English Amiga Board


Go Back   English Amiga Board > Support > support.Hardware > Hardware mods

 
 
Thread Tools
Old 02 February 2018, 07:20   #181
Chucky
Registered User

Chucky's Avatar
 
Join Date: Mar 2015
Location: Karlstad / Sweden
Age: 48
Posts: 615
There have been reports that 3660 have issues on A3000. but I have tried several here at home with no problems. but t is quite known that the 3640 in the 3000 is not really optimal so I guess the issue is there.
Chucky is offline  
Old 02 February 2018, 07:40   #182
ajk
Registered User
ajk's Avatar
 
Join Date: May 2010
Location: Helsinki, Finland
Posts: 1,305
I am also running the A3660 (and an original A3640 before it) on an A3000 with no issues so far. Although I haven't done extensive load testing or anything like that.
ajk is offline  
Old 07 February 2018, 19:22   #183
salocinx
Registered User

 
Join Date: Jan 2014
Location: Zurich / Switzerland
Posts: 79
I meanwhile exchanged the DMAC-02 with another DMAC-02 and did intensive testing with my SCSI HDD. No more problems Although the A3660 is still in my mind
salocinx is offline  
Old 19 February 2018, 07:40   #184
Amiga87
Registered User
 
Join Date: Oct 2017
Location: Sydney/Australia
Posts: 25
Hi,

First up, many thanks to Chucky for open sourcing the A3640r3.3 and A3660.

I've built a handful of these boards now and most parts were easy to come by (except for one) with some patience and industry knowledge.

My tiny contribution to these projects is going to be a a suggestion for a part substitution. As many have discovered, sourcing the F1803 and F803 at reasonable cost is becoming difficult. Alternate part suggestion is the 74F175. Not quite pin-for-pin compatible, but almost. F803/1803 are 14 pin F175 is 16 pin.
  • Cut track to pin 1 (Gnd) of U101 (right near the via is a good spot). Pin 1 of 74F175 is ~CLR, needs to be tied high to Vcc.
  • Remove pins 7 and 10 of the 74F175.
  • Bend pins 8 and 9 towards where the adjacent removed pins were located. These are the Gnd and Clk pins respectively.
  • Solder the IC in place so that pin 1 of the 74F175 is on the pad of pin 1 of U101.
  • You will most likely need a small solder bridge to the 2 bent pins from the pads.
  • Solder a wire link between pins 1 and 16 of the 74F175 (this pulls ~CLR high).

Additional advantage of the F175 is that it is available as a significantly faster part (F803 maxed out at about 70MHz).

Good luck,
A87
Amiga87 is offline  
Old 19 February 2018, 10:24   #185
amigasith
Registered User

amigasith's Avatar
 
Join Date: Jan 2013
Location: Wild South / Germany
Age: 44
Posts: 238
Hi Amiga87, thank you very much for this very nice piece of information! What would make it even more valuable would be a picture Could you post one? This would be awesome
amigasith is offline  
Old 19 February 2018, 12:42   #186
Mathesar
Registered User

Mathesar's Avatar
 
Join Date: Aug 2014
Location: Netherlands
Posts: 302
Yes, Amiga87 Thanks for this!!!!!
I am collecting parts at the moment and was also thinking about alternative parts for the clock divider. I was mainly worried about the skew and fanout of chips like the 74F175.
But you tested it and it works!

I also have a suggestion for the delay line: the DS1100Z-25. This one comes in a SO-8 package and could replace the original DIL delay line. It is readily available from the likes of Farnell, Digikey, etc.

Maybe I can edit the gerbers to support these chips.
Mathesar is online now  
Old 19 February 2018, 13:03   #187
Amiga87
Registered User
 
Join Date: Oct 2017
Location: Sydney/Australia
Posts: 25
Quote:
Originally Posted by Mathesar View Post
Yes, Amiga87 Thanks for this!!!!!
I am collecting parts at the moment and was also thinking about alternative parts for the clock divider. I was mainly worried about the skew and fanout of chips like the 74F175.
But you tested it and it works!
Yes, works fine.

I have run up to between 50 to 66.67MHz (fastest osc I have at the moment).
Attached Thumbnails
Click image for larger version

Name:	IMAG0290_1.jpg
Views:	241
Size:	372.4 KB
ID:	56876  
Amiga87 is offline  
Old 19 February 2018, 13:07   #188
hooverphonique
ex. demoscener "Bigmama"

 
Join Date: Jun 2012
Location: Fyn / Denmark
Posts: 1,093
Quote:
Originally Posted by Amiga87 View Post
My tiny contribution to these projects is going to be a a suggestion for a part substitution. As many have discovered, sourcing the F1803 and F803 at reasonable cost is becoming difficult. Alternate part suggestion is the 74F175. Not quite pin-for-pin compatible, but almost. F803/1803 are 14 pin F175 is 16 pin.
Do we know anything about the inter-flipflop propagation delay variation on this? The 1803 is specified to be within 2ns, but as far as I can see, nothing is spec'd for 175.. I don't know how much of a problem this may present, but I guess there's a reason C= used the (1)803..

