English Amiga Board


Go Back   English Amiga Board > Support > support.Hardware

 
 
Thread Tools
Old 23 April 2019, 20:53   #21
Mathesar
Registered User

Mathesar's Avatar
 
Join Date: Aug 2014
Location: Netherlands
Posts: 197
Found the problem

Update:
As emulation (WinUAE) and reality (vintage A3000 computer) did not show the same behaviour I have done some hardware checks. I think I found the problem, this is how the +3.3V supply to the 060 looks like:
Click image for larger version

Name:	cpu_3.3Vsupply.png
Views:	58
Size:	4.8 KB
ID:	62885
That is not good. The LT1084-3.3 is oscillating!!!! the frequency of the oscillation is approx. 70kHz and the amplitude is 720mV peak-peak.
After checking the datasheet and the A3660 implementation, it seems that the reason is the decoupling capacitors on the +3.3V rails. The datasheet of the LT1084 specifies a minimum capacitance of either 10uF tantalum or 50uF electrolytic. Note that they don't specify ceramic caps. That is because these circuits typically rely on some capacitor series resistance for stability. Now, in the A3660 we have 2*10uF plus some 100nf caps on the +3.3V rails. I have used 10V Murata X7R caps. As ceramics caps must be derated for any voltage above 0V we probably have about 12..14uF on the rail. Not much. Also, because they are ceramic caps the ESR is very low and probably not enough to dampen the regulator hence it oscillates.

To fix it I have added a 100uF electrolytic cap (a cheap one so probably plenty of ESR) like so:
Click image for larger version

Name:	quick_fix.jpg
Views:	78
Size:	41.9 KB
ID:	62886
And indeed, that stopped the oscillations.
And also, it fixed all the instabilities!!!! The computer now runs perfectly stable running adoom, workbench, diropus, hippoplayer, etc, etc.
So, apart from the mathieeesingbas.library and the 68060.library issues there was thus also a hidden hardware problem.

The quick fix I now applied is a bit ugly so I am looking at a more permanent solution. As the lt1084 runs a bit hot (it must dissipate about 1.5 to 2.0 Watts I estimate) I am looking at a switching DC/DC converter on a small PCB that perfectly fits on the LT1084's spot. This is a cleaner solution that would also reduce the power consumption a bit.
I would be surprised if I am the only one with oscillating power supplies as the design seems very marginal. Apperently the 68060 is quite tolerant to bad power supplies as the problem only manifested itself once 68060.library was loaded. Maybe it goes *just* right on other systems.

Any thoughts?

Last edited by Mathesar; 23 April 2019 at 21:18.
Mathesar is offline  
Old 27 April 2019, 21:26   #22
Mathesar
Registered User

Mathesar's Avatar
 
Join Date: Aug 2014
Location: Netherlands
Posts: 197
A quick preview. This is a tlv62569 based 2Amp 3.3V DC/DC converter for the A3660:
Click image for larger version

Name:	A3660_DC_DC.jpg
Views:	57
Size:	111.6 KB
ID:	62929
This converter is designed to fit exactly in the space where the lt1084 would normally be. The efficiency should be >90% reducing heat and power consumption.
A few more checks and I will order some boards.
Mathesar is offline  
Old 13 May 2019, 20:52   #23
Mathesar
Registered User

Mathesar's Avatar
 
Join Date: Aug 2014
Location: Netherlands
Posts: 197
Another update, boards came in today and I assembled one.

It is actually difficult to measure the efficiency and current output of this thing because I don't have the right dummy loads and I have to measure the voltages and currents very accurately. First impressions are good though, it delivered 1.5A into my dummy load quite easily. Before I plug it into my real Amiga I will measure the startup behaviour.

Click image for larger version

Name:	WP_20190513_20_41_04_Pro (2).jpg
Views:	50
Size:	240.5 KB
ID:	63067
Mathesar is offline  
Old 14 May 2019, 02:33   #24
grelbfarlk
Registered User

 
Join Date: Dec 2015
Location: USA
Posts: 1,752
Quote:
Originally Posted by Mathesar View Post
Update:
As emulation (WinUAE) and reality (vintage A3000 computer) did not show the same behaviour I have done some hardware checks. I think I found the problem, this is how the +3.3V supply to the 060 looks like:

That is not good. The LT1084-3.3 is oscillating!!!!
And indeed, that stopped the oscillations.
And also, it fixed all the instabilities!!!! The computer now runs perfectly stable running adoom, workbench, diropus, hippoplayer, etc, etc.
So, apart from the mathieeesingbas.library and the 68060.library issues there was thus also a hidden hardware problem.

