Ina134 and Drv134 Supply bypassing

[JAY X]

Hi!

Some time ago i bought some ina134 and drv134, and would like to know if it is possible to bypass their supply pins (with 100nf or 1uf) rail to rail instead of rail to ground as shown in their datasheet.

I made the experiment with an ne5532 in a fader amplifier, and replaced the two caps to ground for a single 100nf cap rail to rail, and supply noise disappeared completely. Can this be done with dedicated ic’s as line drivers and line receivers??

Thank you for your advise!!

JAY X

 

[Andy Peters]

I made the experiment with an ne5532 in a fader amplifier, and replaced the two caps to ground for a single 100nf cap rail to rail, and supply noise disappeared completely.

Sounds like you had a layout problem. How exactly was "ground" routed on that board?

-a

 

[JAY X]

Hi Andy,

This fader amplifier was made on stripboard, 100nf caps on each rail....to common ground....
It seems that 100nf rail to ground has the risk of injecting supply noise into the circuit.... so i changed these 100nf that go to ground and placed only one in between the positive and negative opamp supply rails... and supply noise  disapeared completely...I powered this fader amplifier from a dual psu (+/- 17v). fader Board ground to the 0v psu conection.

At this point of the story, i would like to know if placing 100nf in between rails can be also done with dedicacted amplifiers like Line Receivers and Line drivers, not only Bipolar opamps as the ne5532 i used, to build the fader amplifier...because most datasheets usually show 100nf or 1uf from each rail to ground... ???

I think it can be done, but i was trying to get some input from people that try to do that with line drivers and receivers...it is all about experimentation...yes but i would like to have some input on this matter...

Thank you for your advise!
JAY X

 

[ricardo]

You are contaminating your Clean ground with decoupling current.  You need to take your Decoupling Caps to a separate Dirty ground.

PCB layout is critical for low noise.  The first rule is to be very clear what your earths are doing.  I always define 3 earths; Clean, Dirty and Chassis.

Consider carefully what you are doing when you connect anything to an earth.  If you connect supply decoupling capacitors to an earth line, you make it Dirty.

Clean earth is used ONLY for input / output earths and feedback components.Dirty earth carries our supply decoupling currents.  The 3 earths are joined ONLY at one point.  Output current must not upset the Clean input and feedback paths; a common source of instability.

Some of you will recognise this as similar to "Star Point" earthing in power amplifiers.
__________________

Decoupling Rail to Rail does NOT replace Decoupling each rail separately to earth.  Don't leave out the Separate Rail to Earths .. just don't dump their garbage onto your clean earth.

And BTW, USE SMALL ELECTROLYTICS FOR DECOUPLING as close as possible to each device.  The 100n are optional.  The Electrolytics are not.

 

[Rochey]

your "Clean" GND is what sounds like shield elsewhere? (thinking of the stuff I've read on the Pin 1 problem)

also, aren't GND currents only an issue if you are dealing with ground traces that have a significant impedance?

Most of the reading I've done of this is by Henry Ott. His suggestions when dealing with high frequency, high accuracy signals is to swamp GND on an entire plane, with as little breaks in the traces as possible (very low impedance... therefore less voltage generated).... along with good system layout partitioning... (extreme example, don't put a low voltage, high impedance analog trace next to the clock line of a DSP).

 

[12afael]

this is interesting. if I understand well it is one ground for power supply decoupling, one for input/output/emiters/feedback and  one for chassis/shield ???
I have seen schematics with two decoupling caps at rails and one from rail to rail, is this a good practice?

 

[Andy Peters]

i would like to know if placing 100nf in between rails can be also done with dedicacted amplifiers like Line Receivers and Line drivers, not only Bipolar opamps as the ne5532 i used, to build the fader amplifier...because most datasheets usually show 100nf or 1uf from each rail to ground... ???

In over 25 years of doing this, I've never decoupled anything between supplies, always from a supply to ground.

Stripboard is a mess. You need to ensure that your power supply common ("ground") is not interfering with your signal reference (also called "ground"). So given a standard inverting op-amp configuration, the non-inverting input usually goes to "ground," so you may need to route that ground to someplace where power-supply noise isn't.

Using planes goes a long way towards cleaning things up, and stripboards don't have planes, so right away you're handicapping yourself.

-a

 

[audiomixer]

your "Clean" GND is what sounds like shield elsewhere? (thinking of the stuff I've read on the Pin 1 problem)

also, aren't GND currents only an issue if you are dealing with ground traces that have a significant impedance?

