For some time (several months) I had what I thought was some intermittent
QRM, where the band was covered in spikes spaced exactly 100 Hz apart.

Here's how it presents itself in the WSJT-X (there are differences on the
waterfall, because I was switching bands, and between digital and CW
modes).

http://arnold.ziffel.one.pl/qmx-wsjtx-swipes.png

When it happens, I can still see the band activity below it.

It came and went and I didn't think much of it, but today, when testing a
new antenna for another band, I noticed the same QRM and started digging.

Here's what I found:

In a speaker it manifests as buzzing. See the video (and pardon the mess
on a desk):

https://youtu.be/k8H-BZrNf8I

It doesn't change when switching bands. It does change (but not go away)
when switching between digi and CW. The same change happens when I enter
diagnostics menu (maybe QMX internally switches to CW when entering this
menu).

It doesn't come from antenna (tested with a disconnected antenna).

It doesn't come from the power supply (tested on batteries).

It doesn't come from a nearby RF field (tested by walking with QMX around
the house without any change).

So it's internal to the QMX.

I started poking and probing, looking for a bad solder joint or something
else intermittent. Everything seemed fine. Then I turned on the
oscilloscope and started poking around the input circuity.

I found some weird ~25 MHz oscillations on outputs and inverting inputs of
two op-amps (two other were fine), both in the same chip (IC406, LM4562).
Look (and pardon the mess once again):

https://youtu.be/SyRv2tueQlk

Mind you, it's not present on any noninverting input.

My plan for now is to desolder this IC and see if it goes away (maybe
PCM1804 is injecting something -- the signal is also present on its VINR+
and VINR-, not present on VINL+ and VINL-), maybe replace this IC, but
it's still strange and puzzling to me.

Why is it intermittent?

25 MHz is the system clock, the output QRM is quite stable, which might
suggest that this signal is in phase with the system clock -- is it
related?

Maybe it's some bad decoupling somewhere, or some other element failed and
the op amp is good?

What am I missing?

Bruce Akhurst <bruce@...> wrote:

So if you have a 2 channel scope verify if the 25MHz is in perfect phase
with the system clock.   If so I'd be looking for a solder blob or a
faulty decoupling capacitor

It is.

Curiously, this signal is there even when the radio receives clearly (the
issue is intermittent, but 25 MHz is there all the time).

I desoldered the op-amp and it stopped the oscillations, so they're not
injected by the PCM1804. Op-amp must be picking them up, maybe from a gate
amplifying the clock signal (IC202) that's very close to it.

I wanted to add some decoupling capacitors here and there, but my scope
probe slipped, shorted something, the board let out a magic smoke (I
couldn't figure out which element smoked) and now it doesn't turn on.

When I press and hold the ON switch, PWR_ON appears, +12V appears,
LIN_REG_EN appears, but VDD doesn't, other signals also don't. 78M33 is
fine.

I'll be diagnosing it further. I just hope the CPU is fine. If not, well,
I guess that's it... eh.

On more than a few occasions I have found that a scope
probe will excite a high speed, high impedance op amp
circuit into oscillating.

-Chuck Harris




On Sun, 29 Dec 2024 17:48:12 -0800 "Stan Dye via groups.io"
<standye@...> wrote:

A probe slip could easily short +12v to Vdd, and it would blow the
D109 zener.  Check D109 by carefully lifting one end of it, and see
if the QMX powers on correctly.

But don't probe around much if it does power up without D109, since
that diode is your protection from blowing the processor and other
chips.  Need to replace it first.

Stan KC7XE

Stan Dye via groups.io <standye@...> wrote:

A probe slip could easily short +12v to Vdd, and it would blow the D109
zener.  Check D109 by carefully lifting one end of it, and see if the
QMX powers on correctly.

Sadly, it's not a diode (nothing changed when I removed it).

Now I suspect Q101 or Q102 (will test them probably tomorrow), because
when I press the button, 12V appears, LIN_REG_EN appears (5.2V), but
measuring on the regulator (78M33):

- Input: 0 -> 0.9V
- Output: 0 -> 0.3V
- Ground: 0 -> 0.3V

78M33 is OK (I tested it after desoldering it). So either something is
drastically pulling it down, or it's one of these transistors (or maybe
C102?)...

I think I'll just remove this board and test it off-circuit (just provide
it 12V, 5V for LIN_REG_EN, and test VDD), and go from there...

