I use headphones directly on the output of power amps to assess the noise. They make a very useful tool when it comes to cable dressing, earth point pickoff location, transformer orientation and cable routing etc. With sensitivities typically 1mW/90 dB you can really dig into the noise quickly and conveniently.
I don't think headphones are useful for providing objective measurements, but they make a great 'differential evaluation' tool.
There is a place for all these tools/techniques - don't knock 'em.
I don't think headphones are useful for providing objective measurements, but they make a great 'differential evaluation' tool.
There is a place for all these tools/techniques - don't knock 'em.
You're right.
At 20kHz it's about 50µohm.
If I do the same calculation as AndrewT
10 A give 500µV
But 10 A give ~60V output
So the ratio is 101dB worst case.
Regards, Hervé.
If you think something is wrong, tell me.
These numbers are to say I don't agree to compare with 1V rms as you did in a previous post.
Anyway this is not realistic since 10A are not flowing between input "chassis & board" ground and output "chassis & board" ground.
Hervé.
I am asking what voltages and currents you are specifying for your ratio comparison.
Regarding the 10Apk I picked out the hat, this takes account of the peak charging current into a smoothing capacitor that typically can be 3 to 6times the rms current drawn through the rectifier.
One might expect the rms current to be around 2Arms (10Apk) indicating an average power delivery of 120W for dual 25Vac secondaries.
Regarding the 10Apk I picked out the hat, this takes account of the peak charging current into a smoothing capacitor that typically can be 3 to 6times the rms current drawn through the rectifier.
One might expect the rms current to be around 2Arms (10Apk) indicating an average power delivery of 120W for dual 25Vac secondaries.
While setting up a preamp I am building, I looked at this thread. I made some measurements.
First, the line amp is a Borbely 797/203 (hybrid tube and MOSFET), that I made from scratch using his 1998 Glass Audio article. It is configured as single-ended (the negative input is to signal ground), with 20-dB of gain. The signal and power supply grounds are connected (as per Borbely's design). All of the RCA jacks are insulated from the chassis, and their barrels (signal ground) are tied together along a solid, 16 AWG straight wire that is solder-mounted very close along the jacks, and one end of it goes to a single point in the middle of the board (star ground). The star ground is about halfway between both channel halves of the PCB. The power supply ground is connected to the star ground. The power supply and tube heater are on a separate board next to the line amp board.
The comparisons I made were between a floating star ground (that is not connected to the chassis,) and with the star ground connected to the chassis, which is steel on the sides, bottom, and top, and aluminum on the front and back.
As usual, the mains (115 VAC) ground is connected to the chassis for safety.
My measurements are as follows:
A clean, 100-MHz sine wave at 0.8 mV peak to peak, both channels, on either grounding system. In other words, no noticeable difference. I could not see any 60-Hz signal, and I usually get a 100-MHz signal on measurements unless I go to a lot of trouble to greatly shorten the scope's ground lead. Also, the chassis was open at the top and right side.
This noise signal did not vary with the volume-gain pot position, or whether I shorted the inputs or not.
Sorry there were no interesting differences. I can describe the lead lengths etc. if anyone is interested, but the layout is not atypical regarding length. Only the leads from the volume pot to the line-amp PCB inputs are shielded, and they route near the tubes.
I also listened with headphones, and could not hear any hum no matter what I did. The headphone jack's shield is insulated from the chassis, and I used a wire from its shield to the star ground.
If anyone has suggestions for more precise measurements, I will give it a try. The scope's minimum calibrated divisions go to 2 mV (with 8 sub-divisions), so it is easy to see about 0.4 mV. The scope's bandwidth is 200 Mhz.
I did connect the buffered vertical scope output back into the other channel to get more gain, but I did not work very hard at it and the wires picked up some added noise.
I did some measurements on a low gain preamplifier in two configurations.
The dut is a 4 channels buffer preceeded by a 4 channels 10k pot. The buffer is mad with 2* 2 channels AD712 AOP.
The "pcb" is a double sided test board with ground plane on top.
The "normal" config is : power supply ground connected to box (steel) ground connected to audio ground.
PCB ground plane is connected to the box by 4 brass spacers , input and output RCA grounds are connected to the box, power supply (+15 0 -15) ground is connected to the box, the 4 "cold pin" of the pot are connected together to the box.
