Again, ANY "enclosed loop area" will transmit and receive when time-varying EM fields or currents are involved:
Are the transformer secondaries twisted with the center tap, ALL the way to the rectifiers, and then is the CT wire twisted with the rectifier output wires, ALL the way to the first filter cap? (This might be very important.)
Are the wires from the input transformer to the 220k resistor tightly twisted (or shielded) ALL the way to each end? And is the resistor connection to the tube pin very short in length?
Are the wires from the input jack to the input transformer tightly twisted together, ALL the way to each end?
How is the input ground wire routed, exactly?
What about the inter-stage signal paths? Are there any corresponding conductors routed with them or are they just forming big loop antennas with something?
Are there any grounds that follow the power supply voltage back to the power supply, or is that all just a big loop antenna? How are the bias grid and cathode grounds routed?
I see big loops made by the anode and bias grid (and grid) wiring. Should those wires all run right up against each other, to minimize the loop areas?
Also, how are your heaters actually wired? Do you have a heater ground-return wire twisted with each heater wire, everywhere, for example?
Some photos would probably be a lot quicker and easier (and better) than typing answers to all of these questions.
Are the transformer secondaries twisted with the center tap, ALL the way to the rectifiers, and then is the CT wire twisted with the rectifier output wires, ALL the way to the first filter cap? (This might be very important.)
Are the wires from the input transformer to the 220k resistor tightly twisted (or shielded) ALL the way to each end? And is the resistor connection to the tube pin very short in length?
Are the wires from the input jack to the input transformer tightly twisted together, ALL the way to each end?
How is the input ground wire routed, exactly?
What about the inter-stage signal paths? Are there any corresponding conductors routed with them or are they just forming big loop antennas with something?
Are there any grounds that follow the power supply voltage back to the power supply, or is that all just a big loop antenna? How are the bias grid and cathode grounds routed?
I see big loops made by the anode and bias grid (and grid) wiring. Should those wires all run right up against each other, to minimize the loop areas?
Also, how are your heaters actually wired? Do you have a heater ground-return wire twisted with each heater wire, everywhere, for example?
Some photos would probably be a lot quicker and easier (and better) than typing answers to all of these questions.
DC for filaments should be better. But if the DC voltage and ground return were routed by separate paths, or not kept always together, so that they enclosed geometric area between them, then they would still have AC currents induced in their loop by stray AC fields in the air (in proportion to the enclosed area, all else being equal), according to Faraday's Law (and Maxwell's Equations).
If all of the AC loops near the power transformer and smoothing caps were also not closed-up, or if fluorescent lights were present and the cover was open, just for examples (RF is another), there could still be significant EMI effects, which could end up being audible since there's so much gain.
If all of the AC loops near the power transformer and smoothing caps were also not closed-up, or if fluorescent lights were present and the cover was open, just for examples (RF is another), there could still be significant EMI effects, which could end up being audible since there's so much gain.
I need to recap a few things; I feel I'm being redundant.
This noise exists. It's not the scope. I fix and built many projects - if they're silent, the scope shows silence. If there's a hum or noise, the scope shows it. This is no exception. Had good thoughts regarding scope, but that's not regarding what's actually happening to the preamp noise. No more scope thoughts!
Grounding:
There is only one point from star ground to chassis.
I'm running every individual ground reference for each stage to what I'll call the 'signal star'.
The power supply grounds of the first, second, third and fourth filter caps are tied together, in that order. The power transformer CT is (now) connected to the first filter cap. I'll call this the 'power ground'.
The signal star and the power ground each have an individual wire running to chassis ground.
I'm using a mix of polyprop ad polystyrene caps. Repostioning the location of these caps makes no difference, soundguruman.
There is nothing hooked up to the input jack - no mic, no other equipment. The secondary of the mic transformer is terminated with a 220k resistor. I've experimented with unhooking the mic transformer, going from shielded to unshielded wire from xfmr -> grid; terminating the grid resistor at a different ground point on the star - no changes. *I've concluded it's not being injected in the grid by termination or wiring.*
gootee - rearranging to this power cap scenario (ct to first, these to second, these to third, these to fourth, connect to ground at fourth) makes no difference. I'd even think that you wouldn't want to hook it up that way, as it technically presents a lower impedance to ground for the third cap (low current draw) and a higher imp to ground for the first and CT (high draw). also, input shorted and input open (yet grid terminated) shows no noise difference (hence 'it's not the grid' conclusion)
let me reply with quotes to your other post in a minute... thanks for the help!
i've never had a problem with AC heaters. even in doing a phono stage (which complete with power amp is even more than 80db gain...), I'm not ready to resign to 'DC or bust' for heaters. not saying it can't be the solution, but i'll run the gamut of other ideas first.
