Toroid transformer noise

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hi All,

I built a really nice tube guitar preamplifier with a toroid power transformer. The unit is extremely quiet except for the transformer spews out EMI (in the form of 120Hz buzz) that gets picked up but the guitar pickups (I'm using a shielded guitar with humbuckers too!). All tube amps do this yes, but this seems excessive. Turning the volume down on the guitar completely silences it.

The whole reason I went with a toroid was because they were supposed to NOT spew out as much EMI as EI types, and many design choices that I made were to reduce and avoid this in the first place.

Some details:
- I overspec'd the trafo to the manufacturer on each winding by about 30%. The manufacturer mentioned they do the same.
- The trafo has a GOSS band for magnetic shielding, which oddly may be making things worse.
- Chassis is thick steel.
- Leads are kept short and twisted tightly.
- Each winding feeds a discrete bridge rectifier with Schottky diodes and a cap across each winding to reduce HF hash from the diodes.
- Using the guitar as a 'sniffer', I notice that the noise is most intense with the pickup right on the trafo, and this dies off rapidly at about 1-2 ft away, but no matter how far I go from there the noise stays the same amplitude.
- Rotating the trafo does nothing.
- Rotating the guitar changes the harmonic spectrum of the noise, but not fundamental amplitude. There seems to only be one spot with a very narrow null in the magnetic field but it's still pretty intense.

How can we reduce this noise? Thank you in advance!
 
Last edited:
If it's coming from the transformer it should be 60Hz, not 120Hz which is the hum of the rectified power. Short the input of the amp. Does the hum go away? If so it's being picked up by the guitar/cable, if not it's from the circuit in which case you might have a bad or undersized filter cap. Do you have the schematic?
 
I have no experience with vacuum tubes or guitars as input source.
From my ancient experience with discrete transistor amplifiers: How is the grounding of the chassis made?

Thank you for your reply! The B+ supply grounds from a point after the last filter cap and ties to a switchable ground lift circuit. Connection is made with a ring washer to #8 PEM stud with a tooth washer to cut into the metal. The powdercoat was masked around the stud for bare contact.

The tube heater supply is elevated via a resistive tap from the B+ and this tap is filtered with a 22uF cap. The ground connection of the tap/capacitor is made to the B+ ground plane.

The Aux supply ground is taken from the last filter cap of the aux supply and it bypasses the ground lift circuit.

A schematic I'm working on will show this in more detail.
 
I can suggest:
- Use a tuned C-RC snubber instead of a cap.
- Use smaller value for first filter cap, and then a buffer (R or L) to larger filtering.
- Use a valve rectifier instead of ss diodes.

Thank you for your reply!

I will try the first 2 suggestions, valve rectifier is not possible.

I did try removing the caps altogether and it does not seem to make a difference. As for the smaller first cap, what is the mechanism at play there? The larger filter draws a bigger charge current from the winding, and thus more noise? For each first filter cap I use 2 rather large caps in parallel (3900uF x 2 on the low voltage, series connected 100uF x 2 for the HV).
 
If it's coming from the transformer it should be 60Hz, not 120Hz which is the hum of the rectified power. Short the input of the amp. Does the hum go away? If so it's being picked up by the guitar/cable, if not it's from the circuit in which case you might have a bad or undersized filter cap. Do you have the schematic?

Well it's more of a buzz than a hum, just like this video, although the video has more of a hum component than I'm getting: YouTube

So yes, it does go away when the input of the guitar is shorted, such as turning the guitar volume down all the way, removing the cable from the amp input (shorting jack) or even turning the amp gain down. It's definitely not the circuit but caused the transformer/wiring.

I'm hesitant that it's a filter problem because ripple on the main caps is where it's supposed to be. What is interesting is the waveforms scoped at the secondary fuses. The high voltage winding looks like your decent clipped-top sine wave that you expect, but the two low voltage windings look like more of a square wave with reverse ringing that is like 50% of the cycle.

Schematic is attached.
 

Attachments

  • PowerSupplyForum.pdf
    135.5 KB · Views: 153
You can record a sample, or set up a live microphone test, and then use free spectrum analyser software to see if it is mainly 60Hz or 120Hz. There are also free spectrum analyser apps for cell/mobile android phones.

If it is mainly 60Hz, then extra shielding around the transformer is a reasnable path to take, if the solid chassis and distance between guitar and amp can't alleviate your concerns.

If 120Hz and higher harmonics are dominant, then reducing the peak of the current waveform (and as a result widening the width of that pulse) is pretty much the path to take (other than shielding).

