opamps and local decoupling of rails, some questions

[living sounds]

Thanks! Not to hijack this thread (moderator, if this should be somewhere else please move my post), but in my console there is a chassis ground normally unconnected to the audio ground, all connected in series from channel to channel to the auxes to the master (file attached). I changed this to a single star ground and solder connected all grounds to lower hum as directed here:
http://www.tangible-technology.com/articles/200b.html

I've still got hum and the noise floor could probably be lower though. I wonder if using the original scheme and dampening all op amp power rails via electrolytics to the chassis ground connection would actually be better... or desoldering the connection between grounds but leaving the star ground... or creating a second star ground scheme for the chassis ground /dirt ground... ?


Also, can I use tantalum caps (they are small and I can get them cheap) for the task?

[edit: forgot the attachment]

 

[living sounds]

And would it be best to also move the other PSU filtering caps to the non-audio ground (I've used much bigger ones there as per Jim Williams suggestions and changed the 27R resisostors to 10R also per his suggestions)?

Would additional local filtering/dampening improve performance of the transistor pair on the input as well?

 

[living sounds]

So I tried this in my console as well as converter (10-47uf caps from power rails to chassis ground) - huge difference in measured and audible performance. Lower THD, reduction in harshness, better details, more low end. Lacking a scope capable of measuring the mhz range  I've no way to find out the cause.

So what's generally the optimum value for the cap? I'm going to order a few hundred! ;-)

 

[Kingston]

Lacking a scope capable of measuring the mhz range.

You have no way of knowing if you are doing more harm than good with just ears and somewhat low resolution RMAA measurements with AD/DA, especially if you also modding the AD/DA at the same time (?!).

Most importantly, you have to understand the grounding and decoupling scheme before planning any changes. Ask yourself if you can you point out where you can make improvements. Otherwise you have no way of knowing if you are again just making things worse or even somewhat broken. I'm glad my mixer learning project inspired but it seems (based on your questions) you are flying blind. Not a good time for rash cap order decisions and just slapping caps to gear at random. Back to theory.

 

[ricardo]

Get a cheap analogue scope.  Even a 10MHz scope will tell you a lot more than soundcard based scopes.  Get some 10x scope probes.


Where is the chassis connection on 200B?

Is p20 connected to the frame of the module?

Is there a schematic that shows the grounding strategy for 200B?

http://www.tangible-technology.com/articles/200b.html

I've still got hum and the noise floor could probably be lower though.

If you got no improvement in hum, you've done something wrong.  Contact Eddie as his article isn't all that clear.


R61/62 = 27R provide damping.

Are there electrolytics near (within 1") of each OPA (esp. 5532) on the modules?

Are they rail to rail or rail to ground?

 

[living sounds]

You're right about the scope, of course.

The chassis connection goes from the point at the connector to the back of a pot, which is mechanically linked to the frame like all other pots.

"p20"?

The whole schematic can be downloaded here:
http://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&cad=rja&ved=0CCUQFjAA&url=http%3A%2F%2Fwww.gearslutz.com%2Fboard%2Fattachments%2Fgeekslutz-forum%2F53786d1206498037-soundcraft-200b-troubleshooting-group-headphones-schematic-200b_schems_only.pdf&ei=Dx1ZUPbBN8fZ4QTh7YDYDw&usg=AFQjCNGkezTZnCfKa9eMS5LPGc19ANv8TA



There is an improvement in hum after the grounding mod, but it's still not great (I opened a thread about this before).

Electrolytics are generally (a lot) further away from the op amps and also shared between several. Would putting back in the old resistors do any good here? I have found that adding 10uf electrolytics from rails to (disconnected from the audio ground) chassis as well as adding another 10uf electrolytic to the 100nf between rails cleared things up quite a bit. But this is probably not the ultimate solution.

Thanks!

Get a cheap analogue scope.  Even a 10MHz scope will tell you a lot more than soundcard based scopes.  Get some 10x scope probes.


Where is the chassis connection on 200B?

Is p20 connected to the frame of the module?

Is there a schematic that shows the grounding strategy for 200B?

http://www.tangible-technology.com/articles/200b.html

I've still got hum and the noise floor could probably be lower though.

If you got no improvement in hum, you've done something wrong.  Contact Eddie as his article isn't all that clear.


R61/62 = 27R provide damping.

Are there electrolytics near (within 1") of each OPA (esp. 5532) on the modules?

Are they rail to rail or rail to ground?

 

[ricardo]

"p20"?

Pin 20

The whole schematic can be downloaded here:

They are a bit vague about earthing but ..

Each module seems to have only 2x47u 25V from Rails to Gnd.  These are located near the 20p connector.  Some modules have the odd 2u2 50V from Rail to Rail.  This is not adequate for stability or low THD.