This noise exists. It's not the scope. I fix and built many projects - if they're silent, the scope shows silence. If there's a hum or noise, the scope shows it. This is no exception. Had good thoughts regarding scope, but that's not regarding what's actually happening to the preamp noise. No more scope thoughts!
Grounding:
There is only one point from star ground to chassis.
I'm running every individual ground reference for each stage to what I'll call the 'signal star'.
The power supply grounds of the first, second, third and fourth filter caps are tied together, in that order. The power transformer CT is (now) connected to the first filter cap. I'll call this the 'power ground'.
The signal star and the power ground each have an individual wire running to chassis ground.
I'm using a mix of polyprop ad polystyrene caps. Repostioning the location of these caps makes no difference, soundguruman.
There is nothing hooked up to the input jack - no mic, no other equipment. The secondary of the mic transformer is terminated with a 220k resistor. I've experimented with unhooking the mic transformer, going from shielded to unshielded wire from xfmr -> grid; terminating the grid resistor at a different ground point on the star - no changes. *I've concluded it's not being injected in the grid by termination or wiring.*
gootee - rearranging to this power cap scenario (ct to first, these to second, these to third, these to fourth, connect to ground at fourth) makes no difference. I'd even think that you wouldn't want to hook it up that way, as it technically presents a lower impedance to ground for the third cap (low current draw) and a higher imp to ground for the first and CT (high draw). also, input shorted and input open (yet grid terminated) shows no noise difference (hence 'it's not the grid' conclusion)
let me reply with quotes to your other post in a minute... thanks for the help!
i've never had a problem with AC heaters. even in doing a phono stage (which complete with power amp is even more than 80db gain...), I'm not ready to resign to 'DC or bust' for heaters. not saying it can't be the solution, but i'll run the gamut of other ideas first.
1) Yes, as much as possible. Secondary CT isn't twisted with anything.
2) input xfmr secondary has 220k terminating right on the transformer / 6" of shielded cable to grid input / have tried unshielded, no change / have tried terminating grid resistor at tube pin instead, no change
3) Nothing connected to primary side of input transformer yet / to be said, disconnecting the input transformer makes no change (keeping the 220k grid resistor to ground connected) - goes with the 'not the grid' conclusion
4) input ground wire, do you mean first stage grid resistor ground? just like the rest - with an individual run to signal star
5) interstage signal paths are all shielded cable. shielded cables are terminated on one end only to the corresponding stage's grid ground; have tried extensive re-routing and even 'removal' of interstage and input cables; no change in hum - 'not in the grids'
6) any grounds in the signal stages go to the signal stage grounds - this includes cathodes, screen bypass, grid resistors.
7) are you suggesting all of my grounds need to be twisted together? twisting only prevents AC induction, right? what does that thought have to do with grounds? all wires to star ground are hugging the chassis tightly, and are away from any small heater loops locally at tube sockets.
8) heaters are wired in parallel in backwards stage order: 6.3v AC from power xfmr (confirmed voltage) -> 12au7 (humbucking) -> ef86 -> ef86 -> 6sj7 -> 6sj7; standard precautions (heavy gauge solid core, tightly twisted, close to chassis, away from anything it can be)
let me attempt some pictures.
Great. Good recap. All sounds pretty good, so far.
But measuring the noise on the ground, while possibly interesting, might not be as helpful as measuring it across each stage's grid-to-ground resistance. (And doing that with input both shorted and open would tell a lot about whether or not loop area in the input circuit was a problem. With input open, one of the loops does not exist. If you get more noise with input shorted, then there is a loop there that is acting as an antenna. This is probably more liklely with a single-ended input, but it's easy-enough to test, anyway,)
If you do measure noise "on the ground", I would think you'd want to measure between star ground to any other ground point, or maybe between two grid input-ground-reference points. But then the probe's ground lead would have to be too long, and would make a loop that would be an antenna, anyway, and it would be difficult to know what was being measured. Maybe the way to do that would be to use differential mode, i.e. if your scope can subtract one channel from another, while using no ground leads with the probes. What make and model is it?
