开云体育

ctrl + shift + ? for shortcuts
© 2025 开云体育
  1. QRPLabs
  2. Topics

QMX SMPS issues, and a possible cure

 

I'd like to know the firmware and hardware rev from anybody who has trouble with the QMX 3.3 and 5.0v power supplies.
Mostly, I'd like to know how often these failures occur with the most recent firmware and hardware.
If that gets noted here in the forum, the pressure increases to find a more robust solution.
?
Symptoms might be a bad diode at D108 or D109, a fried processor, fails to work after jiggling the power connector,
fails to work after accidentally shorting a power supply with a probe, fails to work after multiple short power button presses.
?
As I previously reported, I have some linear regulators on a one inch square scrap of vector board that plugs into JP105,
this seems to work well after some minor mods to the QMX.? Specifically, need to lift the drain lead of Q106 and Q107 to
disable the old SMPS's, need to add 200k in parallel with R112 and R113 so the firmware thinks it's seeing a bit more than
3.3v and 5v (it hangs if it sees a wee bit less) and short drain to source across Q111 so the 3.3v linear can drive ADC_3V3.
?
I'd prefer that Hans change the firmware such that at power up it first checks if LIN_REG_EN is low, and if so skips all the
SMPS start up and voltage checks.? Then the only mod to the QMX would be to add a 10k resistor from LIN_REG_EN to ground.
?
I don't encourage others to repeat my efforts at this time unless this is simply something you enjoy doing.
Odds of blowing up your QMX are greater with my mods and hacks than they are with the old SMPS's.
Success is not guaranteed.? Reports of SMPS failures are no longer very frequent with Hans' latest firmware,
especially given that there are?thousands of QMX's in operation.??
?
As Hans has said, there are many different kinds of failures being reported in the forum.? The majority are due
to bad soldering, especially around attempts to burn the enamel off of the magnet wire.? Fortunately, most?
failures are easy to repair.? Unfortunately, an SMPS failure often blows the 100 pin STM32F446 processor.
?
I feel that the primary issue with the old SMPS's is that the voltage gets checked by the processor only once
every millisecond or so.? That's fine if everything works as it normally does, but something unusual can?
upset things faster than the processor can react.? For example, if the 5v rail is momentarily shorted to ground
by a scope probe, the processor might increase current to maintain 5.0v but not react fast enough when the
short goes away.? Designing an SMPS from scratch (or in firmware) is a difficult exercise in control system theory.
?
I just received my Kicad PCB fab for a one inch square board to replace my linear regulator vector board.
This design has a couple SOT23-6 SMPS IC's.? There's a huge number of different parts from different venders to choose from.
Max current varies from 500ma to 2A. They operate at frequencies varying from 0.5 MHz to 2.2 Mhz.? Cost can be as little at $0.10.
Provision is made to allow the board to be fully shielded using cheap copper foil, held up by L shaped header pins.
I am designing for the cheapest SMPS IC's which operate at an internally determined frequency, more expensive part
(still under $1) can be driven by an external clock source such as a processor PWM pin to eliminate any birdies.
The board can be separately built and fully checked out, including how it behaves with an intermittent or varying Vin, momentarily
shorted outputs, momentary low current demand, and changes in temperature.? It could be temporarily plugged into header pins
at JP105 of the QMX, or permanently soldered in place to those header pins.? If plugged in, I would recommend adding a
soldered wire connection for ground, since if that flakes out momentarily it could be disastrous for the QMX. ?
If it's successful, perhaps we of the forum can make it available as a kit.?
Or perhaps Hans would like to add it to his shop for purchase.
?
Jerry, KE7ER
?
?
I started this thread in response to this post ?/g/QRPLabs/message/139515
?
Re: QMX+ shutdown problem
On Sat, Feb 22, 2025 at 04:10 AM, Kevin (KQ4DTX) wrote:

Jerry, if your question was at me:? My QMX+ is Board V2 with firmware 1-00-027 when my occurrence happened less than a month ago.
?
 

Attached is a photo of my new SMPS board that plugs into JP105 on the QMX+
to replace the firmware controlled SMPS's for VCC and VDD.
An archive of the complete Kicad project including a zip file of Gerber's is at
Check out the attached Readme (also found in that files section archive).
Also a schematic showing how it has been stuffed.
?
If there is sufficient interest and if this can be accommodated in firmware,
then it should be possible to mass produce these cheaply.
There are some minor changes I would like to make, but this initial rev does function properly.
?
Jerry, KE7ER
?
?
On Sat, Feb 22, 2025 at 08:38 AM, Jerry Gaffke wrote:

