Keyboard Shortcuts
ctrl + shift + ? :
Show all keyboard shortcuts
ctrl + g :
Navigate to a group
ctrl + shift + f :
Find
ctrl + / :
Quick actions
esc to dismiss
Likes
Search
QMX+ V2 - I've killed it
|
Thanks for the reply Hans.
?
I think we are going to have to agree to disagree on this.
?
I agree that when this is working as expected, with nothing going wrong, it's fine.? But I do worry about the 1 mS update rate on the SMPS. A lot can happen in that amount of time, 116 PWM pulses for example.
?
Yes, describing the Zeners as fuses is loose terminology, they are more like anti fuses.? They are intended to go low resistance to protect the circuit.? The point I'm making is that they are expected, in extremis, to fail in a way that protects the rest of the circuit.? This seems to happen more frequently than I'd be comfortable with.
?
One of the most common hardware problems that people report is power supply failures, there may be multiple reasons, it may be their own fault. But it seems to happen a lot. At least the supplies seem to be a little fragile.
?
Building the supplies onto the QMX+ main board may help with avoiding construction problems but the supplies haven't changed, they are just as vulnerable to failure but now are more difficult to fix.? It would be possible to use the connections to the dev board to add separate regulators but there's still the problem of disabling the existing supplies, Just unplugging a faulty supply and plugging a good one in isn't possible.
?
I'd rather you devoted your talents to clever innovations such as SSB rather than reinventing the switching regulator.
?
Chris, G5CTH |
|
Hi Chris ?
Yes yes, perfectly OK :-)? ?I understand and respect your position.?? ?
I think a lot CAN'T happen, because the 330uH inductor and 470uF capacitor are BIG. They provide a lot of momentum in relation to the 1ms loop time.?? ?
Actually I'm not expecting them to fail. Failing zeners wasn't my intention. It wouldn't be a reliable way to protect the circuit because while initially they fail short, they're consuming a lot of current and eventually I think other things will start to burn up. My expectation was that in the face of unexpected, un-designed-for events, the zeners would clip any spikes to avoid damage. NOT that they should be driven to the point of failure. I still have not seen any reliable reproducible indication that this occurs, other than by some kind of unreasonable input or constructor failure.?? ?
It could be that they are fragile; but the other possibility is that they are tricky to install, tricky to avoid shorts, get the mechanical alignment correct etc.? ?
That is true, based on the assumption that the vulnerability to failure is due to electrical fragility, a fragile design. If that is NOT the case but the difficulty of assembly and installation is the cause of vulnerability, then having them already on the QMX+ main board does reduce the problems.? ?
AGREED!? 73 Hans G0UPL |
I didn't specify the QMX for direct connection to 13.8 or carsupplies, that's very clear in the documentation. A lot of other equipment has a voltage rating. I know, and in fact I stated that "I am not saying that QMX/QMX+ will or it should...", no misrepresentation on your side, of course. Anyone who has visited my lab knows that it's a terrible mess.At least you could enter in your room...mine can be accessible only if you work in Cirque du Soleil... For running directly from a car, if nervous, why not build a QMX for9V operation and install a 9V converter inline? Or build it for 12V and put some series diodes in to drop the voltage from the car. No plan to put it in a car, but I don't like the possibility even remotely to fry it just because power cord gets disconnected at the wrong time. I don't like the 9V route because should always have a dedicated device to connect it to 12V, that device could be something? incorporated into a QMX+ that is not so tight in space as the QMX, just saying... I just bought a QMX+, and I am tempted to take? an interesting challenge (provided I will be able to assemble it): - setup a quick torture testbench with some relays and one dummy load, one for abruptly providing/cutting the power(12V from a Agilent/Daiwa/el cheapo, your choice), another for the on/off encoder switch and another for one paddle input, plus other to mimic the encoders maybe. If I pull the plug 100 times in a couple of hours, at random while receiving, booting, transmitting, navigating the menus or shutting down are you confident it will survive? Giuseppe Marullo IW2JWW - JN45RQ |
|
Hello Giuseppe Go for it! If you break it I'll send you a new one.? 73 Hans G0UPL On Sat, Oct 26, 2024, 14:38 Giuseppe Marullo[iw2jww] via <giuseppe=[email protected]> wrote: ?>I didn't specify the QMX for direct connection to 13.8 or car |
开云体育Challengge accepted! ? Waiting for it to come, probably I could use a NodeMCU for driving the relays. Xmas project, I may be abroad for some time, not sure yet.
