Layout and Design- How To?

[silverfox]

While I've seen lots of pictures and some builds questions remain. At the moment I'm putting together the power amp section- Fender AB763. In many regards I'm following the original PC board layout and it appears components, even on this high gain amp may be arranged without concern for proximity. That is: the resistors, capacitors and preamp tubes.

Here are the photos of the original and my current rebuild of the power amp driver:



It does not appear that orientation or closeness of the traces on the PCB are a concern relating to feedback or oscillations. Is this correct? If so I assume that Layout problems are most affected by wiring and placement of inductors. I will add, there was a piece of copper clad PC board, grounded, placed over the first gain stages, presumably for shielding purposes.

Thank you for your considerations,

Silverfox.

[g1]

What model amp is that from?

[Chuck H]

I'm not sure how many people will step up for this question because it requires an enormous answer for real clarity. I'm only going to touch on some things.

The trace closeness and proximity absolutely do matter. Or not. It depends on what is close to what and whether or not there is a likelihood of radiant coupling. Since the power supply isn't in the signal chain the main concern would be safety. Are the traces large enough to handle the current? Are they far enough apart to avoid arcing? Another criteria would be keeping noisy power supply circuits and traces away from sensitive signal traces. Another criteria would be heat management. Some components, like electrolytic capacitors, shouldn't be placed near high heat producing resistors. Likewise, placing too many high heat producing resistors in one place can cause problems with heat concentration. Meeting all the criteria is rarely possible and it usually amounts to compromises. If you know what the different circuits are doing and how they behave both physically and electrically you should be able to look at the board and identify efforts to idealize by the manufacturer.

Signal traces require other proximity considerations mostly addressing sensitivity, phase relationships and grounding. All the usual layout rules apply. Don't locate sensitive traces near highly radiant traces, especially those that are of the same phase. Don't locate sensitive traces near noisy traces. Sometimes proximity rules can be fudged. For example, two signal traces of opposite phase can be close together as long as the frequency detriment of any incidental capacitance is acceptable. Grounding can be important. Grounds for sensitive circuits shouldn't share much trace distance with other grounds before either decoupling or confluencing at an established ground plane. Certainly two signal grounds of the same polarity shouldn't typically share long traces. And similar to capacitive coupling for signal traces sometimes signal ground traces of dissimilar phase can safely share some trace as long as the frequency detriment is acceptable. It's common for boards to require testing and redesign because some criteria hasn't been met well enough. Once a working board has been established it's possible for manufacturers to mass produce and recover the R&D cost.

There's WAY more to it than this though.

BTW, it's nearly impossible for someone to look at your photos and know good from bad with reasonable consideration (unless your Rain Man).

[R.G.]

I've done PCB layout and design for fun and as a day job for nearly 40 years now. Not all the time, but along the way. I do a lot more of it these days than ever. Some weeks are spent entirely on layouts and issues relating to them.

I laid out a similar-concept for the PCB for the Workhorse amps. It took me about a month to get it where I thought it should be and where I'd commit it to copper clad. There were two prototype/checking/revision cycles after that. I had no oscillation or interference problems at any point, but then if you review my posting here, I'm obsessive to the point of mania about how grounding and power distribution should be done, as I did have a full time job designing computer system power supplies for some years.

The reason I blather on about my history is not self-aggrandizement, but to qualify what I'm about to say. I've been there, and not only got the T-shirt, but worn out several of them.

Designing a PCB for a tube amp is not particularly difficult in the way that RF layouts are for VHF and microwave, nor in the practices demanded by high speed logic boards. The motherboard in the computer you type on is a MARVEL of the layout practitioners' art, by the way. I stand in awe of what we get in these things for throwaway prices. But the quirks of tubes and audio signals demand that you *know* what the current and voltage is on every trace, and what the driving and driven impedances are along the way, and what it passes near.

High impedance traces - traces where both the driving impedance and the driven impedance, like a plate coupling line into a tube grid, are going to pick up capacitive interference easily. It only takes a few pF from a 100+ signal line to a sensitive grid to make an ugly response bump that just sounds bad, but is almost impossible to find. High gain only makes it worse. High gain, high impedance, and large signal levels are a layout designer's notion of hell.

So, I mean it when I say that I cannot tell whether your layout is great or junk. I did write some of my own internal guidelines down about tube amp PCBs. They're on my web page, geofex.com.

One of the things to do is use non-standard PCB stock. See if you can get some 0.090" to 0.125" thick stock with 2oz copper on it. Mechanically support the board no farther than every 4 inches. I ran steel stiffeners down the length of the Workhorse amps for this reason. It had 16 (!) holddown screw locations on top of that.