Would putting back in the old resistors do any good here?

Won't make much difference to stability.  Even Williams's 10R provide a lot of damping.

I have found that adding 10uf electrolytics from rails to (disconnected from the audio ground) chassis as well as adding another 10uf electrolytic to the 100nf between rails cleared things up quite a bit. But this is probably not the ultimate solution.

Chassis should not be used to carry signal or decoupling currents.

You really want to take another set of Dirty Gnd leads to the New Bus for all the decoupling including the original 47u.  You need 10u 25V from each rail to the Dirty Gnd near each OPA.  You might get away with just having a separate net on each module to take all these decoupling currents including the original 47u.  The separate net is essential, at least at module level.

This is where you need to check stability & distortion for each earthing system.  eg whether you need a separate set of Dirty Gnd leads to the New bus for this Dirty Gnd net or whether you can just connect Dirty net to the original Clean net at the 20p connector.

This may not be ultimate solution but is an ESSENTIAL part of the ultimate solution.

 

[living sounds]

Thanks! I can get 22uf 25V caps for the same price, any reason not to use those?

I don't think there are rail-to-rail caps in the original schematics used at all. I've replaced all the 47uf 25V with 470uf.

To have those connect to the dirty ground as well makes sense.

So a brute force approach would be to create a dirty ground for each module and make a lead from each to a bus bar connected to the chassis next to the audio ground channel bus bar, right?

What about the ceramics? Leave them in? I had a ceramic-to-ground configuration before, which was audibly worse than the rail-to-rail ceramic only currently installed. Probably because of pwr line contamination.
In an effort to get cleaner power to start with I tried adding a global low pass filter to each DC line after the PSU once, but it did not do what I expected it to do. Guess I'll also end up building a better PSU from scratch with discrete regulators as Kingston suggested.

"p20"?

Pin 20

The whole schematic can be downloaded here:

They are a bit vague about earthing but ..

Each module seems to have only 2x47u 25V from Rails to Gnd.  These are located near the 20p connector.  Some modules have the odd 2u2 50V from Rail to Rail.  This is not adequate for stability or low THD.

Would putting back in the old resistors do any good here?

Won't make much difference to stability.  Even Williams's 10R provide a lot of damping.

I have found that adding 10uf electrolytics from rails to (disconnected from the audio ground) chassis as well as adding another 10uf electrolytic to the 100nf between rails cleared things up quite a bit. But this is probably not the ultimate solution.

Chassis should not be used to carry signal or decoupling currents.

You really want to take another set of Dirty Gnd leads to the New Bus for all the decoupling including the original 47u.  You need 10u 25V from each rail to the Dirty Gnd near each OPA.  You might get away with just having a separate net on each module to take all these decoupling currents including the original 47u.  The separate net is essential, at least at module level.

This is where you need to check stability & distortion for each earthing system.  eg whether you need a separate set of Dirty Gnd leads to the New bus for this Dirty Gnd net or whether you can just connect Dirty net to the original Clean net at the 20p connector.

This may not be ultimate solution but is an ESSENTIAL part of the ultimate solution.

 

[Kingston]

Maybe you saw this already. http://www.groupdiy.com/index.php?topic=49877.0

 

[living sounds]

Maybe you saw this already. http://www.groupdiy.com/index.php?topic=49877.0

Yes, but I tried out the original grounding scheme again some time ago, made a whole new fitting flat cable and unsoldered all the grounds, but the mod does work better. But I wonder if it would be best to go back to the original scheme and just use the chassis ground as the dirt ground. All the modules have their own connection to the chassis ground, and all the modules also connect to the chassis via the screws, so there is plenty of room for the current to flow, and I guess low impedance. But I don't know enough theory to make educated judgements, obviously.

 

[ricardo]

I can get 22uf 25V caps for the same price, any reason not to use those?

Small 22u 25V Aluminium electrolytics are good.

So a brute force approach would be to create a dirty ground for each module and make a lead from each to a bus bar connected to the chassis next to the audio ground channel bus bar, right?

This isn't brute force.  This is good earthing practice.  All these leads go to Ciletti's New Earth bus at the centre of the rack, same as all the clean earths.  This is the new reference point for the whole mixer.

I've replaced all the 47uf 25V with 470uf.

This is brute force.  Doing this on the original just puts more rubbish on the Clean earth.

I had a ceramic-to-ground configuration before, which was audibly worse than the rail-to-rail ceramic only currently installed. Probably because of pwr line contamination.

Yes.  The same with 470u decoupling to Clean earth.  All decoupling caps dump Sewage.

