps sampling rate.

In the past I've made custom measurement devices for these kinds of things. Basically put two current ranges side by side. Let the uA clip once the A 'needle' is finally moving off center. With 2 channels you don't need 20-bit accuracy but can get away with e.g. 14 or 16-bit.

If purchasing off the shelf.. I'm afraid either make a compromise on resolution/accuracy or sampling rate, or price.

E.g. normally I would recommend the Nordic Power Profiler, but that doesn't go up to 2A.

[Jorge]

A DMM could be kinda slow (a few ksps). On our use case where the narrow current spikes will make a hell of a difference on a 3-to-5-year battery life span, and I’m afraid of the slow data sampling.

Those R&S with 500ksps seems okish, of course it’s an awesome DMM, but I need it for the data logging.

[thm_w]

DMM's have two issues.
uA to A is a dynamic range of 6 orders of magnitude (120dB, 20-bit). This is not out of the ordinary for 6,5 dig or more DMMs, but then the 2nd issue pops up: usually those devices only reach a high measurement resolution with multiple line cycles. For an IoT product the current profile can be rapidly switching, therefore also requiring ksps sampling rate.

Was referring to using a DMM to cover the lack of long-term logging only.
The threads linked go into more detail.

A DMM could be kinda slow (a few ksps). On our use case where the narrow current spikes will make a hell of a difference on a 3-to-5-year battery life span, and I’m afraid of the slow data sampling.

Those R&S with 500ksps seems okish, of course it’s an awesome DMM, but I need it for the data logging.

You mean the NGU401 SMU? That's $7k.

[tszaboo]

This is very hard to do. A DMM6500 can log fast, but the uA accuracy is bad even in the 100mA range.
A 2280S power supply has the resolution, but the uA accuracy is bad and it is slow.
A 34465A is OK for average current measurements, but slow.
So far, the best way I found was to measure the peaks (scope/DMM6500) and the average current (DMM) separately, and calculate the data offline. I plan to build something for my use case, but that's not going to go to 2A.

[hans]

Perhaps also something to consider.. just a brainwave that hit.. What is the purpose of the measurement?

Is it logging/profiling? E.g. debugging what part of radio transmission etc. is taking the longest? Then a uCurrent hooked up on to a scope is ideal. The more bandwidth the better for scoping.

If you want to calculate the battery life, you basically want a figure of J/s, or uA RMS measurements at DCV, to know how much energy the device is taking. This is a territory where a DMM excels, but only at a limited bandwidth.

Perhaps increasing or decreasing the filtered bandwidth of the power network on the DUT (if it can handle it), is a possibility. A higher bandwidth (=less bulk cap) allows to see sharper current spikes of consuming components. OTOH a large amount of bulk cap could smooth power draw in such a way that the current measured at the power terminals is pretty much all within the RMS bandwidth of a DMM.

[rejoetech]

Here is another option from us, EMK850S+ https://www.tindie.com/products/rejoetech/rejoe-power-monitor/

Output voltage range from 0.5 V to 12.0 V
Maximum current measurement up to 2 A
Measurement resolution down to 1 μA
10 ksps sample rate

Technically unlimited duration for logging when you activate 'auto save waveform', maximum 10 hours logging for each digifile, but new logging file will be automatically created every 10 hours, and data will be kept recording.

[Jorge]

Thanks alot for the replies guys!

Does anyone have some kind of experience with that EMK850S+? Seems cheap...

The idea is to profile the consumption of a device on a medium term log to be able to develop an accurate simulator.
Our device has very random current spikes that need to be accounted.

Starting to think that's probably better to invest on a fully fledged SMU...

[PlainName]

Does anyone have some kind of experience with that EMK850S+?

I've been  hoping someone like you might get one to find out and let us know 

[LogicalDave]

I don't have experience with the EMK850S, but I have used several other devices for dynamic power measurement.  The best is the Joulescope; for something much less expensive, you could try the Nordic Power Profiler Kit (NRF-PPK2: https://www.nordicsemi.com/Products/Development-hardware/Power-Profiler-Kit-2) but I believe it only works up to 1A (so you'll have to measure transmit power separately and turn it down for profiling if you use the PPK2). 

One of the major differentiators for these types of devices is the sample rate.  If you're looking for a fast event, a device that samples at 10ksps or 100ksps might not capture it.  The Joulescope samples at 2Msps and advertises a bandwidth of 300kHz.  how fast and how large are the random current spikes you are trying to capture?  There might be other approaches.

[coppice]

I haven't used the joulescope or the ZS1100A, but they appear to be doing more or less the right thing for measuring total energy in bursty applications. The devil may be in the details, though. E.g. how well are they able to control things like ringing? Most tools, regardless of price, are pretty useless for these bursty loads, with huge dynamic range, where the high bursts are not the long term dominant energy usage. The other approach which works well is in some of the tools for the MSP430 family from TI. These feed energy into the supply in well controlled shots. If you count the shots you have an accurate energy measurement over a wide dynamic range. However, if you need to go up to multiple amps, the TI tools I know of will be overloaded. I do like that approach, though.

[PlainName]

where the high bursts are not the long term dominant energy usage

Depends on the application. I'm currently trying to sort a supply for a (third party) light sensor which connects via WiFi and goes through three AA batteries in a couple of days. I think the periodic WiFi connection, brief though it is, massively overwhelms the sleeping power draw.

Similarly, Bluetooth can be a bummer if, between connections, the kit is just kicking its heels waiting for something to connect. Just spacing out the advertising a bit can make a big difference in life.

Obviously, a slight change in background power draw can make a big difference because it's there all the time, but I would be surprised if it's a generalism that the bursts aren't dominant, at least with 'low power' kit.

[coppice]

where the high bursts are not the long term dominant energy usage

Depends on the application. I'm currently trying to sort a supply for a (third party) light sensor which connects via WiFi and goes through three AA batteries in a couple of days. I think the periodic WiFi connection, brief though it is, massively overwhelms the sleeping power draw.

Similarly, Bluetooth can be a bummer if, between connections, the kit is just kicking its heels waiting for something to connect. Just spacing out the advertising a bit can make a big difference in life.

Obviously, a slight change in background power draw can make a big difference because it's there all the time, but I would be surprised if it's a generalism that the bursts aren't dominant, at least with 'low power' kit.

If the high bursts dominate the power consumption there are numerous techniques that are effective to find them, and try to reduce them. However, you still need something like the joulescope to get an good assessment of what you battery consumption ends up looking like, without waiting for a real battery to wear out in real use.