Most of the reading I've done of this is by Henry Ott. His suggestions when dealing with high frequency, high accuracy signals is to swamp GND on an entire plane, with as little breaks in the traces as possible (very low impedance... therefore less voltage generated).... along with good system layout partitioning... (extreme example, don't put a low voltage, high impedance analog trace next to the clock line of a DSP).

shield would go to chassis - so that exterior RF garbage has not even the chance to get in....

as JR has put it in quite some posts you might prefer to follow the currents also when thinking of ground return currents. that is why I do prefer controlled traces instead of ground planes. You can have a dedicated return path for the decoupling caps and dedicated 0V reference tracks for all the other (critical) stuff like inverted opamp positive input, filter networks, DC bias. These will not carry any significant amps so there is negligible voltage drop and garbage on these nodes. while ground planes might have low impedance they still carry real currents along real paths. the current will not flow into a corner just because you put copper there. bad layout is much more difficult to trace in ground planes.

In over 25 years of doing this, I've never decoupled anything between supplies, always from a supply to ground.

Stripboard is a mess. You need to ensure that your power supply common ("ground") is not interfering with your signal reference (also called "ground"). So given a standard inverting op-amp configuration, the non-inverting input usually goes to "ground," so you may need to route that ground to someplace where power-supply noise isn't.

Using planes goes a long way towards cleaning things up, and stripboards don't have planes, so right away you're handicapping yourself.

well one thing you can do with strip board easily is to run two dedicated grounds, no problemo.


one thing I want to test in the near future is the nature of supply and capacitor ground returns in a real world circuit. must remember to put test points into my next board....

- Michael

 

[ricardo]

your "Clean" GND is what sounds like shield elsewhere? (thinking of the stuff I've read on the Pin 1 problem)

No 'Shield' in the Pin 1 problem is my 'Chassis'.

also, aren't GND currents only an issue if you are dealing with ground traces that have a significant impedance?

Err.rh!  In a low noise application ALL ground traces have 'significant impedance'.

Most of the reading I've done of this is by Henry Ott. His suggestions when dealing with high frequency, high accuracy signals is to swamp GND on an entire plane, with as little breaks in the traces as possible (very low impedance... therefore less voltage generated).... along with good system layout partitioning... (extreme example, don't put a low voltage, high impedance analog trace next to the clock line of a DSP).

Ott is very good.

One of the most important things he explicitly states (that other RFI guru hardly mentions) is that even with a ground plane, you need to know where your currents are flowing .. especially if there are breaks in the ground plane.  If you can't route an earth track on the ground plane layer such that it goes where you want it .. closely following the power trace enclosing minimum area .. then the ground plane doesn't change that.

Circulating currents enclosing a large area are truly evil. 

I have never found a ground plane improved on a careful layout for LN.  In most cases, the plane mucks up my obsessive earth returns.

 

[abbey road d enfer]

It seems that 100nf rail to ground has the risk of injecting supply noise into the circuit.... so i changed these 100nf that go to ground and placed only one in between the positive and negative opamp supply rails... and supply noise  disapeared completely...

Tt looks like your rails are noisy and what you inject into the ground is current resulting from this noise voltage via the impedance of the decoupling caps, which is quite low at typical hiss frequencies.
That's one of the reasons why console manufacturers use small value resistors (typically 22-100r) in series with the rails - the other reason is they also double as fuses in case of something shorting out.
Obviously cleaning the rails is a welcome improvement.
That increases significantly the rail's impedance, which could generate all sorts of instability problems; that's why large value (>47uF) capacitors must be added across the rails.
As to using only rail-to-rail caps, I would not recommend it. It may work with non-critical circuits (VU-meter amp) but I would never use that to run noble signals.

 

[JAY X]

Hi!

Abbey!, good point!....I have found some soundcraft 6000 schematics... and they show rail to rail bypass on all the NE5532 opamps, furthermore, they also show 47uf electros with 10R resistors across supply rails....maybe on the channel cards... It starts to make sense...These 47uf with 10r resistors ¿Must be placed right at the supply input of the card?

In any case i will try both aproaches : RC filter with rail to rail and with rail to ground.

JAY X

 

[JAY X]

Hi!

Another idea crossed my mind: ¿would it be ok to mix some circuits with rail to ground bypass with others with rail to rail bypass? on the same pcb...

JAY X

 

[abbey road d enfer]

Hi!

Another idea crossed my mind: ¿would it be ok to mix some circuits with rail to ground bypass with others with rail to rail bypass? on the same pcb...

JAY X

That may probably work, depending on the actual track inductance. But my understanding is that you want to use r2r decoupling because of the apparent benefit in noise. However, if you mix r2r and r2gnd, you will have to face the problem of noise current injection in the ground - and there I would emphasize on the notion that "ground" is an elusive concept (cf. Ricardo and JR).
You will find in the end that you need r2gnd caps anyway. You may find that some opamps are more forgiving than others (in that respect the MC33178 is mindblowing) and in the end the number, value, arrangement and location of the decoupling caps is not something we can design for you. The layout, as well as the circuitry determine the final arrangement.
If r2gnd decoupling caps pose a noise problem, it's not the concept that's at fault, it's the implementation and/or a noisy PSU.

 

[bruce0]

At this point of the story, i would like to know if placing 100nf in between rails can be also done with dedicacted amplifiers like Line Receivers and Line drivers

You poor man... all these brilliant guys helping you but your question remains somewhat unanswered.

