Efficiency of Voltage Boosters
-
Well, I thought I would re-post the data and try to summarize the thread that was lost.
Objective: Run a sensor node for a year on batteries. Wake period ~2 - 5 minutes.
Approach: Use a voltage step-up converter to boost the voltage from 1 or 2 AA batteries to 3.3 V or 5 V required by the two flavors of pro-mini clones.
Problem: Even though folks have discovered methods to put the sensor node into a deep sleep mode which draws ~100 - 200 microA; the efficiency of the step-up modules prevent realization of such a low current draw; rather, at that low current draw the efficiency of the step-up converters yields ~1-2 mA. See figures,
The good news here is that the cost efficient step-up converters from China are just as efficient as the SparkFun NCP1402 when the LED is disabled.
Solution: The search was on to find a more power efficient step-up converter. Here are some that I and other folks found.
TPS61222 - These are pretty small though.
TPS61097 - These are sort of expensive.
XC9140AXX1 - These look promising.
SC120 - Also cost effective, but may not be as efficient as the above.I have ordered the parts to construct boosters using the TPS61097-33 and XC9140A331 and will report back to the forum with the results.
Everyone - does that summarize the original post sufficiently?
-
Holds the text of the old posts:
-
@a-lurker Well done!
-
Nice summary. I'd like to add the booster chip that axillent was using for his project, the MCP1640.
Also, since this project is utilizing radio communication, both EMI and the quality of the boosted power could be an issue. There were some other post, probably gone now, that talked about connecting batteries directly and separately to the radio to get around such issues. Probably had something to do with ebay regulators or the regulator on the arduino?
Looking forward to more test results. Don't forget to mention what brand / type of inductor and capacitors you go for in the circuits.
-
The tests recorded the radio supply Voltage under various conditions and demonstrated that a low ESR (Equivalent Series Resistance) capacitor gave the best results. It seemed highly likely that the other capacitors tested were plain old electrolytics? They are not suitable for use as decoupling capacitors for the radio. Either tantalum or aluminum polymer capacitors should be used, as they typically have the desirable characteristic of a low ESR value. They should be mounted physically at the radio and wired between the power pins - plus and ground. Their leads should be as short as possible. Any oscilloscope measurements should have the probe connected directly across the capacitor. A couple of the tests used 220 uF caps, perhaps a little unnecessarily large. A lower value of 47 uF should be adequate.
It was suggested that the cleaner power supply Voltage may have reduced electromagnetic interference (EMI) at the radio receiver. This may be the case but I'm inclined to think that it just improves the operation of the receiver IC itself. Either way, a clean power supply that has a bit of grunt when needed, is always a good thing.
Hope I have remembered this correctly.
-
@a-lurker Hmm. Perhaps the store shouldn't recommend electrolytic capacitors then for the radio decoupling.
Comparison between tantalum and ceramic capacitors:
http://www.avx.com/docs/techinfo/mlc-tant.pdfPertinent conclusion table:
Texas Instruments recommends (fancy SMD) ceramic capacitors for the TPS61097 and TPS6122x. Empirical evidence would suggest that cheap Ebay caps may not be the best idea. Nor would it be prudent to replace them with Tantalum, though that could also be attributed to cheap Ebay goods. The post is unclear on this.
From the TI datasheet on the TPS61097:
-
What about being able to turn on / off the voltage regulator via output pin and mosfet and powering the arduino from a cap while in standby? and in awake mode it turns on regulator and charges up the cap again.
-
@bjornhallberg said:
@a-lurker Hmm. Perhaps the store shouldn't recommend electrolytic capacitors then for the radio decoupling.
Yep, we should probably do that. Need help finding a good candidate. Found this. Any good?
Search: "tantalum 47uf -smd"
-
Yes, you got the result from the booster voltage ripple test correct, the 4.7 microF is the one recommended from the MySensors store, and the other 220 microF is a high-end audio electrolytic. The low ESR is here. Here is the data,
Now, with that being said; why power the radio from the booster at all? I would submit, for battery powering a sensor node, that one should simply connect the radio directly to the battery (2xAA). The Arduino seems to be less sensitive, but still need to test. Connecting the battery directly also is more efficient, no power loss through the booster.
