Interesting hybrid mic pre thread

[gyraf]

so like kinda of a "limited time offer"..?

I think of this forum more like an archive for future diy'ers to dig into, which make me feel that the time-limited approach to design sharing is a bit counter-productive. But tastes differ..

Jakob E.

 

[emrr]

Why give us a half arse schemo to analyze  ???.

That's a very silly question, which shouldn't need explaining.  Post the PM660 publicly yet?

When I first designed it I posted it like everything else that I generously give away...where were you?

If you posted it, it was for 3 or 4 seconds.  I have several of your related pieces saved, which you pulled down.  My point is he may be headed for commerce, and wishes to leave it obscure to some degree.  For those reading with comprehension, he was dragged into this thread, and didn't start it himself. 

 

[abbey road d enfer]

"Low Z is not a guarantee of low noise; very often ground arrangement becomes more critical."

So true as I well know... I use a single star ground but keep the rectifier loops very short in the power supply. I basically follow the practices given by Bill Whitlock and find they work well. A star ground by definition is zero potential because it is the reference.

Such a common misconception.
With starground, every single stage reference potential is different of the other. Hierarchical ground arrangement is the key for low-noise in unbalanced configuration.
Hum induction is much more sensitive in star ground because flux is circulated on a greater length of conductor.
Star ground is for electricians; makes sure the discharge path doesn't induce dangerous voltage on other points.

 

[mjk]

I have a number of reasons to share my designs with the community. The main reason is that I get a lot from the community that goes into my designs, and I want to share my learning back. A more selfish goal is that I get a lot of good design review feedback and suggestions for improvement.

At what level to share designs? Clearly there is a wide range of level that different folks here are comfortable working with. I am used to working at a somewhat conceptual level, sharing circuit ideas as I build, measure, and refine. I assume that anyone using circuits I publish would know how to achieve their own design goals, i.e. I may not be publishing "DIY-ready" designs but I do build and test the circuits for the most part.

As for the micpre, I have no problem publishing more details, but the actual product is in the final stage of development and there are still some areas being tweaked. I don't worry too much about someone selling cheap knockoffs, as my stuff is pretty much handmade and "high-touch".

My actual schematics are broken up into PCB sections and discrete wire sections so are hard to follow online, hence the redrawn signal path schematic with simplification. I'm planning to create a high level schematic, which I will publish, after my design updates.

Cheers,

Michael

 

[mjk]

With starground, every single stage reference potential is different of the other. Hierarchical ground arrangement is the key for low-noise in unbalanced configuration.
Hum induction is much more sensitive in star ground because flux is circulated on a greater length of conductor.

Hi,

Thanks for the comments; this is one of my favorite topics and I would like to know more about how to make a distributed ground work.

I'm not sure what is represented by the capacitors in your diagram; is it the output current path or power supply AC noise coupling or both? At what point and how does the ground rail connect to power supply returns, pin 1 returns, chassis, and safety ground?

In my version of a single point ground, the input reference path, output current loops, shunt regulator returns, etc. go to the single point separately. All input references would thus be at the same potential assuming negligible current flow in reference loops. Certainly there is an opportunity for the magnetic loop area of a circuit to be larger with a star ground, but with careful layout I have not had any problem with induced current due to magnetic coupling. In the micpre, the longest signal path connection to the star ground is about 5 inches. I think the total end-to-end loop area of a ground bus would be similar or even greater, depending on layout.

Cheers,

Michael

 

[Kingston]

My actual schematics are broken up into PCB sections and discrete wire sections

If I may ask, why? The overview you published is a very simple design with not even a PSU included, even if expanded so that all parts are shown. What would be the point of breaking up a simple thing like this into sections? some kind of limitation of your design software?

 

[Kingston]

I'm not sure what is represented by the capacitors in your diagram; is it the output current path or power supply AC noise coupling or both? At what point and how does the ground rail connect to power supply returns, pin 1 returns, chassis, and safety ground?

The image just represents the fact that in an optimal layout situation, total resistance of your ground reference(s) will be higher for all "star ground" based implementations. More noise than well implemented hierarchical ground. I've personally found this to be a non-issue in simple p2p tube builds, but it is of very high importance in more complex high speed designs. The caps represent individual hierarchical nodes that you should not mix, but are not absolutely necessary.

For this kind of layout, there is only "one point" where your whole unit connects to star ground, and that's somewhere in the B+ PSU. Usually B+ PSU output if there is some kind of regulation going on instead of first filtering cap right after rectifier (only rectifier crud present at that point, no use sending it to chassis). This first-filter-cap-to-star is an oddly popular implementation. Maybe people just don't know it's not really optimal.