Having such high peak current levels can exacerbate the diode turn-off related noise issue - which is why a tuned winding snubber may assist, or go back to the future and use valve diodes.

You may be able to simply test the amp using a battery for valve heater supply and control power - that would remove charging pulse currents from the heater windings (given you use DC heater powering, and seem to have a substantial control power requirement - and may help localise where your main noise source is coming from. The heater windings could well be outermost on the toroid, and so not as shielded as the HT secondary winding.

Some background on the rectifier related noise issue is in:
https://www.dalmura.com.au/static/Power%20supply%20issues%20for%20tube%20amps.pdf

Ciao, Tim
 
Last edited:
You can record a sample, or set up a live microphone test, and then use free spectrum analyser software to see if it is mainly 60Hz or 120Hz. There are also free spectrum analyser apps for cell/mobile android phones.

First, thank you so much for your excellent reply!

I just did this, there's no 60Hz component. It's all 120Hz+, looks to be 120Hz intervals throughout the entire audio bandwidth of the amp at pretty consistent amplitudes.

The rest of your post was a goldmine of information!

So I tried to install snubbers and ran into some very interesting things. First, I observed the LV supply at the fuse and saw the same nasty ringing I saw yesterday (pic 1). Then I observed the HT supply which looks pretty clean (pic 2). I then removed the HT fuse and observed the LV supply and most of that nasty ringing is gone, but you can see some ringing now on the front side of the peak whereas before the ringing was on the back side (pic 3). Finally, I added a 100nF + 100ohm snubber with a Cx of 10nF and this made the ringing worse (pic 4).

This is all very confusing...Can someone explain what is going on?

I'm now attempting to figure out the snubber values based on the Quasimodo method mentioned in that excellent article that was attached.
 

Attachments

  • LowVoltageSupply.jpg
    LowVoltageSupply.jpg
    570.9 KB · Views: 227
  • HTSupply.jpg
    HTSupply.jpg
    601.9 KB · Views: 232
  • LV-NoHTFuse.jpg
    LV-NoHTFuse.jpg
    705.1 KB · Views: 223
  • LV-Snubber.jpg
    LV-Snubber.jpg
    732.9 KB · Views: 220
Nice to see a real world, albeit complicated, situation as you have multiple rectified current pulses passing through different windings, and having different conduction durations and leakage inductances :eek:

For clarity, where were the probe ground and tip connected for the scope photos? And were pic 1 and pic 3 at the same probing locations, but just with CH1 gain and offset different?

More insight may also be gained by putting a trace of winding conduction current (eg. a sense resistor between bridge neg and 0V) on the screen along with the related winding arm voltage.

The low resistances related to the toroid windings and ss diodes and large filter caps appear to be showing up diode 'turn-on' ringing effects as well as the more typically observed turn-off effects. The issue is effectively the same - a current wants to step change. There is also a chance that diode turn-on recovery is also coming in to play.

Anyway, it would be easier to discern behaviour details if you could obtain dual trace setups. That would also be good to set up, as a way to show any change from introducing a tuned snubber as compared to perhaps the base case of no local snubber cap, or your default snubber cap.

Ciao, Tim
 
For clarity, where were the probe ground and tip connected for the scope photos? And were pic 1 and pic 3 at the same probing locations, but just with CH1 gain and offset different?

Ground was attached to PCB screw attached to the chassis via a PEM standoff, and the measurement was made with the tip was on the fuse itself, on the rectifier side, in both cases. And ya, the gain was different, I was trying to really capture the ringing but never took a pic for a side by side comparison.

More insight may also be gained by putting a trace of winding conduction current (eg. a sense resistor between bridge neg and 0V) on the screen along with the related winding arm voltage.

The low resistances related to the toroid windings and ss diodes and large filter caps appear to be showing up diode 'turn-on' ringing effects as well as the more typically observed turn-off effects. The issue is effectively the same - a current wants to step change. There is also a chance that diode turn-on recovery is also coming in to play.

Anyway, it would be easier to discern behaviour details if you could obtain dual trace setups. That would also be good to set up, as a way to show any change from introducing a tuned snubber as compared to perhaps the base case of no local snubber cap, or your default snubber cap.

Ciao, Tim

I could likely do both experiments tomorrow. I'm a bit confused I have to admit. For the most part I get the basic concepts at play here (charging currents, different conduction durations and leakage inductances), but I am struggling to put it all together into how to capture more evidence required for a solution. It sounds like this may be a much bigger problem than just slapping a snubber on each winding and calling it a day, no?