I wondered about the CT and smoothing caps grounding method that was given, too, actually. It seems like grounding the first cap, and the CT, completely separately, would keep the CT current and the bulk of the charging pulse currents from sharing any conductor with anything else, on the way to the star ground. It's the sharing that's bad. The nastier currents induce voltages across the inductance (and resistance) of the conductor, which are seen (to some extent at least) all the way back up the chain, by anything that eventually goes through any bit of the shared conductor. The subsequent caps should have much-less-obnoxious currents in their ground pins so it seems like it would be wrong to have them share their ground conductor with the first cap and the CT. But I could be completely wrong. For some reason I keep forgetting how that part is typically done.
Carry on.
But measuring the noise on the ground, while possibly interesting, might not be as helpful as measuring it across each stage's grid-to-ground resistance. (And doing that with input both shorted and open would tell a lot about whether or not loop area in the input circuit was a problem. With input open, one of the loops does not exist. If you get more noise with input shorted, then there is a loop there that is acting as an antenna. This is probably more liklely with a single-ended input, but it's easy-enough to test, anyway,)
If you do measure noise "on the ground", I would think you'd want to measure between star ground to any other ground point, or maybe between two grid input-ground-reference points. But then the probe's ground lead would have to be too long, and would make a loop that would be an antenna, anyway, and it would be difficult to know what was being measured. Maybe the way to do that would be to use differential mode, i.e. if your scope can subtract one channel from another, while using no ground leads with the probes. What make and model is it?
I wondered about the CT and smoothing caps grounding method that was given, too, actually. It seems like grounding the first cap, and the CT, completely separately, would keep the CT current and the bulk of the charging pulse currents from sharing any conductor with anything else, on the way to the star ground. It's the sharing that's bad. The nastier currents induce voltages across the inductance (and resistance) of the conductor, which are seen (to some extent at least) all the way back up the chain, by anything that eventually goes through any bit of the shared conductor. The subsequent caps should have much-less-obnoxious currents in their ground pins so it seems like it would be wrong to have them share their ground conductor with the first cap and the CT. But I could be completely wrong. For some reason I keep forgetting how that part is typically done.
Carry on.
In bringing all of the power supply caps together on one 'power ground' and then running that star to chassis - that was an improvement in the ground/chassis noise seen on my scope. I'm starting to move away from thinking that the hash and very small ripple I mentioned that I see on the chassis is the problem.
(Pictures aren't turning out. I'll borrow a camera soon....)
So...
if it isn't the grid...
if it isn't the grounding scheme itself..
perhaps this is something to do with the transformer?
describing this as best i can: the transfomer is mounted flat, with the lams running parallel to the chassis. so, one half of the windings is outside (covered), and the other half of the windings is inside the chassis, and not covered by a bell or shield.
could this be doing something?
(pictures, i know...)
Then does it at least run right next to the secondaries? Each secondary flows to/from the CT. Each makes A LOOP with it (or maybe lots of them). If a loop has enclosed geometric area (space between the in/out conductors, anywhere), in this case it's a transmitting antenna. After the rectifiers it's also transmitting 120 Hz (maybe the harmonics you saw?).
Try disconnecting that 220k resistor, when the transformer is removed. That breaks the loop. If the noise goes away, you will KNOW which loop is involved. Small probability but big payoff if it occurs = worth considering.
No. I meant the one that comes in with the input cable, which is labeled "at jack" on the schematic. What does "at jack" mean? Please say it doesn't mean that the jack is grounded to the chassis.
Can't agree! "Signal" is only one conductor, correct? Using a shield, that I assume is "grounded" (to where?!), and connecting the shield on one end only, would be great, IF you had twisted pair inside the shield, and the signal's corresponding "ground return" on the other wire in the pair.
It's about LOOP area. But if you have even REMOVED the signal and input wiring, then maybe it's in the power/gnd system.
I'm not smart-enough about tubes to know how those should be routed, to not form any enclosed loop area. Just running wires to star ground could still be forming large loops with the SOURCE conductors. Everything should be paired and then run next to each other or twisted. That is NOT always possible. (Maybe consider adding a big blank PCB for ground and glue all the wires to it. (grin))
See my comment for number 6. Unless the corresponding signal or power supply current for each of those ground wires is running in the chassis (which it's not), having them hug it doesn't lower the enclosed loop area, although it might provide a little shielding. The current in each ground wire also COMES FROM somewhere, and they eventually form a loop (a "circuit"). Ideally, where it comes from and where it goes should not have any space between them, anywhere. Otherwise, any changing magnetic field (which is included in every changing electromagnetic field) WILL induce a current in the loop, which will induce a voltage across any impedance in the loop.