Anyway, the B+ PSU out is your one and only ground reference for all audio circuits.

Of course there are still separate connections from all pin1's and heater PSU (unless biased positive) to the star point. They don't connect anywhere on the "audio ground" nodes. For pin1 it could be debated that if everything else is well implemented, it won't matter if they connect directly to chassis. I mean, it is essentially just a shield, and if it acts as anything else it's a design error.

The above applies just the same for all bipolar audio designs.

 

[mjk]

My actual schematics are broken up into PCB sections and discrete wire sections

If I may ask, why? The overview you published is a very simple design with not even a PSU included, even if expanded so that all parts are shown. What would be the point of breaking up a simple thing like this into sections? some kind of limitation of your design software?

The complete preamp has relay switched P48, a 6 way input selector with transformer ratio switching, a transformer DI, a DPA 130V mic interface, and output tap switching (and metering). The DPA circuit and the 1st stage vacuum tube each have their own CCS+shunt regulators, i.e. it doesn't all fit on a single sheet.

The limitation of my design software (Eagle) seems to be that it wants to put everything on the schematic on the board, but parts of the circuit are discrete wired on the front panel and back panel. The schematic therefore has connectors that go off to the various tube sockets, switches, pots, etc.

What's needed for me to communicate this properly (I think) is a single sheet redrawn to show core of the amp including the selectors and shunt regulators, and a second sheet with the DPA circuit, DI, ribbon mic buffer, etc. The power supply is trivial but it's a single sheet also.

 

[abbey road d enfer]

I'm not sure what is represented by the capacitors in your diagram; is it the output current path or power supply AC noise coupling or both?

These caps are the necessary local decoupling caps. Indicating how any B+ noise is injected in the ground connection.

  At what point and how does the ground rail connect to power supply returns,

For hierarchical ground, the reference point of the last stage goes to B+ reference.

pin 1 returns, chassis, and safety ground?

Almost irrelevant in terms of hum, but in terms of RFI, very often the reference of input stage is connected to pin 1 of input connector and to chassis.

In my version of a single point ground, the input reference path, output current loops, shunt regulator returns, etc. go to the single point separately. All input references would thus be at the same potential assuming negligible current flow in reference loops.

You certainly don't have the right to assume them negligible, because that's exactly where internally generated hum originates from. Hum is a matter of microvolts, generated by milliamps circulating in milliohms.
Ohm's law rules!

Certainly there is an opportunity for the magnetic loop area of a circuit to be larger with a star ground, but with careful layout I have not had any problem with induced current due to magnetic coupling. In the micpre, the longest signal path connection to the star ground is about 5 inches.

5 inches can be a pretty long path for magnetically-induced hum. In a mains transformer, 5 inches is enough to produce 300 mV.

I think the total end-to-end loop area of a ground bus would be similar or even greater, depending on layout.

Yes, but since the signal travels parallel to the ground (that's the whole principle of hierarchical ground), any voltage induced in the ground path is compensated by the same in the signal path.
And end-to-end is not what matters, it's stage-to-stage that counts.

 

[Kingston]

pin 1 returns, chassis, and safety ground?

Almost irrelevant in terms of hum, but in terms of RFI, very often the reference of input stage is connected to pin 1 of input connector and to chassis.

I have seen this oddity plastered elsewhere as well, with no clear explanation.

How would this ever work? we already have reference coming from B+ end (itself connected to chassis). Connecting input stage to pin1 which is also connected to chassis would create the longest ground loop humanly possible. say hello to 50dB of hum.

 

[tmuikku]

Here is cool document on grounding which helps to visualize hierarchical ground and current flow. There is also some text on grounding the input at the end of the document.

http://www.freewebs.com/valvewizard2/grounding.html

 

[abbey road d enfer]

pin 1 returns, chassis, and safety ground?

Almost irrelevant in terms of hum, but in terms of RFI, very often the reference of input stage is connected to pin 1 of input connector and to chassis.

I have seen this oddity plastered elsewhere as well, with no clear explanation.

How would this ever work? we already have reference coming from B+ end (itself connected to chassis). Connecting input stage to pin1 which is also connected to chassis would create the longest ground loop humanly possible. say hello to 50dB of hum.

In that case, B+ reference should not be connected directly to chassis. This arrangement may transfer noise problems to the output instead of the input. If the output is balanced, it's not a problem.

 

[mjk]

So the trick is to isolate the signal path grounds and create a ground path which is physically in parallel with the signal flow. Doing this allows the noise and signal to be "locally" connected at each stage because the reference noise will subtract from the signal noise at the input of each stage. Is that the deal?

