It is currently Sat Sep 21, 2019 7:08 pm

All times are UTC - 7 hours





Post new topic Reply to topic  [ 100 posts ]  Go to page Previous  1 ... 3, 4, 5, 6, 7
Author Message
PostPosted: Mon May 13, 2019 2:07 pm 
Offline
User avatar

Joined: Tue Feb 26, 2019 6:36 pm
Posts: 21
On the top loader NES-101, I was able to clean up the +5V signal by replacing almost all electrolytic caps with polymer electrolytic capacitors (I replaced the ones that mattered in this situation - Power). I replaced the main filter cap and the cap near the cartridge slot. I also added a 47uF in place of C35 (which was just a decoupling capacitor located way too far away from the PPU Pin 40 to do anything). This reduced the noise on Pin 40 of the PPU, but DID NOT change the noise on the video output. The noise on the Pin 40 was reduced from ~60mV pk-pk to ~20mV pk-pk. I picked the 47uF cap due to its impedance curve. Pin 40 now looks much like the measurement I took on the AV Famicom.

Before (same pic as in one of my posts above):
Attachment:
5V-Before.png
5V-Before.png [ 755.14 KiB | Viewed 10849 times ]


After:
Attachment:
5V-Clean.png
5V-Clean.png [ 472.11 KiB | Viewed 10849 times ]


Board:
Attachment:
Poly Caps.png
Poly Caps.png [ 631.37 KiB | Viewed 10849 times ]




I also tested PPU revisions by swapping in my -H from the AV Famicom into the NES-101. Noise was still there. In fact, the -G from the NES-101 I installed in the AV Famicom produced clean video output. Which concludes that the PCB layout is the issue, not the PPU revision.


Top
 Profile  
 
PostPosted: Mon May 13, 2019 7:51 pm 
Offline
User avatar

Joined: Tue Mar 22, 2016 8:27 pm
Posts: 347
Location: Minnesota, USA
I agree that it sounds like a layout thing. What could it be about the layout that causes the problem? From my experience in noisy environments, I have found that there is always a possible kludge once the problem is understood. Sometimes the kludge comes first before we understand it! That's kind of what I have been looking for as I poke around. Sticking resistors and caps at different places, adding sockets, yanking/swapping chips, etc. Something is bound to have an interesting effect.

We know that the jailbars are stationary and synchronized to when the PPU accesses memory. That is a really big important fact. What path does it take to make its way out to the video signal? In terms of EMI, is it conducted (through wires) or radiated (through the air)?

You and I have both shown that better decoupling caps greatly improves the situation. Is there still something on the power rails somehow? Ground loops? Inductance of the PPUAD traces? I am at a loss.

Ultron wrote:
This reduced the noise on Pin 40 of the PPU, but DID NOT change the noise on the video output.

This is a very interesting observation. It points to the jailbars being picked up IN or AFTER the video amp, which could include the television set itself. Does the NES-101 have a metal shield like the AV famicom?


Top
 Profile  
 
PostPosted: Mon May 13, 2019 8:55 pm 
Offline
User avatar

Joined: Tue Feb 26, 2019 6:36 pm
Posts: 21
Ben Boldt wrote:
I agree that it sounds like a layout thing. What could it be about the layout that causes the problem? From my experience in noisy environments, I have found that there is always a possible kludge once the problem is understood. Sometimes the kludge comes first before we understand it! That's kind of what I have been looking for as I poke around. Sticking resistors and caps at different places, adding sockets, yanking/swapping chips, etc. Something is bound to have an interesting effect.

We know that the jailbars are stationary and synchronized to when the PPU accesses memory. That is a really big important fact. What path does it take to make its way out to the video signal? In terms of EMI, is it conducted (through wires) or radiated (through the air)?

You and I have both shown that better decoupling caps greatly improves the situation. Is there still something on the power rails somehow? Ground loops? Inductance of the PPUAD traces? I am at a loss.


I believe it's the way the ground plane is laid out. The PPU does heavy switching on it's output pins. The large changes in current due to the switching need a large ground plane to help dissipate these spikes. I don't think the ground plane in the area of the PPU is large enough. I believe this to be the problem in both the Famicom and NES-101.

