Somewhat tweaked 5E3 layout - request for comments.

[Wombaticus]

Hi guys,

I'm getting set to help a friend do a more-or-less traditional 5E3 build. However, I haven't been well pleased with any of the layouts on the Web for the 5E3 (none of which seem to stray far from the Fender original).

So I've come up with my own (using a 9" board), and I'd appreciate a couple of more eyes on it before drilling turret holes. The main differences are as follows:

1. Includes optional turrets at the beginning for dropping in an amplified Zener B+ drop if necessary. The MOSFET itself would be in a fully insulated package and bolted directly to the chassis beside the turret board.

2. Includes a voltage divider for heater elevation.

3. Moves the 68K input grid stops off the board and onto the pins of the tube where they belong.

4. Includes an optional extra dropping resistor/PSU cap to feed the input 12AY7 (these are shaded on the schematic and can just be omitted).

5. I've attempted a rational "bussed stars" ground scheme with ground returns close at the feeding cap (or, if passing through a coupling cap, to the next cap in sequence).

I haven't tried to include the trannies or power tubes in this sketch. Anyway, please point out anything I might have overlooked or screwed up:

http://wombatamps.com/Alternate_5E3.png

[Wombaticus]

Many thanks Fyl...but really what I was looking for was some commentary as to the sufficiency or potential for problems with the proposed layout...

[fyl]

Single ground bus with secondary legs (noisy), 250R cathode R stuck between two caps (reduced lifetime), opposite phase signals too close (possible oscillations), not enough filtering before the heater elevation (ripple on top of heating voltages), noisy Mosfet/Zener section (should go to a dirty ground and be isolated from the rest).

[Wombaticus]

Single ground bus with secondary legs (noisy), 250R cathode R stuck between two caps (reduced lifetime), opposite phase signals too close (possible oscillations), not enough filtering before the heater elevation (ripple on top of heating voltages), noisy Mosfet/Zener section (should go to a dirty ground and be isolated from the rest).

OK, now we're getting somewhere, this is more helpful. If you can explain in more detail a couple of things it would really help my understanding:

1. Why is a secondary leg off of a bus inherently noisy, so long as everything returns to where it should? As for it being a single ground bus, I *think* that what I've done is really treated each cap as a star point, and the bus is just essentially connecting a little daisy chain of star points. Is this problematic? If so, why?

2. Cathode resistor position is a compromise, I know. 105degree caps will help somewhat.

3. Can you describe specifically where you think the opposite phase stuff is too close? In general, how far apart should components with opposite phase signal be kept? This seems to be one of the harder things to keep in mind when doing a layout.

4. As for the filtering on the heater elevation, I was going by Merlin's recommendation in his book that the heater elevation voltage divider can be placed at "any convenient place" in the PSU chain. A quick LTSpice simulation seems to confirm that even when there is very substantial ripple present in the B+, there is none worth speaking of at the junction of the voltage divider (~ 1/100th of a volt). How much is too much?

5. Coming as it does in the main heavy current loop around the first PSU cap, I'm not seeing how having a separate ground for the Zener/Mosfet would make any difference. Where would you attach it, if not to the reservoir cap ground, which is essentially the main star point?

I *could*, just as easily, reconfigure the amplified Zener so that the current setting resistor doesn't even go to ground, but this way seemed to be easier with the layout.

Finally, I'm wondering out loud if there is any reason/benefit to putting, say, a 100nF cap across the Zener? So long as it's breakdown voltage is exceeded, it's not really switching on and off is it? Or is it?


Anyway, many thanks for helping me out with this, and apologies for all of the questions.

Cheers,

Paul

[fyl]

Is this problematic? If so, why?

The best results are obtained in a tweed-style amp (as well as in many others) by using three grounds: earth, dirty and clean.

Earth should be located as close as possible to the mains inlet. No other connexion there.

Dirty should be app. 1" from earth, and as close as possible to C1, which is the noisiest point in a standard amp (ground from a mosfet working on pulsated DC is another very noisy point). C1, C2 as well as power tubes returns (Rk/Ck ground) should connect here.