? From: [email protected] <[email protected]> On Behalf Of Hans Summers
Sent: Sunday, October 27, 2024 3:06 PM To: [email protected] Notification <[email protected]> Subject: Re: [QRPLabs] QMX+ V2 - I've killed it ? Hello Giuseppe ? Go for it! If you break it I'll send you a new one.? ? 73 Hans G0UPL ? On Sat, Oct 26, 2024, 14:38 Giuseppe Marullo[iw2jww] via <giuseppe=[email protected]> wrote:
|
|
On Sat, Oct 26, 2024 at 04:37 AM, Giuseppe Marullo[iw2jww] wrote:
I just bought a QMX+, and I am tempted to take? an interesting challenge (provided I will be able to assemble it):I was thinking of doing the same thing, but not being random, choosing sequences of power switching using something like a MOSFET as a drive controlled by a microcontroller.? Try sequences of cycling the power at 1 millisec intervals or less and various on/off ratios for a second, then applying full power.? It's the sort of thing that an intermittent contact followed by a good contact will give.? Maybe just a power interrupt for times around a few millisecs, possibly with a few hundred uF to help hold the power up. ?
Use a really low resistance power supply, a good battery for example.
?
Hans wants a reproducible way of destroying these supplies, this could do the trick
?
Chris, G5CTH. |
开云体育Chris, I have several power supplies, but nothing too extreme. One switched? Lafayette, about 25A countinous, a traditional Diamond 25A, a couple of Agilent (not lab grade I assume AFAIK 3A top),? and a GPIB programmable one (don't remember the brand). Last one could be the one needed, unluckily it is the least
powerful but could be "scripted". Using A Amica V2 NodeMCU (or a Teensy 4.1) and relays, there is a limit in terms of ms of what can be implemented. I don't know much about MOSFET,s should you help with that I
could implement it using MOSFET for the power supply, or even also
substitute all the relays. My Idea: MCU--> Opto --> Driver --> Relay Be my guest and we could arrange something fancier. I need also to name the beast,so far it will be QMX Nemesis... >I was thinking of doing the same thing, but not being random,
choosing sequences of power switching using something like a
MOSFET as a drive Giuseppe Marullo On 10/28/2024 4:42 PM, Chris wrote:
|
|
Whatever you guys decide to do, please make sure you know what type of power “glitch” took out exactly what part of the QMX. Since you’ll likely sacrifice a QMX to the magic smoke gods, provide Hans with a clear scenario so hopefully he can fix it in a future firmware revision. Tony On Mon, Oct 28, 2024 at 2:13?PM Giuseppe Marullo[iw2jww] via <giuseppe=[email protected]> wrote:
|
|
Could get the same information with just the separate 3.3v switcher board and a power resistor to take a typical QMX load from it.
No need to burn up a perfectly good rig.
?
Jerry, KE7ER
?
?
On Mon, Oct 28, 2024 at 12:36 PM, Tony Scaminaci wrote:
|
|
I performed a similar experiment on 15 3.3V SMPS boards, but with no way to simulate the corrections from the MCU, it was inconclusive. What I did find was that any current above 100 mA through Q111 and D109 allows the 3.3V output voltage to climb as high as 4.2V *without* melting the diode. I stopped at that current level so I can’t say how much higher the voltage could climb before D109 actually melts and (hopefully) shorts out. Regardless, 4V is the absolute maximum limit for the MCU 3.3V bank so it’s clear that the MCU could be damaged by overvoltage before D109 actually performs its intended protection function. What Giuseppe is doing is testing the resilience of the QMX during power supply glitches caused by a loose power plug, a plug accidentally pulled out, etc. The goal is to quantify how these glitches are being handled by the firmware’s correction algorithm and hopefully determine where failures are occurring as a result. Tony On Mon, Oct 28, 2024 at 2:59?PM Jerry Gaffke via <jgaffke=[email protected]> wrote:
|
|
Just an engineering guess based on our 40+ years of experience in blowing up radios at our respective companies Hans. I called it “learn by doing” but my managers didn’t find that phrase humorous. Still “better it fail in our engineering labs than in the field at a customer site”. A direct quote from the founder, CEO, and Chairman of the Board at Motorola, Robert W Galvin. Tony On Mon, Oct 28, 2024 at 4:07?PM Hans Summers via <hans.summers=[email protected]> wrote:
|
|
?
Tony wrote:
"I performed a similar experiment on 15 3.3V SMPS boards, but with no way to simulate the corrections from the MCU, it was inconclusive."
?
