Is my MCF5307 dead or not?

[conorcaulfield]

What signals should still be present on the pins of a MCF5307 even when it cannot boot properly? For example, should PSTCLK still appear on pin 184? A coldfire processor is used in a synthesiser I am working on. The synth fails to boot. The main CPU may be suspect. So I am trying to find a way to verify whether the CPU operational but 'confused' or entirely dead. Are there functions which an improperly booted or unbooted MCF5307 will still try to execute. My apologies if this sounds like a stupid question, I am trying to help a whole community of people who have problems with this synth. Many thanks.

[miduo]

Hi,

Thank Tom's detailed information on this regard.  " What signals should still be present on the pins of a MCF5307 even when it cannot boot properly?". There no answer on this "should" . I also recommend customer check the reset vector on the data lines during reset with a scope or logic analyzer. And also check the data lines based on your schematic.

I attached a quite old document on the PLL configuration of MCF5307 for your reference.

[conorcaulfield]

Fang Li

Thank you, that PDF does provide a useful quick reference.

[TomE]

MCF5307? That's an old one. The User Manual (Rev 2.0) is dated August 2000. That's a good thing as it should be a lot simpler than the later ones, with less things to go wrong (in setup, configuration, booting etc).

> I am trying to help a whole community of people who have problems with this synth.

I could read that as a community, all of whom use one single synth, but I assume you mean there's a lot of people with a lot of these things, all over a decade old, and all showing similar problems. If that's the case, check all the power supplies with an oscilloscope. The electrolytic capacitors dry out in these things and that usually puts ripple on some of the power supplies. That may be dropping low enough to trigger whatever is used for the power-on-reset. Given it is a Synth with analog outputs, there should be other power supplies for the analog side. Make sure none of them have ripples and are upsetting some other circuitry and generating lots of interrupts or something.

This CPU is so simple it only has two power supplies - VCC, which is 3.3V and PVCC for the PLL. Later models have all sorts of different lower core voltages. Check those supplies for level and ripple. Make sure the capacitor on the PVCC supply hasn't dried out or shorted. or something.

You HAVE to use an Oscilloscope. You can't do this properly with a multimeter.

Make sure the JTAG pins aren't at the wrong levels. That can force the CPU into a debug mode. Make sure TRST, TMS and TDI are at the proper levels, or pulled to the correct supplies as per Chapter 19. Check HIZ.

The only other important inputs for the CPU are the RSTIN and Clock Input (CLKIN) from an external oscillator. Make sure the input clock is OK and the right frequency. it is possible that crystals may get damaged, "wear out" or suffer voltage drift. Read the PLL chapter to see what multipliers the CPU is set to apply. There should be pullup or pulldown resistors on Data Bus Pins D0-D7 (or perhaps a tri-state buffer enabled by RSTIN or RSTOUT) that set the initial clock frequencies and other bus parameters like the bus width. See 17.5.5. If these are wrong the CPU won't boot. If the board has jumpers or switches to allow some of these parameters to be changed, then maybe the jumpers or switches are corroded, a jumper has fallen off or someone has fiddled with them.

RSTOUT should follow RSTIN as per the diagram in 7.4.2, as long as the clock is OK.

BCLKO is "compulsory". That one should be toggling. The PSTCLK signal is only required if you are debugging, and being a high frequency clock it is often a problem for EMC Emissions. So it is normal to turn this one OFF. That is done in code by the "CSR[PCD]" bit and register, see Figure and Table 5.8. Then notice an 18 year old BUG in the manual, where "PCD is listed in the Table, but is missing from the Figure, so you don't know what its reset value is!

Fortunately there's an Errata for the manual. Unfortunately it doesn't fix this bug. Looking in manuals for later chips (like the MCF5329) shows the Figure properly filled in and showing the initial value of "zero". So PSTCLK should be toggling after reset, but the chip may execute some code that turns it off later - if it gets that far.

Assuming all that is working, watch the chip select signals. The CPU starts executing by fetching locations 0 and 4 in main memory and then jumping to the start vector. The Chip Select Module defaults to enabling CS0 for all accesses until programmed otherwise. So CS0 should be toggling. If the code runs from ROM/FLASH, then it should keep toggling. If the system loads the code from FLASH into RAM and runs it from there, then CS0 may only run for a a fraction of a second before one of the other chip selects (connected to RAM) takes over.

Make sure nothing is pulling the BR (Bus Request) signal down or the CPU will give up the bus.

If it is getting that far and going off the rails, make sure nothing is doing something stupid on any of the interrupt request lines.

Tom

[conorcaulfield]

Tom

Fabulous! There's plenty of good steers in that reply! I'll check out those pins and parameters and get back to you if I may.

The Power On Reset is handled by a Maxim DS1813. The 'First Byte' on the data bus IS handled by a buffer with hard wired config pins. I'll check out ripple and caps first, then move on to the more esoteric signals. Many thanks for your help, it's heartening to have something to go on.

Yes, there are people all over the world who have this synth. It's a very expensive synth and it's regarded as one of the most sophisticated and controllable synths ever made. Maybe that's why it has to run at a full 90Mhz!

Thanks again, Conor.

[TomE]

I won't mention the name (because you haven't), but is it the 2000 or the 2500 that is giving problems?

