Project: Noob makes car stereo keep-alive (boost power supply)

[tinkerbear]

I barely know what I'm doing on this project, so I was hoping I could get some experienced eyes for a design review?

Plus, there MUST be somebody else who needs this sort of thing.

This started off as a simple-sounding project: Keep the stereo playing uninterrupted when cranking the engine.  A modern car stereo with Bluetooth/USB playback has 20-30 seconds of startup time - irritating enough by itself, even more so when bluetooth needs to re-sync and playback has to be restarted just because you need to start the engine.

The obvious problem: It's designed to do that.  Toyota's keyswitch cuts power to the ACC circuit while the engine cranks.*  A diode OR gate from the IGN and ACC lines to the stereo's switched power lead should do it.  (Size of diodes depends on whether stereo uses switched power lead for actual power, or just as a signal.  For my stereo, it's only a signal.)

*[Why?  Some folks have said it's to ensure the maximum power for cranking, but that doesn't ring true.  Power to rear defog, heater fan, seat heaters, headlights - nothing else is cut during engine cranking, just ACC.  The best explanation I've heard is that Toyota can't control what's plugged into the ACC circuit, so they're protecting it from the voltage dropout by just shutting it off.]

Aaaand FAIL.  Stereo still shuts off during cranking.  Ok, so not that simple.  The voltage drop caused by the starter is resetting the stereo.  (Aside: adding a capacitor to the switched power lead kept it above 12v, so it's the drop on the BATT power lead that's causing the reset.)

Then I found this guy's solution: http://areksnotes.blogspot.de/2013/05/blaupunkt-melbourne-120-vs-voltage-dip.html

Testing shows my stereo resets if input voltage drops below ~10.5v, and draws 2-6 amps depending on volume setting.  My car spent 500ms below or not-consistently-above 10.5v during cranking.  (Worst case - I needed a new battery as it turns out.)



After two voltage drops on hefty schottky diodes, I'd need something like $80 worth of supercaps to make that work.  That struck me as kinda wrong - I know there exist car stereos that can deal with a voltage drop like this, and no car stereo manufacturer is going to use $80 in specialized parts to solve this problem.  You could do something similar with a much cheaper sealed lead-acid battery, but they're bulky and difficult to stuff in the dash behind the stereo.

What would a proper EE do?  I'm a CS grad so there's some overlap, but analog stuff was never my strength.  While I knew of the existence of voltage boost things, my understanding was a bit vague.  Surely a boost power supply could do this?

[Insert montage of two weeks of me plodding through the first couple chapters of a book on switching power supply design and learning the basics of LTSpice, so I don't have to breathe too much of the magic black smoke...]

I came up with this:



It's mostly cobbled off the datasheet, with a bit of logic (U2 voltage comparator) to shut it down (pulling the LT1270A Vc low) and bypass it (P-channel MOSFET) when voltage is good.  R4 is the load, R10 is the switched output.  D5/D6 are the diode OR gate for IGN/ACC - just connected to fixed voltages for the simulation.  U4 - the other half of the voltage comparator - is being used as an inverter for the ORed IGN/ACC signal.  There's a transient voltage suppressor slapped on the output, just for paranoia.

It does this:


So I built it up in Eagle:


To the original simulated circuit, I added:

• An input TVS (D3), for paranoia
• Two different inductor footprints (L1/L2)
• An LED, to tell me when it's boosting, for setting the voltage trigger
• A 5A polyfuse, which takes ages to trigger even if we hit the 10A the LT1270A/inductor/diode are capable of
• A capacitor (optional) on the ACC/IGN ORed signal, for delayed power-off

And ordered a board:


It's sized to build into an extruded Aluminum box http://www.rfsupplier.com/aluminum-enclosure-case-433201150lwh-p-1178.html using the box side for heat-sinking.  Probably not enough for full-load constant use, but hopefully good for 500ms at a time.

I built it, and it worked on the bench!  Installed in the car, it worked great... for about 4 days.

First "oh, duh" problem: Q3, a BSS138 n-channel mosfet, was getting a raw 12-14v on input, which should be fine because it'll take 20v.  Except cars electrics are a bit rough, and little spikes apparently matter to mosfets.  Fixed by using a resistor/zener diode voltage divider thingy on the input voltage (bodged horribly onto the board).

As of now, it's been in the car for two weeks, and seems fine.  Surely there are some more mistakes waiting to bite me?

Could it have been made smaller/lighter?  How do you gracefully handle an overload?

[zapta]

I had similar problem with my car. It has a Valentine 1 installed and whenever I crank voltage drops down to 8V which causes it to reboot and do the startup been sequence. It's even worse if the car is on auto shutoff mode where it turns off the engine when waiting in red light. Solved it by installing their brownout protector http://www.valentine1.com/v1info/bob.asp .   I assumed that it just has simple circuit with a diode and a capacitor but it actually has a DC/DC that can go to 6V or so input voltage.