The quick fix I now applied is a bit ugly so I am looking at a more permanent solution. As the lt1084 runs a bit hot (it must dissipate about 1.5 to 2.0 Watts I estimate) I am looking at a switching DC/DC converter on a small PCB that perfectly fits on the LT1084's spot. This is a cleaner solution that would also reduce the power consumption a bit.
I would be surprised if I am the only one with oscillating power supplies as the design seems very marginal. Apperently the 68060 is quite tolerant to bad power supplies as the problem only manifested itself once 68060.library was loaded. Maybe it goes *just* right on other systems.

Any thoughts?

That is very interesting, I have some 060 adapters and they mostly work fine. However I have had a couple where certain combinations of this 060 board with this 060 chip works in the A3640, but not in the Warpengine. I figured it was slightly bad soldering or the sockets where not perfectly clean or something to that effect.


I have one system with an 060-66MHz WarpEngine and a Mediator board with a lot of stuff in the system. It's stable running just the 060 but when I start the PCI PPC card I start getting 060 crashes.



I'm going to give this modification a try.
grelbfarlk is offline  
Old 14 May 2019, 09:45   #25
utri007
mä vaan
 
Join Date: Nov 2001
Location: Finland
Posts: 864
Prety muc all 68040/68060 programs and libraries suppose that there is a FPU.

Havin 68060 without FPU is not a smart. Remember that if you buy full 68060 CPU with 120€, you can get sama amount of money it maybe even more latter. It is a investment, unlike many other (new) computer related stuff.
utri007 is offline  
Old 14 May 2019, 10:51   #26
Mathesar
Registered User

Mathesar's Avatar
 
Join Date: Aug 2014
Location: Netherlands
Posts: 197
Quote:
Originally Posted by utri007 View Post
Prety muc all 68040/68060 programs and libraries suppose that there is a FPU.

Havin 68060 without FPU is not a smart. Remember that if you buy full 68060 CPU with 120€, you can get sama amount of money it maybe even more latter. It is a investment, unlike many other (new) computer related stuff.
Have you read the entire thread? Although I did have some initial problems related to not having an FPU it turned out that there was a problem with the hardware.
I am now setting up my system with the LC060 and the solution to the problem you describe is pretty simple: do not install the 060 versions but install the 020/030 versions.
I made of habit on that anyway because when for whatever reason I have to revert to the A3000's onboard 030, nothing will work anymore!

Last edited by Mathesar; 14 May 2019 at 13:33.
Mathesar is offline  
Old 16 May 2019, 11:23   #27
Kin Hell
Registered User
Kin Hell's Avatar
 
Join Date: Mar 2009
Location: Cornwall, UK
Posts: 644
Quote:
Originally Posted by utri007 View Post
Prety muc all 68040/68060 programs and libraries suppose that there is a FPU.

Havin 68060 without FPU is not a smart. Remember that if you buy full 68060 CPU with 120€, you can get sama amount of money it maybe even more latter. It is a investment, unlike many other (new) computer related stuff.

Quote:
Originally Posted by Mathesar View Post
Have you read the entire thread? Although I did have some initial problems related to not having an FPU it turned out that there was a problem with the hardware.
I am now setting up my system with the LC060 and the solution to the problem you describe is pretty simple: do not install the 060 versions but install the 020/030 versions.
I made of habit on that anyway because when for whatever reason I have to revert to the A3000's onboard 030, nothing will work anymore!

Quote:
Originally Posted by Kin Hell View Post
....Alternatively, use a Full 060 with FPU & all will be Golden.
LC CPU's are Failed FPU Fabrications & were binned accordingly at the time of Manufacture.

The ground work done by Chucky for the Amiga Community is amazing, but perhaps there are one or two Foibles with the A3660 and certain HW configurations??

Kudos to you for finding a hardware hack/modification, to rectify your issue though. - Truly amazing work fella & well done.
Kin Hell is offline  
Old 16 May 2019, 15:28   #28
Mathesar
Registered User

Mathesar's Avatar
 
Join Date: Aug 2014
Location: Netherlands
Posts: 197
Quote:
Originally Posted by Kin Hell View Post
The ground work done by Chucky for the Amiga Community is amazing, but perhaps there are one or two Foibles with the A3660 and certain HW configurations??

