95 replies to this topic

[mlord]

There have been more than a few reports lately, of motor failures on Evolution mounts.  Today, I had opportunity to dig a bit into these reports, with a disturbing conclusion.  Read on!

 

At some point around 2020, Celestron began shipping Evolution mounts with the "Rev.K" main PCB (electronics) inside.  And anyone sending an Evolution to Celestron for "repair", would most likely have received it back with a brand-new "Rev.K" main PCB installed.  The Rev.K features the 3rd-gen WiFi revision for the Evo, along with other changes to the main PCB (electronics) of the mount.

 

One "feature" of Rev.K is becoming rapidly apparent:  a high frequency occurrence of failures to the L293DD motor driver chip on the board.  As in, a REALLY HIGH frequency of occurrence.  Scour posts in these forums over the past 5-6 months, and you'll find a number of them.  And these forums are only the tiny tip of a very large iceberg when it comes to owners/users of Celestron mounts, including the Evolution.  Here is the Rev.K, Evolution mainboard:

 

 

A month or so ago, I acquired this Rev.K from another CN user who had experienced failure of the L293DD chip.  They simply purchased a replacement board from Celestron, and sold me the "dead" one.  I replaced the faulty chip here, and all was good.

 

Then, other users reported similar failures.  And one of those users is from Ontario Canada, where I also reside.  So I invited him to visit along with his broken Evolution mount.  He had a different issue with it originally, and had sent it at great expense to Celestron for out-of-warranty "repair".  They simply replaced all of the electronics with "new", including swapping his older Rev.F mainboard for a fresh Rev.K mainboard, and shipped it back to him.  That's when the real problems began.

 

The "repaired" mount quickly failed, with one of the motors no longer responding.  Dead L293DD chip, for sure.

 

So.. today, we finally got together.  We opened up the mount, verified everything else inside was in great condition, and so removed the PCB to replace the dead chip.  Fine.  Reinstalled it, and then did an unusual thing:  unlike Celestron, we actually tested it!  We hooked up a StarSense AutoAlign, and told it to go ahead and do a full automatic alignment.  Indoors, daytime, so it never actually succeeds, but keeps trying.

 

Now, StarSense AutoAlign is a bit different than a regular hand-controller alignment.  When one uses a Nexstar+ hand-controller, most people slew in a single direction at a time, and then perhaps remember to slew back slightly using UP and RIGHT for final centering of an object.  Whereas StarSense does it all by itself, slewing in both directions at the same time, and then quickly doing the UP and RIGHT at the end of each slew.

 

Within the first 30-seconds of the StarSense AutoAlign, the newly-replaced L293DD chip went.. POOF! .. up in smoke.

 

Crap.  Bad chip, I guess.  Having a bag full of them on hand, we just replaced it, again.

And.. within 30-45 seconds of testing, POOF! .. up in smoke, AGAIN. 

 

One definition of insanity, is repeating the same action over and over, expecting a different result, despite strong evidence that the same result will repeat itself.  Which in this case, it did.  POOF! yet AGAIN!

 

Well, no problem, perhaps just a bad PCB then.  So I got out the other, recently repaired Rev.K Evolution PCB, and we ran the same test on it.  POOF! within 30 seconds.  Whoa.. that's not good. 

 

The first unit was running the latest 7.18.5028 firmware.  The second unit was still on the older 7.17.0098 firmware.  It made no difference.  We repeated the test again, with a heat-sink added to the chip.  POOF!  Again.

 

It wasn't until much later, that I thought the evidence should be recorded, so I dumpster-dived into the trash can, and found most (but not all) of the discarded chips from our session.

 

 

Having fried more than a few of them, we switched to an older, Rev.F PCB, and ran the same StarSense "stress test" on it.  After perhaps 30 minutes of no failures, we stopped it.  This unbothered Rev.F board happened to be running the latest 7.18.5028 firmware , same as the first repeatedly failed Rev.K board.

 

Many mounts, not only Celestron ones, sometimes (rarely, but it happens) experience L293DD failures.  But not like this!

 

EDIT: Update: Celestron are aware of this, and working on the issue.  Hopefully there will be a firmware fix at some point.