From the photos I've found, the 2500 looks to have a better packaged power supply, while the 2000 has a power supply that looks like the ones that give problems in old Topfield PVRs. I've replaced caps in at least three of these, and I don't even own one!

> Maybe that's why it has to run at a full 90Mhz!

It is likely to be running a full 32-bit wide data bus, and at least 24 bits of address. There has to be RAM, ROM and there are even expansion slots. That means it is a very wide bus, running fast, and covering a lot of that board. That means a lot of dynamic current and noise that needs very good power supply filtering to keep in check.

The 2500 board looks to only have about six electrolytics on it. I'd guess a lot of them should be for keeping the analog side quiet, so that's not much left for the CPU and other stuff on the bus. They must be relying on the small ceramics and the main power supply for the bulk capacitance - on the other end of the power supply loom.

I'd add more capacitors on the main board if there's an obvious set of pads to solder to.

Tom

[conorcaulfield]

Tom

     It's the Alesis Andromeda A6. But your other guesses are bang on. It uses pretty much everything the Coldfire has to offer.

    I have a service manual with full schematics and PCB layouts. I could send it to you if you're interested, but I don't want to impose on you. In addition I don't know if attachments would pass Freescale's moderation and filters.

    Right now it's off the bench and I'm waiting for some ribbon connectors. It has serious ribbon problems too, simply bad manufacture. So I can't fire it up again until I've made those ribbons.

    Yes, you would have thought 60w ribbon connectors would be easy to get from RS, but I like ribbon connectors that expose the pins (to a very sharp and narrow probe), which aren't in stock so I'm at a standstill until they show up.

    Regards, Conor.

[TomE]

> Alesis Andromeda A6

http://www.deepsonic.ch/deep/htm/alesis_andromeda_a6.php

The pictures show plenty of capacitors near the analog stuff, very little visible near the CPU and its memory.

It also shows a "standard" switch mode power supply. First you swap it with a "known good one" and see if the problems go away. If you now suspect you have a bad one and need to repair it, check all the output voltages, and maybe even increase the load (whatever you do to the unit to maximise power drain, or add some load resistors) and if you see any ripple or sag.

If you see any, the first suspect is the big capacitor on the high voltage. That may have lost sufficient capacitance that it is letting ripple through to the other supplies. Suffice to say these things are dangerous and you don't go anywhere near it unless you know what you're doing. Then there's another six capacitors on the output visible, potted in goo to stop then vibrating, and to make getting them out to check and to replace really difficult. My preference is to pull all of these out, measure on a professional capacitance meter (multimeters can't measure capacitance this high) and replace the ones that are low. You can't use "any old capacitors" but should use specialist low ESR ones. Unless the design won't work with low ESR ones (some depend on the resistance for stability). So good luck.

Ribbons? Lots of ways to go wrong there. I hope they're not sending any fast CPU bus-type signals over those.

It looks like the CPU is in the middle with its FLASH (looks like 8 or 16 bits wide, so slow to fetch code and data from), "Fast SRAM" and "Slow "SRAM". Then there seems to be "buffers to the left of them" and "buffers to the right of them" where bus signals (or maybe latched I/O ports) head off to the rest of the unit. Through ribbon cables.

> I like ribbon connectors that expose the pins (to a very sharp and narrow probe)

In a pinch you can always get a probe onto the wire where the ribbon has been cut on the edge of the crimp connector.

If you're making up new cables, why not crimp an extra FEMALE plug onto the cable (or onto ALL cables) facing up to provide test points that will hold an oscilloscope probe and also hold multiple wires.

Tom

[conorcaulfield]

Tom

I've repaired a LOT of SMPS. I tend to use Panasonic FC, FM, or FR series caps, or Nichicon PW. They're the best I can find anywhere. I've spent plenty of time picking out the glue from the output caps and inductors. Why oh why do they glue the components most likely to need replacement!

I've got plenty of quiet and stable TTI bench PSUs. So I can set up a stand-in for the A6's own PSU.

Power decoupling? Yes indeed! Not a lot of it near the CPU, SRAM or Buffers and none at all on the PLL supply pins. What would you recommend here?

My ribbon, fortunately, had faults on LED_Row lines so there isn't much of an issue for the CPU. Other people's ribbons may have faults on switch lines which will likely cause a constant SW_INT which does have consequences for booting.

No, there are no fast signals on the front panel ribbons. Comms from the front panel boards to the CPU board is serial, from a separate front panel CPU (87C52), and does run on a short ribbon. The serial link is switch, LED, and display data. There's also a 'staircase' of pot values from each of the three front panel boards.

An extra connector is a good idea, but I want to leave the synth looking as original as possible inside.

Yes, buffers all over the place! For the analog masters, analog ASICs, PC Card, ADC, Control MUX, FX ASIC/Keyscanner, et al.

To be fair to it, busses to the analog boards are buffered just before and just after a 100mm ribbon cable with some interleaved grounding.

Could I tempt you with a full service manual and plenty of better photos than the Deepsonic site? It'll all start easily, soon you'll be as hooked as the heroic dude who set out to rewrite the OS!

Thanks Tom, any input is greatly appreciated.

Regards, Conor.