Here is the distortion at 100Hz.
The levels of 2H and 3H are very low... -132 and -135dB. However, I cant get a good reading on THD+N due to the 60-120-180Hz line freq getting into the picture.
I looked at the output of the freq gen without the USB control input and the line harmonics are still there at same level. Only the other computer's USB to the A-P was connected to freq gen.... power running on battery only.
The +/- 15 supply is very quiet but I added 10,000 mfd across each PS output with No measured change in power line harmonic levels. So, Tentatively conclude it is coming from supplies inside the osc chassis? What do you have in the way of internal power?
But, I'll keep looking.
-Richard Marsh
The levels of 2H and 3H are very low... -132 and -135dB. However, I cant get a good reading on THD+N due to the 60-120-180Hz line freq getting into the picture.
I looked at the output of the freq gen without the USB control input and the line harmonics are still there at same level. Only the other computer's USB to the A-P was connected to freq gen.... power running on battery only.
The +/- 15 supply is very quiet but I added 10,000 mfd across each PS output with No measured change in power line harmonic levels. So, Tentatively conclude it is coming from supplies inside the osc chassis? What do you have in the way of internal power?
But, I'll keep looking.
-Richard Marsh
Last edited:
The power supply harmonics can only come from your power supply or near by power sources.
Is there any higher Z node inside, well the entire oscillator core. There is no internal regulation at the oscillator core that would be included in the power supply. The peak gain of an SVO at resonance is about 60dB or more. If the op amp PSRR is 100dB and at a peak gain of 60dB then there is only a 40dB margin. If your supply is good to -80dB - 40dB margin that puts it around -120dB. Which is about what you have.
Measure the supply directly and see what it is.
My measurements of power supply hum and harmonics were quite a bit lower.
Is there any higher Z node inside, well the entire oscillator core. There is no internal regulation at the oscillator core that would be included in the power supply. The peak gain of an SVO at resonance is about 60dB or more. If the op amp PSRR is 100dB and at a peak gain of 60dB then there is only a 40dB margin. If your supply is good to -80dB - 40dB margin that puts it around -120dB. Which is about what you have.
Measure the supply directly and see what it is.
My measurements of power supply hum and harmonics were quite a bit lower.
Last edited:
Does the AP have a 60Hz rejection filter. If the 60Hz can be attenuated the 180 might also drop. What's the 60Hz level at?
Try grounding the output ground of the PS if it's floating.
Richard if you measure from the Bandpass output then you attenuation on both sides of the fundamental. The low frequency noise should be lower but the oscillator distortion will be higher. If the AC power noise doesn't change then it not passing through the oscillator. If it does change then it's AC noise ingress. There are only three ways it can get in. Directly from the power supply, ground loops or magnetic field induction. Use what you have to figure which it is. Moving the DUT away from sources shows the amount of magnetic induction.
Power supply ingress can be directly measured or at least justified through calculation.
Ground loops are the most difficult.
The input from the power supply hits a very low Z ground plane. The BNC's are isolated from the chassis. The BNC ground is connected to the low Z ground plane. The chassis is just a Farady shield. No current passes through it. The USB and the entire control board are galvanic isolated from the main board. There's nothing to loop.
If the ground screw to the shield (chassis) becomes loose all hell breaks out.
Cheers,
Power supply ingress can be directly measured or at least justified through calculation.
Ground loops are the most difficult.
The input from the power supply hits a very low Z ground plane. The BNC's are isolated from the chassis. The BNC ground is connected to the low Z ground plane. The chassis is just a Farady shield. No current passes through it. The USB and the entire control board are galvanic isolated from the main board. There's nothing to loop.
If the ground screw to the shield (chassis) becomes loose all hell breaks out.
Cheers,
Something else that came to mind is that placing a large amount of capacitance on the output of a low Z regulator won't do anything to ripple or hum or noise for that matter. It's redundant. That's why we don't see much more that 10uF to 100uF just enough to stabilize the regulator depending on type and requirements. Adding some resistance between the PS and added capacitance might help some. An LC would be better.
When I see any change.... adding cap or moving distance from other gear..... it tells what direction to go. If no change, I look else-where. I have plenty of parts from filter R&D and plan to do the LC next. The HP PS is run by its self from a 1KW TOPAZ Ultra-Isolation transformer.
THx-- RNM
THx-- RNM
Last edited:
This hum is a problem I'm always fighting. My array of viktor oscillators are in a cast aluminium box with batteries to deal with the potential hum. I still need to run aseperate ground to the chassis of an analyzer and try different grounds on the analyzer to find the lowest hum. Often I'm shutting down the lights. If there is current in power cords under the bench they show up. At these levels every measurement is a challenge and a treasure hunt. I doubt its a power supply problem. This will show up if there are currents across the grounds but ripple almost never shows up if the caps don't couple it across the ground network.
My simple principle is to use a separate ground to shift any ground currents from electrostatic coupling away from the signal cable.
Getting hum down to -130 dB is hard. Lower is really hard. Good luck.
My simple principle is to use a separate ground to shift any ground currents from electrostatic coupling away from the signal cable.
Getting hum down to -130 dB is hard. Lower is really hard. Good luck.
Thanks.... I will find the source/cause..... the cure is something else.
1. What about the USB power? It isnt likely to be super low noise.
2. Is its ground isolated from other circuits and grounds?
Anyway, the distortion is super low and settles quit fast... you dont even notice the time because it is so short.