Another thing is, in circuit design, you learn to keep analog and digital ground paths separate from each other, and only tie them together at the power source (or power connection). Since the PPU is doing double duty by reading in pixel data and producing a video signal, and only has one GND pin, you would point to this. But, the "jailbar" noise is much less on a toaster NES (it is there, I have measured the video signal), and non existent on the AV Famicom, which both use the same PPU. So, it's safe to say the PPU itself isn't the source of the problem. What would happen if you removed the PNP transistor for buffering and tied it's collector (which should be tied to "analog" ground) to the regulator ground pin, or a different section of the ground plane? I wonder if this would help.

I think the real fix would be an adapter board, with the PPU mounted on it, some decoupling MLCC caps, and a nice large ground plane. The board could be mounted where the PPU normally sits, with the EXT and GND pins on the main PCB tied to the ground plane on the adapter board. The question would be how to fit it into the Famicom case.

Quote:
This is a very interesting observation. It points to the jailbars being picked up IN or AFTER the video amp, which could include the television set itself. Does the NES-101 have a metal shield like the AV famicom?


Yes it does, but those metal shields wouldn't help in this case. All my tests were with them removed, and the AV Famicom still comes clean.


Top
 Profile  
 
PostPosted: Mon May 13, 2019 9:00 pm 
Offline
User avatar

Joined: Tue Feb 26, 2019 6:36 pm
Posts: 21
Just to show you the noise on the GND pin of the PPU on the NES-101 (using the heat sink as a reference):

Attachment:
PPU GND Noise.png
PPU GND Noise.png [ 474.63 KiB | Viewed 10816 times ]


Meanwhile, the CPU GND pin is clean as a whistle:

Attachment:
CPU GND Noise.png
CPU GND Noise.png [ 448.08 KiB | Viewed 10816 times ]


You can see the ~676 kHz noise on the GND pin, which is every 8th pixel, the same noise seen on the video.


Top
 Profile  
 
PostPosted: Mon May 13, 2019 9:17 pm 
Offline
User avatar

Joined: Tue Mar 22, 2016 8:27 pm
Posts: 347
Location: Minnesota, USA
Wow, I think you are really onto something with that. This high frequency stuff can be sneaky though, depending how you were probing it, you may have formed a loop large enough to pick up radiated stuff. But it really does prove a point.

When I wrapped one of my PPUs in copper foil, I soldered all GND pins directly to the foil and put ceramic bypass caps directly from the 5V pins to the copper foil as well. This forms a very low impedance local ground plane. I did all of the soldering on the top of the chip, so it does not interfere with any of the pins going into the socket. None of that seemed to have any visual effect but I did not measure the composite signal at the time that I did that. When I swapped the shiny copper PPU to the bare one from my front loader, again it is visually the same on the screen. Sorry for not taking measurements lately.

To your point, I didn't connect my copper foil to any sturdier ground anywhere. I wonder if something good might happen if we connected some thick wire between various GNDs and 5Vs. Some SNES video improvement guides suggest adding some wire like this.


Top
 Profile  
 
PostPosted: Tue May 14, 2019 1:49 pm 
Offline
User avatar

Joined: Tue Feb 26, 2019 6:36 pm
Posts: 21
Ben Boldt wrote:
Wow, I think you are really onto something with that. This high frequency stuff can be sneaky though, depending how you were probing it, you may have formed a loop large enough to pick up radiated stuff. But it really does prove a point.

When I wrapped one of my PPUs in copper foil, I soldered all GND pins directly to the foil and put ceramic bypass caps directly from the 5V pins to the copper foil as well. This forms a very low impedance local ground plane. I did all of the soldering on the top of the chip, so it does not interfere with any of the pins going into the socket. None of that seemed to have any visual effect but I did not measure the composite signal at the time that I did that. When I swapped the shiny copper PPU to the bare one from my front loader, again it is visually the same on the screen. Sorry for not taking measurements lately.

To your point, I didn't connect my copper foil to any sturdier ground anywhere. I wonder if something good might happen if we connected some thick wire between various GNDs and 5Vs. Some SNES video improvement guides suggest adding some wire like this.


