Turbo 040 in SE/30 cold start up boot failure

[ttb]

I am trying to get a Daystar Turbo 040 working in my SE/30 using one of @Bolle's short adapters. I've already managed to work through other issues, but this one has me stumped. Basically, the machine will not find any startup disks when it boots from a cold start (e.g. flipping the power switch after it's been off for a second or more). I have tested this with two different working SE/30 logic boards, two different AB/PSUs, and two different SCSI2SDs. Also tried a different accelerator adapter, different floppy and SCSI cables, tried it with six different sets of RAM in different arrangements, swapping ROMs, etc.

The test matrix is basically accelerator installed vs not installed on one axis, floppy vs SCSI2SD on the other:

		Floppy		SCSI2SD
Not installed	(1)		(2)
Installed	(3)		(4)

The boot floppy has a System 6.0.8 system folder on it. The SCSI2SD has a 7.5.5 system folder. The Turbo 040 has firmware 4.11. The SE/30 has the stock ROM. Notes about each configuration:

1) Works fine, no issues.

2) Works fine, no issues.

3) The system shows a gray screen with cursor. After a little bit, it shows the Turbo 040 happy mac briefly before ejecting the floppy. It then goes to a gray screen with no cursor an stays there. If I reinsert the floppy and press the reset button, the same thing happens. If I reinsert the floppy and blip the power switch (turn it off for less than a second), the same thing happens. Basically there is no way to get it to boot from the floppy.


View attachment IMG_1448.mp4





4) This would be the normal operating configuration (accelerator installed, boot from SCSI2SD). From a cold start it goes straight to a blinking question mark. The amount of time it takes to show the blinking question mark can vary a bit. It can be almost instant like below, or take several seconds. If I insert the boot floppy while the question mark is there, it goes into the same state described in note 3 above. If I hit the reset button, it will just give the blinking question mark again.


View attachment IMG_1449.mp4





The real weirdness (and hopefully a clue as to what's wrong) happens when I blip the power switch. This means interrupting power for less than a second (I haven't tried to find the lower bound, yet). When I do this, the system boots up without any hesitation. Additionally, once it is booted up, I can restart it at will and it remains fine. Hitting the reset switch also reboots it without issue.


View attachment IMG_1450.mp4





My initial thought was that the PSU is overloaded. I've only scoped the 5V line but did not see any sag on startup. I have not checked the other rails and I'm not sure whether the Turbo 040 would even use them. My other thought (before I did a lot of testing with the boot floppy) was that it was related to some setting in the SCSI2SD. I tried adding delay, different modes, etc but it always produced the same results.

Any thoughts on where I should start? I've basically done all of the component swapping I can think of to narrow it down. I'm hoping someone with a better idea of what the boot process entails and how the Turbo 040 integrates with that might be able to point me in a direction. Thanks!

 

[ttb]

Alright, I have an interesting data point to share. I just procured a PowerCache P33 and swapped it in for the Turbo 040. Same result! Flashing question mark initially, but blipping the power causes it to boot.

So the good news is that my accelerators either both work fine or are miraculously both broken in the same way. My current best guess is that maybe the PSU rise time is too slow for the accelerators. I've measured it at about 50ms from off to full 5V at the cache card connector. If the motherboard components started working at a lower voltage than the accelerator (or vice versa), this could cause them to get out of sync.

The closest I've come to verifying this is that when blipping the power switch the CPUCLOCK signal from the cache connector will fully drop out but the 5V supply to the cache connector drops to 2V. Green is 5V line and yellow is the CPUCLOCK signal:



I'm going to keep probing the other signals on the cache connector. If anyone has a schematic for the PowerCache and/or Turbo 040 or can tell me what I should look at to determine whether it's "alive" that would help out too since then I can pinpoint when it comes up in the startup process.

 

[Crutch]

Are your PSUs stock?  If so are they recapped?  I have seen all kinds of weird issues running accelerators with old PSUs, just yesterday I tried a known-good Diimo in an SE/30 that otherwise works fine and would randomly see either (1) just a black screen on startup or (2) 2 black columns per 8 bits on the display (vertical bars) but otherwise normal operation.  In my case it turned out the 5V rail was dropping low with the accelerator installed, so easy to diagnose, but I’m sure other more nefarious things could arise (like your suggestion of a slow rise time).  I tried swapping out the PSU with another old PSU and it didn’t help — but when I swap in my modern SEASONIC-stuffed modded PSU it all works fine.