The isolated config : pcb, supply ground, rca, pot disconnected from the box.
For the measurement, input preamp is grounded before the pot which is in max attenuation position.
I used a sound card (Terratec DMX 6 fire usb) with a 48kHz sampling frequency and I record a shorted sound card signal.
On the plots, the isolated config (1) is in red, all ground to the box (2) in blue and the SC shorted input (3) in green.
The difference between are not very important but we can notice :
There is a peak at 50Hz (~1.5dB) in config 1 not present in the other ones
Another one at 250Hz in config 1, also present in config 2&3 1dB smaller
Another one at 500Hz in config 1, also present in config 1&2 0,5dB smaller
The peaks at 5kHz are ~the same.
Peak at 8kHz is smaller by more than 1.5dB in config 2&3
Peaks at 14,4kHz are ~the same.
Peak at 16kH is smaller by more than 3dB in config 2&3.
Preliminary conclusion : Connecting the audio ground to the chassis of a low gain preamp is a very very bad idea !
It is simpler than isolating all the connectors, ground plane etc.. to the box
It is more quiet than isolating, and it is not worse if box is in plastic.
The dut is a 4 channels buffer preceeded by a 4 channels 10k pot. The buffer is mad with 2* 2 channels AD712 AOP.
The "pcb" is a double sided test board with ground plane on top.
The "normal" config is : power supply ground connected to box (steel) ground connected to audio ground.
PCB ground plane is connected to the box by 4 brass spacers , input and output RCA grounds are connected to the box, power supply (+15 0 -15) ground is connected to the box, the 4 "cold pin" of the pot are connected together to the box.
The isolated config : pcb, supply ground, rca, pot disconnected from the box.
For the measurement, input preamp is grounded before the pot which is in max attenuation position.
I used a sound card (Terratec DMX 6 fire usb) with a 48kHz sampling frequency and I record a shorted sound card signal.
On the plots, the isolated config (1) is in red, all ground to the box (2) in blue and the SC shorted input (3) in green.
The difference between are not very important but we can notice :
There is a peak at 50Hz (~1.5dB) in config 1 not present in the other ones
Another one at 250Hz in config 1, also present in config 2&3 1dB smaller
Another one at 500Hz in config 1, also present in config 1&2 0,5dB smaller
The peaks at 5kHz are ~the same.
Peak at 8kHz is smaller by more than 1.5dB in config 2&3
Peaks at 14,4kHz are ~the same.
Peak at 16kH is smaller by more than 3dB in config 2&3.
Preliminary conclusion : Connecting the audio ground to the chassis of a low gain preamp is a very very bad idea !
It is simpler than isolating all the connectors, ground plane etc.. to the box
It is more quiet than isolating, and it is not worse if box is in plastic.
Attachments
I also did measurements on a high gain preamplifier.
Preamp is a discrete MM Phono stage, gain is ~66dB at 1kHz.
Each channel has its own double side pcb with complete ground plane on top.
The normal config is :
power supply (+15 0 -15) ground (5 pin DIN connector) connected to steel box on one side
each ground plane connected to box with 4 metal spacers
output connector ground connected to the box near the power supply connector
input connector ground connected to the box on opposite side
Isolated config :
all ground disconnected from the box
The first plot show the response of the preamp with pseudo random pink noise in "normal" config.
Spectrum are done with input grounded with a shorted RCA plug, average on 100 acquisitions.
Isolated config is on left (red & green).
Normal config is on the right (gray & blue).
Results are clearly better on what I called the normal config where all ground are connected to the box in order to benefit of the steel box shielding and of the low impedance ground path.
As for the low gain preamplifier, connecting all ground to the box is simpler than isolating and work at least as well as the isolated way.
Preamp is a discrete MM Phono stage, gain is ~66dB at 1kHz.
Each channel has its own double side pcb with complete ground plane on top.
The normal config is :
power supply (+15 0 -15) ground (5 pin DIN connector) connected to steel box on one side
each ground plane connected to box with 4 metal spacers
output connector ground connected to the box near the power supply connector
input connector ground connected to the box on opposite side
Isolated config :
all ground disconnected from the box
The first plot show the response of the preamp with pseudo random pink noise in "normal" config.
Spectrum are done with input grounded with a shorted RCA plug, average on 100 acquisitions.