My observation was, at a range of ~few feet, that I would get ~20-70% radio communication fails in the 'pingpair' sketch with the radio powered by the booster, and 100% success radio powered from 2xAA batteries, and 100% success radio powered by booster with the low ESR capacitor mentioned above. I didn't have any lower capacitor values to test (actually, I think I might have a 150 microF as well, but no lower values).
I also tested the 5 V -to- 3 V step-down regulator from the MySensors store and that is dead flat, no voltage ripple (~1-2 mV noise). But, that is also power inefficient, however it is a good option for a plug-in node or gateway.
-
So, you are saying that it looks like there is no capacitor or anything else needed when powering the radio separately?
And, when using a capacitor the 220 microF is better in any case (good or less quality)?
-
@bjornhallberg said:
Looking forward to more test results. Don't forget to mention what brand / type of inductor and capacitors you go for in the circuits.
I'm currently doing tests with prototype of future MySensors board.
It is started today from a single AA battery.
The radio part is not yet soldered but will be soldered as soon as I will compile and test a correct bootloader
-
As for capacitors.
Do not mix different things.old-style electrolitic and tantalum can be used to decouple radio module power.
this capacitor is needed to reduce drop of the voltage at a time of receiving and sendingas for step-ups/step-downs/LDO there is no ideal capacitor in general. You should always refer to the recommendations from the datasheet.
Most modern power chips are required multil-layers ceramic capacitors. Such capacitors they have stable characteristics and very low ESR (much lower than tantalum have). In most cases it is X7R or X5R (the difference between this two is temperature stability)
But for some power chips ceramic will be not recommended. Refer to the datasheet.
-
@marceltrapman for radio the capacitor is recommended always. For some good quality modules this capacitor is build on the board, but in most cases it is not. 220uf is too much. 10-47uf is sufficient, tantalum is better than electrolytic
-
I've (perhaps naively) built my battery powered MySensors pcb's around the 3 pin configuration of the ebay cheapie. If many others have done the same then I can see some value in someone creating a drop-in replacement board based on the more efficient chips. I'd definitely buy the more efficient version if someone created it, whether pre-made or solder-it-yourself.
-
Good to see new activity on this thread again.
@axillent Unless there is some EMI coupled into the power leads of a battery connected radio, or current draw from the battery can't keep up with demand, I don't see why a capacitor is needed. In fact, the product brief for the nRF24L01+ states that it is designed with an on-board voltage regulator and permits operation with a coin-cell battery. Is there something I'm missing?
Also, regarding capacitors, I think you mean 220 microF is 'overkill' or 'more-than-enough', rather than 'too-much'. Right?
Anyway, here is a comparison of aluminum polymer capacitors of different capacitance. Note how the 220 microF cap is cheaper and has lower ESR than either the 10 microF or the 47 microF, and is only slightly larger in physical size.
Tantalum on the other hand has a quite high ESR, on the order of ohms, not good for suppressing voltage ripple according to this document.
That being said, ceramics may be the way to go for sub-100 microF as they are lower cost, smaller, and very low ESR; above 100 microF ceramics get very expensive, ~$4 for a 220 microF.
-
@Bandra Great, would you be able to share your run times so that we can get some real world data? That is, how long can you run a node before the battery is discharged?
I was thinking about making a board using one of the high efficiency boosters listed in the first post, if they prove their worth. I get my components in on Saturday, so sometime next week I should have preliminary results.
-
@axillent Unless there is some EMI coupled into the power leads of a battery connected radio, or current draw from the battery can't keep up with demand, I don't see why a capacitor is needed. In fact, the product brief for the nRF24L01+ states that it is designed with an on-board voltage regulator and permits operation with a coin-cell battery. Is there something I'm missing?
I'm not an EMI or decoupling freak but I knew from practice own and others that communication distance is very vary with and without capacitor soldered very close to the radio.
You can try yourself. My PA+LNA modules are having tantalum soldered near power pins while cheap $1 modules do not have a capacitor on board
-
@axillent I totally agree with the radio being very sensitive to noise and voltage ripple on the power pins. I have had great improvements with running the radio on battery directly.