Other measurements, just for completeness (D109 removed, button released
-> pressed):

Module 1:

- PWR_ON 12V -> 0
- PWM_5V 0
- V_IN 12V
- HOLD 0
- ADC_5V 0
- VCC 0
- +12V 0 -> 12V

Module 2:

- PWM_3V3 0
- LIN_REG_EN 0 -> 5.2V
- +12V 0 -> 12V
- VCC 0 -> 0.285V
- ADC_3V3 0

Thanks,
Adam

If you shorted Vdd to 12V, the processor might've gone. What's the resistance from Vdd to ground?

Paul -- AI7JR

Stan Dye via groups.io <standye@...> wrote:

A probe slip could easily short +12v to Vdd, and it would blow the D109 
zener.  Check D109 by carefully lifting one end of it, and see if the 
QMX powers on correctly.

Sadly, it's not a diode (nothing changed when I removed it).

Now I suspect Q101 or Q102 (will test them probably tomorrow), because 
when I press the button, 12V appears, LIN_REG_EN appears (5.2V), but 
measuring on the regulator (78M33):

- Input: 0 -> 0.9V
- Output: 0 -> 0.3V
- Ground: 0 -> 0.3V

78M33 is OK (I tested it after desoldering it). So either something is 
drastically pulling it down, or it's one of these transistors (or maybe 
C102?)...

I think I'll just remove this board and test it off-circuit (just provide 
it 12V, 5V for LIN_REG_EN, and test VDD), and go from there...

Other measurements, just for completeness (D109 removed, button released 
-> pressed):

Module 1:

- PWR_ON	12V -> 0
- PWM_5V	0
- V_IN		12V
- HOLD		0
- ADC_5V	0
- VCC		0
- +12V		0 -> 12V

Module 2:

- PWM_3V3	0
- LIN_REG_EN	0 -> 5.2V
- +12V		0 -> 12V
- VCC		0 -> 0.285V
- ADC_3V3	0

Thanks,
Adam

-- 
Paul -- AI7JR

Adam via groups.io <qrp-labs@...> wrote:

Sadly, it's not a diode (nothing changed when I removed it).

An update.

The diode (D109) was in fact damaged, but not only the diode. Something
else on the 3V3 board was wrong as well. When trying to diagnose it, a
capacitor on the board blew up (I'm almost sure I wired everything
correctly... turns out I didn't), and as I first wanted to quickly test if
the radio works, I made a linear regulator, provided artificial ADC_3V3
signal just to satisfy the CPU, and it works. That's great.

Another issue is that L101 (the one on the 5V board) gets very hot (it got
to 110°C before I stopped the test). D108 (protection diode on the 5V
board) was damaged as well, but when I removed it, nothing changed. Radio
works, L101 gets insanely hot, and I'm afraid of running it for too long
(longer than ~20 seconds it takes L101 to reach this temperature).

Any ideas on what else could I check?

Or maybe this temperature is normal (I highly doubt)?

Stan Dye via groups.io <standye@...> wrote:

 If your receive is working properly, it's probably not a component on
the main board - most of the receive section components are on Vcc. 
Does receive still work?

Hard to tell. There's normal noise, so it probably works at least
partially, but I don't want to stress the board more...

I would measure the main-board resistance between the 5V rail and gnd,
without the power supply in place.  This should have several Kohms of
resistance, if not, there is a failure somewhere on the main board. 
Hopefully it is OK and the only failure is in the 5V supply.

23 ohms, measured on the main board with power supply boards removed.
No wonder the inductor is getting bloody hot.

I isolated it by lifting the Vcc pins of the ICs one at a time until the
short disappeared.  But this IC being damaged also destroyed all signal
receive functionality.

I think this is what I'll do -- make a list of possible components and
just disconnect them one at a time, measuring the resistance.

An option to consider is to order a set of the little power supply
boards from qrp-labs (on their spare parts page) - they are very
inexpensive, and may save you some troubleshooting and repair.

When (if) I fix the issue with the short on the main board and verify that
everything works, I'll think about it -- right now I have a linear
regulator instead of a 3V3 buck one, which seems to work just fine.

Another option I considered is designing a custom board with just a power
switch and two linear regulators instead of these two boards. I don't like
the idea that the control loop goes via the CPU. Of all design decisions
in the QMX, and despite the trick with adjusting the switching frequency
so the harmonics don't interfere with reception, I'm not a big fan of this
solution.

Thanks!