This is the input voltage I get during crank

[tinkerbear]

I would guess the brown-out protector must be the same sort of thing internally - though it can be a lot smaller, as it doesn't have to deliver a possible 6A of current.

[T3sl4co1l]

No overvoltage protection...

Automotive stuff is always rated for at least 60V, and often up to 80 or 100V.

Or actually, I should take that back.  You do have D3 and D5 -- however, these won't weather a load dump.  Somehow I doubt you ever intend to have your battery disconnect while the engine is running -- and I really have no idea how rare these events really are, they don't sound common -- but those are the conditions which cause a load dump: without a beefy load (battery) to regulate system voltage, the alternator goes batshit crazy.  The standard is a swell of about 80V over 10ms, then decaying over 100ms, with an equivalent source resistance of 0.5-2 ohms.  So, potentially up to 160A short circuit, lots of energy (10s of J?), or... you're probably better off not shunting all that and just handling the 80V instead.  Then put some 80V MOVs or a self-resetting crowbar or something in there to still have an absolute limit.

The ACC and IGN1 inputs don't have any protection, just a big cap, which sounds kind of vulnerable.

Also, if R8 is set all the way to zero, you could potentially have unlimited power applied to IC1 through the input pins, with its supply lagging behind.

You probably don't need the full range anyway -- protip with trimmers is to calculate exactly what range you require (I'm guessing it's actually a 10-14V adjustment, or something like that), tack on just a little bit of excess (so you can be sure component tolerances don't conspire to squeeze your range lower than you need), and use exactly that amount.

Is Q1 backwards?  It doesn't "do" anything, in that it will never have more than -0.7V across it.

Regarding the overall system: avoid nonlinearity anywhere you can.  Now, you've slapped in a pair of comparators here, so... so much for that.  And there's not even any hysteresis on them, so a marginal supply voltage will just sit there and chatter and screw itself up...  Will it actually fry?  Probably not.  Will it be irritating?  Maybe (extra supply ripple / noise).  Does it matter in the grand scheme of things?  Doubtful.  Is it ugly?  You better believe it!!  Not good engineering practice.

Ya know, let's reflect on what a boost converter is...
When (U1?) FB pin is < VREF (whatever it is, 1.24V I'm guessing? haven't looked), it turns up the peak current.  When FB > VREF, it turns it down, presumably until it stops switching altogether.

But wait, isn't this exactly what we want?  It does nothing if voltage is in the nominal range, but turns on when it drops out...

Really, all you need is exactly the boost converter, with the voltage setpoint being the minimum voltage your stereo can safely operate on (10-12V..?).  Add some filtering and voltage protection stuff and you're fine.

Of course, that doesn't address the delayed-off option, for which you'd need to diode-OR the battery and accessory lines together, and add a timed switch.  But that can then be a seperate, isolated function, which needs absolutely no interaction with the converter -- so it can do its thing (or not, as it sees fit) without additional consideration.

Tim

[tinkerbear]

No overvoltage protection...

Automotive stuff is always rated for at least 60V, and often up to 80 or 100V.

Or actually, I should take that back.  You do have D3 and D5 -- however, these won't weather a load dump.  Somehow I doubt you ever intend to have your battery disconnect while the engine is running -- and I really have no idea how rare these events really are, they don't sound common -- but those are the conditions which cause a load dump: without a beefy load (battery) to regulate system voltage, the alternator goes batshit crazy.  The standard is a swell of about 80V over 10ms, then decaying over 100ms, with an equivalent source resistance of 0.5-2 ohms.  So, potentially up to 160A short circuit, lots of energy (10s of J?), or... you're probably better off not shunting all that and just handling the 80V instead.  Then put some 80V MOVs or a self-resetting crowbar or something in there to still have an absolute limit.

The ACC and IGN1 inputs don't have any protection, just a big cap, which sounds kind of vulnerable.

Somewhere I have picked up the notion that those TVS's could handle a load dump, and I was depending on that being true.  I'll go learn the math and see how badly I missed the mark - because you're right that 80v is going to take nearly everything in the circuit out.

The capacitor on ACC/IGN1 is only there for delayed shutoff, so no protection at all really.  I was thinking they were tied to BATT via the switch, and it's got a TVS clamping it.  Wooly thinking.  Actual protection is a better idea.

Also, if R8 is set all the way to zero, you could potentially have unlimited power applied to IC1 through the input pins, with its supply lagging behind.