 

Conclusions:

• Rev.K PCBs can quickly blow out the L293DD chip -- very easy to reproduce.
• Firmware version makes no difference.
• The older Rev.F PCB has no such issue.

Advice?

• If you have a Rev.K Evolution within its return period, wait for a fix, or RETURN IT.
• If you have an older revision that has a fixable failure, do NOT send it to Celestron for repair, because they may replace it with a Rev.K.
• If you don't yet own an Evolution (excellent mount, by the way), don't buy a new one for now.

 

EDIT: A particularly DANGEROUS aspect of these failures, is that the Evo has an internal battery pack, which continues to feed high-current to the PCB after the L293DD chip fails (POOF!).   So the chip continues to burn and smoke internally for a while, and the battery continues to discharge at a high rate, until the battery itself finally dies.  And without dismantling the tray area of the mount, there's no way to disconnect the battery to prevent this.

 

The "power" switch on an Evo does not actually control the power feed -- it's just a signal to the firmware to turn stuff on or off, which only works if the firmware is still up and running, which in this case it is not.   And Celestron don't seem to have discovered protective "fuses" yet.

Edited by mlord, 18 April 2025 - 05:08 PM.

[Xeroid]

ouch  

[DMR6124]

Not good news!!!

 

I hope Rev L boards don't have the same defect - just ordered one from Agena.

[mlord]

Ahh... Rev.L  so there's some hope!

[mlord]

We did spend some time examining the Rev.K PCB next to the Rev.F, under magnification to see the layout and components more clearly.  There have been some subtle changes for the Rev.K, including omission of a U3 "transient suppression" chip for the outputs of the rotary encoders:

 

 

So at one point, we transplanted the "missing" chip from a Rev.F donor board to one of the Rev.K boards.  At first, this appeared to "fix" things.  The stress test took perhaps 10-minutes before the inevitable POOF! repeated itself.  So, some improvement, but not enough.  There must also be other relevant changes that we haven't found.

Edited by mlord, 18 April 2025 - 09:54 AM.

[mlord]

One more thing on this topic.  Our stress test was always done with the mount's clutches disengaged, which does affect the power/current curves for the two motors.  One would expect this to present less of a load than it would with clutches engaged, but it could also affect how long the motors' "free run" after movement is "stopped" at the end of a slew.  Either of those things could be a factor here.

 

The working theory here, is that the MC firmware is reversing polarity to the motors, for a change in movement direction, without first ensuring that they have 100% ceased motion from the opposite direction.  This rookie error is the easiest way to kill the L293DD chip.  Motion is detected through monitoring the rotary encoders, thus the potential significance of changes in related circuitry there.

 

My own mount is also an Evo. Rev.K, and it is always used with StarSense, and thus far has survived for several years.  I suppose I'm going to have to run this exact stress test on it now, both with and without the clutches engaged.  But with the tray area opened up in advance, so I can more quickly disconnect the battery when the L293DD goes POOF!  It's main Rev.K PCB dates from early 2020, whereas the other two we were testing are likely from a more recently manufactured batch.  There could be slight component variations between those, without any intentional board changes.

Edited by mlord, 18 April 2025 - 09:55 AM.

[mlord]

Today I re-ran the simple StarSense "stress test" to deliberately fry another L293DD chip, for the purpose of recording the event (video with sound) to pass along to Celestron.  This time it took just over 2.5 minutes to go POOF!   Except there was no smoke because I pulled the power plug as soon as the mount shut down.  The video isn't that interesting except to people working on the problem, so I haven't made it generally available.

 

After replacing the L293DD chip yet again, I left the (14.5V) VS pin up off the pad, and soldered a wire to it, and another to the pad.  Then I used the two wires to insert my ammeter into the circuit, to measure the total current flowing through the L293DD.

 

According to my meter, the max current (both motors accelerating to speed 9) was 0.369 amps, well within margins for the L293DD.  So motor current draw is not the issue on my test-bed, at least.

 

EDIT: I should have also done the measurement in a real mount with clutches engaged, for higher motor load.  Now on the to-do list for later. 

Edited by mlord, 20 April 2025 - 07:32 AM.

[Martinbruce]

This is scary stuff when these EVO scopes run 2k plus. Furthermore, I got mine post 2020. I’ll be checking on the cost of REV L boards.  Thank you for all your electronic snooping. Not an easy thing to do. 