Easy enough, just power the MCU from a 3.3v linear, but have wires from the SMPS to the MCU for PWM_3V3 and ADC_3V3.
?
?
Hans wrote:
"Why are you guys so sure anything's gonna fry!"
?
I'm not convinced it will fail, but if they are whacking at it 100,000 times per day using relays
to put various jumps on Vin and the load current,? that's a pretty severe test.??
Your confidence is inspiring.
?
Jerry, KE7ER
? |
|
I left some info out… I didn’t have any actual radios, just 50 supposedly bad SMPS boards Jeff Moore sent me to troubleshoot. Of those, about half were returned to him unscathed. The remainder (9 3.3V and 11 5V) are still being looked at in more detail but I do have one 3.3V SMPS with a shorted D109 and one 5V board with a shorted D108 (1/2W before the switch to 5W). I found one 5V with a blown out 470 uF cap (6.3V before the change to 10V). As for the rest, I’m not clear on what’s blown yet. Tony On Mon, Oct 28, 2024 at 4:38?PM Jerry Gaffke via <jgaffke=[email protected]> wrote:
|
|
Hi Jerry,
Could get the same information with just the separate 3.3v switcherboard and a power resistor to take a typical QMX load from it. No need to burn up a perfectly good rig.I just ordered one QMX+ PCB Rev 2, it has the integrated power supplies now, so this is not possible. I don't know if Hans had handed it over to TNT, it says shipped but not yet collected. If it is not too late, he could put into the package a couple,? I have no use for them afterwards (both outcome, with or without magic smoke) but that could? be an interesting test as well. Maybe this is something Hans could? do? Nah, SSB implementation über alles! Tony, Whatever you guys decide to do, please make sure you know what type ofpower “glitch” took out exactly what part of the QMX. Since you’ll likely >sacrifice a QMX to the magic smoke gods, provide Hans with a clear scenario so hopefully he can fix it in a future firmware revision I need to create a testbed script, and maybe a video recording that will show testbed n., LCD output and SWR meter? in the video, else I would need to be there if/when the magic smoke will come out. Hans, Why are you guys so sure anything's gonna fry!We don't know, but this should shed some light: - QRPers against couch potatoes, squeezing the last mA from the battery risking the magic smoke or using more traditional old school 78XX...this is the question. I am open to suggestions but I don't have Keysight's budget, so keep that in mind. Giuseppe Marullo IW2JWW - JN45RQ |
|
I believe I have already put my QMX+ through the torture test!
After constructing it from a kit and then firing it up for the first time without any smoke I proceeded to put it through all the hardware tests using the terminal. All good until I got to the SWR sweep and results were not as expected, the terminal locking up and a reboot of the QMX was required. Tried several times still nothing.
It so happens that my computer is on another bench from the test bench so I had my back to the QMX whilst testing. I had a look over my shoulder during one of the sweeps and to my horror there was the QMX display pulsing on and off at each transmit test point as the fold back on my linear power supply dutifully limited the current to the 250ma I had set it for during initial startup. Panic! A quick reset of the power supply and the SWR sweeps were all good.
Ok, perhaps not the ultimate torture test as I was running around 8v on a 12v build and the SWR sweep does limit power by design. But, it did demonstrate to me the design is robust and the QMX has since been operating successfully on FT8 and WSPR.
Nigel, VK6NI? ? |
|
Giuseppe,
?
Do as you wish.
Hans seems confident you won't get it to fail when doing normal things
that a rig would be subjected to, such as momentary disconnections from the power supply.
Could just perform the test like that.
?
But easy enough to run an extended 3.3v SMPS test without endangering the rig,
and record everything that happens to a disk file using a PC soundcards app.
?
If I were doing this, I'd remove Q111, Q108, and D103.
Buy a bunch of 3.6v zener's, and put one in parallel with C106.
Add linear regulators (with heat sinks) for Vdd and Vcc from an always on power supply
so the STM32F is always taking care of the PWM_3V3 and ADC_3V3 signals.
Pur a load resistor across C106 to draw a current about equal to what the rig normally draws.
?
Now rapidly vary the Vin voltage with step changes,
and perhaps vary the load across C106 with step changes too.
Record Vin, the voltage across C106, and the current to your load,
perhaps using PC soundcard mic inputs.
Could control Vin and the load current using relays or nFET's and a RaspberryPi.
?
Jerry, KE7ER
?
?
On Mon, Oct 28, 2024 at 03:54 PM, Giuseppe Marullo[iw2jww] wrote:
Hi Jerry, |
Daniel Conklin
to navigate to use esc to dismiss