I'm putting together a measurement setup for some tube amps I'm building and having a weird hum issue. I'm using a Focusrite Scarlet 2i2 and REW. Initial setup and calibration went fine using loopback cables. Hooking up to a headphone amp gave a large 60Hz hum (-60dB), though. Some of the things I've tried:
Any thoughts on what I could/should try next?
- I thought maybe magnetic from PS choke to output transformer, so I moved the power supply farther away (3 feet). Zero improvement.
- I thought perhaps ripple, though the amp is fully cascode CCS loaded. Even so I added additional RC filtering. Zero improvement.
- I figured ground loops. So I got an ADUM3166 usb isolator setup, and run the focusrite on battery power. Zero improvement.
- I thought maybe still some weird ground interaction/pickup from the PS. So I got a switcher HV power supply. Zero improvement.
- I thought perhaps the HV supply's rectifier was still causing issues. It has a large output capacitor, so I can literally shutoff and yank the power cord on the supply, leaving the entire amp floating and no operating PS nearby. You can still clearly see the 60Hz hum on the spectrum for 3-4 seconds before the supply drops out. (The filaments are also on battery.)
Any thoughts on what I could/should try next?
I have had a lot of problems with 60Hz hum pick up even with my RTX-6001, balanced input connections can help if you have that option. I have a small shield box, and also a ground plane on my bench.
My sympathies. It sounds like you have looked at a number of causes. Does the USB isolator also isolate the shields? I run my measurement computer on an isolation transformer and that helps with the common mode leakage currents through the computer PSU (The computer is an Intel Nuc 7i5)
I used WE repeater coils (101C?) to break ground loops and still have a couple around for that use. They are surprisingly OK until you get into the -80dBr and lower thd and noise range.
Look for other sources on your bench. I had problems with lights and swapped a number of LED bulbs before I found one that wasn't spewing EMI that I could pick up.
My sympathies. It sounds like you have looked at a number of causes. Does the USB isolator also isolate the shields? I run my measurement computer on an isolation transformer and that helps with the common mode leakage currents through the computer PSU (The computer is an Intel Nuc 7i5)
I used WE repeater coils (101C?) to break ground loops and still have a couple around for that use. They are surprisingly OK until you get into the -80dBr and lower thd and noise range.
Look for other sources on your bench. I had problems with lights and swapped a number of LED bulbs before I found one that wasn't spewing EMI that I could pick up.
I have the eval board for the ADuM3166: https://www.analog.com/en/resources...are/evaluation-boards-kits/eval-adum3166.html It has a complete gap of like 7-8 mm wide in the planes. I'm using balanced in & out on the focusrite scarlet, input transformer and output transformer on the amp. The filaments are normally powered by a small fully isolated DC-DC converter with just a couple pF capacitance, though I've tried battery as well. The 3166 has about 19pF capacitance in/out. The normal PSU I use has an electrostatic shield, and there's just one single connection to earth ground.
It feels like some high impedance node somewhere that I'm missing is picking it up somehow, especially when I still get it with the entire amp floating. I was suspicious too of the CCS's, but they're simple depletion types so they don't really have any high impedance nodes.
I'll try going through step by step grounding each tube grid hard to the star ground and see if I can find anything that way.
It feels like some high impedance node somewhere that I'm missing is picking it up somehow, especially when I still get it with the entire amp floating. I was suspicious too of the CCS's, but they're simple depletion types so they don't really have any high impedance nodes.
I'll try going through step by step grounding each tube grid hard to the star ground and see if I can find anything that way.
That's as good an approach as any. It's maddening, after 10 years or more I too am still working on the problem.
What sort of signal levels are we talking here?
I probably don't fully understand the problem so I might be missing something key. I have had success at times by running fairly high signal levels to the DUT and interposing an attenuator between the source and the DUT. If generator output level is low this can make a big difference. (I typically divide by 40dB or so.)
What sort of signal levels are we talking here?
I probably don't fully understand the problem so I might be missing something key. I have had success at times by running fairly high signal levels to the DUT and interposing an attenuator between the source and the DUT. If generator output level is low this can make a big difference. (I typically divide by 40dB or so.)
It seems like the GNDs are fully isolated, and I can't imagine the capacitance between the domains in the ADUM is sufficient at 60Hz to let anything significant through it. Are you grounding the focusrite to the same GND as the DUT - I believe that is the case based on what you have said. (All the sorts of things I would do btw.)
Can you post an example of one of your measurements with description of test method/setup? The goal probably should be to get the pick up down to the point where it doesn't invalidate your measurements. (I sometimes ignore particularly if the spur(s) is within 5 - 10dB of the measurement system noise floor. (Is that surrender? 🙂 )
Can you post an example of one of your measurements with description of test method/setup? The goal probably should be to get the pick up down to the point where it doesn't invalidate your measurements. (I sometimes ignore particularly if the spur(s) is within 5 - 10dB of the measurement system noise floor. (Is that surrender? 🙂 )
The noise floor is around -100dB when doing a FFT. The 60hz hum shows up at about -60dB, which in this scale is around 700uV RMS. I found that shorting the grid of the input tube to ground does nothing, but shorting the output tube grid to ground does cut the hum to around 250uV RMS (-72dB or so). Changing the input stage from a CCS load to a resistor load also drops the hum, implying the first stage's gain is part of it.