I know that the 4.7 microF electrolytic recommended on the MySensors store does very little to help.
When I powered the radio from the booster I was getting poor reception and many failed messages; I didn't have any tantalum capacitors on hand, so I tried to look at suppressing the voltage ripple of the power to the radio using the beastly 220 microF polymer capacitors and it worked well.
I think we have come to the same conclusion...voltage ripple to the radio is bad, suppress it any way you can and one will have better results.
-
@therik said:
@Bandra Great, would you be able to share your run times so that we can get some real world data? That is, how long can you run a node before the battery is discharged?
I was thinking about making a board using one of the high efficiency boosters listed in the first post, if they prove their worth. I get my components in on Saturday, so sometime next week I should have preliminary results.
I only sent my pcb's off for production a couple of days ago. They have a 5cm x 3cm footprint which is designed to fit perfectly on top of a 2xAA battery holder. If the analysis of the chinese cheapie step up converters is right then it may be a couple of months before I run out of battery and have some real world results. However they have the battery level sensing circuit on them so I may have some results sooner once I tweak the algorithm for the 1.2 volt cells. Will let you know how I go.
-
On the cheap electrolytic capacitors: They tend to have a high leakage current, which is not good in battery powered scenarios.
-
(Note resurrected post) @therik I used to think that you could run the MySensor stuff just using two batteries connected directly to the processor but came to the conclusion that's a bad idea. First, to get the best out of some sensors running them off a fixed Voltage is preferable.
Secondly two alkaline AAs supply about 2.4V to 2.6V (during most of their operational life time) and the 3.3V CPUs have the brownout set to 2.7 V (typ). So if you use this method, you need to ensure the brown out detector (BOD) fuses are set to disable the brownout detection completely. Doing it in software is not sufficient, as the BOD is automatically re-enabled as soon as the CPU comes out of sleep.
The CPU spec shows the maximum Frequency vs. VCC line for between 1.8 V and 2.7 V. The line equation is Vmin = (F-k)/m, where from the spec k = -8 and m =6.6666 so for 8 MHz: Vmin = 2.4V Towards the end of their life two AAs will go below this Voltage. What happens then? With an inverter you can monitor the battery Voltage and trigger an alarm when it gets really, really low.
Also with the inverter you can squeeze every last ounce of juice out of them and everything will still work.
@axillent Totally agree about charging batteries - probably more hassle than it's worth. Depending on the application, just solar charging a super capacitor could work well. Imagine a temperature sensor reporting back every fifteen minutes and it consumes say 30ma at 3V3 for 0.5 seconds while it reports and we use one of these:
https://www.sparkfun.com/products/746
time roughly = (C/I)*(Vmax-Vmin)
so say we use say a Vmax of 2V and Vmin of 1V and use that to power an inverter to 3V3. Also assume we draw 100 mA at 1V, which is about 30 mA at 3V3 for the CPU and radio. We have a ten Farad cap but it's a bit out of tolerance, so it's actually only five Farads.
(5/0.1)*(2-1)= 50 seconds
That allows for one hundred 0.5 second samples to sent before the cap is "flat", which is 25 hours if sent every 15 minutes. So it looks like it could work OK. Should work for @HeK in Sweden where the shortest day is only about 7 hours long, with the main problem is keeping snow off the solar panel.
-
Just adding to the above. It does make sense to run the radio directly off the battery (1V9 to 3V6) but two problems remain: the CPU brownout detector and the minimum Voltage required to run the CPU, which is 2V4 at 8 MHz.. See also:
http://forum.micasaverde.com/index.php/topic,20078.msg164716.html#msg164716
The radio must have a supply Voltage of 2V7 to 3V3 if the input signals are greater than 3V6. you would encounter this when running a CPU at 5V, which hopefully you wouldn't be doing on a battery powered set up.
-
@a-lurker If brownout detection is disabled we could power atmega328p with 1.9V minimum, but frequency of operation will decrease.