Clean should be located close to the input jacks and be used by all low-level stages as well as control pots.

Every element in a ground group can be connected star-style - what KOC call "galactic grounds" - or bus-style - easier and giving IMO as good or better results.

Can you describe specifically where you think the opposite phase stuff is too close?

V1 : signal in crossing plate out because of the coupling cap location - it should be above the socket, not on the side of the board. Can be a problem if you don't use shielded wires there.

V2 : ditto and more.

In general, how far apart should components with opposite phase signal be kept?

Half an inch is usually OK.

As for the filtering on the heater elevation, I was going by Merlin's recommendation in his book that the heater elevation voltage divider can be placed at "any convenient place" in the PSU chain.

I don't agree with Merlin here. The voltage divider should be placed after the screens filtering node and get it's ground return on the clean side. You don't want to pollute heaters with a noisy ground, do you?

Anyway, LTspice is a fine piece of software but far from effective when it comes to design power supplies.

[loudthud]

IMHO the layout is fine with the exception of the output stage cathode resistor. They get quite hot even when you use a 10W part. Best to move it to the end of the board and mount it up in the air slightly. As far as oscillations go, there's two schools of thought. Some say that it's in-phase signals that create problematic positive feedback. A 5E3 simply doesn't have enough gain to be a problem. You might want to split the cathodes on the input stages but you can just use the exsisting eyelets.

[Wombaticus]

The best results are obtained in a tweed-style amp (as well as in many others) by using three grounds: earth, dirty and clean.

Earth should be located as close as possible to the mains inlet. No other connexion there.

Dirty should be app. 1" from earth, and as close as possible to C1, which is the noisiest point in a standard amp (ground from a mosfet working on pulsated DC is another very noisy point). C1, C2 as well as power tubes returns (Rk/Ck ground) should connect here.

Clean should be located close to the input jacks and be used by all low-level stages as well as control pots.

Every element in a ground group can be connected star-style - what KOC call "galactic grounds" - or bus-style - easier and giving IMO as good or better results.



V1 : signal in crossing plate out because of the coupling cap location - it should be above the socket, not on the side of the board. Can be a problem if you don't use shielded wires there.

V2 : ditto and more.



Half an inch is usually OK.



I don't agree with Merlin here. The voltage divider should be placed after the screens filtering node and get it's ground return on the clean side. You don't want to pollute heaters with a noisy ground, do you?

Anyway, LTspice is a fine piece of software but far from effective when it comes to design power supplies.

OK, a bunch of stuff in there, but a few things I think I should comment on.

First of all, I've no doubt that acceptable results can be had by using the type of ground scheme you describe. But the justification for using it seems to me a bit theoretically dubious.

First, if you have two attachment points, the separate ground groups are effectively connected by the chassis. You then have an additional conduction path between the two, and therefore the potential for current to flow through the chassis, which never need happen and is undesirable.

The second thing that gives me pause is that you imply that it's somehow important to have your dirty ground attachment point in physical proximity to the mains earth. But I can't for the life of me figure out why this could be important in terms of the operation of the circuit. The only reason one might attach significance to this is if there is a misconception about the "dirty current" being cleaned up by somehow flowing into the chassis and out the mains ground...
But no current *ever* flows through the mains earth unless there is a fault condition. The *only* job of the circuit ground is to reference the circuit to zero volts - and no current need ever flow through the circuit ground to get that done. The only place the return current should end up in is back in the PT.

I'll grant that it does often make sense to have a central star point at or near the negative side of the reservoir cap, but for most amps this need not and shouldn't be the actual point of ground connection!

The bottom line is that with a single channel amp, there is no reason to ever have more than one circuit ground attachment point. (and of course, the mains Earth, but that's really just part of the chassis, and purely for safety).

There could of course be other grounding issues with this particular layout...

