acceptable ripple limits - versus unacceptably anal ripple reduction

[Enzo]

That is what I am saying, yes.


I have trained many technicians over the years. And I see certain patterns of behavior over and over. One involves scopes. techs not experienced with scopes develop the need/desire to get a waveform on the screen, no matter WHAT. I made up the term "trace chasing" for it, maybe someone has a better term. A guy connects his scope to some point, let us say the input stage B+, and sees a flat line, so he turns the scope vertical up and up until lo and behold, a waveform emerges. it might be a millivolt of ripple, hell it might be the local AM radio station a half mile away. But he got SOMETHING on the screen and so felt like he got somewhere.

I am not accusing you or anyone in this discussion of doing that, but it illustrates my point. If you look hard enough you will find something. I set up my meter or my scope to see a reasonable likeness of the node I am researching. If the DC appears calm, then I accept that it is, rather than go trace chasing to find a problem, any problem.

You want numbers? Go find an amp you think sounds OK and measure its ripple.


Let me put it another way. What is the amount of salt to put on your steak so it tastes best? I want to know in grams or milligrams and maybe break it down into salt grains per square centimeter. No? How about taste the steak, and shake a little salt on there if it is needed. We may never know how many milligrams of salt taste best, nor care. Deciding how much ripple to have on the lower B+ nodes is not something I decide with a meter. If I plug in an amp on my bench and it sounds OK, I am not going to even measure its ripple. If my first stages are hummy, chances are ripple is not the cause anyway. Unless those filter/decoupling caps are absent completely, if i still have ripple down at the end of the B+ chain, I am going to have TONS of it farther upstream to occupy myself. If I have a dried up cap in the end B+ node, it might supposed to be 20uf and it is acting like a 5uf, I am more likely to have instability in the amp than ripple problems. I am going nuts trying to find one now, but there were some ancient amps that used just a 0.1uf cap as a decoupler for the input stages.

[eschertron]

You want numbers? Go find an amp you think sounds OK and measure its ripple.

Exactly what I need to hear.
In my day job I mostly program industrial controllers, so the need to codify with as few degrees of freedom as possible comes with the mindset. I try to visualize and model real phenomena in terms simple enough for me to understand. I find it frustrating when I come to the realization that the physical universe has more variables in it than I want to allow. The more I read the textbooks, though, and the more posts I read from you and a few others - well, slowly I find I can handle the concept of more variables even if I can't express them numerically in a model. So, when I'm slow, sometimes a kick in the pants is what I need!

I started my first amp build in 2012. I think I encountered - and probably not yet solved - every problem there could be with the little unit. Yes, I had motorboating (had to figure out what the phenomenon was before I could look for a solution!), and high-freq oscillation, and then too much rolloff, and... well, good practice anyway. And I blame MEF for feeding my obsession.
Thanks, Enzo

[Enzo]

I used to work in the coin-operated amusements industry, and before 1976 all our pinball machines were relay logic systems. Bladed contacts had to close to complete a circuit to energize a relay, that relay in turn might energize several more, etc. Even when the machines went digital, as in computerized, we still had open blade contacts all over the playfields, activated by the ball hitting them.

I had a new tech come to me wanting to know the adjustment for the contact point gap in the blade switches. Kinda like setting the gap on a spark plug for your car, I guess. Imagine the "deer in the headlights" look on his face when I said "Adjust them so they work."

[eschertron]

Imagine the "deer in the headlights" look on his face when I said "Adjust them so they work."

Gratefully, I can see both sides of that.
Good one!

[R.G.]

I hope I was clear; I meant ripple in the first gain stage, not the first power supply node.

So did I. But then I switched gears on you. Unless your wiring is a problem and letting ripple sneak through when it should not, then reducing the ripple on the power supply node which feeds the first gain stage consequentially reduces the ripple IN that stage by the same amount. If there is nearly no ripple in the power feeding the first stage, then the gain stage can have ripple rejection down to zero and still have negligible ripple. If you kill the ripple source, you kill the ripple.