OK. But the point was whether or not "the outbound" and "the inbound" have much space between them, anywhere. Sounds like they don't... but be sure. So let's see, does that mean that you run the heater winding's center tap "ground" anywhere? Or, where does it go, exactly?
And have you measured for noise across the heater connection, at each tube?
Once again, about measuring noise: When you try something, like removing the input transformer, just as an example, and report "no changes", what does that "no changes" mean? You need to be measuring the noise across a grid resistor (or pot), or across the output. You CAN'T gauge the effects of what is tried by measuring the noise on the ground. That can't be trusted. Why? Well, for one thing, I could put the same noise on the ground AND the signal lines, and it wouldn't affect the output, as long as it was in phase where it mattered (like at the top and bottom of the grid resistors). So please standardize on measuring the noise between two points, where it's meaningful, to everyone.
A simple list of the noise levels across each stage's grid-to-ground input resistor, and across the final output connector, would be appreciated. Technically, that is to be done with the inputs SHORTED. That means that you should jumper the two input conductors or pins together, to guarantee close to zero volts, there.
Thanks. That would sure save a lot of words, I'm guessing.
Last edited:
The shielded cables for interstage signals should probably NOT have their shields terminated (on one end only) to the grid-resistor ground-reference points. They could pick up noise and inject it into the ground, there, where it would essentially arithmetically sum with the grid signal (probably after inducing a voltage across the conductor to star ground). Either terminate them at both ends (not sure WHERE, on the source end, but whatever corresponds to the same signal's ground, there), or, to the chassis or star ground. I'd probably try terminating both ends, first, while measuring the noise across the output of that stage while trying different terminating ground points.
Edit: I'm signing off. I have an early meeting tomorrow morning (in seven hours, after an hour drive) and have to sleep.
Edit: I'm signing off. I have an early meeting tomorrow morning (in seven hours, after an hour drive) and have to sleep.
Last edited:
let me investigate the secondary CT getting closer to the secondary AC lines.
oh - when i say 'no change', i'm reporting back of the amplified noise after the plate coupling cap. (sorry, thought this was clear!)
for grid 220k talk: disconnecting the grid leak resistor just causes a large amount of 60cycle hum to appear on the output of the tube in question. about 2.6v p-p.
there isn't any input cable right now.... forget the 'xlr' part, as no wire is actually going to the XLR (or anything) from the primary side of the input xfmr
interconnect signal shielded cables - the shield on each is connected to signal star ground via it's own path. locally grounding them to the potentiometer casings (for fun) actually didn't cause any additional hum to appear.
"noise across the heater connection"? as in, signal sum and difference between the heater leads at each tube? or?
oh - when i say 'no change', i'm reporting back of the amplified noise after the plate coupling cap. (sorry, thought this was clear!)
for grid 220k talk: disconnecting the grid leak resistor just causes a large amount of 60cycle hum to appear on the output of the tube in question. about 2.6v p-p.
there isn't any input cable right now.... forget the 'xlr' part, as no wire is actually going to the XLR (or anything) from the primary side of the input xfmr
interconnect signal shielded cables - the shield on each is connected to signal star ground via it's own path. locally grounding them to the potentiometer casings (for fun) actually didn't cause any additional hum to appear.
"noise across the heater connection"? as in, signal sum and difference between the heater leads at each tube? or?
Had to re-read and get back to this premise to keep in mind.
Switched 6SJ7 from pentode to triode; gain of noise decreased accordingly.
Trying a 'staged star' approach, so that grid and cathode are connected together before going to the star ground... perhaps...
PROGRESS!
connecting the grid's ground to the cathode and running both to star improved amplified noise down to ~12mv from 20mv.
thinking in induced loops... especially with the stage ground wires to star travelling over lots of real estate... although it's not a closed ground loop, it seems to still induce as such through the tube's internal resistances / capacitances (can't forget about those)
spell check says there's no such word as 'capacitances'. bah!
connecting the grid's ground to the cathode and running both to star improved amplified noise down to ~12mv from 20mv.
thinking in induced loops... especially with the stage ground wires to star travelling over lots of real estate... although it's not a closed ground loop, it seems to still induce as such through the tube's internal resistances / capacitances (can't forget about those)
spell check says there's no such word as 'capacitances'. bah!