Kudos to you for finding a hardware hack/modification, to rectify your issue though. - Truly amazing work fella & well done.
Thanks. And yes, Chucky is a hero!
Mathesar is offline  
Old 16 May 2019, 19:01   #29
Mathesar
Registered User

Mathesar's Avatar
 
Join Date: Aug 2014
Location: Netherlands
Posts: 197
A3660 dc/dc

Final update, I have soldered the DC/DC converter on the A3660 and it works! The converter runs stable and stays really cool.

Before I soldered the converter on the A3660 I checked the startup behaviour. The 68060 cannot tolerate voltages higher than 4V above VCC(+3.3V) on any of it's IO pins. The power supply must guarantee this and Motorola recommends putting 3 bypass diodes in series over the voltage regulator. The 040/060 adapter and thus the A3660 don't have these diodes but in practice the +5V of the A3000 ramps up so slowly that the difference between the +5V and +3.3V rails never goes beyond 4V when you use the LT1084. With a switching regulator like this however it could be a problem as most switching regulators have features like startup delays, undervoltage lockouts, soft starts and so on. So, I added the series diodes as per Motorola's recommendation. This is the startup behaviour when the +5V ramps up in about 10us (on a real A3000 the +5V rail ramps up much slower):
Click image for larger version

Name:	A3660_DCDC_Startup.png
Views:	33
Size:	5.0 KB
ID:	63101
The regulator starts after 600us but the diodes keep the difference well below 4V while the converter is not yet switching.

This is how the converter looks on an A3660:
Click image for larger version

Name:	WP_20190516_13_17_50_Pro.jpg
Views:	52
Size:	474.8 KB
ID:	63102

Here are the KiCad files for anyone who is interested. Feel free to use/modify it as you see fit. Use at your own risk! I have some boards left so if you are interested send me a pm.
A3660_DC_DC_rev1.0.zip

EDIT: added BOM and PDF schematics to zip archive for those who do not have KiCad installed.

Last edited by Mathesar; 17 May 2019 at 13:18.
Mathesar is offline  
Old 17 May 2019, 11:20   #30
Mathesar
Registered User

Mathesar's Avatar
 
Join Date: Aug 2014
Location: Netherlands
Posts: 197
Oh, forgot to mention:
I have designed the DC/DC board in such a way that you can solder it directly to the A3660 without any pins or wires.

Pre-tin the A3660 LT1084 solder holes with solder (make a nice small bump) and position the holes in the DC/DC board over the pretinned LT1084 solder holes on the A3660. Now apply plenty of solder. The underside of the DC/DC board must not touch the A3660 main board so I just put a piece of paper inbetween during soldering. Check for continuity and shorts using a multimeter and you should be good
Mathesar is offline  
Old 17 May 2019, 13:42   #31
Thomas Richter
Registered User
 
Join Date: Jan 2019
Location: Germany
Posts: 204
Quote:
Originally Posted by Mathesar View Post
Thor's mmulib 060 library crashes on me with a recoverable alert (error 3f02803e) with or without fpu. This happens as soon as setpatch is executed. Maybe this library doesn't like something about the 3000. I will investigate more.
This error means that the mmu.library could not read back its own data. I've seen this before: it happens if there are bit-flips on the bus, and the data that returns from the memory is different than what was written to memory.

IOWs, you will not get happy with any library before you resolve this hardware problem - don't shoot the messanger, the P5 library will not make your board working either, it just will not detect this problem.
Thomas Richter is offline  
Old 17 May 2019, 18:29   #32
esel
Registered User
 
Join Date: Jul 2012
Location: Kassel
Posts: 129
Quote:
Originally Posted by Mathesar View Post
Final update, I have soldered the DC/DC converter on the A3660 and it works! The converter runs stable and stays really cool.


Attachment 63102

Here are the KiCad files for anyone who is interested. Feel free to use/modify it as you see fit. Use at your own risk! I have some boards left so if you are interested send me a pm.
Attachment 63115

EDIT: added BOM and PDF schematics to zip archive for those who do not have KiCad installed.

