[TriEmbed] Powering ESP32 from an 8v golf cart battery
John Wettroth
jwet at mindspring.com
Thu Nov 19 13:37:22 CST 2020
Pete,
If the high current peaks are a small fraction of the time, you can "bypass"
your way around it- I don't know if the ESP gadgets do this- maybe you or
others do.
There are several systems that do this in either simple or pretty complex
ways. We called these techniques "BFC" for big fat capacitor.
If you have a high current spike that is drawn for some time at some
current- the product iT is charge- coulombs. Charge is also CV, so you can
charge up a cap and dump into your load. The simplest way to do this to
have big cap at the load fed by a low value resistor to limit recharge
current to something you can live with. If the duty cycle is low, you can
blast out a big pulse and then have plenty of time to charge back up before
its needed. The problem is usually that in this simple case, you need to
use a much larger cap 10x than just CV would dictate because while the cap
is discharging- the charge is depleting and its voltage is falling (I/c
=dv/dt). To limit the droop to something liveable, you uses an 8x or 10x
sized cap- this way it only droops 10 or 12% over the burst.
The first time that I ever saw this trick was in two way pagers way back
when. The reciever on a pager drew nothing but when 2 ways came out, every
now and then, they needed to send a little text message and that drew a
bunch of current. This was the techniuque. In two way pagers, they
engineered in the droop and started with transmit power kind of dialed back
and as the cap discharged, it turned up the relative power in order to keep
a constant total power output. Pretty clever, this was early days of
supercaps but they were high impedance, the newer ones would be great- like
from Maxwell. I think they just used a few thousand uF tantalum or some
kind of nice electrolytic like a Sanyo OS-Con.
There are quite a few systems that use this trick- I've seen a lot of hobby
(and pro) spot welders that do this. A lot of energy harvesting systems
take in energy as the can until they get enough and then do their high
powered thing and start the cycle again. Walmart had these little solar
powered LCD price tags (about 2 x 1" cell) that worked this way. The host
system had IR emitters in the ceiling that sent command to the tags. The
tags could always receive but they communicated acknowldgements back to the
system with an IR led when they could. The time budget was something like
an hour or something- better than sending a person with a price gun etc.
I saw a fancier system at Rockwell Automation- AKA Allen Bradely. A common
problem in industrial is "last gasp" mode on power failures. In these
little PLC's they need to shutdown in an orderly way and it took someting
like a minute. Batteries weren't an option in most cases due to life. They
came up with a clever little discrete circuit using maybe 4 transistors and
an inductor. It consited of a boost converter that took the 24v and made
something like 300v to charge a HV cap. The same little circuit could turn
around and become a buck converter that could take the 300v and bring it
back to 24v when the power failed. This takes a different tack and uses a
pretty nice bit of physics. The energy in a capacitor is 1/2 CV^2, by
boosting up the voltage they could get this squared energy. Energy is Power
x Time, (watt-seconds or joules). When the engineer showed me his solution,
I was amazed and we couldn't make anything any cheaper or better really.
I guess the ultimate in this is Xenon flash strobes though I don't think
phones have to rely on this for using LED flash- the battery has such a low
impedance, you just hammer it for a little bit.
I have a little Ryobi cordless nailer- I don't know exactly how it works
but the final action is a piston that runs from an air resevoir. I don't
know if it has a little tiny air compressor that fills up the air resevoir
or some kind of piston run by solenoid to recharge the air. It recycles
really quickly. There is a tear down by AVe on the web. I need to watch it
again.
If you know the peak current, duration, etc- you can see if this works for
the ESP32. The charge storage IT is good but the energy storage with the
1/2CV^2 with a buck boost could work probably. I would think that a
regulators like Nick found would be simplest- its kind of ideally suited
excecpt for the 5v max Vin.
Regards,
John M. Wettroth
(984) 329-5420 (home)
(919) 349-9875 (cell)
_____
From: TriEmbed [mailto:triembed-bounces at triembed.org] On Behalf Of Pete
Soper via TriEmbed
Sent: Thursday, November 19, 2020 10:47 AM
To: triembed at triembed.org
Subject: Re: [TriEmbed] Powering ESP32 from an 8v golf cart battery
Thanks for sharing, this, John. The other day trying to gauge how much
Paul's golf cart battery would be drained with a linear vs switching
regulator a very quick glance of ESP32 stuff gave a guestimate of about
4.7V*50mA (but I guessed this wasted power wasn't relevant to Paul). But I
hadn't thought at all about peak currents as they relate to regulator
choice, and more importantly how much time is spent at those higher
currents.