You certainly don't want long connections to a far away star point in low-noise circuits. As the input uses a transformer it is easy to arrange for the first valve grid and cathode to use the same ground connection - then they are guaranteed to be working from the same reference. Don't run 'both' to the star; run just one wire to the star otherwise you have introduced an unnecessary loop.
Never blindly follow a grounding scheme, whether star or bus or anything else. Always think about where the currents go, where the loops are. Eliminate all ground loops, and minimise the area of all signal and power loops.
Never blindly follow a grounding scheme, whether star or bus or anything else. Always think about where the currents go, where the loops are. Eliminate all ground loops, and minimise the area of all signal and power loops.
Like Tom says, transformers can get the noise in.
So do 1N4007 btw. Try 22 ohm resistors between them and the first cap to lower current pulses. Or better still
So do 1N4007 btw. Try 22 ohm resistors between them and the first cap to lower current pulses. Or better still
An externally hosted image should be here but it was not working when we last tested it.
PROGRESS!
connecting the grid's ground to the cathode and running both to star improved amplified noise down to ~12mv from 20mv.
themagicmanmdt,
How goes it?
There might be a little better explanation of the loops, in post 16, here:
http://www.diyaudio.com/forums/tubes-valves/215237-hum-homebuilt-amp-2.html#post3074815
Refer to the schematic linked in post 1, there, or here it is at: https://sites.google.com/site/jonandersonmn/5902.png
Also, there is one error about the original grounding scheme in post 16. But it is corrected here, in post 21:
http://www.diyaudio.com/forums/tubes-valves/215237-hum-homebuilt-amp-3.html#post3075791
P.S. If you also look at the power supply wire at the top of your schematic, and consider caps to be shorts for AC and its harmonics, there are at least three completely-closed loops (not through tubes) that pass through each pot, where induced voltages would directly feed the grids.
Cheers,
Tom
Last edited:
so, i implemented the 'raised ground' by connecting the joint power ground/signal ground point via a resistor and cap in parallel. i tried different combinations, yet they all yielded a louder noise level.
however, that's not what was enlightening.
i was able to more clearly see the induced noise on ground, thanks to the induced ground voltage over the resistance added. the 'elevated' ground showed a clear AC line, and the higher resistance i used for the star-to-ground connection (up to 100 ohms, didn't matter how much cap was used to bypass it), showed a larger waveform...
100ohms yielded 20mv peak, so there's, somehow, a .2mA leak to ground that is outside of my star. not only that, but it seems to be at line frequency.
i'm thinking that it's in the power transformer windings. my other observations of the ground noise at the power transformer mounting bolts being higher than the rest of the chassis seems to help this. moving the star ground further from the power transformer also decreases the noise on the star and hence in the circuit.
i'm going with my initial question and title of the post - due to power transformer. i haven't insulated the main bolts from the chassis (as was mentioned), which may eliminate this leakage and connection... or, i just need a different power transformer.
any more thoughts on all of this?
however, that's not what was enlightening.
i was able to more clearly see the induced noise on ground, thanks to the induced ground voltage over the resistance added. the 'elevated' ground showed a clear AC line, and the higher resistance i used for the star-to-ground connection (up to 100 ohms, didn't matter how much cap was used to bypass it), showed a larger waveform...
100ohms yielded 20mv peak, so there's, somehow, a .2mA leak to ground that is outside of my star. not only that, but it seems to be at line frequency.
i'm thinking that it's in the power transformer windings. my other observations of the ground noise at the power transformer mounting bolts being higher than the rest of the chassis seems to help this. moving the star ground further from the power transformer also decreases the noise on the star and hence in the circuit.
i'm going with my initial question and title of the post - due to power transformer. i haven't insulated the main bolts from the chassis (as was mentioned), which may eliminate this leakage and connection... or, i just need a different power transformer.
any more thoughts on all of this?
Given that it higher closer to the transformer . I would stainless steel bolts and some washers and get the tranny off of the chassis . Teflon , polyprop, rubber or fiber will work . My guess is the the frame of the tranny is inductively coupling to the chassis distance and non magnetic (stainless steel ) or brass bolt will help.
Last edited:
If you have a ferrous or magnetic material chassis and mount a EI lamination style transformer using "Z" orientation i.e. with the laminations in the same plane of the chassis and the bottom of the winding going through the chassis, then you will definitely induce some AC voltage into the chassis. Insulation may help but I'm not sure if it can be totally alleviated as it's due to magnetic leakage i.e. so-calle "stray field". There still needs to be a return path somewhere but your sections of wire from each star to chassis may create a circuit somehow. Pictures???
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.