Ground loops and TTL supply from PSU questions

[Jon Danniken]

Hello,

I have two questions.

First, do I need to be concerned with ground loops for computer circuits? I
am planning on taking 5V and Ground from a molex off of the PSU to power a
small circuit that I am building inside of the computer case. The circuit
will use that ground as a reference for a voltage (2V or so) coming off of
the video card adaptor (AGP port).

Am I going to be okay referencing the voltage from the video card to the
ground wire from the PSU, or should I reference it to something on the video
card itself? If I do need to reference it to something on the card itself,
would it be okay to tie the ground on the video card to the ground on the
molex from the PSU?

My second question concerns using the 5V output from the PSU as the Vcc for
a TTL chip. Should I use a voltage regulator to ensure that the supply is at
5V (and not too far over), or is it generally okay to just use the 5V as it
comes off of the molex?

Thanks for any insights into these concerns,

Jon

 

[Jan Panteltje]

Hello,

I have two questions.

First, do I need to be concerned with ground loops for computer circuits? I
am planning on taking 5V and Ground from a molex off of the PSU to power a
small circuit that I am building inside of the computer case. The circuit
will use that ground as a reference for a voltage (2V or so) coming off of
the video card adaptor (AGP port).

At the risk of being wrong (as always possible), I'd get the signal from
that AGP differential.
Then you can use normal ground from the connector, and any ground current
will not interfere with the input.

Am I going to be okay referencing the voltage from the video card to the
ground wire from the PSU, or should I reference it to something on the video
card itself? If I do need to reference it to something on the card itself,
would it be okay to tie the ground on the video card to the ground on the
molex from the PSU?

No, use differential input, ground side diff input on AGP card.

My second question concerns using the 5V output from the PSU as the Vcc for
a TTL chip. Should I use a voltage regulator to ensure that the supply is at
5V (and not too far over), or is it generally okay to just use the 5V as it
comes off of the molex?

Yes.

 

[Rich Grise]

On a sunny day (Thu, 23 Nov 2006 22:16:31 -0800) it happened "Jon Danniken"


Yes.

I disagree here. Jon should either use the +5 from the PSU, and just
capacitate it smooth, or run a regulator off the +12. Any +5 regulator
with only a +5 supply will put out less than +5V, because there is no
headroom.

Thanks,
Rich

 

[Jan Panteltje]

I disagree here. Jon should either use the +5 from the PSU, and just
capacitate it smooth, or run a regulator off the +12. Any +5 regulator
with only a +5 supply will put out less than +5V, because there is no
headroom.

Thanks,
Rich

Oh I agree, when using the 5V you should filter / decouple it locally.
If the app is purely digital that PSU 5V will be OK, if it involves
any analog processing you could use the 12V or a switcher on board from some
other supply.
To make it short, the 'molex' IS the connector from the PSU I think?
Or did you mean one of thsoe power connectors as to the IDE drive?

 

[Jon Danniken]

At the risk of being wrong (as always possible), I'd get the signal from
that AGP differential.
Then you can use normal ground from the connector, and any ground current
will not interfere with the input.


No, use differential input, ground side diff input on AGP card.

Thanks, Jan The reason I want to tie the grounds together is to be able to
use a transistor (NPN) to switch in the circuit. I would drive the
transistor from a 4013 flip/flop, and connect the 2V coming from the AGP
card to an appropriate resistor, and then to the Collector of the
transistor. When switched on, the transistor would complete the circuit
through the shared ground, and change the reference voltage from the AGP
card (desired effect)

Otherwise, I'm thinking I would have to use a quad bilateral switch.

Here is a crude ascii drawing:

|--/\/\/---- From AGP voltage
|
|
|/
From switching circuit --/\/\/---|
|>
|
|
===
GND

This is why I am hoping to tie the grounds in together, so that I can just
use a transistor to switch in the circuit.

Thanks,

Jon

 

[Jon Danniken]

Oh I agree, when using the 5V you should filter / decouple it locally.
If the app is purely digital that PSU 5V will be OK, if it involves
any analog processing you could use the 12V or a switcher on board from some
other supply.
To make it short, the 'molex' IS the connector from the PSU I think?
Or did you mean one of thsoe power connectors as to the IDE drive?

Thanks Jan, and Rich.

Yes, the voltage will be coming off a molex connector from the PSU.. The
chips involved will be a MAX232, a 4013 (flip flop), and possibly a 7404
(hex inverting buffer) if I need it. There will also be a few transistors
switching things in and out..

I would definitely toss in a few caps across the supply as it enters the
board, but will that be enough?