.. and just use the chassis ground as the dirt ground

I'm really unhappy about using Chassis for anything except shielding.  I always define 3 earths; Clean, Dirty & Chassis.  They are connected ONLY at the Star Point.

Calrec desks had a single connection to chassis.  It was a thick Cu strap from the Star Point; like from car battery to chassis.  Part of final test was to disconnect this strap and check for ANY inadvertent connection to chassis from the various earths.

Abbey on http://www.groupdiy.com/index.php?topic=49877.msg631531 talks about Bus Bars and Star Points.  But really a Bus Bar is a practical attempt at a Star Point.  What you want is something which has 0R to all attachments but this is Unobtainium.  Note Ciletti's New 'Bus Bar' is as small as possible.

I wish JR would join in as his experience of high quality mixing desk design is more recent.  My experience is more than 30 yrs old though IMHO, it was at the very highest level.

It's really these details that separate great mixing desk design from the merely adequate.  Agonising over the choice of OPA is pointless if the earthing system won't allow a practical signal path with less than 0.1% THD.

 

[living sounds]

Thanks, I've ordered 10uf now from a different vendor during the forum downtime, but I guess this isn't really critical anyway.

Since this all seems to be getting as low resistance as possible does the diameter of the cable going to the bus bar matter?

I've made some experimentation, and found that effectively disconnecting the chassis ground is  not a good idea in terms of hum and interference (even with it connected the cell phone mast signal noise is slightly audible above the onise). But simply adding the capacitors to the connected grounds makes the sound much better. Now I have to try what the added dirty ground does.

I can get 22uf 25V caps for the same price, any reason not to use those?

Small 22u 25V Aluminium electrolytics are good.

So a brute force approach would be to create a dirty ground for each module and make a lead from each to a bus bar connected to the chassis next to the audio ground channel bus bar, right?

This isn't brute force.  This is good earthing practice.  All these leads go to Ciletti's New Earth bus at the centre of the rack, same as all the clean earths.  This is the new reference point for the whole mixer.

I've replaced all the 47uf 25V with 470uf.

This is brute force.  Doing this on the original just puts more rubbish on the Clean earth.

I had a ceramic-to-ground configuration before, which was audibly worse than the rail-to-rail ceramic only currently installed. Probably because of pwr line contamination.

Yes.  The same with 470u decoupling to Clean earth.  All decoupling caps dump Sewage.

.. and just use the chassis ground as the dirt ground

I'm really unhappy about using Chassis for anything except shielding.  I always define 3 earths; Clean, Dirty & Chassis.  They are connected ONLY at the Star Point.

Calrec desks had a single connection to chassis.  It was a thick Cu strap from the Star Point; like from car battery to chassis.  Part of final test was to disconnect this strap and check for ANY inadvertent connection to chassis from the various earths.

Abbey on http://www.groupdiy.com/index.php?topic=49877.msg631531 talks about Bus Bars and Star Points.  But really a Bus Bar is a practical attempt at a Star Point.  What you want is something which has 0R to all attachments but this is Unobtainium.  Note Ciletti's New 'Bus Bar' is as small as possible.

I wish JR would join in as his experience of high quality mixing desk design is more recent.  My experience is more than 30 yrs old though IMHO, it was at the very highest level.

It's really these details that separate great mixing desk design from the merely adequate.  Agonising over the choice of OPA is pointless if the earthing system won't allow a practical signal path with less than 0.1% THD.

 

[JohnRoberts]

I wish JR would join in as his experience of high quality mixing desk design is more recent.  My experience is more than 30 yrs old though IMHO, it was at the very highest level.

It's really these details that separate great mixing desk design from the merely adequate.  Agonising over the choice of OPA is pointless if the earthing system won't allow a practical signal path with less than 0.1% THD.

Sorry. I have been having trouble posting for last couple days... I had at least one post disappeared yesterday. Lets see if this one sticks.

My experience with console design is already a little dated, since it predates these modern uber-opamps, but I am always glad to share general philosophy. FWIW the higher gain bandwidth opamps can revel problem that didn't exist in old legacy designs.

I am a big advocate of using differential referencing, between numerous local circuit blocks. Pretty much every opamp gain stage has + and - inputs so most can be configured as differential, to ignore ground potential differences between local areas.

Brute force ground approaches, just can't possibly deliver 0V across a few meters of distance.

=====
Regarding going into somebody else's design after the fact, to clean it up, is a similarly difficult task. Popular products may have published tweaks that are the results of people spending time and a lot of trial and error at the design.

In general to add differentials where the original design did not anticipate will get complicated.

Most such tweaks are like the lump of dirt under the carpet,,, you can often stomp it down some and spread it around but a lot of work to finesse it completely away. 