I'd like to know the firmware and hardware rev from anybody who has trouble with the QMX 3.3 and 5.0v power supplies.
Mostly, I'd like to know how often these failures occur with the most recent firmware and hardware.
If that gets noted here in the forum, the pressure increases to find a more robust solution.
?
Symptoms might be a bad diode at D108 or D109, a fried processor, fails to work after jiggling the power connector,
fails to work after accidentally shorting a power supply with a probe, fails to work after multiple short power button presses.
?
As I previously reported, I have some linear regulators on a one inch square scrap of vector board that plugs into JP105,
this seems to work well after some minor mods to the QMX.? Specifically, need to lift the drain lead of Q106 and Q107 to
disable the old SMPS's, need to add 200k in parallel with R112 and R113 so the firmware thinks it's seeing a bit more than
3.3v and 5v (it hangs if it sees a wee bit less) and short drain to source across Q111 so the 3.3v linear can drive ADC_3V3.
?
I'd prefer that Hans change the firmware such that at power up it first checks if LIN_REG_EN is low, and if so skips all the
SMPS start up and voltage checks.? Then the only mod to the QMX would be to add a 10k resistor from LIN_REG_EN to ground.
?
I don't encourage others to repeat my efforts at this time unless this is simply something you enjoy doing.
Odds of blowing up your QMX are greater with my mods and hacks than they are with the old SMPS's.
Success is not guaranteed.? Reports of SMPS failures are no longer very frequent with Hans' latest firmware,
especially given that there are?thousands of QMX's in operation.??
?
As Hans has said, there are many different kinds of failures being reported in the forum.? The majority are due
to bad soldering, especially around attempts to burn the enamel off of the magnet wire.? Fortunately, most?
failures are easy to repair.? Unfortunately, an SMPS failure often blows the 100 pin STM32F446 processor.
?
I feel that the primary issue with the old SMPS's is that the voltage gets checked by the processor only once
every millisecond or so.? That's fine if everything works as it normally does, but something unusual can?
upset things faster than the processor can react.? For example, if the 5v rail is momentarily shorted to ground
by a scope probe, the processor might increase current to maintain 5.0v but not react fast enough when the
short goes away.? Designing an SMPS from scratch (or in firmware) is a difficult exercise in control system theory.
?
I just received my Kicad PCB fab for a one inch square board to replace my linear regulator vector board.
This design has a couple SOT23-6 SMPS IC's.? There's a huge number of different parts from different venders to choose from.
Max current varies from 500ma to 2A. They operate at frequencies varying from 0.5 MHz to 2.2 Mhz.? Cost can be as little at $0.10.
Provision is made to allow the board to be fully shielded using cheap copper foil, held up by L shaped header pins.
I am designing for the cheapest SMPS IC's which operate at an internally determined frequency, more expensive part
(still under $1) can be driven by an external clock source such as a processor PWM pin to eliminate any birdies.
The board can be separately built and fully checked out, including how it behaves with an intermittent or varying Vin, momentarily
shorted outputs, momentary low current demand, and changes in temperature.? It could be temporarily plugged into header pins
at JP105 of the QMX, or permanently soldered in place to those header pins.? If plugged in, I would recommend adding a
soldered wire connection for ground, since if that flakes out momentarily it could be disastrous for the QMX. ?
If it's successful, perhaps we of the forum can make it available as a kit.?
Or perhaps Hans would like to add it to his shop for purchase.
?
Jerry, KE7ER
?
?
I started this thread in response to this post ?/g/QRPLabs/message/139515
?

QMXPSr00.JPG
Readme
Readme
qmxps01.pdf
qmxps01.pdf

 

From the readme:? "Behaves well when output is suddenly shorted and then the short is removed."??
Try this:
The most brutal power supply torture test I saw was to connect the supply output with heavy leads to two 3/8" copper bolts closely spaced in an insulating block.
Energize the supply, then take a coarse file and file the ends of the bolts simultaneously.??
Watch the sparks fly and monitor the supply for instabilities. ?The fast, random short/open transitions, with bonus negative-resistance arcs, will challenge any control loop.
I love getting paid to do stuff like that.
73, Don N2VGU
 

Don (aka El Destructo),

Can you move that file at 700,000 files per second? Lol… your job sounds like a ton of fun. Hope you’re using the proper PPE. :-)

Tony AC9QY

On Thu, Apr 3, 2025 at 3:19?PM Donald S Brant Jr via <dsbrantjr=[email protected]> wrote:
From the readme:? "Behaves well when output is suddenly shorted and then the short is removed."??
Try this:
The most brutal power supply torture test I saw was to connect the supply output with heavy leads to two 3/8" copper bolts closely spaced in an insulating block.
Energize the supply, then take a coarse file and file the ends of the bolts simultaneously.??
Watch the sparks fly and monitor the supply for instabilities.? The fast, random short/open transitions, with bonus negative-resistance arcs, will challenge any control loop.
I love getting paid to do stuff like that.
73, Don N2VGU
 