I guess my point is, enough of these old PSUs are losing their oomph now that seeing the same issue with two PSUs doesn’t mean you should rule out the PSU.  But you already knew that  

 

[ttb]

The PSUs are both stock Sony and both recapped except for the one large 400V cap that I bought in the wrong size (too tall). I recapped them both with the same caps, so it's quite possible I did something consistently stupid, though I'm thinking I'll try to rig up a lab power supply to the AB so I can eliminate that as a possibility.

There doesn't seem to be too much voltage sag during power up, but maybe my timebase is too short here. It seems pretty stable at 5V once it finishes rising. The yellow trace is measured at the floppy connector and the green trace is at a 5V pin on the cache connector. I'll try measuring it over some seconds instead and see if anything interesting pops out.

 

[ttb]

Alright, this is (somewhat) solved. As @Crutch pointed out, there's no reason to trust two power supplies, especially since I recapped both with the same parts.

I first tried splicing in a lab power supply between the AB and LB, but it appears the LB either looks for the -5 and -12V lines before booting or there's something with the power on sequence that's lacking with this setup so it didn't work.

Next, I tried it with a third AB/PSU combo, this time both unrecapped and the PSU is an Astek rather than a Sony like the others. No issues! Boots both the PowerCache and the Turbo 040 from a cold boot every time. The Turbo 040 config freezes once booted, but I believe that's because it's taxing the old PSU beyond what it can take.

So what's causing this? Well, the risetime of the 5V line (measured at the cache connector, as above) is less than 1/5th as long as with the recapped Sony PSUs. I'm measuring it at 9ms. The span is the same in this window as the above to get an idea of the remarkable difference:



Now I guess I'll have to dig into some analog circuits theory to figure out what I can do to decrease the risetime on my two Sony PSUs. I'd be curious to know what the 5V risetime is for other working configurations with the PowerCache or Turbo 040.

 

[hfrazier]

@ttb I Actually experienced this same problem. I have a few 040's and a P33 and P34 cache. Tested all the accelerators on my iisi, so I knew they were good. The symptoms are exactly as you experienced, and IIRC the P34 worked fine. (I was playing with this back in January, so the exact details are a bit hazy but more or less that's how it behaved.) Interested to see what you uncover.

... maybe a power manager board that powers up the logic board once all the analog board caps have charged?  }

 

[ttb]

That's another good point. Both of my faulty-config ABs are also recapped (with the same caps) so I shouldn't rule them out, either. On the 5V rail itself it looks like there's just a 100 uF cap in what looks like an RC filter for the AB, but it's certainly possible the other circuits on the AB using the 5V rail could be impacting things.



I'll do some PSU and AB swapping to see what the effect is. If it does seem isolated to the PSU I'm somewhat tempted to be lazy and swap a modern PSU like the Seasonic into my dead Astek enclosures, but it would be nice to get the Sonys working in my application.

 

[ttb]

Alright, mark this as solved. I took the nuclear approach and swapped in a new power supply, a Seasonic SSP-300SUB as described in this thread:





This resolved my issues with the Turbo 040 and PowerCache. I have not measured the 5V rise time with the new power supply but I would expect it's much shorter than the 50 ms I was seeing on my recapped Sony.

 

[Bolle]

I just experienced the exact same thing when testing my PowerCache and upcoming Turbo040 clones.

Sometimes I would have to reset the machine before it would boot with a Turbo040 in place.

With adapters, network card and accelerator in place the 5V line was sitting at 4.78V. That seemed to be enough for the PowerCaches to work reasonably well, but once they warmed up some of them started to get unstable.

The Turbo040s would crash nearly instantly once the control panel loaded and the cache got activated.

I turned the power supply up to give me a nice 5.05V under load and all problems magically disappeared... PowerCaches were stable running benchmarks for an hour and Turbo040s didn’t crash with the caches activated.

The problem with turning up the voltage is that now my 12V line is sitting at 12.8V... that’s probably ok for the sweep supply because that’s sitting behind another linear regultor but I am not sure how the hard drive is going to like that in the long term.

 

[ttb]

That's pretty interesting. With my recapped Sony PSUs I had adjusted the output to be 5V on the dot with a calibrated meter. As you can see from the o-scope trace earlier it's still bang on with the accelerator loading it up. Conversely, my (now dead) Astec was only putting out 4.84 V which sounds pretty close to the limit you were at yet that one seemed to work okay. Granted I didn't run it for more than a few minutes and it's now toast, but at least it didn't show the same problem.

I guess the moral of the story is that these accelerators are generally sensitive to power supply parameters including 5 V voltage and rise time. Hopefully that gives the next person frantically searching old threads for a solution a good starting point.

EDIT: And to really put a bow on it, here is the 5 V rise time for the Seasonic, measured at the external floppy port with the accelerator and ethernet card installed:



8 ms is even faster than the Astec!