The tubes are kind of microphonic (old DHTs) so I wonder if it could be picking up a vibration even? There's no physical transformers sitting on the same workbench plugged in, but I can try isolating the test board with some vibration damping pads. The fact that it's at 60hz and not 120hz for the main peak makes me think it's definitely from the mains itself (magnetic coupling, vibration, etc...) and not power supply ripple. I'll also see if there's any non-DHT non-microphonic tubes I can try out in roughly the same circuit layout.
The tubes are kind of microphonic (old DHTs) so I wonder if it could be picking up a vibration even? There's no physical transformers sitting on the same workbench plugged in, but I can try isolating the test board with some vibration damping pads. The fact that it's at 60hz and not 120hz for the main peak makes me think it's definitely from the mains itself (magnetic coupling, vibration, etc...) and not power supply ripple. I'll also see if there's any non-DHT non-microphonic tubes I can try out in roughly the same circuit layout.
Most likely electrostatic (capacitive) coupling to surrounding objects based on the scenario you have described. Can you make a simple faraday cage with some wood dowels and aluminum window screen, ground it and place over the test fixture? Alternatively a grounded foil shield around the input DHT in particular - noting I have to shield the DHTs in my line stage for exactly this reason. You can extend it to the power tube as well, but likely at the point of diminishing returns.
What does the loop back look like?
What does the loop back look like?
I tried isolating physically with some dampening pads, to no effect. So not microphony.
I found out unplugging two LED lights (one my under-shelf lights for my desk, and one my ring light for my microscope) dropped the hum from ~700uV to a mere ~130uV. They were off, so this wasn't a magnetic field caused by current draw, but just some 60Hz field from their little switching plugs being radiated out through their rather long wiring. I was expecting possibly HF noise from the LEDs so I had them off, but not 60Hz hum just from being plugged in. There might be other devices that I cannot disconnect during my testing that would get me that last bit.
It's sort of a good news bad news thing. Good news, I finally have an explanation, I'm not chasing a measurement ghost. Bad news, means the circuit is pretty sensitive. I was hoping to use this amplifier hooked up by my PC where I work from home, and I'm sure there would be all kinds of garbage radiated by the PC, two monitors, a laptop, all their countless switching supplies, etc... Maybe having it all boxed up in a metal case would help enough? If I had to have the tubes covered in shields that would be a bummer though from an aesthetic standpoint.
I'll do some playing around with gain and listening levels. There's a good chance I could go down to just the lower-mu output tube, even though the circuit would actually end up with a small amount of net attenuation.
EDIT: To answer your question, there's absolutely no hint of the hum in the loopback test, so the focusrite scarlet isn't picking it up at all on its own.
I found out unplugging two LED lights (one my under-shelf lights for my desk, and one my ring light for my microscope) dropped the hum from ~700uV to a mere ~130uV. They were off, so this wasn't a magnetic field caused by current draw, but just some 60Hz field from their little switching plugs being radiated out through their rather long wiring. I was expecting possibly HF noise from the LEDs so I had them off, but not 60Hz hum just from being plugged in. There might be other devices that I cannot disconnect during my testing that would get me that last bit.
It's sort of a good news bad news thing. Good news, I finally have an explanation, I'm not chasing a measurement ghost. Bad news, means the circuit is pretty sensitive. I was hoping to use this amplifier hooked up by my PC where I work from home, and I'm sure there would be all kinds of garbage radiated by the PC, two monitors, a laptop, all their countless switching supplies, etc... Maybe having it all boxed up in a metal case would help enough? If I had to have the tubes covered in shields that would be a bummer though from an aesthetic standpoint.
I'll do some playing around with gain and listening levels. There's a good chance I could go down to just the lower-mu output tube, even though the circuit would actually end up with a small amount of net attenuation.
EDIT: To answer your question, there's absolutely no hint of the hum in the loopback test, so the focusrite scarlet isn't picking it up at all on its own.
It's electrostatic as suspected, the source forms one plate of the capacitor and the plate in your DHT and components on the bread board the other. Anything connected to the AC mains whether powered or not can contribute to the problem. The good news is that you can probably reduce the pick up by enclosing the chassis with perforated sheet stock. Build using a fully enclosed aluminum chassis and you might be good enough even without any additional shielding.
I'll have to consider a temporary metal/foil enclosure around the internals to see if that's enough or if I need to shield the tubes themselves as well. Sitting next to a PC seems like a pretty hostile environment for old DHTs. It seems like it should be possible to measure that field but most spectrum analyzers don't go that low.
Thanks for all the help @kevinkr
Thanks for all the help @kevinkr
You can use a sound card based analyzer or an old HP. I use an RTX-6001. My Picoscope 5242D has a 70MHz spectrum analyzer function and works well at audio frequencies, dynamic range isn't much over 80dB, but this is not a big deal for this sort of use.
Shielding the circuitry inside an enclosed metal chassis will make a significant difference.
Shielding the circuitry inside an enclosed metal chassis will make a significant difference.