You probably don't need the full range anyway -- protip with trimmers is to calculate exactly what range you require (I'm guessing it's actually a 10-14V adjustment, or something like that), tack on just a little bit of excess (so you can be sure component tolerances don't conspire to squeeze your range lower than you need), and use exactly that amount.

Huh.  Cool.  Will do that.

Is Q1 backwards?  It doesn't "do" anything, in that it will never have more than -0.7V across it.

Q1 is an attempt to bypass the voltage boost bits and avoid the diode voltage drop on D4 -- and all the worries about dissipating heat from it when the boost circuit isn't running.  When the output of either of the voltage comparators is low, the LT1270's Vc is pulled low (standby) and Q1 should conduct.  When the voltage comparators are both high, the gate is pulled high by a pullup, and it needs to block the 12v boost output from feeding back to the lower-voltage input.

Without Q1, D4 is going to be turning up to 5W of power into heat all of the time the stereo is on.  That feels like a lot to me.

Now that I think about it though, I've seen switching power supplies that use mosfets instead of diodes, to avoid that voltage drop... so maybe I can throw out the diode and the voltage comparators.  Hmmm.

Regarding the overall system: avoid nonlinearity anywhere you can.  Now, you've slapped in a pair of comparators here, so... so much for that.  And there's not even any hysteresis on them, so a marginal supply voltage will just sit there and chatter and screw itself up...  Will it actually fry?  Probably not.  Will it be irritating?  Maybe (extra supply ripple / noise).  Does it matter in the grand scheme of things?  Doubtful.  Is it ugly?  You better believe it!!  Not good engineering practice.

Check.  Add hysteresis.  If the comparators stay.

Ya know, let's reflect on what a boost converter is...
When (U1?) FB pin is < VREF (whatever it is, 1.24V I'm guessing? haven't looked), it turns up the peak current.  When FB > VREF, it turns it down, presumably until it stops switching altogether.

But wait, isn't this exactly what we want?  It does nothing if voltage is in the nominal range, but turns on when it drops out...

Really, all you need is exactly the boost converter, with the voltage setpoint being the minimum voltage your stereo can safely operate on (10-12V..?).  Add some filtering and voltage protection stuff and you're fine.

I was trying to keep the standby current low, and the LT1270 pulls 7mA unless you pull Vc low and put it into standby (100µA).  Though now that I've built it, the circuit as a whole still pulls 10mA when doing nothing, probably from the voltage dividers and comparators, so if I hit 7mA it'd be an improvement.  I was hoping for less.

Of course, that doesn't address the delayed-off option, for which you'd need to diode-OR the battery and accessory lines together, and add a timed switch.  But that can then be a seperate, isolated function, which needs absolutely no interaction with the converter -- so it can do its thing (or not, as it sees fit) without additional consideration.

Tim

Thanks Tim!  That was very helpful.  I've got some reading to do.

[T3sl4co1l]

Q1 is an attempt to bypass the voltage boost bits and avoid the diode voltage drop on D4 -- and all the worries about dissipating heat from it when the boost circuit isn't running.  When the output of either of the voltage comparators is low, the LT1270's Vc is pulled low (standby) and Q1 should conduct.  When the voltage comparators are both high, the gate is pulled high by a pullup, and it needs to block the 12v boost output from feeding back to the lower-voltage input.

Right, cuz the output gives it a few volts... don't know why I didn't see that before.

Now that I think about it though, I've seen switching power supplies that use mosfets instead of diodes, to avoid that voltage drop... so maybe I can throw out the diode and the voltage comparators.  Hmmm.

Yes, for higher efficiency you may consider a synchronous rectifier, probably a controller chip with external switches rather than an integrated device.  Or one of their uModule things, if you're feeling rich...

I was trying to keep the standby current low, and the LT1270 pulls 7mA unless you pull Vc low and put it into standby (100µA).  Though now that I've built it, the circuit as a whole still pulls 10mA when doing nothing, probably from the voltage dividers and comparators, so if I hit 7mA it'd be an improvement.  I was hoping for less.

Yeah, a good plan for battery-op stuff.  If you can find a micropower controller that also has beefy drive outputs (1A peak gate drive should be enough for the kind of transistors needed here), you should be good there.

Thanks Tim!  That was very helpful.  I've got some reading to do.

Cheers!

Tim

[David Hess]

I think you are on the right track.  I have seen this type of circuit used to power amateur radio gear where sagging of the automotive 12 volt supply may cause problems.

If I were doing it, I might implement a continuously running push-pull transformer buck converter with a nominal input range of 8 to 16 volts and fault protected output of 13.8 volts or whatever is desired.  The transformer ratio sets the relationship between duty cycle and input to output voltage ratio.  The power supply would be more complicated but no switching between modes would be necessary and the output could be arbitrarily free of electrical noise and even isolated if desired.