Last edited by hooverphonique; 19 February 2018 at 15:54. Reason: replace "latency" with "propagation delay" for clarity
hooverphonique is offline  
Old 19 February 2018, 14:38   #189
Mathesar
Registered User

Mathesar's Avatar
 
Join Date: Aug 2014
Location: Netherlands
Posts: 302
Quote:
Originally Posted by hooverphonique View Post
Do we know anything about the inter-flipflop latency variation on this? The 1803 is specified to be within 2ns, but as far as I can see, nothing is spec'd for 175.. I don't know how much of a problem this may present, but I guess there's a reason C= used the (1)803..
I also worry about that, it's called skew. Apperently it works though. Maybe it isn't such critical after all. There also exist parts like the 74ACT175 which are more modern (being CMOS) and thus maybe show less inter-flipflop varation.
I also looked at the 74ACQT574 by fairchild that has a guaranteed inter-flipflop skew less than 1.0ns. Even better than the 1803. On top of that it also has a large fanout. Ideal for a clock driver.

One other worry I have is the clock to output delay of the flipflops. This should be less than 10ns (For a 25MHz 040) otherwise the bclk and pclk have too much skew between them.

But hey, the proof is in the pudding
Mathesar is online now  
Old 19 February 2018, 15:57   #190
hooverphonique
ex. demoscener "Bigmama"

 
Join Date: Jun 2012
Location: Fyn / Denmark
Posts: 1,093
Quote:
Originally Posted by Mathesar View Post
One other worry I have is the clock to output delay of the flipflops. This should be less than 10ns (For a 25MHz 040) otherwise the bclk and pclk have too much skew between them.
Well, let's hope they are within the typical delay spec then, because otherwise you'll be in trouble
hooverphonique is offline  
Old 19 February 2018, 20:01   #191
Mathesar
Registered User

Mathesar's Avatar
 
Join Date: Aug 2014
Location: Netherlands
Posts: 302
Quote:
Originally Posted by hooverphonique View Post
Well, let's hope they are within the typical delay spec then, because otherwise you'll be in trouble
Yes, but maybe that can be improvedl. I think there are two free gates on the 74F04 so we can delay the pclk (040/060 cpu clock) by running it through the 2 free NOT gates.
The extra delay will be between 3ns and 12ns (worst case, 0..70deg).
The 74F175 has a delay between 4ns and 9.5ns (worst case, 0..70deg)
So absolutely worst case skew of pclk and bclk is between -8.0ns and +6.5ns. And that would probably never happen because both IC's will have approx. the same temperature. Typical values are probably only a few ns then. What about that?
Mathesar is online now  
Old 20 February 2018, 00:14   #192
Amiga87
Registered User
 
Join Date: Oct 2017
Location: Sydney/Australia
Posts: 25
Quote:
Originally Posted by Mathesar View Post
And that would probably never happen because both IC's will have approx. the same temperature. Typical values are probably only a few ns then. What about that?
That was my thinking. 74F175 on the data sheet was 2ns worse than the 74f1803 in a worse case scenario (over temp and voltage ranges).

If there is concern about the skew between the flip-flops:
- clk90 runs through the delay line anyway, which is configurable.
- bclk and cpuclk are inverted, if propagation skew of the flip-flops is really an issue, you can use the inverted output of the bclk for the cpuclk straight from the 74f175, but then you would lose a little of the delay from the 74f04 which may be desired?

Anyway, works for me (I used the TI part) and was stable during a 8 hour load test.

Worse case scenario if it doesn't work for someone, you'll need to source a 74f(1)803.

Regards,
A87

Last edited by Amiga87; 20 February 2018 at 00:22.
Amiga87 is offline  
Old 20 February 2018, 11:52   #193
Amiga87
Registered User
 
Join Date: Oct 2017
Location: Sydney/Australia
Posts: 25
Just an additional note:

I built the a3660, so the propagation delay of the flip-flops to the bclk is not critical as it is for the 040.

Nevertheless i did check 2 boards where i'm using the 74f175, prop delay is approx 5ns on the samples i have (input to output) and skew of the 3 clock outputs was less than 1ns.

A87
Amiga87 is offline  
Old 20 February 2018, 14:08   #194
Mathesar
Registered User

Mathesar's Avatar
 
Join Date: Aug 2014
Location: Netherlands
Posts: 302
Thumbs up

Quote:
Originally Posted by Amiga87 View Post
Just an additional note:

I built the a3660, so the propagation delay of the flip-flops to the bclk is not critical as it is for the 040.