I'll play around and see if I can come up with anything that appears useful. Should I even bother with my plan to inject a square wave with a function gen with all the other windings shorted and attempt to tune a snubber for each winding?

Thank you so much!
 
I can appreciate it is a long haul when its you doing all the work. I can profer that it should be a good learning curve, and should allow a better understanding of the influences at play.

The tuned snubber is just about minimising the disturbance that can become noticeable at diode turn-off. By circulating that disturbance energy within the winding in the easiest damped path, it can reduce transfer of some of that disturbance energy to other secondary windings, as well as escaping in to circuitry. It also has the advantage of cleaning up scope waveforms, so as not to distract from the bigger picture.

The magnitude and shape (harmonic structure) of charging pulses won't be changed by the snubber - those pulses will always be there. The concern then rises if they are complicit in your 120Hz noise transfer hassle. It looks like that is the main concern to focus on here. How to improve that situation comes I suggest from a better understanding - starting with the waveforms of individual windings, and whether some windings cause more noise egress than others (as identified by removing the B+ winding fuse).

For probing, you may get a cleaner (less prone to other parasitic signals) waveform by connecting the ground clip to the bridge negative 0V. That is elevated for the heater winding (should still be ok to clip there - or temporarily ground that bridge neg end for testing).
 
Last edited:
I can appreciate it is a long haul when its you doing all the work. I can profer that it should be a good learning curve, and should allow a better understanding of the influences at play.

I completely agree and I have no problem doing the leg work; spent all day researching and experimenting. My problem has always been when I don't fully understand something, I brute force it with will and determination, which is exhausting, lol.

The magnitude and shape (harmonic structure) of charging pulses won't be changed by the snubber - those pulses will always be there. The concern then rises if they are complicit in your 120Hz noise transfer hassle. It looks like that is the main concern to focus on here. How to improve that situation comes I suggest from a better understanding - starting with the waveforms of individual windings, and whether some windings cause more noise egress than others (as identified by removing the B+ winding fuse).

So, I started thinking about the current pulse, and my first thought was to simply remove a cap from each supply, cutting the reservoir in half. But I realized that would increase the charge pulse's current because the load stays the same (but the pulse width would be wider). The only way to reduce the charge pulses and to increase their width is to add resistance to the XFMR/BR/Reservoir loop. This problem seems to be a case of the dark side of trying to be too perfect: toroid transformer, schottky diodes, high value, low ESR caps in parallel (which halves the ESR). I likely designed this problem in! :(

For probing, you may get a cleaner (less prone to other parasitic signals) waveform by connecting the ground clip to the bridge negative 0V. That is elevated for the heater winding (should still be ok to clip there - or temporarily ground that bridge neg end for testing).

I plan to do this in a bit. I actually removed the heater elevation wondering if the elevation could have been causing some DC to flow in the coil but it appears to not. At least this will be safe!
 
How far away do u go till tha hum is tolerable? The angle of your pickup should also alter the hum intensity. Something is radiating 120Hz really well which implicates a big loop in there.
Itd be interesting to see your layout in there.

It's interesting, at about 1-2 feet away the noise drops off considerably but most of that seems to be higher harmonics. From that point I can get about 6-8 ft away (limited by headphones) and it will not decrease further. I can angle a bit and it appears the best position is 45 deg facing the unit on the left side.

I just hooked up another preamp (professionally made with EI core XFMRs) and it's buzzing is about 30% less, but the higher frequency harmonic content is drastically less than my unit. It seems like the problem actually is more above 120Hz! Oddly, with the guitar volume down or unplugged, my unit has far less noise and no buzz (the professional unit does still buzz in this case).

As far as loop areas, I tried really really hard to minimize every single loop area. Loops are either routed right next to or directly on top of one another for the most part, even the transformer leads are super short (3.75" for the primary, 1.75" secondaries).

Attached are the supply layouts. First is HV, second is the heater supply (left) and Aux supply (right).
 

Attachments

  • HVSupplyLayout.png
    HVSupplyLayout.png
    90.1 KB · Views: 194
  • LVSuppliesLayout.png
    LVSuppliesLayout.png
    162.7 KB · Views: 63
Last edited:
Thinking about the decrease in buzzing with distance element of your problem . Have you checked the preamp is not oscillating .

I just checked and no, there's no evidence of oscillation audibly or on a scope. Sound is very clean and smooth. It's not to say it's not, but it would likely have to be 50 MHz +. I was pretty diligent about using grid stoppers right on the valve bases with resistor vias on the pads of the tube. The power supply nodes are all super-clean too, the issue appears to be related solely to the transformer and immediate connections.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.