Thanks,

Jon

 

[Jan Panteltje]

Thanks, Jan The reason I want to tie the grounds together is to be able to
use a transistor (NPN) to switch in the circuit. I would drive the
transistor from a 4013 flip/flop, and connect the 2V coming from the AGP
card to an appropriate resistor, and then to the Collector of the
transistor. When switched on, the transistor would complete the circuit
through the shared ground, and change the reference voltage from the AGP
card (desired effect)

Otherwise, I'm thinking I would have to use a quad bilateral switch.

Here is a crude ascii drawing:

|--/\/\/---- From AGP voltage
|
|
|/
From switching circuit --/\/\/---|
|>
|
|
===
GND

This is why I am hoping to tie the grounds in together, so that I can just
use a transistor to switch in the circuit.

I would do it something like this (discrete), differential amp.
_________________________________________ +
| |
[ ] [ ]
| |--------------------------> signal
d d
-- g JFETs g---------
AGP s s |
SIG |-------------| |
| |
AGP -------------------------- +
GND | |
| [ ]
c |
b----------------------|
e NPN ---
| \ / temp comp diode
| current source ---
[ ] |
| [ ]
| |
------------------------------------ GND
No need to tie any ground together,

 

[Jan Panteltje]

Thanks Jan, and Rich.

Yes, the voltage will be coming off a molex connector from the PSU.. The
chips involved will be a MAX232, a 4013 (flip flop), and possibly a 7404
(hex inverting buffer) if I need it. There will also be a few transistors
switching things in and out..

I would definitely toss in a few caps across the supply as it enters the
board, but will that be enough?

The PC PSU will have to work between about 4.75 and 5.25 I think, mine goes down
to 4.81 according to bios....
That is within the TTL spec, CMOS HC[T] may have even bigger range, look it up.
It should work!

 

[Jon Danniken]

I would do it something like this (discrete), differential amp.
_________________________________________ +
| |
[ ] [ ]
| |--------------------------> signal
d d
-- g JFETs g---------
AGP s s |
SIG |-------------| |
| |
AGP -------------------------- +
GND | |
| [ ]
c |
b----------------------|
e NPN ---
| \ / temp comp diode
| current source ---
[ ] |
| [ ]
| |
------------------------------------ GND
No need to tie any ground together,

Thanks, Jan, but if I am reading that correctly, you have the AGP actually
turning the FETs on and off.

In the application I am doing, the switching is done by the remote circuit
(by a flip flop), which connects the AGP voltage to ground via a resistor.

The reason for this circuit is to fool a voltage sensing chip on the AGP
card into thinking that certain voltages on the AGP card are too low. This
is done by connecting a parallel resistance to the existing resistor on the
AGP card. By doing this, certain voltages on the AGP card will be raised,
which is what I am trying to accomplish. (this actually works, I am just
trying to automate it with a switching circuit so it isn't left on when it
isn't needed)

I like the idea of keeping the AGP ground seperate, though, the safety of
which is my main concern.

Thanks,

Jon

 

[Jon Danniken]

:

Thanks Jan, and Rich.

Yes, the voltage will be coming off a molex connector from the PSU.. The
chips involved will be a MAX232, a 4013 (flip flop), and possibly a 7404
(hex inverting buffer) if I need it. There will also be a few transistors
switching things in and out..

I would definitely toss in a few caps across the supply as it enters the
board, but will that be enough?

The PC PSU will have to work between about 4.75 and 5.25 I think, mine goes down
to 4.81 according to bios....
That is within the TTL spec, CMOS HC[T] may have even bigger range, look it up.
It should work!

Thanks, Jan, that will save me some needed space.

Jon

 

[Jan Panteltje]

I would do it something like this (discrete), differential amp.
_________________________________________ +
| |
[ ] [ ]
| |--------------------------> signal
d d
-- g JFETs g---------
AGP s s |
SIG |-------------| |
| |
AGP -------------------------- +
GND | |
| [ ]
c |
b----------------------|
e NPN ---
| \ / temp comp diode
| current source ---
[ ] |
| [ ]
| |
------------------------------------ GND
No need to tie any ground together,

Thanks, Jan, but if I am reading that correctly, you have the AGP actually
turning the FETs on and off.

In the application I am doing, the switching is done by the remote circuit
(by a flip flop), which connects the AGP voltage to ground via a resistor.

The reason for this circuit is to fool a voltage sensing chip on the AGP
card into thinking that certain voltages on the AGP card are too low. This
is done by connecting a parallel resistance to the existing resistor on the
AGP card. By doing this, certain voltages on the AGP card will be raised,
which is what I am trying to accomplish. (this actually works, I am just
trying to automate it with a switching circuit so it isn't left on when it
isn't needed)

I like the idea of keeping the AGP ground seperate, though, the safety of
which is my main concern.

Thanks,

Jon

Ah now I get what you want.
Well in that case it is the other way around, the diff amp goes on the AGP
card and perhaps drives a power MOSFET or transistor.
Your driver signal is the differential.
But likely only a simple MOSFET will do, 5V on enough noise margin.