Nevertheless i did check 2 boards where i'm using the 74f175, prop delay is approx 5ns on the samples i have (input to output) and skew of the 3 clock outputs was less than 1ns.

A87
So, no issues there. Nice mod!
I am collecting parts to build one of these cards and will also use the 74F175. It would be nice if we could modify the board to use this chip instead. Also, did you look at the DS1100Z-25 delay line? Same function, cheaper, better availability but sadly SO-8....
Mathesar is online now  
Old 21 February 2018, 23:00   #195
Chucky
Registered User

Chucky's Avatar
 
Join Date: Mar 2015
Location: Karlstad / Sweden
Age: 48
Posts: 615
Interesting. I will soon order a new batch of PCBs for me to have .. (while developing Rev2) I think I will replace the 74F1803 with this solution instead..
Chucky is offline  
Old 22 February 2018, 00:41   #196
Amiga87
Registered User
 
Join Date: Oct 2017
Location: Sydney/Australia
Posts: 25
Quote:
Originally Posted by Chucky View Post
Interesting. I will soon order a new batch of PCBs for me to have .. (while developing Rev2) I think I will replace the 74F1803 with this solution instead..
Could easily support both with a couple of extra 0R resistor links.

Quote:
Originally Posted by Mathesar
I also have a suggestion for the delay line: the DS1100Z-25. This one comes in a SO-8 package and could replace the original DIL delay line. It is readily available from the likes of Farnell, Digikey, etc.
Would work fine. Might be room for a dual footprint inside the existing DIP footprint if reworing the board. Or, just get a 8-pin SOIC to DIP board and wire to the existing PCB footprint.

Thanks again,
A87
Amiga87 is offline  
Old 22 February 2018, 08:42   #197
Mathesar
Registered User

Mathesar's Avatar
 
Join Date: Aug 2014
Location: Netherlands
Posts: 302
Thumbs up

Quote:
Originally Posted by Chucky View Post
Interesting. I will soon order a new batch of PCBs for me to have .. (while developing Rev2) I think I will replace the 74F1803 with this solution instead..
Sweet! Is this the A3660 and are you going to make it a revision 1.1 then ?

I am on the point of ordering A3660 PCB's but I am still contemplating to do the 75F175 / DS1100Z change myself but you can probably do it better.

In the meantime I will put out an order to Mouser
Mathesar is online now  
Old 23 February 2018, 10:04   #198
Chucky
Registered User

Chucky's Avatar
 
Join Date: Mar 2015
Location: Karlstad / Sweden
Age: 48
Posts: 615
I will check on a solution taking both 74F175 and the old (1)803 solution. as a rev 1.1 board.

As for development. I would need a good way of tapping into signals from GALs ertc to use on my rev2 with FPGA as glue-logic replacement. simply making some vias larger so I can solder in a wire.. so after this weekend I would have that done.

(solder a wire going to my devboard so I can replace chip for chip until all is in FPGA)
Chucky is offline  
Old 23 February 2018, 20:16   #199
Mathesar
Registered User

Mathesar's Avatar
 
Join Date: Aug 2014
Location: Netherlands
Posts: 302
Adapter

Quote:
Originally Posted by Amiga87 View Post
Or, just get a 8-pin SOIC to DIP board and wire to the existing PCB footprint.
Drawn up a quick board, I need to clean up the silk but... would anyone be interested in a DS1100Z-25+ to DS1000-25 adapter?
Attached Thumbnails
Click image for larger version

Name:	Nieuwe afbeelding (8).jpg
Views:	65
Size:	51.5 KB
ID:	56973  
Mathesar is online now  
Old 23 February 2018, 22:02   #200
Chucky
Registered User

Chucky's Avatar
 
Join Date: Mar 2015
Location: Karlstad / Sweden
Age: 48
Posts: 615
I have now implemented the 74F175 as an option (there is 2 U101 footprints now)

but will do some changes, I will do some vias large enough to solder a wire to it. (and via will be shown) this so I can put wires in where signals is and test my FPGA progress. while doing changes anyway
Chucky 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
Commodore A3640 Rev. 3.1 watchdog mod trixeo Hardware mods 2 01 March 2014 11:45
A3640 rev 3.2 bpco MarketPlace 1 27 May 2010 09:15
FS: Minimig PCBs and Keyrah FS Kristian95 MarketPlace 7 03 October 2007 12:33
Ebay at 1 USD : A3640 Rev 3.1 for Amiga 4000 running at 32 Mhz !!! Effy MarketPlace 21 01 August 2005 15:05
A3640 Rev 3.0 accelerator card problem Tony Landais support.Hardware 3 23 December 2003 09:41

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 14:36.


Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2020, vBulletin Solutions Inc.
Page generated in 0.10025 seconds with 16 queries