• Please review our updated Terms and Rules here

Maximum power supply ripple?

S

shirsch

Veteran Member
Joined
Aug 17, 2008
Messages
873
Location
Burlington, VT
I'm trying to troubleshoot some QBone crashes on my pseudo-11/73 system. It's built on a Heath H-11 backplane that I've added the four upper address bits to. Using just the M8192 CPU and QBone (emulating 4M memory, MSCP storage and SLU) I get persistent DMA-related crashes in the QBone emulation software. Joerg suggested I check the power supply. Voltages are spot on, but I'm seeing +/- 300mv P-P ripple on the 5V bus at the power supply (probably a bit worse at the backplane due to IR drop). Is this enough to create problems? I cannot seem to find any specs to compare with.
 
How are you measuring the ripple? Scope, DVM, or some other method? Chinese PSU maker MeanWell says:

"Acceptable ripple voltage is about 100mV peak to peak. Majority of good power supplies have ripple and noise figures of better than 10mV rms, while SMPS figures of 50mV or less are possible, however, higher current supplies are likely to have slightly higher values."

I presume the 'crashes' are a recent problem? Your system used to be stable? Sometimes higher than expected ripple can be a symptom of excessive current draw. Can you tell if any of the components are hotter than usual?
 
I used a scope set for AC coupling and simply observed for a while. The problems I'm seeing are not new and have happened with both the original 11/03 16-bit CPU and the 11/73 CPU. Now that I'm running BSD 2.11 on the 11/73 there are a lot more disk intensive activities so it's happening more frequently. I can crash it 100% of the time by running fsck on the root file system. I'll put a gun to my head and setup the logic analyzer to trigger on the bad DMA operation (hack to QBone code thanks to Joerg) and see if anything jumps out.

Not sure what to do about the power supply beyond checking caps for value and ESR. 300mV does seem noisy enough to be suspect as the root cause.
 
300mV ripple sounds way too high. For example, the H7864 (microVAX) supply is spec'd at 50mV P-P on +5V and 75mV P-P.

What is the part number of the supply you are using? I'd be tempted to hack an external supply (PC or bench) in and see if the problems go away before trying to track it with a logic analyzer.

CW
 
Yeah, I had a feeling that was excessive. The power supply is a Heath analog design with pass regulation that's specific to the H-11 system. Seeing as the unit is 40-something years old I wouldn't be surprised if the caps were all dried out. Problem with an external supply is generating DCOK and POK.
 
Ripple is ripple. 100mv is suspect and 300mv is problematic. You don't need a scope as DVM will do.
 
From the KDJ11-A (M8192) CPU Module User's Guide:

"The +5 Vdc must be regulated to ±5% and the maximum ripple should not exceed 100 mV peak-to-peak. The +12 Vdc must be regulated to ±3% and the maximum ripple should not exceed 200 mV peak-to-peak."

I couldn't find schematics for the QBone, so it's hard to say if it has tighter power supply requirements or not. I would guess the beagle-bone has an regulator, but the misc logic (FPGAs? hard to tell from the photos) probably don't.

CW

Note: You can dummy the DCOK/POK signals by hard-wiring or switches for testing purposes. I doubt the timing on them is critical to operation.
 
Last edited:
One place to start looking is for failing capacitors in the power supply, which lead to ripple like that.
 
The capacitors at the +5 out were my first thought. I pulled them and checked with C-meter and ESR tester. They were fine. Maybe 10% high in value and ESR of about 0.01 ohm.
 
Bitly said:
I couldn't find schematics for the QBone, so it's hard to say if it has tighter power supply requirements or not. I would guess the beagle-bone has an regulator, but the misc logic (FPGAs? hard to tell from the photos) probably don't.
Click to expand...
All logic except QBus transceivers is 3.3V and fed from on-board regulation. Ditto for the Beaglebone. The transceivers are certainly a potential issue. Theyr'e obsolete and hard to source, but I think I have a few good ones that can be swapped in.
 
You must log in or register to reply here.
Back
Top