I know you get it right, i just want to standout that there is no problem of operation at low voltage if we know 2 thinks:- disable BOD (no soft)
- we get lower frequency
-
There are a few subjects in this valuable thread.
One is the best type of capacitor to use across the power and ground of the nRF24L01+ module. I'm very interested in that. My earliest tests were frustrating, I was busy trying to figure out what was wrong with the software (before I came here by the way), but it turned out to be power and/or too long SPI wires. A cap across the power at the module made a great difference. So I'm tending towards always doing that in the future, and I'd like to know what kind to stock up on.
I'm seeing some opinion that tantalum would be better, some that it has too high ESR. So a low ESR electrolytic (probable aluminum polymer) would be better. Or maybe a ceramic is better still, and cheap.
Are there any recommended caps below 50 cents in small quantities? Is there anything from eBay or Tayda or are those all junk for our purposes? If not, Digikey or Mouser is OK, tho the minimum shipping raises the effective cost per unit substantially for small quantities (and I may not have a large order to combine it with for months).
There is a variation of this question which is about caps for use with noisy boost converters. I am thinking that a good cap for that purpose would also be a good cap for use with the tranceiver in general - low ESR and with an appropriate value, with low leakage. (In another case the noise of a sensor or actuator might be what needs to be filtered out rather than a boost voltage converter).
-
I changed the capacitor that I used for my radios from 4.7 mF into 220 mF and I can only say thank you for this discussion, it is really immediately much better
-
@marceltrapman mF as in milli or micro?
µF I suppose..
-
@marceltrapman there is a convention to mark micro by µ if you have a special symbol or by u if you don't
this way 220 micro Farad should be market as 220 µF or 220 uF@Damme it is a good point)
-
-
@marceltrapman You're welcome
I tried 220µF myself now (EU keyoard can write µ with alt-gr+m or (ctrl+alt m if alt+gr is missing))
much better, I first used 33µF and then 100µF but 220µF is a winner.
The wierd thing is that I cant measure any difference with 33µF or 220µF with my oscilloscope. but with no cap at all my arduino nano 3.3v looks like this: (AC coupled)
And wit cap (33µF or 220µF makes no difference at all..) But it works much better..
-
@Damme strange, disturbance is around 50hz... Could it be caused by mains supply?
-
@Yveaux I triggerd on transmission.
-
@Yveaux As mentioned I am still learning about this stuff.
The question I have right now is, a bit off topic, at which pin do you measure this?
Is that simply vcc on the radio?
-
@marceltrapman I guess. If you want to measure stability of supply of the nrf then you measure on vcc & gnd of the nrf.
-
-
@Damme said:
much better, I first used 33µF and then 100µF but 220µF is a winner.
The wierd thing is that I cant measure any difference with 33µF or 220µF with my oscilloscope.So I think you are saying that you are not seeing a visual difference (on the o'scope) between 33uF and 220uF, but you are seeing better real world performance with the latter?
Do the different capacitors have the same spec's? I'm wondering if the capacity is the only significant change, or if the different values of cap you tested have different specs (eg: ESR) or even diff technology.
-
@Zeph Exacly, I first used 3.3µf and there I see voltage drop. But almost no measurable difference with 33µF or 220µF. But I get less failed transmission with the larger cap.
They are all the same brand (some cheap Chinese unknown (to me) brand, "Chong" (yay!))
All 16v, and all electrolytic. I don't have any ESR meter, thinking of trying to measure it with oscilloscope and function generator.. (Or I'll just bu one :))I bought them from ebay in a large 1800pcs asorted pack.
-
New data on the XC9140A331 3.3 V voltage booster. This looks like a nice IC with low quiescent current and an improvement over the stock "china" 3.3 V booster.
This option can be purchased from Digikey for $0.90, with a total bill of materials of $1.34 (in quantity, minus a PCB). It is also a nice size SOT23-5, not too small to hand solder.
Thoughts?
-
@therik thanks, it is an interesting chip
-
I look forward to a test of the TPS61222 used in the MySensors Battery board - I see the chip on your list and hope you will be checking it.
Felix at lowpowerlab (http://lowpowerlab.com/blog/2014/06/08/powershield-r2-released/) has switched from the TPS61220 to the LTC3525 for stability reasons. That's a 5v chip, tho but maybe there's a related one as a contender.