In my star ground scheme the filter caps and stage output current (shunt regulator or cap) each have separate returns from the input references to the star point, not common as shown on the sketch. There is no galvanic coupling due to shared return paths. Also with my plate feedback circuit the AC signal reference is derived from the output so I think already the noise and signal are somewhat in parallel. E.g. the noise increases when I ground reference a stage.

I don't think I'm coupling much hum into the return wiring with my star ground in the first place (input transformer seems to be more sensitive than the wiring), but I should be able to try the other ground scheme out by disconnecting the signal path grounds from the star and running a ground wire from point to point from the B+ return at the output back to the input. I will need to ground the input end and leave the B+ supply return "floating" with the output transformer primary, because of the shared ground with the DPA input. I have already been thinking that if I can isolate the signal path grounds from the star point there may be some further improvement available.

I already isolate the rectifier loop as much as possible using a daisy chain filter ground; it's the rediated EMI from the rectifier switching in the 180Hz-4KHz range that's my biggest noise issue. It's mostly radiated from the transformer windings so shielding the transformer eliminates this source. But I still would like the unit to be  insensitive to other external fields as well. I guess it doesn't need to operate near Xray machines or arc welders though...

Cheers,

Michael

 

[abbey road d enfer]

So the trick is to isolate the signal path grounds and create a ground path which is physically in parallel with the signal flow. Doing this allows the noise and signal to be "locally" connected at each stage because the reference noise will subtract from the signal noise at the input of each stage. Is that the deal?

That's exactly it. Think of each stage as a differential amp, its ground being the negative input.

In my star ground scheme the filter caps and stage output current (shunt regulator or cap) each have separate returns from the input references to the star point, not common as shown on the sketch.

That means that the filter caps reference and the output stage reference may develop some differential voltage. Fortunately, as you have properly made a hierarchical ground between the transformer, rectifiers and caps, this should not happen.

There is no galvanic coupling due to shared return paths. Also with my plate feedback circuit the AC signal reference is derived from the output so I think already the noise and signal are somewhat in parallel. E.g. the noise increases when I ground reference a stage.

I don't think I'm coupling much hum into the return wiring with my star ground in the first place (input transformer seems to be more sensitive than the wiring), but I should be able to try the other ground scheme out by disconnecting the signal path grounds from the star and running a ground wire from point to point from the B+ return at the output back to the input. I will need to ground the input end and leave the B+ supply return "floating" with the output transformer primary, because of the shared ground with the DPA input. I have already been thinking that if I can isolate the signal path grounds from the star point there may be some further improvement available.

Hence, you don't have a purely star ground. You have a combination of hierarchical and star. This is the arrangement that has to be taken in anything that is more complex than a simple gain stage + PSU. Phantom powering is one thing that forces compromises.

I already isolate the rectifier loop as much as possible using a daisy chain filter ground;

That's the most important single action to do in order to optimise ground noise circulation.

it's the rediated EMI from the rectifier switching in the 180Hz-4KHz range that's my biggest noise issue. It's mostly radiated from the transformer windings so shielding the transformer eliminates this source. But I still would like the unit to be  insensitive to other external fields as well. I guess it doesn't need to operate near Xray machines or arc welders though...

Cheers,

Michael

In order to make it less susceptible to external fields, you must minimise the footprint of the whole unit. Not always feasible and convenient.

 

[mjk]

In my star ground scheme the filter caps and stage output current (shunt regulator or cap) each have separate returns from the input references to the star point, not common as shown on the sketch.

That means that the filter caps reference and the output stage reference may develop some differential voltage. Fortunately, as you have properly made a hierarchical ground between the transformer, rectifiers and caps, this should not happen.

Sorry, I meant B+ signal decoupling caps and B+ shunt regulator return paths go to the star ground. I already have the B+ filter caps on a hierarchical ground on the power supply board, and floating B+ output to the star ground. I think if I now create a parallel ground running from the output transformer primary return at the star point back to the input transformer secondary side return I will eliminate a few star connections. Each star return is a separate PC track so I can just cut them. I guess the P48 and the DPA buffer should return at the input end of the hierarchical ground also. The DPA is OK because it uses a CCS and shunt regulator but the P48 has an existing return at the star point I can't easily move. A floating P48 supply could be returned at the input end.

I like the differential analogy. I was looking for a way to hum balance the signal path, which I thought was feasable due to transformer coupling at both ends (modulo P48 issues). This looks like the way to do it.

Thanks,

Michael

 

[abbey road d enfer]

You must leave the center tap of the transformer alone. The reference point of B+ is the negative pole of the smoothing cap.

 

[mjk]

You must leave the center tap of the transformer alone. The reference point of B+ is the negative pole of the smoothing cap.

That's correct. It keeps the charging current pulses local to a short loop.