Don,
?
These are pretty dinky supplies, no need for your 3/8" copper bolts.
But the file certainly makes for lots of crud to study.
?
I loaded up the supply with 50 ohms on 5v, 25 ohms on 3.3v (100ma and 132ma respectively).
When scraping the 12v source to the supply across a file, I saw an occasional triangular peak
on 5v that got up to about 5.7v, the base of that triangle was about 100us.
This was overshoot as the supply recovered from a quick dip in 12v, and only when the?
5v rail didn't go below about 3v.? Scope photo attached.
?
This is exactly the case that was killing QMX's when 12v power was intermittent,?
though I can guarantee the overshoot then was far far worse.? That has mostly
been addressed through firmware changes, though I suspect issues remain.
?
In my Readme I say:
> Next rev of this board will have:
>? Larger inductors than the now 3x3mm, with pads large enough to do rework.
>? Inductors (1uH?) added at input to each switcher to reduce noise on supply.
?
Larger inductors, both physically larger and more uH.
Because the current through the inductors now rises and falls very fast,
it very nearly reaches 0 current in the valleys on VCC.? Planning to go from 4.7uH to 22uH.
That will slow the rate at which current rises, and I'd expect it to cure the spike.
Adding the 1uH up front of the switchers (cap still across the switchers Vin) should also
help, meant primarily to reduce noise on the incoming rail but may also help avoid
sudden dips on Vin to the IC's.
?
I also tried using the file as a load on both VCC and VDD, going from no load to 0 ohms.
Again saw spikes, the worst was still around 5.7v peak on VCC.
?
The 3.3v rail was relatively clean in all cases, in part because it is supplied from VCC.
?
Anyways, if there's enough interest and I have a few of these auto-stuffed,?
I'll plan on shipping a couple off to you for further tests.?
Also to Jeff Moore and Hans if they are interested.
These inductors are short and robust, the boards are small and cheap.
I'm confident the boards can go out as first class letter mail here in the US.
?
Attached is a scope photo of the VCC spikes when under load and?
running the 12v supply across a file.
?
Switching at 2mhz and a 100ma load on Vcc, 130ma on Vdd (typical for a QMX),
the efficiency is around 84%.
I don't expect switcher noise to be an issue, if so we can reduce it to around 500khz
by using a different variant of these IC's and significantly improve efficiency.
We may not quite reach the efficiency of the old scheme which operates at 100khz
but I for one am more interested in a robust design than the last few percent in efficiency.
?
Jerry, KE7ER
?
?
On Thu, Apr 3, 2025 at 01:18 PM, Donald S Brant Jr wrote:

From the readme:? "Behaves well when output is suddenly shorted and then the short is removed."??
Try this:
The most brutal power supply torture test I saw was to connect the supply output with heavy leads to two 3/8" copper bolts closely spaced in an insulating block.
Energize the supply, then take a coarse file and file the ends of the bolts simultaneously.??
Watch the sparks fly and monitor the supply for instabilities. ?The fast, random short/open transitions, with bonus negative-resistance arcs, will challenge any control loop.
I love getting paid to do stuff like that.
73, Don N2VGU

QMXPS_Vcc_Scope.JPG

 

To be clear, that spike to 5.7v under worst case 12v supply interruptions would not have damaged anything on the QMX+.
Also, the QMX+ has 470uF on VCC at C107, when I repeat the test with 470uF added the spike is pretty much gone (about 0.1v).
This board would work safely in the QMX+ as originally built.
?
My board did not have any ringing on normal power up, only with extremely short and sharp peterbations on the +12v supply.
It could be that most SMPS designs have some minor trouble here, not a condition that is normally tested for.
?
Just increasing the VCC inductor from 4.7uH to 10uH did not reduce the spike much, it needed the larger 470uF capacitor.
?
Performing these tests where load or supply are interrupted by scraping a wire across a file would likely kill?
a stock QMX, even with the latest firmware.? Also, it would be a difficult test to conduct without risking the STM32F.
This new board has the advantage of being fully testable without being tied to a QMX.
?
Designing a robust feedback loop on a switching power supply from scratch is not trivial.
There's a huge range of fault conditions to account for, and tradeoffs to be made between quick response and undesired ringing.
?
Jerry, KE7ER
?
?
?
On Thu, Apr 3, 2025 at 05:26 PM, Jerry Gaffke wrote:

I loaded up the supply with 50 ohms on 5v, 25 ohms on 3.3v (100ma and 132ma respectively).
When scraping the 12v source to the supply across a file, I saw an occasional triangular peak
on 5v that got up to about 5.7v, the base of that triangle was about 100us.
This was overshoot as the supply recovered from a quick dip in 12v, and only when the?
5v rail didn't go below about 3v.? Scope photo attached.
?
This is exactly the case that was killing QMX's when 12v power was intermittent,?
though I can guarantee the overshoot then was far far worse.? That has mostly
been addressed through firmware changes, though I suspect issues remain.
?

1 - 6 of 6