These days with miniature SMD technology you could actually add a bunch of circuitry piggy backed into an old design but I am not sure it is worth the effort and expense. At the end of the day you are still left with an old console.

====
I hope this thread is not putting too much emphasis on PS design. From my perspective, PS design (including decoupling et al) is adequate or not. I am not familiar with a continuum of marginally better performing and worse performing results.  maybe I never looked close enough with a high resolution test bench.

Note: I have experienced unexpected issues with large console PS ranging from one that resonated in sympathy with a nearby AM broadcast tower, and another large console where noise from 3 terminal regulators, ended up in the bus noise floor (long story).

Each console design is kind of unique but as I like to repeat consoles and power amps, are the most difficult simple circuits to do well. Very simple in concept, but the details can eat your lunch.

good luck and sorry for another non-answer.

JR 

 

[joaquins]

I'm building a mixer, I want to ask if my grounding idea seems right to you...

12 or 16 channels to 4 stereo groups (anyone could be used as master) 2 slots for 500 and 51X devices and one for mixing.

I'm thinking in using a dirty ground for each slot parallel with supply rails all long the mixer, a clean ground going down in a channel taking it's 3 slots and connected to a buss ground all long the mixer (heavy bar) and a chasis.

Someday, when I have some comps on it I'll like to use some little GR vu, but no more than that, I'm thinking in using dirty ground for it, will it be right?

The other thing is disturbing me is the chasis, 500 and 51X has a chassis pin, but if I use that pin and connect it to the module chassis I could be getting a ground loop across the chassis it self, I mean, frontplate, screw, frame, screw, frontplate and all over again, how does it works? How is this pin used?

Thanks

JS

Other Q I forgot, the summing buses are unbalanced, it's ok to use the heavy clean ground buss as summing bus? or should I use another one? I'm not crazy about the noise performance, just wanted to keep this mixer simple and with reasonable performance. The studio I ussualy make pro analog mastering works with 70dB S/N, so I think 80dB will be fine for me.

 

[tonedude]

I use rail-to-rail decoupling all the time, with excellent results. Typically I have 2x 100 nF caps to ground, and 10-100 uF rail-to-rail for each chip...

...

Samuel

Maybe this is a stupid question but is the "rail to rail" supposed to be a BP or NP electrolytic, or is a regular "polar" one ok?

 

[Monte McGuire]

There's a clear polarity between the rails, so why not a polarized cap? Polarized el-caps are generally smaller for a given size and capacity, and you have many more types to choose from.

 

[JohnRoberts]

I use rail-to-rail decoupling all the time, with excellent results. Typically I have 2x 100 nF caps to ground, and 10-100 uF rail-to-rail for each chip...

...

Samuel

Maybe this is a stupid question but is the "rail to rail" supposed to be a BP or NP electrolytic, or is a regular "polar" one ok?

From that same post

Bottom line is that decoupling must be engineered, not left to a few "rule-of-thumbs" and guesses. No approach will be optimum in any case. 

When decoupling across significant DC potentials polar capacitors are designed for that application so typically used.

Decoupling just rail to rail in a bipolar (+ and -) PS wrt ground, there needs to be decoupling from rails to ground also.

I like to say follow the current, and signal current leaving an op amp output generally drives into ground. To complete the current's return path to the PS decoupling caps, the current must travel from ground back to the supply rails.  At LF it will always get there eventually but for HF the paths are local and generally through caps.

JR

 

[sahib]

The other thing is disturbing me is the chasis, 500 and 51X has a chassis pin, but if I use that pin and connect it to the module chassis I could be getting a ground loop across the chassis it self, I mean, frontplate, screw, frame, screw, frontplate and all over again, how does it works? How is this pin used?

When we were at the beginning of designing the 51X-511 rack I specified 9 pin connector for the power. This had the Audio Ground, Power Ground and Chassis all separately going to the power supply and bonding together at a single point. This was overruled by  Volker and Jeff  on the basis that even API 500 did not have this scheme and we stuck with 7 pin Neutrik which only had one Ground and Chassis. 

However, on my new rack TAC 511 I have the 3 grounds, Audio, Power and Chassis all separately going to the power supply and bonding at the Chassis stud. Hence my agreement with Ricardo.

Edit. My apologies for missing to respond to your question.  On the 51X backplane PCB the Audio and Power Grounds are  connected together permanently. Therefore you have only GROUND and CHASSIS coming out of the PCB. Therefore if you use  7 pin connector you will have two options in terms of the CHASSIS.

You'll join the CHASSIS that comes from the PCB and the case chassis at the connector and take it to the power supply.
Or you'll take only the CHASSIS from the PCB to the connector and float the case (which is definitely not a good idea).