Yes, I agree with picking up noise with "how you probe it". Most of my measurements were done with a ground spring on the o-scope probe, which is the best way to do it. The GND measurements obviously were not done that way, but you would figure that the noise would show up on both grounds if the probe was picking up interference.

I'm not sure if the copper foil would do the trick. Maybe it is too thin?

I did try adding a ~22 gauge wire between pin 20 and a different spot on the ground plane. This seemed to make a little difference, but not as much as I would have hoped.

I think I'm giving up on the NES-101 for now, and will move on to the Famicom to see what I can do.


Top
 Profile  
 
PostPosted: Tue May 14, 2019 10:28 pm 
Offline
User avatar

Joined: Tue Mar 22, 2016 8:27 pm
Posts: 347
Location: Minnesota, USA
Ultron wrote:
The GND measurements obviously were not done that way, but you would figure that the noise would show up on both grounds if the probe was picking up interference.

You can test that if you connect the probe tip directly to probe GND and nothing else, and hold it approximately where you were probing before. Whatever doesn't show up this way was your true difference in GNDs.

Ultron wrote:
I'm not sure if the copper foil would do the trick. Maybe it is too thin?

Good point. This is some pretty thin stuff I used. I cut it easily with a scissors and X-acto knife.

Ultron wrote:
I did try adding a ~22 gauge wire between pin 20 and a different spot on the ground plane. This seemed to make a little difference, but not as much as I would have hoped.

Was this a visible difference on the screen, or was it a difference on your scope?


Top
 Profile  
 
PostPosted: Thu May 23, 2019 1:45 pm 
Offline
User avatar

Joined: Tue Mar 22, 2016 8:27 pm
Posts: 347
Location: Minnesota, USA
Ugh, I don't know how I did this but I got my CY7C128A chip last night to find that it is not even the right number of pins! Here I was thinking I could just plug the thing in and it would work. Silly me. We kind of disproved the RAM chip anyway but just thought I would post an update on that...

Edit:
Heyyyyyy! The datasheet shows a 24-pin chip and this is 28-pin! LOL I wonder what the heck is going on here.

Edit 2: The picture of the item is 24-pin. Hmm...

Image


Top
 Profile  
 
PostPosted: Sun Jun 23, 2019 10:32 pm 
Offline

Joined: Sun Jun 23, 2019 10:30 pm
Posts: 1
I think tantalum might not be the best choice here, they work their magic best on voltage rails and things that always stay positive krogerfeedback.


Last edited by Conrad7K on Mon Jun 24, 2019 9:49 pm, edited 1 time in total.

Top
 Profile  
 
PostPosted: Mon Jun 24, 2019 9:23 am 
Offline
User avatar

Joined: Tue Mar 22, 2016 8:27 pm
Posts: 347
Location: Minnesota, USA
Conrad7K wrote:
I think tantalum might not be the best choice here, they work their magic best on voltage rails and things that always stay positive.

For the bypass cap (i.e, the one directly from 5V to GND), I found that using a tantalum produces the best results, and there is nothing that should ever make that cap go negative, so I really do recommend tantalum for that cap.

With the large series cap, I found that using a tantalum isn't really any better than using an electrolytic. From what I understand, this cap never goes backwards either though. Think about it like this: the signal before this cap ranges from 0V to +1V. After the cap, the signal ranges from -0.5V to +0.5V. So, at all times, the cap basically has a constant DC -0.5V across it, then the whole cap floats along with the signal, not really seeing the signal internally. Since it is holding a flat -0.5V, you put it in backwards and it should always be happy. To your point, I would recommend electrolytic for this one, especially for unknowns like when you plug and unplug the video jack, etc, but I think in general a tantalum would actually still be OK.


Top
 Profile  
 
Display posts from previous:  Sort by  
Post new topic Reply to topic  [ 100 posts ]  Go to page Previous  1 ... 3, 4, 5, 6, 7

All times are UTC - 7 hours


Who is online

Users browsing this forum: No registered users and 7 guests


You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot post attachments in this forum

Search for:
Jump to:  
Powered by phpBB® Forum Software © phpBB Group