Mystic Died. Looking for an MC88916DW80 chip.

[JDW]

Somehow I killed my Mystic merely by making some 4-wire resistance measurements in the vicinity of the XC88916DW55 chip on my LC575 motherboard. I know it's a problem on the motherboard, not the analog board. I believe the 88916DW chip is now bad because when I put my scope on it, the Q0 and 2X-Q outputs are higher than they out to be compared to the SYNC input from the motherboard.



Most of the Chinese sellers are shutdown for 2 weeks thanks to Chinese New Year holidays, but I'm also concerned about the lack of Motorola markings on some chips shown on AliExpress, making me feel they are fakes.

Would any of you happen to have a stash of known good MC88916DW80 chips?

I'm seeking the DW80, rather than the DW70 or DW55, because I will be running my Mystic at 40MHz or higher. (I actually had it running at 50MHz via Spicy O'Clock before I killed it with my resistance testing.)

Thanks.

 

[This Does Not Compute]

You might have done so already, but try reaching out to Mouser. They may still be able to get a few -- apparently the part is being discontinued, but there may be some left in the supply chain. https://www.mouser.com/ProductDetail/NXP-Semiconductors/MC88916DW80?qs=WhRErS6oET%2BCZr2YQaNLxw==

 

[trag]

I have a stash of the DW55

I have some DW80s as well and could spare you a couple, but I've never confirmed that they're good. I can't remember the source at this point. Too many years. Seems like it was a 68KMLA member, but I'm not sure.

Not sure about shipping to Japan though.

In case that wasn't clear, you're welcome to as many 55s as you could practically need. Pretty sure this is a legitimate reel and not a fake.

 

[JDW]

You might have done so already, but try reaching out to Mouser. They may still be able to get a few -- apparently the part is being discontinued, but there may be some left in the supply chain. https://www.mouser.com/ProductDetail/NXP-Semiconductors/MC88916DW80?qs=WhRErS6oET%2BCZr2YQaNLxw==

Thank you for the advice, Colin!

I spent a half hour speaking to Mouser via Chat. The best they could do was recommend alternative parts, non of which are pin compatible:

• 8-pin PLL
• 16-pin PLL 1
• 16-pin PLL 2
• 16-pin PLL 3

In contrast, the 88916DW is a 20-pin part, specifically made for use with the 68040.

I have a stash of the DW55

I have some DW80s as well and could spare you a couple, but I've never confirmed that they're good.

Thank you, Jeff! I had sent a PM to @Kai Robinson a few days ago but didn't hear back from him, so I decided to create this thread today. Kai had a stash of the DW80 chips back in 2021, so I was checking with him to see if he still had them. He replied back just now to say he's checking his inventory.

If Kai ends up no longer having the DW80 chips, I would like to buy those couple you're willing to spare, in addition to one of the DW55 chips too, just in case the DW80 chips don't function. I'll update this thread to let everyone know the situation after I hear back from Kai.

Of course, that doesn't help people who may be in my shoes in the future. Pulling the chips off working boards isn't a great idea. Besides, most of those are the DW55 anyway. Even the DW80 is being used beyond its 80MHz spec when you overclock your 040 processor beyond 40MHz. That's because the clock driver outputs a normal speed clock like 40MHz, but also a double-speed clock, like 80MHz, which is required for the 040.

Anyway, I'll update this thread when I hear back from Kai.

Thank you both very much for your help!

—James

 

[trag]

Sounds like a plan. Let me know if you need the chips, JDW. I'm happy to help if I can.

Is there any experience out there on whether the DW55 and DW80 are interchangeable? If they were actually tested and speed binned then it's less likely, but if they got the manufacturing to the point where they were all coming off the same wafer and all with good performance and then just marked some of them slower than others for marketing reasons....

 

[JDW]

Sounds like a plan. Let me know if you need the chips, JDW. I'm happy to help if I can.

Is there any experience out there on whether the DW55 and DW80 are interchangeable? If they were actually tested and speed binned then it's less likely, but if they got the manufacturing to the point where they were all coming off the same wafer and all with good performance and then just marked some of them slower than others for marketing reasons....

I've not heard back from Kai, so it could be he can't find his stash. I emailed you just now about acquiring some of your chips.

My DW55 chip should theoretically only handle up to 55MHz on the 2X output. Meaning, there's also a 1X output. For example, if you run an 040 CPU at 25MHz (1x output) you must also feed it a double speed clock of 50MHz. A DW55 would accommodate that just fine, because again, it's rated for up to 55MHz on the 2X output.

In my case, I had been driving the DW55's 1X output at up to 50MHz, which means the 2X output was 100MHz. It handled that just fine until I used my Fluke 8845A to make 2X4W (4 wire, using 2 probes) resistance measurements, then the chip suddenly started outputting 2X on the 1X output, and 4x on the 2X output. All the outputs are now twice as fast as they should be, which is driving me insane because I don't know why. You'd think if you'd fried the chip, it would stop working. Well, in my case, the DW55 chip is working twice as fast as before. I have no idea why. And even when I remove Spicy and use the motherboard's stock 33MHz speed, the DW55 chip is outputting 66MHz on the 1x output and 120 MHz on the 2X output. And I think that when you overclock the CPU too high, it won't chime and the screen remains dark, which are the symptoms I have now. Even so, I don't know if the DW55 chip is the ONLY problem.

But I must start replacing something, and since I know the outputs are twice what they out to be, I need to change the DW55 chip. But because I can clocking the chip to 40MHz and higher, using a DW80 chip (which has a 2X output that is confirmed by the manufacturer to work at 80MHz) is prudent. Overclocking the 040 beyond 40MHz would put the DW80 beyond spec, but like I said, I've been doing that with the DW55 chip, and it's been working fine for about 2 years.

 

[JDW]

I removed the XC88916DW55 chip this evening. It's all ready for a replacement. And I do hope that is the fix. Even so, I won't know until I try a chip swap.

 

[trag]

then the chip suddenly started outputting 2X on the 1X output, and 4x on the 2X output.

I spent a little time reading the datasheet for the 88916 and didn't see anything that should cause that behavior. I was thinking there might be a pin that sets a mode or some such. Admittedly, I did some skimming and didn't read everything in detail. There's some complication there with the reset in and reset out pins.

My thinking was that maybe you somehow caused one of the control pins to be shorted or open or some such. I can't figure out what the MR pin even does though. Not without more attention.

I notice there's a 31.44 oscillator near the 88916. Is it possible that something shorted a trace so htat the Sync input is getting the 31.44 signal instead of the 16.667 or whatever it is supposed to get?

 

[JDW]

@trag
I've been exchanging PMs with someone who asked me to do various tests. I made a couple videos of those tests, where I had the XC88916DW55 chip desoldered from the LC575 motherboard and placed inside an SMD to DIP socket for my breadboard. Even without any SYNC input, I get roughly a 30MHz output on Q/2, roughly 60MHz on Q0 and roughly 120MHz on 2XQ. But even when I try to inject a signal at the SYNC input, nothing on the outputs change, although that could be because I was using a 3.1MHz test signal rather than the minimum 5MHz signal the data sheet specifies. All the details and the scope signals are in my two videos below...

 

[Kai Robinson]

@trag
I've been exchanging PMs with someone who asked me to do various tests. I made a couple videos of those tests, where I had the XC88916DW55 chip desoldered from the LC575 motherboard and placed inside an SMD to DIP socket for my breadboard. Even without any SYNC input, I get roughly a 30MHz output on Q/2, roughly 60MHz on Q0 and roughly 120MHz on 2XQ. But even when I try to inject a signal at the SYNC input, nothing on the outputs change, although that could be because I was using a 3.1MHz test signal rather than the minimum 5MHz signal the data sheet specifies. All the details and the scope signals are in my two videos below...

I've not had a chance to have a look at my component selection until today, i'll see what i have in stock - also, for clock signal generation - do you have any spare 4-pin DIP frame Oscillators to hand to generate a fixed frequency which would be suitable? IIRC i sent you some 15.6672MHz ones for the SE Reloaded a while ago - it's an oddball frequency, sure, but still, should be suitable for an input to that clock generator?

 

[JDW]

I've not had a chance to have a look at my component selection until today, i'll see what i have in stock - also, for clock signal generation - do you have any spare 4-pin DIP frame Oscillators to hand to generate a fixed frequency which would be suitable? IIRC i sent you some 15.6672MHz ones for the SE Reloaded a while ago - it's an oddball frequency, sure, but still, should be suitable for an input to that clock generator?

Thanks for the reminder, Kai. I also just remembered that I could use the output from Spicy O'Clock too. So I will give both of those a try and see what happens.

 

[JDW]

Here's my test using Spicy O'Clock, which allows me to satisfy the XC88916DW55 chip's minimum input frequency requirement.

So basically, this is proving the chip is bad. I will provide another update when the chips from @trag arrive. Thanks again, @trag!

 

[JDW]

I would like to once again humbly thank @trag for having so kindly shipped me some MC88916DW80 and DW55 chips. They arrived today and I was able to put a few to the test in my test jig.




I measured current consumption first. I found that a good DW55 draws 75mA while a DW80 draws only 67mA, while being fed by Spicy O'Clock at 19MHz. Both of those current consumption values are much lower than the defective DW55 I removed from my LC575 board, which drew 109mA. And thankfully, unlike the defective DW55 chip, the good DW55 and DW80 @trag sent to me show no output when nothing is connected, show a Q/2 output frequency that matches the Sync input, show a Q of twice the Sync input, and show a 2X_Q output that is correctly twice the Q output. So all seems to be working well on my bench test.

No time to solder a chip into my motherboard tonight, but I will see if I can do that tomorrow night after work. I will of course use one of the DW80 chips since my intention is to overclock using Spicy O'Clock. Fingers crossed that single chip swap will get my motherboard working again!

@trag you're amazing!

 

[Kai Robinson]

Don't forget when overclocking to set the SCSI clock to /2 isntead of 1:1 ratio

 

[trag]

@trag you're amazing!

You're very welcome, JDW. I'm am happy I was able to help.

Do you have an oscilloscope? I am wondering what kind of comparison you would see on your test jig between the DW55 and DW80 chips as you increase the input frequency.

Oh, but remembering earlier posts, I guess the limiting factor is the input (sync) frequencies available to you.

 

[JDW]

I apologize for my delayed reply. I've been extremely busy lately, and on top of that, I've been feeling rather terrible all month. No time to visit the doctor until early March. It is also the reason I've not had a single moment of time to do any video work for my YouTube channel.

I did not have time to solder the DW80 chip into the motherboard. Sadly, I doubt I will have time tonight either.

But in answer to your question, @trag , below are the screenshots made on my Rigol DHO804 (hacked to 100MHz bandwidth, with 150MHz rated probes, done with short "ground spring", but done on a breadboard which causes waveforms to look a bit funky)...

YEL trace: Q_DIV_2 (pin 18) — should match freq. of SYNC, and it does!
BLU trace: SYNC IN (pin 8)


YEL trace: Q0 (pin 10) — should be twice the freq. of SYNC, and it is!
BLU trace: SYNC IN (pin 8)


YEL trace: 2X_Q (pin 19) — should be 4x SYNC, and it is!
BLU trace: SYNC IN (pin 8)


Again, this was a breadboard test, which is not ideal, so waveforms aren't shaped perfectly. I did not put any termination resistors on 2X_Q either. I measured everything straight out of the chip.

The above waveforms are the DW55, but I found the DW80 to look pretty much the same. It's just that I measured the DW55 chip to draw 75mA, while the DW80 draws only 67mA under the same test conditions.

The real test will come when I get the DW80 chip solder onto the LC575 motherboard.

 

[JDW]

@Kai Robinson
In answer to your comment about the SCSI clock, my LC575 motherboard already has a 0Ω resistor on at R120, with nothing at R106. The schematic below is for the LC575 and the gray parts are infuriatingly hard to read, but you can see that it says "Stuff R120 for 33/40MHz systems" and like I said, I have a 0Ω resistor already there. Which of course raises the question about what happens when you over-clock well past 40MHz!

I actually made these measurements using my 2X4W probes and which ultimately led to the death of my motherboard:

• R4: 62.03Ω (within its 1% tolerance)
• R80: 140.9Ω (board was 33MHz originally?)
• R120: 0Ω
• R106: pads open
• R107: Marked "474" (470kΩ) but measures 54KΩ (must remove to get accurate measurement)
• R95: Marked "301" (300Ω) and measures 304Ω (within 1% tolerance)

​

LC575 SCHEMATIC DOWNLOAD​

 

[JDW]

I have been feeling absolutely terrible all month, and completely exhausted every day after work, which is why this has take so long. But I was able to solder in the 88916DW80 chip that @trag so graciously sent to me, and I am pleased to report that I get a bong and boot!




Spicy is disconnected, no card is installed, and I've not tested external SCSI booting yet, but those tests will come tomorrow night after work.

I cannot tell you folks what a TREMENDOUS RELIEF this is!

I also can emphatically say that you should NEVER do 2x4W resistance measurements on your LC575 board!
That is what killed the stock DW55 chip. Hopefully further testing will show nothing else was damaged. But oh, what a fabulous relief it was to hear the bong after soldering in that DW80 chip. Wow!

A MILLION THANKS, @trag !

 

[JDW]

I am happy to report that I was able to boot into the following OS's, both from my internal spinning platter HDD and an externally connected BlueSCSI v1:

• System 7.1
• System 7.5.5
• System 7.6.1
• OS 8.1

I then put the SD card into my BlueSCSI v2 and retested, and all was well with that.

I am still using the stock 33MHz CPU clock speed. Tomorrow I will reinstall Spicy O'Clock and start testing at 40MHz. This should be interesting because I had freezing issues in the past with my BlueSCSI v2 and when booting into 7.6.1 and 8.1. So those issues were either due to the CPU being clocked too high, or the DW55 chip had issues. This DW80 might do better. We shall see.

That's all for today.

 

[JDW]

As I mentioned earlier in this thread, there are TERMINATION RESISTORS used on the 2X_Q output of the 88916DW chip, which leads to PCLK on the 68040 CPU. So for example, when the CPU clock speed is 33MHz, PCLK will be twice that at 66MHz. And that is really the reason the 88916 chip is used -- to provide multiple clocks.

But I can't quite figure out Apple's termination scheme.

This is from the LC575 schematic (gray parts are VERY HARD TO READ):

And this is from the Quadra 605 schematic (much easier to read):

So except for the Resistor number, the info in that little box toward the right seems to be the same for both motherboards.

Now the data sheet for the MC88920 is exactly the same as the datasheet for the MC88916 when it comes to the FORMULA that let's you calculate the two resistances...



So if we examine those LC575 & Q605 schematics, we see the following:
Rs = 61.9Ω 1%
= Zo-7Ω (as per the datasheet)

And for a 33MHz CPU clock, we see this:
Rp = 140Ω 1%
= 1.5xZo (from datasheet)

So, what is Zo?

Using Rp=140Ω...
Zo = 140 / 1.5 = 93.33Ω
Therefore, Rs = 93.33-7 = 86.33Ω
But Rs = 61.9Ω!!!

Using Rp=110Ω...
110 / 1.5 = 73.33Ω
Therefore, Rs = 73.33-7 = 66.33Ω
But Rs = 61.9Ω!!!

Rs is a fixed 61.9Ω value.
And we know Zo=Rs+7Ω.
So Zo=61.9+7 = 68.9Ω
Therefore Rp=1.5Zo=103.35Ω

But that cannot be right because Zo will change according to the 2X_Q output clock frequency!

So the question is, what should Rp be for clock speeds higher than 40MHz?

I never changed Rs or Rp before, and I had been testing the full range of clock speeds from 33Mhz (stock) all the way to 50MHz (stable, but some freezing when booting 7.6.1 and 8.1 and even 7.5.5, although 7.1 always worked. But the fact remains we have a HF transmission line which requires proper termination via resistors, hence this is a valid question to ponder.

The 88920 datasheet has this Application Note pertaining to Termination of the 2X_Q output:

Two specs (tRISE/FALL and tPULSE Width 2X_Q output,
see AC Specifications) guarantee that the MC88920
meets the 20MHz and 25MHz 68040 P–Clock input
specification (at 40MHz and 50MHz). For these two specs
to be guaranteed by Motorola, the termination scheme
shown in Figure 3 must be used. For applications which
require 1.5V thresholds, but do not require a tight duty
cycle the Rp resistor can be ignored.

Bear in mind the 88920 chip is only rated for 25MHz CPU operation (50MHz on the 2X_Q output).

The 88916 datasheet has this Application Note pertaining to Termination of the 2X_Q output:

The pulse width spec for the Q and 2Q_X outputs is
referenced to a VCC/2 threshold. To translate this down
to a 1.5V reference with the same pulse width tolerance,
the termination scheme pictured in Figure 3 must be
used. This termination scheme is required to drive the
PCLK input of the 68040 microprocessor with the 88916
outputs.

The 88916DW80 allows for a CPU clock speed of 40MHz, which means 80MHz on the 2X_Q output.

Thoughts?