PGD PGC Crosstalk

Status
Not open for further replies.

3v0

Coop Build Coordinator
Forum Supporter
The first time I seen the ICSP cables with PGC and PGD adjacent to each other I suspected it was not a good idea. As the technology is applied to ever faster chips crosstalk has got to be a limiting factor.

If anyone is designing a PIC programmer I suggest changing

VPP, VDD, GND. PGC, PGD
to
VPP, PGC, GND, VDD, PGD

This should eliminate crosstalk.

There is a short discussion on the subject at
http://www.embedinc.com/picprg/icsp.htm
Under the section header PGC to PGD Crosstalk.

You can toss rotten fruit at me if you like...
 
To me
VPP, Vdd, PGC, Gnd, PGD

Could be a better option, keeping programming pins away of "higher voltage" signals.
 
mister_e said:
To me
VPP, Vdd, PGC, Gnd, PGD

Could be a better option, keeping programming pins away of "higher voltage" signals.
Click to expand...

Could be. Here was the logic I used.
I know a 5 pin ICSP can be inserted backward with a ICD2 or PICkit2 without damag. Switching VDD and PGC preserves this.

Reversing your cable will switch VDD and GND and I think that would be a problem.
 
The reason most use VPP,+5,GND,PGD,PGC is it's the order Microchip uses on their programmers. Short cables keep the problem to a minimum.

Microchips Real ICE suggests a differential cable at clocks above 40MHz.
 
The rate of flash memory burning is the only true problem. When flash burn is not the bottle neck, the very programming device is capable of pushing a clock above 40MHZ is not the bottle neck, then I will be concerned.

Since a 60 inch cable has not ever caused a programming failure, any talk of crosstalk is snake oil in my ears. Off to the bath house I go.
 
blueroomelectronics said:
The reason most use VPP,+5,GND,PGD,PGC is it's the order Microchip uses on their programmers. Short cables keep the problem to a minimum.

Microchips Real ICE suggests a differential cable at clocks above 40MHz.
Click to expand...
I understand your reason. With the addition of a simple adapter your programmer is compatible with Microchip. What I do not understand is Microchips reason. These things are getting faster and it is starting to bite them. It would have been easy to do better.

At would be nice to use a longer ICSP cable. It could possibly fix that occasional programming error.

Like I said. Feel free to throw fruit. I know I am going against the grain.
 
We may be able to get the same result with a simple cable mod.

One could quite possibly get the same benefit by using a cable that switched the signals at each end. With some care one could cut the 3-4 and 7-8 pairs loose and swap them at the IDC connectors. I will make up an 18 inch cable and give it a try. If it works it is the best of both worlds.

blueroomelectronics said:
When I designed the Inchworm I sweated over the pinout and connector.
Click to expand...
Both the inchworm an Junebug are great programmers. I like the 2x5 connectors a lot. They have redundant circuits and are dead easy to make. The choice is/was a good one. In terms of sales compatibility is a big issue. I understand that.

Still I think Microchips decision to put PGC and PGC next to each other is a poor one. IFF switching the two signals allows the use of longer cables and more reliability that would be a good thing. The down side is that to connect to other peoples targets you would have to use an adaptor with crossed wires.
 
35" ICSP cable works. large image alert

I have built a 35 inch twisted ICSP cable. It seems to be working.

The construction is easier then I first indicated. All you do is cut the outside two wires loose and move each pair to the other side of the cable. Both ends. Rip the wires down far enough such that you can lay them flat under the strain relief.

I know a few downloads and verifies are not enough to declare success. I will let you know in a week if it works well or sooner if not. The other thing that needs to be pointed out is that I did not make a 35 inch standard cable to compare it against. Given that the Microchp instructions suggest a cable under 6 inches that does not sound like a good idea. Try it if you like.

The picture is large to show the cable ends.
**broken link removed**


conductors
9 10 3 4 5 6 7 8 1 2
1 2 3 4 5 6 7 8 9 10
...
1 2 3 4 5 6 7 8 9 10
9 10 3 4 5 6 7 8 1 2
 
Last edited:
3v0,

Your image isn't showing and the IP address is 192.168.... which is a local address.

Mike.
 
Pommie said:
3v0,

Your image isn't showing and the IP address is 192.168.... which is a local address.

Mike.
Click to expand...
Dag Nabit Anyway Fixed it.
 
Though difficult, let us try shielded wire for either PGC or PGD (one of them would do)wires with shield grounded at one end. this would help I feel.
 
we cant go on changing the standards while we can shield one wire by a bypass cable and still maintain the order. screened interconnect cable patches are not uncommon in RF circuitry, i remember. entire ribbon is embedded in shield. it wont help. At times a 22E resistor in series to clock wire and a 22-47pf cap to ground close to load side,reduces the crosstalk
 
Last edited:
The cable mod I have SEEMS to be working. I am thinking that if it does not the chance of the shielded cable working is small. Reason is that Microchip claims the problem is reflections rather then cross talk. That can be fixed too but not as easily.
 

What are you taking about?
My cable mod uses the same pinout as Bill uses on his programmers and is easy to construct.
 
3v0 said:
T.............
If anyone is designing a PIC programmer I suggest changing

VPP, VDD, GND. PGC, PGD
to
VPP, PGC, GND, VDD, PGD

This should eliminate crosstalk.

............
Click to expand...
This is from your earlier post. and Bill has indicated that he followed Micrchip order. and it is

VPP, VDD, GND, PGD,PGC, (aux)(PICKIT2 version.)

so I felt the order is suggested to be changed for a new designer. Yes I do agree that it serves reduce the crosstalk. but then on that programmer doesn't become compatible to meant ready boards using Microchip version. This is what I tried to indicate, please.
 
blueroomelectronics said:
Well the programmer has a pair of 4.7K pulldowns, would another pair of 4.7K on the target balance the cable and reduce reflections?
Click to expand...

If I remember. Termination with Z0 at one send would reduce reflection. 4.7K and all are seen from the DC side. At 40MHz or 20MHz the Zo would be far less.
 
Status
Not open for further replies.
Menu
Log in
Register
Cookies are required to use this site. You must accept them to continue using the site. Learn more…