But I'm writing to wonder how much help decoupling the regulator and
minimizing inductance might help with transient scenarios like you describe
below? Or would the cap dumping current just fake out the regulator and
simply make the droop happen a little later?
-Pete
On 11/19/20 8:53 AM, John Wettroth via TriEmbed wrote:
That's a nice part Nick. Like a lot of modern parts on small process, its
kind of designed around one LiIon cell operation. I like the "green mode"
stuff. LTC and Maxim make some parts like that. The difficulty with
switchover type parts is when you have a big load transient (eg- a transmit
burst). If the part is in the low quiescent mode, it can droop and cause
havoc.I noticed it has a forced green mode pin which could alleviate this
with some thought in software. A lot of times, its easier to have a
seperate regulator for the high current case or if they can be split up.
I didn't realize that these ESP devices require so much current, I've only
played with one on the bench pretty casually. For the 8v input case, you
really might want to consider a buck converter unless the high current only
runs very intermittently. The power wasted in a linear could create
potential heat problems. You're talking about getting rid of 1.5 watts from
your 8v source.
I agree with Nick on those DFN packages- awful, you're into hot air or oven
reflow. I like parts that come in a variety of packages including through
hole, this is getting rare these days.
Regards,
John M. Wettroth
(984) 329-5420 (home)
(919) 349-9875 (cell)
_____
From: Nick Edgington [mailto:nickedgington at edgingtonlabs.com]
Sent: Wednesday, November 18, 2020 6:44 PM
To: Josh Wyatt
Cc: jwet at mindspring.com; TriEmbed Discussion
Subject: Re: [TriEmbed] Powering ESP32 from an 8v golf cart battery
personal I like the STM
<https://www.st.com/resource/en/datasheet/ld39130s.pdf> LD39130SJ30R which
is a good match for esp32 it will do 300ma with a 300mv drop the quiescent
current is a remarkable 1 µA in green mode, 45 µA in normal mode which
matched with the ULP mode on the esp32, Not a problem with a golf cart but
important for low power sensor. and to top it of STM will send you a couple
for just the shipping cost,
The package a somewhat of a pain.
Nick
I have a number of the regulator board I pictured earlier should you be near
Apex and want one.
nje
On Wed, Nov 18, 2020 at 1:17 PM Josh Wyatt via TriEmbed
<triembed at triembed.org> wrote:
This is awesome info John, I'm enjoying the read.
>From personal experience, the ESP32S (and ESP8266) can be pretty power
hungry when the radios are on, and are particularly sensitive to
brownouts... I try to use something with at least 300mA and with good, stiff
caps.
Thanks,
Josh
On Wed, Nov 18, 2020 at 9:42 AM John Wettroth via TriEmbed
<triembed at triembed.org> wrote:
Shane,
There are a ton and its unfortunately one of these "it depends" kind of
things. But here are a few and why.
Most of the stuff I do is pretty small, low current stuff and 2.7 or 3.3v-
an 8 bit uP and a display with some RF. Very low quiescent parts suitable
for circuits with sleep mode, etc. A lot of my stuff lives outside and runs
on a 12v battery so I try to shoot for -40C operation and 16v Max vin. Temp
range matters for capacitors mostly and dropout. I buy almost exclusively
from Digikey. Their search engine and service is amazing even if they cost
a bit more. I tend to design very low power things and like low Iq for
sleep operation. Generally very low Iq means poor HF rejection (you need
loop gain)- in RF stuff, I'll compromise on Iq and shut the block down, etc.
I prefer newer parts, there has been so much progress in the last 15 years,
its amazing- there is no reason to use a 7805 for any real design- even
cheap stuff. There are better and even cheaper alternatives if you're
building more than a few hundred. At low volumes, 7805's can be awfully
cheap but they're really only designed as 60/120 Hz type regulators in a
traditional AC supply.
Some old favorites-
Micrel (Mchip owns Micrel linear now) MIC5203, 5205 series of BiCMOS types.