-RM
1. What about the USB power? It isnt likely to be super low noise.
2. Is its ground isolated from other circuits and grounds?
Anyway, the distortion is super low and settles quit fast... you dont even notice the time because it is so short.
-RM
Last edited:
Some useful numbers -135 dB from 3V is 178 nV. That would be 17.8 nA across .1 Ohm of ground wire. If the source is 120V 60 Hz I calculate about .6 pF is all that is needed to get that much hum. To help identify the source look att he spectrum. 60 Hz is power leakage, 120 and 180 are from ripple in a power supply and could be magnetic coupling from a transformer.
Demian the entire control board micro, usb and all is isolated. The grounds are completely separate. It was the only way this control system would work.
There are three shun regulators on board to isolate the currents from the ALC and timing circuits. Two +5V and one -5V. Thes supply the ADC and one Mdac in the ALC.
The one other +5V series regulator supplying the relays, tuning Mdacs and control interface.
There is absolutely no digital activity on the tuning during measuring time. All circuits are static with holding registers. During tuning the relays and Mdacs receive spi communication at 4MHz and 12MHz respectively. Not much time there. Then silence. All rails are RC filtered and then local low Z bypassing at each IC. I've had it running on batteries but the hum and ripple and harmonic were still there. Most of it seem to be coming from the monitor but of course if turned it off I couldn't see. The active TT is worse than the oscillator for pickup and it runs on batteries.
We live in a noisy environment.
Well the cost just kept going up.
Cheers,
Last edited:
The USB noise raises the floor up so high you could only measure the very first Phase Linear soup can heat sink power amplifiers. The ones sold as kits. Made in Vancouver.
Okay, I just gave my age away.
I built a low noise series regulator from Demian's design and that didn't make any different either. So I went back to the crappy Micronta bench supply I have. This was right about the time Demian was demonstrating stunning measurement results using some crappy power supply. A simple current source feeding a zener diode with a big capacitor on the output would work just as well. I did this using an LM317 as a current source and the noise was surprisingly low. Darn near 1u wide band. So what's all the power supply fuss?
Ego. Mine is better than yours and I can always hear it. 
Power supplies can be easy circuits and a fools errand. The supply is the easy part. Application so it really does what is needed is hard. You must understand and follow where every current goes and what circuits are sensitive. Most decent opamps have a lot of careful balancing inside to increase PSRR. Single ended transistor amplifier, especially without feedback, will have almost no supply rejection. Most good crystal oscillators are in this class and need really low noise supplies.
If your load has the dynamics of a resistor (none) you do not need a low impedance source with excess current capacity. If your supplying an opamp feeding 3V to an external load everything about the current loops will matter.
I think the problems are from leakage currents. After the basics are taken care of its either leakage currents or magnetic fields to deal with. As I illustrated above it takes very little capacitance to degrade the noise floor from -150 to -135. Often rerouting the ground connections make a difference. I reduced the 60 Hz modulation of a crystal reference oscillator in an instrument by moving a ground return to the other side of the rectifier diodes. Seeing the modulation was quite difficult but once you see it you need to get rid of it.
Power supplies can be easy circuits and a fools errand. The supply is the easy part. Application so it really does what is needed is hard. You must understand and follow where every current goes and what circuits are sensitive. Most decent opamps have a lot of careful balancing inside to increase PSRR. Single ended transistor amplifier, especially without feedback, will have almost no supply rejection. Most good crystal oscillators are in this class and need really low noise supplies.
If your load has the dynamics of a resistor (none) you do not need a low impedance source with excess current capacity. If your supplying an opamp feeding 3V to an external load everything about the current loops will matter.
I think the problems are from leakage currents. After the basics are taken care of its either leakage currents or magnetic fields to deal with. As I illustrated above it takes very little capacitance to degrade the noise floor from -150 to -135. Often rerouting the ground connections make a difference. I reduced the 60 Hz modulation of a crystal reference oscillator in an instrument by moving a ground return to the other side of the rectifier diodes. Seeing the modulation was quite difficult but once you see it you need to get rid of it.
Any hum modulation in the oscillator core will show up as side bands, +/- 60 Hz or whatever, on either side of the fundamental. You wouldn't be able to see this with the AP because it removes too much of the fundamental. But you can see it with the ADC sound card - notch method. Samuel illustrated this in his thesis. I don't know if this was done with an AP or something else. Can the notch depth be controlled on an AP?
The switched integrator in the PI controller is very sensitive to pick up being low pass in nature and the gain increases as frequency approaches dc. The gate time is very short. Roughly 40us for this one. Most of the time the switch is open and the integrator input is floating. Integrators are very fussy and it doesn't take much to get a ripple going in the control system from leakage discharge of the integrator capacitor. I can see the hum modulation at the output of the Mdac that converts the samples to dc. This all settles once the oscillator board is in a grounded chassis. It's one of the problems I observed with a switched integrator. The ADC's are just as sensitive.
Maybe we should stick it in a iron vault and bury it in the ground.
The switched integrator in the PI controller is very sensitive to pick up being low pass in nature and the gain increases as frequency approaches dc. The gate time is very short. Roughly 40us for this one. Most of the time the switch is open and the integrator input is floating. Integrators are very fussy and it doesn't take much to get a ripple going in the control system from leakage discharge of the integrator capacitor. I can see the hum modulation at the output of the Mdac that converts the samples to dc. This all settles once the oscillator board is in a grounded chassis. It's one of the problems I observed with a switched integrator. The ADC's are just as sensitive.
Maybe we should stick it in a iron vault and bury it in the ground.
Last edited: