SE/30 Y3 Crystal Replacement Recommendation?

[ComputerCompulsive]

Greetings,
Wondering if someone can help me to replace the Y3 Crystal one of my three broken SE/30 boards I'm trying to repair.
Afraid that the legs just couldn't take an ultrasonic cleaning and broke off right at the base; I tried replacing it with one of these https://www.mouser.ca/ProductDetail/774-ATS169-E units I had from mouser as I thought that the frequency was correct but it doesn't appear to work.

I get no sound when I try this crystal in the board with known working sound (immediate death chime) and all it did was crackle. If I put an original crystal in I get normal sound on the same board so obviously I've missed something.

Anyone know what will work and maybe a link to mouser or digikey please?

Thanks for any assistance!

EDIT: Okay I think I found my error in the Crystal I ordered. Was talking with a local electronics expert and he suggested that I check the capacitors around the oscillator on the schematic and see what pf they are vs the pf of the Crystal I ordered. Sure enough I ordered a 20pf Crystal but the schematic suggests I need 33pf. Mouser doesn't seem to have any 33pf Crystals though (closest I found was 30pf). So does anyone know of an alternative that works?


Note the board I'm working on has video but no sound atm (well some sounds sometimes if I hit restart and interrupt a few times I almost hear the death chimes)

 

[naruse]

hi there! -- where you able to figure this out? I am actually in the same exact situation with the Y3 crystal :/

 

[JDW]

I realize you aren't doing an SE/30 Reloaded build, but sometimes those builds have BOMs with a modern parts list you can buy today. The SE Reloaded build's BOM is like that. Even so, I've never done an SE/30 Reloaded build myself and therefore have no experience selecting parts such as the Y3 Crystal, but I would have thought there would be a complete BOM at the Reloaded Github created by @Bolle and yet I see that no Y3 crystal is mentioned there. Maybe the SE/30 Reloaded build expert, @This Does Not Compute , may have some ideas? Or perhaps, @Kai Robinson ?

 

[naruse]

yeah, the one from @Bolle is for the bottom part AFAIK -- indeed @This Does Not Compute shared his BOM (https://docs.google.com/spreadsheets/d/1Nx-KRaF5cdhcJSdrX42SqrtQcKcN2UZhHbEiThW19gc/edit?usp=sharing) (which is the one I am guiding myself from (thanks again! )) and indeed for Y3 it just says

CPU crystal

Y3

1

TH

Transplant

16.9344MHz, seemingly not available

Which is true -- you can find some parts on gidikey but I think the pf that it requires (which is 33pf) is not existant, the closest one is 30pf as @ComputerCompulsive mentioned.

I havent delved too much into it yet but I think (from what ChatGPT said) that if you can change the capacitance of the nearby capacitors you might be able to change a bit the crystal (but this is not my area of expertise -- I'm. a SW dev after all )

 

[chuck.dubuque]

yeah, the one from @Bolle is for the bottom part AFAIK -- indeed @This Does Not Compute shared his BOM (https://docs.google.com/spreadsheets/d/1Nx-KRaF5cdhcJSdrX42SqrtQcKcN2UZhHbEiThW19gc/edit?usp=sharing) (which is the one I am guiding myself from (thanks again! )) and indeed for Y3 it just says

CPU crystal

Y3

1

TH

Transplant

16.9344MHz, seemingly not available

Which is true -- you can find some parts on gidikey but I think the pf that it requires (which is 33pf) is not existant, the closest one is 30pf as @ComputerCompulsive mentioned.

I havent delved too much into it yet but I think (from what ChatGPT said) that if you can change the capacitance of the nearby capacitors you might be able to change a bit the crystal (but this is not my area of expertise -- I'm. a SW dev after all )

30pf "should" be close enough. Should. If the 30pf crystal does not work, or does not work reliably, you can try soldering a 3pf non-polarized capacitor between the pins of the 30pf crystal. This would effectively create a 33pf crystal.

 

[JDW]

Examining my stock SE/30 motherboard and the BOMARC schematic, I see this...

Looking at Bolle's Github Gerber files (via JLCPCB website viewer), I see on the top side of the board, he has designed it to accept either the old cylinder-shaped crystal (shown above) or the kind more commonly used today:

And on the bottom of Bolle's PCB, we see pads for the required two capacitors:

So in terms of what should work on an SE/30 Reloaded Motherboard, if one wishes to use the aforementioned ATS169-E on Mouser, which clearly shows a Load Capacitance of 20pF (which is a very common value, by the way), we need to calculate what each of the two external capacitors should be (C33 & C34).

I believe the board is using a Pierce Oscillator circuit , which is typically used with MCUs.

We unfortunately must make some assumptions before proceeding. There is something called parasitic (stray) capacitance inherent to any circuit, which pertains to how the circuit is designed (trace length, width, proximity to other traces, etc.) We also have input and output capacitances of the microcontroller (Cin & Cout). Unfortunately, Cin & Cout are almost never given in datasheets (even if we had one), and the parasitic capacitance is also an unknown.

Assuming stray capacitance (inherent to the board design) of 3pF is reasonable (but it could be more because thru-holes are used).

Assuming Cin & Cout to be 5pF each, is also reasonable.

We can combine those two assumptions into Cstray = 8pF.

The total Load Capacitance (CL, as defined in the XTAL datasheet, which in our case is 20pF) is Cstray plus the external capacitor value you must choose, which we can call Cext in the following formula:

CL = Cext/2 + Cstray
20pF = Cext/2 + 8pF
Cext = C33 = C34 = 24pF

If the assumptions are correct (and only testing would show if they are), you would need 2pcs of 24pF C0G with a voltage rating anywhere from 16V to 50V (whatever physical size will fit the pads on the bottom of the motherboard).

If your C33 & C34 are currently the stock 33pF values, we can see why using the ATS169-E (CL=20pF) is not producing the oscillation expected by the circuit. The frequency is being shifted too far off due to C33 & C34 being non-optimal values. But because we've made some assumptions, it's best to have a range of capacitor values to solder-in and test. Once the correct value is found for C33 & C34, that helps everyone because you will know what XTAL p/n and what C33, C34 values to use.

​

 

[JDW]

But now that I ponder it further, it very well could be ESR that's the problem here, rather than Load Capacitance.

You see, those cylindrical CAN oscillators are typically used in clocks and watches, and in those applications the XTAL ESR is very high, on order of 50kΩ, 90kΩ or higher. But if you read the ATS169-E datasheet , ESR is only 40Ω. Normally, you cannot use a 40Ω XTAL in a circuit as a substitute for one that is 90kΩ.

You might try a resistor to see if you can get the ATS169 to work. But there are caveats to that. Normally, a series resistor of a modest size is placed on the "OUT" pin of the XTAL to reduce current a bit. See diagrams below.

SE/30 SCHEMATIC:

CAVEAT
You would possibly need to add 50kΩ of resistance, which is rather uncommon and could degrade the quality of the oscillation or even prevent oscillation from happening altogether. Starting out with a smaller value like 10kΩ and going up from there to about 50kΩ is reasonable for testing. If it works reliably and consistently, fantastic. If not, there would be no alternative but to try and find a cylindrical 2-pin oscillator at the required frequency. Even then if the ESR is off too much related to the stock Y3 part (197-39) , you may have trouble.

 

[Bolle]

I have used these before as drop in replacements: ABLS-16.9344MHZ-B4-T
The load capacitance doesn't match the original either but it has always worked fine for me so far (on both original logicboards that needed the crystal replaced because it got eaten by cap fluid, or when building SE/30 and IIfx reloaded boards)

 

[This Does Not Compute]

I have used these before as drop in replacements: ABLS-16.9344MHZ-B4-T

Thanks, I've gotten this one added to my spreadsheet. Do you happen to have a recommendation for Y2 as well?

 

[Bolle]

Do you happen to have a recommendation for Y2 as well?

Looking through my mouser history it seems I have ordered 38-HX5F12.5-32.768KHz before and they again seem to be fine as drop in replacements.

 

[JDW]

@ComputerCompulsive & @naruse

When you've had opportunity to test the drop-in replacement recommended by @Bolle and/or try the other information provided in this thread, please be sure to let us know if your Y3 crystal issue was resolved and what specifically resolved it.

Thanks!

 

[GreenBar0n]

Wondering if my Y3 has been damaged from a leaking cap. My UE10 had four traces severed, I had to remove UE10 and reinstall it to fix the traces.

My Y3 issue looks like this:

Leg farthest from UE10:

Leg closest to UE10:

Should both legs of Y3 be identical in wave form, like Y1 is?

No Chime and simasima - CPU is not walking the bus correctly according to the Repair Guide.

Thanks.

 

[JDW]

Before discussing the details of waveform shapes, we need to assess if your scope probe is using a short or long ground lead, which is discussed in the following video:

 

[GreenBar0n]

Before discussing the details of waveform shapes, we need to assess if your scope probe is using a short or long ground lead, which is discussed in the following video:

I was using a very long ground lead, I'll make it as short as possible and measure again at Y3, UE10 and C33 and C34, to make certain of what I'm seeing here. I'll do this now. Thanks for the reply!

 

[GreenBar0n]

I shortened my probe ground lead, or got rid of the extension I was using - this green one:

And used only the ground lead the probe came with:

This made the ringing less on the closest Y3 leg to UE10, which was the really odd pattern before:

But now the Y3 leg farthest from UE10 isn't even showing a wave. Measuring this way:

Result is no wave form at all without that long green ground lead:

Not sure how this is possible, or what I'm doing incorrectly at this point.

Thanks for your help!

 

[JDW]

Adding a long green extension lead like that for Ground is a huge no-no when making HF measurements, but I must go further to say that your current black ground lead is still too long...

In the video I linked for you, you will see the video creator shows he initially used the same black ground lead as you are now using, and his use of that black ground lead was causing all kinds of waveform distortions. But then he switched to a very short ground lead which eliminated much of the distortion. Something like this:

Problem is, you'd need to find a solid GND connection in close proximity to your test point. But the reality is, that is the only way to get a clean waveform reading on a scope when measuring high frequency (MHz, not kHz).


As to not getting a scope reading, it's a matter of using the wrong ground or not having a solid connection to ground or a solid connection to your test point.


Lastly, be sure to download the SE/30 Repair Guide from @Mac84 here because it shows you interesting things like what those CRT display artifacts signify:

https://tinkerdifferent.com/threads/macintosh-se-30-vertical-horizontal-bars-on-display-simasima-repair-guide.4705/#post-40662

 

[GreenBar0n]

I have been working with the Simasima Repair Guide for about three days now, I'll post some things I found, approaching it from a new users perspective in a moment.

I definitely watched the video you posted regarding the shortest ground probe lead possible, and I got a bit confused by the buffer circuit, but I understand how having long leads could cause ringing, thank you for that.

But here's something strange to add to this, I put the long green probe ground lead back in:

And measured the Y3 leg that had the strange waveform I posted earlier:

- and now it's an even cleaner wave form than with the short ground lead shown above, just moments ago:

But now the Y3 leg that is farthest from UE10, that was a good wave form earlier, is gone, like nothing is there.

Should I try to replace Y3? Pin 24 and 25 on UE10 shows the same wave conditions, as well as C33 and C34.

I have a Reloaded V5 board that I was going to build, but I'm trying to make sure all of my components are in good order before swapping them all over.

Thanks again!

 

[JDW]

When you watch that video I linked for you, ignore his use of a buffer and hex invertor. That isn't relevant. I merely linked it for you to show how shortening the GND lead worked wonders in cleaning up his measured waveforms.

I can only say that lengthening a scope probe's GND lead should not normally lead to cleaner signal measurements! It not only adds resistance, but the long ground acts like a radio antenna to pick up interference which ultimately gets into your scope in the form of waveform distortions.

I'm going to tag @Willj because he has done many SE/30 RELOADED boards, including one from me, which works perfectly. Maybe Will might have some insights to share about the Y3 XTAL. Not sure if he always uses the stock Y3 XTAL or if he has ever used the substitute XTALs mentioned earlier in this thread. I will send him an email now to ensure he sees this thread.

 

[GreenBar0n]

I made the shortest ground I could.

Stuck it in the main power connector ground:

Wave form is about the same as the long green extension, somehow:

But the Y3 leg that is farthest from UE10, that had the cleanest wave form yesterday, doesn't show anything with the short ground:

Look what happens to that same leg when there's no ground at all, it's a ringing mess, but shows the correct wave form and speed.

I'll keep at this and try to figure out where I'm going wrong. Looks like Y3 is good though.

 

[JDW]

Will is across the globe and already in bed, but he said he needs to do some testing tomorrow morning his time so he can reply to this thread then.