[HalfEazy]

So I made a post about the firmware update recently. I had my evolution mount replaced by celestron because it was only 2 months old. They sent me a new mount with the firmware update. I am now getting this message everytime I try and connect.

 

https://imgur.com/a/ut1HqqP

[mlord]

I’ll be checking on the cost of REV L boards.

Hold off on that for now.  I think the Rev.L ones are also affected.

 

[mlord]

I am now getting this message everytime I try and connect.

That message is regularly seen by people new to the SkyPortal/SkySafari apps.

 

It means your phone/tablet has probably switched back to a different WiFi, one with internet access, rather than the Celestron WiFi.  This is so they can track you better.

 

Please do NOT post further about that in this thread.  This thread is specifically about the Evolution Firmware burning up the L293DD motor driver chips, not "how to use the apps".

Edited by mlord, 20 April 2025 - 07:35 AM.

[rrpallechio]

How do I determine which Rev.x I have?

[Martinbruce]

Here is the REV L board on Agena Astro. Not cheap and with the current tariff environment hard to say what the cost will be going forward. Cheers

 

 

https://agenaastro.c...er-f00-004.html

[HalfEazy]

connection_failure.jpg

That message is regularly seen by people new to the SkyPortal/SkySafari apps.

 

It means your phone/tablet has probably switched back to a different WiFi, one with internet access, rather than the Celestron WiFi.  This is so they can track you better.

 

Please do NOT post further about that in this thread.  This thread is specifically about the Evolution Firmware burning up the L293DD motor driver chips, not "how to use the apps".

I'm not new to the app. I thought this might be related in someway because of how u were explaining the wifi is connected internally

[mlord]

How do I determine which Rev.x I have?

If the built-in WiFi calls itself "SkyQLink-xx", then you most probably have a Rev.F PCB.

If it calls itself "Celestron-xx" (two hex digits), then I think (not sure) it is a Rev.J.

If it says "Celestron-xxx" (three hex digits), then it's Rev.K or higher.

 

Cheers

[mlord]

According to my meter, the max current (both motors accelerating to speed 9) was 0.369 amps, well within margins for the L293DD.  So motor current draw is not the issue on my test-bed, at least.

So, if it's not total motor current that is killing the L293DD chip, then what are the other possibilities?

 

Well, the first thing that comes to mind might be a firmware bug that reverses motor polarity while the motor is still in motion.  This could happen during a direction change, where the firmware first slows the motor down to a full stop, and then starts it again in the new direction.  If the firmware doesn't pause long enough in the middle there, it could blow the chip.

 

When I videoed the latest deliberate failure, the motor sound in the video indicates that the chip failed during high-speed slewing of both motors.  No sound of it slowing down for a direction change at the point of failure.  So probably not happening due to that kind of timing error while changing direction.

 

But it still could be due to a bug where the firmware accidentally reverses polarity, just for an instant, mid-slew.  When slewing, I do hear the odd lack of smoothness in motor drive, and always wondered what the cause might be -- a possible audible clue here.

 

To find out if this is what is happening, I'll need to hook up my venerable Logic Analyzer (LA) to all of the inputs of the L293DD and try to figure out what condition to trigger it on.  Still thinking on that problem (how to trigger it), as my LA has a limited size trace buffer.

 

EDIT: Using the LA on this could be complicated by the fact that PWM is used on the inputs to produce varying motor speeds..

 

Another possible cause could be voltage spikes from the motor brushes, feeding back into the L293DD and killing it that way.  This is easier to look at with just an oscilloscope.  On the list of things to-do here.

Edited by mlord, 20 April 2025 - 07:06 AM.

[Ice Cube]

Here is the REV L board on Agena Astro. Not cheap and with the current tariff environment hard to say what the cost will be going forward. Cheers

 

 

https://agenaastro.c...er-f00-004.html

i'm a little worried that AA's photo shows a REV K board...

[mlord]

i'm a little worried that AA's photo shows a REV K board...

Not uncommon -- just an old photo.  But Rev.L does not have a fix for this issue either.