The standard power supply for tube amps generates a DC level at the first filter cap of some 5-10%, depending on the size of the filter cap and the loading. This is then run through a choke and shunted by a second filter cap. The choke is typically some henries, maybe 10, and the cap is 10-20uF. Just looking at voltage dividers and assuming the lowest frequency is 120Hz in the ripple, then Xl = 2*pi*F*L = 6.28*120*10 = 3536 ohms for 10Hz inductor, and Xc is 1/(2*pi*120*20E-6) = 66.3 ohms. So the 120Hz part of the ripple is attenuated roughly to about 66.3/(7536+66.3)= 0.0087 of its value on the first filter cap. Harmonics of 120Hz are attenuated by a factor of their harmonic number more than this. This goes to the screens and to the next Pi section. This is often a 10K-ish resistor and another 10-50uF cap. So the attenuation is another 66.3/(10066.3) = 0.0066, or 0.000057 of its value at the first filter cap. If the ripple at the first filter cap was 25V, then after a 10Hy choke and a 10K/20uF attenuation, it ought to be down to about 1.4mV (!), and be in the solid "doesn't matter' range, especially since there are usually at least one and often two more Pi sections after this to the first gain stage.

If you're seeing more than millivolts of real, no-fooling 120Hz ripple on the power supply feeding the first plate, there is something wrong in the power chain leading to that stage. If you're seeing more than millivolts of real, no-fooling 120Hz ripple on the PLATE of the first gain stage, it's getting in through some wiring or signal-bleed issue, or bad filter caps.

My suggestion was to feed the plate of the first gain stage from the source of a power MOSFET with its gate tied to 200-300V of zeners to ground, pulled up with a biggish resistor, and a cap across the zeners to eat the zener noise. The drain of the MOSFET is fed from a higher B+ stage, and the source has perhaps a mike or so of cap to ground on it. This ought to whack any signal on the B+ line by 20-30db below wherever it is.

I took a gander at the mosfet follies and more... Among the drawings I've looked at there's a section that show a zener/mosfet circuit that reduces the B+ arbitrarily, but in those drawings (I think) the ripple will simply ride on top of the reduction and not be eliminated. Can you point me to the circuits you had in mind?

Just described. It's a source follower fed a zener voltage in the amount of the B+ you want for the plate supply of the stage. The zener *conductance* and bypass capacitance in parallel are multiplied by the MOSFET. This is sometimes called a "capacitive multiplier" when a bipolar is used instead of the MOSFET. It's a really, really stupidly-simple series regulator. There may be other minor parts for protection, like a gate zener.

As Enzo and others note, it's not likely to be "ripple" in the sense we get when you use the word. Ripple starts as a sawtooth on the filter filter cap, and it's cause by the effect of periodic charging of the cap and the voltage rundown between charges. It's filtered by the parts between there and the stage using it.

So yes - you may have hum, not ripple.

And there is one particularly pernicious form that is subtle. If your wiring of the CT of the transformer to the first filter cap is wrong, you can get 120Hz spikes spread through the grounding to all stages, and it will be impossible to remove with any amount of work on the B+. It's caused by the rectifier charging pluses through the resistance of the wires, and wobbles the whole ground network up and down relative to the B+.

...so "for decoupling" you're saying that the primary purpose is to prevent oscillation, like motorboating?

Well, "decoupling" is for just that - removing paths through which one stage can couple *anything* to another. If it's a low frequency coupling, it can cause motorboating; if it's high frequency coupling, it can cause high frequency oscillation. The right amount of coupling in most cases is zero, which is not attainable; so we settle for all the DE-coupling we can get.

[eschertron]

If you're seeing more than millivolts of real, no-fooling 120Hz ripple on the power supply feeding the first plate, there is something wrong in the power chain leading to that stage. If you're seeing more than millivolts of real, no-fooling 120Hz ripple on the PLATE of the first gain stage, it's getting in through some wiring or signal-bleed issue, or bad filter caps.

Actually, I'm not 'seeing' any ripple just yet as I'm still in the designing and second-guessing phase. The numbers I get for ripple at the first gain stage come from Duncan's PSUD for the proposed design. From what I've gathered in this thread and reading through Merlin's chapter on noise and hum, I concur that I should be OK with millivolts (millivolt?) of ripple at that node. I understand that radiated 'interference' from other parts of the amp can be affecting the input more than the residual noise on the PS node or the generated noise in the input stage. That's also why I want to try DC heaters on stages 1 and 2 of this design.

And there is one particularly pernicious form that is subtle. If your wiring of the CT of the transformer to the first filter cap is wrong, you can get 120Hz spikes spread through the grounding to all stages, and it will be impossible to remove with any amount of work on the B+. It's caused by the rectifier charging pluses through the resistance of the wires, and wobbles the whole ground network up and down relative to the B+.