Nice ! is it possible to add a Jumper to select between 3,3 and 5 Volt ?
so it would be easy to switch between a 040 an a 060.
esel is offline  
Old 17 May 2019, 19:09   #33
Mathesar
Registered User

Mathesar's Avatar
 
Join Date: Aug 2014
Location: Netherlands
Posts: 197
Quote:
Originally Posted by Thomas Richter View Post
This error means that the mmu.library could not read back its own data. I've seen this before: it happens if there are bit-flips on the bus, and the data that returns from the memory is different than what was written to memory.

IOWs, you will not get happy with any library before you resolve this hardware problem - don't shoot the messanger, the P5 library will not make your board working either, it just will not detect this problem.
I won't shoot you… without your loadmodule tool my system wouldn't work at all.
The last time I tried your library I hadn't resolved the power supply problem yet. Also, I remember the 68060.library only guru'd when mmu.library was not installed. (At first I thought mmu.library was optional). I will give it another try now that the power supply is fixed.
Mathesar is offline  
Old 19 May 2019, 12:19   #34
Mathesar
Registered User

Mathesar's Avatar
 
Join Date: Aug 2014
Location: Netherlands
Posts: 197
Quote:
Originally Posted by esel View Post
Nice ! is it possible to add a Jumper to select between 3,3 and 5 Volt ?
so it would be easy to switch between a 040 an a 060.
Feel free! Should be possible using some heavy duty jumpers.
Mathesar is offline  
Old 19 May 2019, 12:22   #35
Mathesar
Registered User

Mathesar's Avatar
 
Join Date: Aug 2014
Location: Netherlands
Posts: 197
Quote:
Originally Posted by grelbfarlk View Post
That is very interesting, I have some 060 adapters and they mostly work fine. However I have had a couple where certain combinations of this 060 board with this 060 chip works in the A3640, but not in the Warpengine. I figured it was slightly bad soldering or the sockets where not perfectly clean or something to that effect.

I have one system with an 060-66MHz WarpEngine and a Mediator board with a lot of stuff in the system. It's stable running just the 060 but when I start the PCI PPC card I start getting 060 crashes.

I'm going to give this modification a try.
I am very interested in your results. Like I said, I can't imagine I am the only one with this problem.
Mathesar is offline  
Old 19 May 2019, 14:57   #36
esel
Registered User
 
Join Date: Jul 2012
Location: Kassel
Posts: 129
Quote:
Originally Posted by Mathesar View Post
Feel free! Should be possible using some heavy duty jumpers.

i cant do it by myself thats why i am asking...
esel is offline  
Old 19 May 2019, 19:10   #37
Mathesar
Registered User

Mathesar's Avatar
 
Join Date: Aug 2014
Location: Netherlands
Posts: 197
Quote:
Originally Posted by esel View Post
i cant do it by myself thats why i am asking...
Understood.
When I think a bit more about how to do such a jumper in a *safe* way I realise it is actually not so straightforward. Better leave the converter as it is.
Mathesar is offline  
Old 19 May 2019, 19:14   #38
Mathesar
Registered User

Mathesar's Avatar
 
Join Date: Aug 2014
Location: Netherlands
Posts: 197
Quote:
Originally Posted by Thomas Richter View Post
This error means that the mmu.library could not read back its own data. I've seen this before: it happens if there are bit-flips on the bus, and the data that returns from the memory is different than what was written to memory.

IOWs, you will not get happy with any library before you resolve this hardware problem - don't shoot the messanger, the P5 library will not make your board working either, it just will not detect this problem.
Update: I have installed the latest mmulib from aminet and it is rock stable. In fact, I am typing this on my A3000 while the machine has already been running for 6 hours+ running ibrowse, sysspeed, shell and a bunch of commodities.
Mathesar is offline  
 


Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)
 
Thread Tools

Similar Threads
Thread Thread Starter Forum Replies Last Post
New A3660 cards on eBay macce2 Amiga scene 26 11 March 2019 10:33
Building a A3660 kit Cav Hardware mods 1 07 January 2019 21:26
Let me present the A3660 Chucky Hardware mods 51 13 July 2018 09:08
How to add 512KB ROM to A3660 ? salocinx Hardware mods 8 17 April 2018 01:39
Pushing the A3660 to the next step Chucky support.Hardware 15 27 November 2017 22:44

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump


All times are GMT +2. The time now is 00:50.


Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2019, vBulletin Solutions Inc.
Page generated in 0.09412 seconds with 16 queries