As for the phase issues - were you referring just to the wire runs, rather than component placement? If so, I'm not too worried. I'll point out that the 12AX7 needs to be moved an inch or more to the right anyway, and rotated a bit. I in any case use shielded runs to the grids.


As for the heaters, I need to think on this a bit.....

Thanks again for stimulating the think juices!

[Wombaticus]

IMHO the layout is fine with the exception of the output stage cathode resistor. They get quite hot even when you use a 10W part. Best to move it to the end of the board and mount it up in the air slightly. As far as oscillations go, there's two schools of thought. Some say that it's in-phase signals that create problematic positive feedback. A 5E3 simply doesn't have enough gain to be a problem. You might want to split the cathodes on the input stages but you can just use the exsisting eyelets.

Any particular reason to split the cathodes? Having a common cathode should maybe produce a bit more 2nd harmonic distortion when playing with bridged inputs, but otherwise not too different from split resistors of double the size.... I suppose the real advantage would be the opportunity to bias the tubes differently for more tonal variation?

[Steve Conner]

I agree, the layout looks fine and I think Fyl et al. are overthinking this. It's a low-gain amp, so ground loops, heater ripple etc. aren't mega-important. The main thing is where you return the PT CT to, and you got that right.

The other important thing is the input jack ground, but you haven't shown that in your layout at all. Douglas Self, in a hi-fi context, recommends grounding the input jack directly to the chassis and grounding the preamp bus to that. The idea is that if ground loop current, or RF pickup on the guitar lead's screen, comes at you from outside, it can't get inside the chassis.

For the cathode resistor, you really want to keep it away from the electrolytic caps. How about one of those aluminium housed Dale resistors bolted to the chassis?

Dale / Vishay - RH01010R00FE02 - Resistors - Passive Components - Allied Electronics

Also whenever using a MOSFET, RG and I recommend a 12V zener diode between gate and source (diode's cathode to gate). It helps avoid mysterious MOSFET deaths by ensuring the gate voltage can never get excessive. Some power MOSFETs have it built in. The other vital thing is a gate stopper resistor, but you remembered that.

[fyl]

First, if you have two attachment points, the separate ground groups are effectively connected by the chassis. You then have an additional conduction path between the two, and therefore the potential for current to flow through the chassis, which never need happen and is undesirable.

Have you looked at a tweed chassis lately? Chromed steel with input and output jacks mounted directly to it. Is there a built-in ground path between the input and output jacks?

Have fun building and debugging this amp.

[Steve Conner]

Have you looked at a tweed chassis lately? Chromed steel with input and output jacks mounted directly to it. Is there a built-in ground path between the input and output jacks?

In the case of the 5E3, the ground reference for the speaker output jack doesn't matter a bit, because no NFB is taken from it.

Just make sure that you return the cold lead from the OT secondary straight to the jack, so that the speaker current doesn't flow in the chassis.

In general, a change of ground reference between the speaker jack and the rest of the amp can lead to hum injection into the NFB loop, but the effect is very small (it has a gain of 1 to the speaker)

And, it's always technically been bad form to use the chassis as part of your signal grounding scheme. Especially if it's steel. But Fender et al just did it anyway to save money.

[fyl]

In the case of the 5E3, the ground reference for the speaker output jack doesn't matter a bit, because no NFB is taken from it.

If you say so.

[Steve Conner]

Yes, I do say so.

[Wombaticus]

Have you looked at a tweed chassis lately? Chromed steel with input and output jacks mounted directly to it. Is there a built-in ground path between the input and output jacks?

Have fun building and debugging this amp.

Good point, but I either use isolating Cliff jacks, or an insulative washer with the Switchcraft types. I never, ever, use an uninsulated input. This ain't my first rodeo.

[Steve Conner]

Douglas Self, in a hi-fi context, recommends grounding the input jack directly to the chassis and grounding the preamp bus to that.

Sometimes the uninsulated jack is better. But if you're unsure, the insulated one is the safer option. I've built amps both ways and none of them hums.