PNP pass element but controlled drop out current. 16v 50/150 mA, Iq 1 mA
max. Micrel invented BiCMOS pretty much and its good to see that Microchip
is keeping a lot of their parts.
Microchip 1791 is a great HV regulator- 30v max in, 70 mA out, load dump
(48V), Iq 70 uA. Microchip makes tons of cheap analog parts these days and
lot of good linears. They acquired Telcom semi many years ago which was a
big CMOS linear company (like Maxim). Newer CMOS stuff is good but the older
stuff is not so good- done on large processes and traded Low Iq for
performance- very slow load and line transient recovery, no PSRR, etc.
Toko TK11625 and TK1150, Digikey dropped Toko but I still have a lot of
these around- they're officially obsolete but they're plentiful everywhere
and there are newer alternatives. Available in TO-92, 100 mA, Tk71150 is 5v
LDO with good HF rejection for low noise for a post after a switcher,
Quiescent is OK at 300 uA, Seiko makes similar BiCMOS parts, can be hard to
find these days. Microchip basically copied these regulators to create
their line and DigiKey wants to keep Mchip happy.
TI TLV1117 A "special" very low Iq LM1117 variant, quiescent of 100 uA.
Better PSRR and dropout. Good in 3 Alkaline of 1 LiIon to 2.5v apps. Max
Vin is 5.5v, only downside.
Some favorites lately (doing low cost stuff)
Diodes Inc AP-7381 series. Available in TO-92 option still for quick
perfboard builds and breadboard. Cheap. Very low Quiescent, 50/150 mA
variants. Digikey large stocks always.
ABLIC S-812Cxx series. 1uA Iq, 10-100 mA output depend on voltage.
Quiescent useful for running a real time clock or deep shutdown on a HV
input.
Anything that Maxim or LTC/ADI makes are invariably awesome but low volume
pricing (<10k) is awful. Real customers pay nothing like those prices
believe me. I have odds and ends of Maxim leftovers but never have what I
need. TI has better pricing but isn't innovating much in this area- the
TLV1117 is an exception- excellent.
In your Golf Cart app, what's your load current min and max and vin min and
max. Any big line or load steps? Temp range and size could help too. Any
special operation needs like sleep?
After 25 years in Standard Products at Maxim, I can talk Linear IC's more
than anyone cares to listen. Take care- shoot me a private mail or call if
you have specific questions.
Regards,
John M. Wettroth
(984) 329-5420 (home)
(919) 349-9875 (cell)
_____
From: Shane Trent [mailto:shanedtrent at gmail.com]
Sent: Tuesday, November 17, 2020 9:57 AM
To: jwet at mindspring.com
Cc: Pete Soper; TriEmbed Discussion
Subject: Re: [TriEmbed] Powering ESP32 from an 8v golf cart battery
John,
Thank you for your breakdown on voltage regulators. Would you mind sharing
some of your favorite part numbers in the "Modern BiCMOS LDOs"?
Thanks!
Shane
On Mon, Nov 16, 2020 at 5:52 PM John Wettroth via TriEmbed
<triembed at triembed.org> wrote:
I'll put my 2 cents in here for a few subtleties. I defined probably 1000
different linear and switching regulators at Maxim in my 25 years.
Vout plus dropout is pretty good overall. But don't forget that dropout is
defined where the output voltage drops 100 mV. Its coming out of regulation
and all the goodness that linears give you stops happening. You also want
to do this at max load, max output tolerance and worst temp. The drop out
for bipolars decreases for higher temps which helps but at very cold temps,
it can grow- a lot a very cold. This is not allways well specified.
There are several classes of dropout that are driven by the design of the
ouput stage
Vdropout
2v standard bipolar linears like the 7805 use an NPN darlington output-
nice low impedance and easy to use.
1v LM1117 type bipolar linears use a Sziklai modified darlington with
an NPN follower driven by a PNP- pretty good comprimise
.5v LM2940 PNP pass element parts have low dropout but some squirelly
stability issues at times and can have high quiescent at dropout
.1v PMOS or charge pumped NPN pass element types that looks like a small
resistance in dropout. Quiescent can be very low.
.
Depending on the type of regulator, there are subtlties that happen around
dropout.
Old bipolar regulators like the 7805 have a drop out of about 2v
conservatively. The output is an emitter follower darlington stage which is
2 vbe's (.7v each) and 2 Vce sat (about .2v each). This is about 1.8v.