The answer is YES.

PSU bypass simply stabilize the voltage of the rails they control by preventing them from varying due to resistance as the current used varies.  It is done everywhere on digital and analog circuits, and (risking life and limb in this smart crowd) I can think of no condition under which more PSU isolation caused a problem with the chip.

You can, and it is sometimes worthwhile, run them rail to rail, rails to ground, or both.

All of the other advice you are getting here is great, and actually the other answers are more important than your question, in that they deal with how to avoid power supply loads and noise from effecting signal integrity.

The issue with these bypass caps is that they act like coupling capacitors and carry the PSU noise and voltage variations from the rail TO the ground.  So that is why you are getting answers about ground cleanliness.  Which ground you use, how it is coupled to the other parts of the circuit, and it's nature (ground plane, trace, capacitance inductance) can all matter even in low speed analog circuits (and more so in high speed digital circuits where the issues become even more complex).

 

[[silent:arts]]

http://groupdiy.com/index.php?topic=55887.0

 

[JAY X]

Hi Bruce!

Thanks for the advise...i see...it all matters... it is time to experiment a bit more...!!

As always, there are never easy answers for this...in any case now i have more clear the direction to try..

Thank you all!!

JAY X

 

[JAY X]

Hi All!

I'm back messing with my mixer circuits, and things improved a bit. ( a bit less noise) ;D

Basically in my circuit i applied the ground scheme yo can see in this link:
http://home.comcast.net/~zenkai-audio/gndtss/pages/s3_chsh.htm#top

it is described as Balanced i/o with shields terminated to chassis.

You can see that the center tap of the transformer secondary is bonded (non switchable), from the center tap sedcondaryto the main AC inlet earth ground. In my case i made this bond from the first filter capacitor.

the question is: ¿what if i disconnect off this bond? ¿may improve noise? ¿Could be dangerous?

Remember that i am not asking to disconnect AC main ground!!! NOOOOOO!!. Just the bond  from the transformer secondary to main AC inlet ground that goes to chassis.

thanks for your advise!
JAY X

 

[abbey road d enfer]

You can see that the center tap of the transformer secondary is bonded (non switchable), from the center tap sedcondaryto the main AC inlet earth ground. In my case i made this bond from the first filter capacitor.

Both are equally incorrect. The point that should be connected to the chassis ground is the reference of the regulators (or the last smoothing caps if there's no regulator).

the question is: ¿what if i disconnect off this bond?

It won't prevent the unit from working, but the safety rules would not be met.

¿may improve noise?

Actually it may, or may not. If the audio ground is not connected to chassis ground at all there would probably be noise.

¿Could be dangerous?

In case the transformer becomes leaky, the audio ground may carry lethal voltage, that could be transmitted to the ins and outs. If both ins and outs are transformer-isolated, that could be non-consequential. If the ins or outs are not transformer-isolated, the audio connections could carry lethal voltages.

Remember that i am not asking to disconnect AC main ground!!! NOOOOOO!!. Just the bond  from the transformer secondary to main AC inlet ground that goes to chassis.

That is well understood.

 

[JAY X]

Hi Abbey!

I picked the reference point of regulators to chassis ground as you explained and It worked!. This improved noise even a bit more.

Now for the last screw turn..I think what i'm going to describe/ask, was explained in previous post in this thread but i go for a "remix".. ;D

Let's see...,

1. I have bypassed the opamps rail to rail.
2. The opamps take the power from the psu output rails.
3. The opamps take their GND reference from the psu board output rails, ¿right?...as they don't have a gnd reference on the audio pcb
    boards....

3.1 I have some schematics from a soundcraft 6000 mixer. They supply the opamps thru an RC filter: 47uf/10R.  or 220uf/3k3.
      The ground point in between the two 47uf caps is described as MAIN GND.

3.2. ¿Could that MAIN GND be chassis gnd? i think it is... but i'm not sure..

4. I don't know how to upload images here...

JAY X

 

[abbey road d enfer]

3. The opamps take their GND reference from the psu board output rails, ¿right?...as they don't have a gnd reference on the audio pcb
    boards....

Correct. It is very important that the 0V distributed ot the modules along with the rails be taken from this reference point.

3.1 I have some schematics from a soundcraft 6000 mixer. They supply the opamps thru an RC filter: 47uf/10R.  or 220uf/3k3.

The latter is only for low noise discrete stages where te current draw is <1mA.

      The ground point in between the two 47uf caps is described as MAIN GND.

  It's always difficult to describe what a "ground" does. In tha case, "MAIN" describes the 0V reference that goes from left to right and is used also as audio signal reference.

3.2. ¿Could that MAIN GND be chassis gnd?

No. Chassis ground is distributed via pin 4 of the longitudinal loom. MAIN ground is distributed via the rearcon PCB's and the small loom that joins it to the module.

4. I don't know how to upload images here...

Click on Attachments and other options at the bottom of your post.