[Hikescdnrckys]

Today I re-ran the simple StarSense "stress test" to deliberately fry another L293DD chip, for the purpose of recording the event (video with sound) to pass along to Celestron.  This time it took just over 2.5 minutes to go POOF!   Except there was no smoke because I pulled the power plug as soon as the mount shut down.  The video isn't that interesting except to people working on the problem, so I haven't made it generally available.

 

After replacing the L293DD chip yet again, I left the (14.5V) VS pin up off the pad, and soldered a wire to it, and another to the pad.  Then I used the two wires to insert my ammeter into the circuit, to measure the total current flowing through the L293DD.

 

According to my meter, the max current (both motors accelerating to speed 9) was 0.369 amps, well within margins for the L293DD.  So motor current draw is not the issue on my test-bed, at least.

 

EDIT: I should have also done the measurement in a real mount with clutches engaged, for higher motor load.  Now on the to-do list for later. 

Hi Mark. Did we actually try just using the hand controller and pressing an up and left button together four or five times to see if that blew the chip?

[mlord]

Doing that has never blown any chip here.  I have tried it.

StarSense AutoAlign has a lot more patience than I do though,

which is how it does manage it.

 

Cheers

[mlord]

.. I'll need to hook up my venerable Logic Analyzer (LA) to all of the inputs of the L293DD and try to figure out what condition to trigger it on.  Still thinking on that problem (how to trigger it), as my LA has a limited size trace buffer.

The L293DD is internally partitioned into two, independent halves, one for each motor.  Each half has two logic inputs (one for each wire going to the motor) to control the polarity of the outputs, two high voltage outputs for the motor, and an enable pin.  The enable pin is where one would feed a PWM signal to control the actual speed of the motor.

 

So.. I can hook up the Logic Analyzer (LA) to the two logic inputs for each motor; four inputs in total.  Simple enough.  And then perhaps attach the +5V power to another input on the Logic Analyzer.  For triggering, just have it stop collecting data when the +5V dips below 2V, which is what happens when the chip dies and shorts itself to ground.  This is all easy to do, and safe enough for the LA.

 

With that setup, the LA will run continuously, keeping track of all 0<-->1 transitions on the four logic inputs, with timing information, until the +5V line dips.  At which point it stops, and I can then examine the history of the four logic inputs, to see if any of them changed state inappropriately leading to the failure.

 

That's my plan.  Just waiting for more ChipQuik removal solder to arrive (tomorrow?) so that I can easily swap L293DD chips again -- running very low on the stuff at the moment from all of the recent trials.

Edited by mlord, 21 April 2025 - 12:55 PM.

[mlord]

The L293DD is internally partitioned into two, independent halves, one for each motor.  Each half has two logic inputs (one for each wire going to the motor) to control the polarity of the outputs, two high voltage outputs for the motor, and an enable pin.  The enable pin is where one would feed a PWM signal to control the actual speed of the motor.

Before getting out the Logic Analyzer, I first hooked it up to my Digital Oscilloscope (DSO).

Well well well..

 

They are NOT using the two Enable pins for PWM.

Instead, they are switching one of each pair of IN (input logic) pins at a time to do PWM.

Okay, not the way I would have designed it, but that can work.

 

However, the PWM rate is a constant 23.44KHz.

The L293DD chip clearly specifies "up to 5KHz" in the datasheet,

so Celestron are driving it way beyond specification.

 

This could explain a few things.. 

 

I also noticed that the motors are being driven continuously,

even when not "in use", with a low percentage PWM output.

This explains why my thermal imaging camera shows the L293DD chip

getting warm even when the mount is completely idle.

 

Passing this info along to Celestron as well as documenting it here.

 

EDIT:  Some good news, is that the PWM lines into the L293DD come directly from GPIO pins of the main processor.  So it should be possible to bring that aspect "into spec" with a firmware change.  This may or may not help with the issue at hand.

Edited by mlord, 21 April 2025 - 05:29 PM.

[mlord]

Some geek entertainment:  A pre-beta firmware release on the Rev.K, killing the L293DD chip in just over a minute..

 

https://www.youtube....h?v=2D-WdI8jBB0

[Lee D]

Some geek entertainment:  A pre-beta firmware release on the Rev.K, killing the L293DD chip in just over a minute..

 

https://www.youtube....h?v=2D-WdI8jBB0

"This video is private"

[mlord]

?? Try now?