Worry *a lot* about what ground currents flow in what traces. Putting one ground buss the length of the whole amp may be fine - or it may be intolerable and an oscillation disaster. I ran a complete clustered-star ground system for the amp, which I think accounted for why it was not only not very oscillation resistant, but remarkably hum-free.

Don't mount tubes on a PCB. Mount the tube sockets on a holder and run flying wires. Ditto controls and switches. Worry a lot about how you'll get to those wires to replace them when (not if) a control needs replaced.

The list is very long. And it's part of why Chuck said that the considerations are enormous. Really, you have to know what drives, receives, and is near every wire, and what that does to the signal on the wire.

Or be lucky. Luck is good, if you have it.

[silverfox]

Layout and Design- Clearifications

g-one- The board is from a Peavey Road Master. High gain, 160 watt amp.


R.G.- "Or be lucky. Luck is good, if you have it." I've always liked what Napoleon said of choosing Generals for the Military. “I only chose lucky Generals, and great Generals always make their own luck.”

War is like processor board layout design. Hmm and truth to the analogy in some ways... The logistics of suppling units in varing forms of opposition. And others exists- Traffic management. I believe your experience may be transposed with great value.

Okay enough about war, piece and traffic management.

I cut holes for the tube socket and will wire those pins to the components.

These points are well taken. I'm not doing a PCB but rather positioning the components similarly and then using point to point wiring on the Perf. Board. These voltages, due to it being only the phase inverter, are lower. However I will try to incorporate the cautions suggested as I wire the components up. I did this once before and didn't like the fuzziness in the sound. Perhaps I was getting problems related to your cautions. The Perf board is the layout similar to the PCB, however the wires point to point may be where the trouble comes in.



I'll look more closely at the traces for clues to success. I would also say, I am very appreciative for the information found on this site,

Silverfox.

[silverfox]

AB763 Power Amp Circuit

While putting the finishing touches on Peavey's version of the AB763 phase inverter circuit last night, I realized the negative feedback coupling cap is an electrolytic. After checking several times this is in fact the way it was designed. I did some research and found it to be a legitimate use for the cap but don't understand the advantage over a polyester or film cap that is non polar. Now that I think of it nonpolar caps are used on crossover networks.

Any ideas on this? See "C38"



Silverfox.

[Chuck H]

If you're talking about C38 the reason it's electrolytic is so it won't be the size of a lemon and cost fifteen bucks. uf's that high are hard to come by in film caps.

[loudthud]

C38 keeps DC from going through the Presence pot (and Resonance control if the amp has one). This prevents the pots from being scratchy as they are adjusted.

[Alan0354]

The question is really too broad to answer. I did a lot of experiment with the placement of components on amp and I did years of pcb layout for EMI, ground noise and RF layouts. I found the components placement is not critical in these kind of audio frequency amps, just go by logical order and some precaution described below. Make the input lines short. Try to put the grid resistor, coupling cap close to the grid so the grid line is not strung all over the place.

The more important thing is the grounding. Pay attention the high current notes. The return of the reservoir cap has to be close to the cathode of the power tube, preferably at the same star ground only with the power tubes. The PI and it's filter cap should be on one star. Then the preamp and it's filter cap should be a separate star away from the high current circuit.

Actually, if you do like the old Fender layout from preamp on one side and propagate down to the power amp. Then ground the grounding point to the chassis along the way, that is the safest way. All the filter caps for the section stay in that section and ground in that section, you should not have grounding issue. People like to play with star ground, that can be a double edge sword that more likely to cut yourself. You lift the ground and put onto a star, you really better know what you are doing. I trouble shoot for a guy building a kit Twin reverb and had hum problem. We trouble shot down to find he put one ground return of a preamp onto the star for the PI stage. Moving that one wire solve the problem. When in doubt, layout like Fender, drop the ground along the way.

On the power amp, the critical point is the NFB summing junction. Power amp is like a discrete opamp with +ve and -ve input. The feedback resistor goes from the OT back to the -ve input of the PI stage. The -ve input IS THE MOST CRITICAL POINT. Make it short. All the resistors that connect to the -ve input need to be close to the -ve input of the PI. The open loop gain of the power amp is not very high, I don't think oscillation is going to be a big problem.

The power amp is like an opamp with -ve feedback with a feedback resistor from the OT and a gain setting resistor. A lot of people use a pot for the gain resistor and connect the cap on the wiper. For max presence, the cap is shorted directly to the -ve input of the PI. In real design, this is dangerous as anyone design opamp circuit know that this is asking for oscillation. This put a pole on the feedback loop and cause 90 deg phase shift and you kill the phase margin. The safe way is to have a small resistor in series with the pot to form a lag lead network to compensate the phase.