I've read and re-read what you and others have written about grounding, and am still trying to get a warm, fuzzy feeling about the different 'kinds' of ground currents. Until then I am obediently following suggestions about grounding, including using the first filter cap negative lead as the ground reference for the PT and the amp. Shortest path and all that.

I appreciate your time, R.G. Thanks!

[loudthud]

In PSUD you can look at current in the transformer. You would like to not infect the ground with small voltages from that current. Remember that filter capacitors are fairly low impedance at line frequencies so any noise on the ground gets conducted to the B+ side without much attenuation. You must let the resistors or choke in the B+ line isolate the noise upstream and keep the preamp ground quiet.

[R.G.]

I've read and re-read what you and others have written about grounding, and am still trying to get a warm, fuzzy feeling about the different 'kinds' of ground currents. Until then I am obediently following suggestions about grounding, including using the first filter cap negative lead as the ground reference for the PT and the amp. Shortest path and all that.

There is a good way to remember what to do with grounds.

==> EVERY GROUND WIRE IS REALLY A RESISTOR. <==

It generates a voltage across it if a current flows through it. The only way a resistor can have zero volts between its ends is to have zero current flowing.

A triode stage gets its "ground reference" from the voltage on the cathode. It's the voltage between the gate and cathode that makes it amplify. If the ground wire at the cathode is waving around because some other circuit is pouring current pulses through it, then the voltage caused by those extraneous current pulses interacts with the signal at the grid to put a representation of the current pulses into the tube. If the grid signal is zero, then only the extraneous pulses can be heard.

It's not possible to have zero current in a live circuit (although grid leakage current comes close) so the right amount of other currents flowing through the ground wire for a triode stage is just the current for that one stage. It still generates a ground voltage, but it generates it in the same way that a tiny cathode resistor does, and causes only a tiny reduction in gain. The more other circuits which share that ground return wire, the more unrelated voltages are generated on the ground wire and injected into the tube via the grid-cathode voltage.

The power supply first capacitor is especially bad. The pulses from the CT of the power transformer are BIG, as 100% of the current needed to run the amplifier sections powered by B+ has to come in through those pulses, and they are short in time duration. So they are BIG currents compared to anything else in the amp, even the power tube cathode currents, which are in second place. If the power rectifier pulses flow on a wire that can cause a ground voltage shift, they will cause hum/ripple that cannot be gotten rid of any way except rewiring. This is why the only place the PT center tap can go is directly to the minus terminal of the first filter cap. From the first filter cap, currents representing the power tube cathodes and all the rest flow. It makes sense to make the first filter cap negative terminal be where all the other "ground" wires return, and this is the theory behind star grounding. Whatever else is done, star grounding, or buss grounding, or whatever, the minus of the first filter cap is where the CT of the power transformer (or (-) side of the bridge rectifier if that's how it's wired) goes. This keeps those big, ugly pulses off the wires that carry signal.

[Chuck H]

++++

That's why I use three ground points. I've been bashed here for it a few times but I'm not fixing what isn't broken! A single star ground means that all the grounds share equal resistance at one point on their way to 0V. I have one ground near the input for all the preamp grounds. I have another about center chassis for moderate current and non preamp grounds like reverb drivers, tremolo oscillators, presence shunts, etc. And then a ground near the power supply for all higher current grounds. I never daisy chain a single ground (like joining two cathode circuits to the same grounded eyelet). I always try to place decoupling filters near the circuits they decouple. My amps are usually quiet enough that I get spooked on the initial turn on (Is it working?). Unless I do something stupid like locate a reverb transformer too close to the recovery stage (@#$%!) Still have to fix that one

But this isn't REALLY a grounding thread, is it. Still, as Enzo noted, the ripple at the preamp HV nodes in properly working amps of any reasonable design is small enough to be neglected. The more likely culprit for noise and hum will be layout and/or grounding.

[eschertron]

But this isn't REALLY a grounding thread, is it.

Yes, but... So my understanding is that the single ground wire that collects currents from the first stage (or first couple) goes directly to the first filter cap neg terminal for most effective noise rejection. Yah?

[Chuck H]

Yah! Ideally you want that filter close to the ground point and anything with a 0V reference that it decouples should be grounded there via it's own ground lead to that joining with no daisy chains on a board or terminal strip. It can look pretty untidy that way but it works.

[kg]

It's the voltage between the gate and cathode that makes it amplify.

there goes rg, blurring lines again!!!