Since the output is a follower, it has a gain of 1 and are generally very
well behaved with very little thought given to bypassing and stability.
High frequency rejection is poor and accuracy is somewhat poor.
First generation bipolar LDO's like the LM2940 etc, used a PNP output stage
with a grounded NPN pulling its base down. These parts have two Vce sats in
the dropout path (about .5v). These transistors have gain on top of the
error amp gain and get unstable without following the the bypassing
instructions closely. The output cap becomes the dominant pole and the ESR
of the output cap has to in a specific range- neither two small or two
large. The other annoying feature of this class is as you approach dropout-
the beta provided by the PNP pass element goes south and they can draw lots
of current at or near dropout trying to keep the PNP in saturation. In low
power circuits, this can cause a sort of latching action and flatten a
battery in no time.
The LM1117 type second gen bipolar LDO's have a NPN follower ouput. These
have the benefit of a follower but only moderate dropout performance. They
were basically invented to make 3.3v from 5v which a 7805 couldn't do. Good
for point of load but kind of mediocre otherwise.
Modern BiCMOS LDO's generally have a PMOS pass element and some MOS and
Bipolar circuits.. These can have very low quiescent, very low noise and
the lowest dropout possible. They are also pretty stable with most loads
but take a signicant cap on the output usually. Something like a 10 uF
ceramic. Microchip make some good low cost parts in this class. Probably
my favorites.
There are all CMOS LDO's that share most of features of the last category
but don't get the low noise and high accurancy generally.
One other issue is your 8v battery. The charging voltage on a Lead Acid
could be over 10V which is a common abs max for many linears.
My 2 cents.
Regards,
John M. Wettroth
(984) 329-5420 (home)
(919) 349-9875 (cell)
_____
From: TriEmbed [mailto:triembed-bounces at triembed.org] On Behalf Of Pete
Soper via TriEmbed
Sent: Monday, November 16, 2020 10:46 AM
To: triembed at triembed.org
Subject: Re: [TriEmbed] Powering ESP32 from an 8v golf cart battery
On 11/15/20 10:34 PM, The MacDougals via TriEmbed wrote:
_______________________________________________
Triangle, NC Embedded Computing mailing list
To post message: TriEmbed at triembed.org
List info: http://mail.triembed.org/mailman/listinfo/triembed_triembed.org
TriEmbed web site: http://TriEmbed.org
To unsubscribe, click link and send a blank message:
mailto:unsubscribe-TriEmbed at bitser.net?subject=unsubscribe
--
photo
Shane D Trent
Patent Agent
919-348-0061 | shanedtrent at gmail.com
ShaneTrent.com | Skype: skype:shane.trent1 <>
Raleigh, North Carolina
<http://www.linkedin.com/in/shanetrent> Social
icon <http://twitter.com/sdtrent> Social
icon
Create your own
<https://www.wisestamp.com/create-own-signature/?utm_source=promotion&utm_me
dium=signature&utm_campaign=create_your_own&srcid=> email signature
_______________________________________________
Triangle, NC Embedded Computing mailing list
To post message: TriEmbed at triembed.org
List info: http://mail.triembed.org/mailman/listinfo/triembed_triembed.org
TriEmbed web site: http://TriEmbed.org
To unsubscribe, click link and send a blank message:
mailto:unsubscribe-TriEmbed at bitser.net?subject=unsubscribe
_______________________________________________
Triangle, NC Embedded Computing mailing list
To post message: TriEmbed at triembed.org
List info: http://mail.triembed.org/mailman/listinfo/triembed_triembed.org
TriEmbed web site: http://TriEmbed.org
To unsubscribe, click link and send a blank message:
mailto:unsubscribe-TriEmbed at bitser.net?subject=unsubscribe
_______________________________________________
Triangle, NC Embedded Computing mailing list
To post message: TriEmbed at triembed.org
List info: http://mail.triembed.org/mailman/listinfo/triembed_triembed.org
TriEmbed web site: http://TriEmbed.org
To unsubscribe, click link and send a blank message:
mailto:unsubscribe-TriEmbed at bitser.net?subject=unsubscribe
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.triembed.org/pipermail/triembed_triembed.org/attachments/20201119/840a12e9/attachment.htm>
More information about the TriEmbed
mailing list