-
How do you guys measure <mA accuratly? Using mr Jones's µCurrent?
-
@Damme Yup. Build some myself (and have a few spares, for who'se interested)
-
Any news on this front?
Started looking at the LTC3525 as well, but it ended up being a bit more expensive than I had wanted (looking at Digikey it is basically the most expensive DC-DC step-up IC they stock). For the 3.3V version there is like one (1!) AliExpress seller that has them in any quantity. 10pcs $23 and I'm currently pondering that.
Like @Zeph said, both lowpowerlab and harizanov have switched to the LTC3525.
Btw. Found an Eagle pcb for the LTC3525 here:
https://github.com/meigrafd/boost_converterAnyone found any other IC? TPS61016? TPS61006? TPS61201? ISL9111? MAX1724? UCC3941?
Or is the good old TPS6122x still the best choice at the end of the day in terms of price / performance? Looking at AliExpress it is basically half the price of LTC3525. And they're both the same pesky small package.
Also, still completely in the dark when it comes to finding suitable inductors in particular (sourced from China). If you're willing to buy from Digikey et al sure, but if you're trying to cut some corners things get a lot harder.
-
I have designed a PCB shield for the pro-mini, radio, and various sensors based on the the XC9140 chip. I will be powering the radio directly from 2xAA batteries. Boards should be here any day now and I'll check out the design. The XC9140 is not a bad chip from what I can tell, and it is cheap from Digikey, see posts above for BOM.
-
@hek Can you update the website for the decoupling capacitor recommendation? I wasted a lot of time trying to figure out why I have a lot of lost ACKs with my voltage boosted sensor until I found this topic. This made the sensors cry for a new parent quite frequently and it kept them awake more than necessary.
I had about 35% ACK packet loss with the 4.7uF capacitor which went down to 0.5% with a 68uF one (though the data got through). I will try to buy an even bigger low ESR one to make it zero, but the 4.7uF definitely doesn't do a good job if the signal is actually dirty.
-
Ok, I'll could increase the recommendation to 47uF in the next update.
But i'm not sure which works best. A electrolytic low ESR or if it is good enough with a cheap ceramic variant. If someone has the time and/or the equipment I would really appreciate some research.
-
47uF is what it took for my setup to work properly. Mine are just run of the mill electrolytic.
-
The "Efficiency of Voltage Boosters" thread is full of useful info but has been sleeping for a year so this is an effort to bring it back to life again
I have built a few battery powered sensors based on the 328P-PU running at 8MHz as well as 1 MHz. Batteries used are common NiMH size AA + cheap Chinese 3.3V step-up converters. Both the processor and the radio are powered from the 3.3V converter which I understand can be a bad practice. Still I decided to try because I did not want to tamper with the fuses/bootloader at this stage and luckily I have so far not experienced any of the here described connectivity problems.
More or less out of curiosity I made a few tests of this setup using different sizes of same brand electrolytic capacitors (nothing fancy, bought as a Velleman high-Q kit, labeled "made in Europe"). The results does not give more info than already available here but it helped me understand and hopefully can help others too.
For the test setup a 10µF is soldered to the radio pins and I can add another in parallel through a socket. The unit tested is also equipped with both a 3.3 and a 5V step-up. A photo of the setup as well as a few screen dumps from an oscilloscope are attached as a pdf (hope it works). Readings are taken using a 8 MHz chip programmed as a motion sensor and while the sensor is at rest.
Here is a short summary and a curve of same data:
10 µf dV 76 mV
10 + 22 µF dV 43 mV
10 + 47 µF dV 28 mV
10 + 100 µF dV 14 mV
10 + 220 µF dV 9 mV
10 + 470 µF dV 5 mV
A bigger capacitor of course lower the amplitude of the ripple (not unexpected :-)) and the rapidly falling curve shows that the current recommendation of 47 µF is a good choice. Adding more will lower the ripple but not at all proportionally.
I am not experienced neither in building these battery powered sensors nor in measuring them why any comments/corrections will be appreciated.