Isolated config is on left (red & green).
Normal config is on the right (gray & blue).
Results are clearly better on what I called the normal config where all ground are connected to the box in order to benefit of the steel box shielding and of the low impedance ground path.
As for the low gain preamplifier, connecting all ground to the box is simpler than isolating and work at least as well as the isolated way.
Attachments
I finally did a third set of measurements with a modified "isolated" configuration closer to "classical" version.
Circuit is a dual ~20dB gain amplifier with opa134 aop, pcb is a double side test board with ground plane on top.
Isolated version : Steel box is connected to power supply ground on one point at power supply connector level. PCB is mounted with isolated spacers and connected to ground on one point. Output RCA ground connectors -on power supply side - and input RCA ground connectors -on opposite side- are isolated from the box.
Connected version : PCB is mounted in the steel box with metallic spacers, then connected to the box, input and output RCA ground connectors are connected to box and to ground plane.
Two measurements have been performed :
1) spectrum with a -100dB input signal level
2) spectrum with shorted input, with a shorted RCA plug.
Power supply is smps +/- 15V. A signal generator, a scope and a solder iron are on at ~50cm of DUT
Results : "classical" config is in red, "all connected" is in blue.
1) -100dB input :
1kHz is at same level ~-80dB
50Hz peak is greater by ~7dB on "classical" (-102, -109dB)
100hz peak is greater on "all connected by 3dB (-113, -110dB)
a 121Hz peak ( -115dB) is present on "classical", not on "all connected" (-123dB)
150Hz peak is greater on "all connected" by 3dB ( -105, -102dB)
The 2 main differences are at 50Hz and 120Hz at the advantage of the "all connected" version.
2) shorted input
50Hz peak is at same level (-130dB)
90Hz peak is ~at same level (-131dB)
120Hz peak is greater on "classical" by 10dB (-123, -133dB)
Once again, the main difference is at the advantage of "all connected" version.
Conclusion : "all connected" ground strategy including ground connection to chaasis works at least as well as all other ground cabling methods.
These results are not in contradiction with electromagnetism laws. The more you make low impedance ground loops in your chassis, the more you'll be immune against variable electromagnetic field.
Circuit is a dual ~20dB gain amplifier with opa134 aop, pcb is a double side test board with ground plane on top.
Isolated version : Steel box is connected to power supply ground on one point at power supply connector level. PCB is mounted with isolated spacers and connected to ground on one point. Output RCA ground connectors -on power supply side - and input RCA ground connectors -on opposite side- are isolated from the box.
Connected version : PCB is mounted in the steel box with metallic spacers, then connected to the box, input and output RCA ground connectors are connected to box and to ground plane.
Two measurements have been performed :
1) spectrum with a -100dB input signal level
2) spectrum with shorted input, with a shorted RCA plug.
Power supply is smps +/- 15V. A signal generator, a scope and a solder iron are on at ~50cm of DUT
Results : "classical" config is in red, "all connected" is in blue.
1) -100dB input :
1kHz is at same level ~-80dB
50Hz peak is greater by ~7dB on "classical" (-102, -109dB)
100hz peak is greater on "all connected by 3dB (-113, -110dB)
a 121Hz peak ( -115dB) is present on "classical", not on "all connected" (-123dB)
150Hz peak is greater on "all connected" by 3dB ( -105, -102dB)
The 2 main differences are at 50Hz and 120Hz at the advantage of the "all connected" version.
2) shorted input
50Hz peak is at same level (-130dB)
90Hz peak is ~at same level (-131dB)
120Hz peak is greater on "classical" by 10dB (-123, -133dB)
Once again, the main difference is at the advantage of "all connected" version.
Conclusion : "all connected" ground strategy including ground connection to chaasis works at least as well as all other ground cabling methods.
These results are not in contradiction with electromagnetism laws. The more you make low impedance ground loops in your chassis, the more you'll be immune against variable electromagnetic field.
Thank you.
Unfortunately, it seems that "isolated connectors" supporters, who all agree that we all have to do measurements to make comparison, think now they have not to do it, and don't want to discuss on it except to say that it is not the right measurements.
The main arguments of the "isolated" way are :
-
Why ?
and : ground loops
And when I ask for "ground loops" explanation, this is what I obtain :
It is curious to note that it seems to be easier to make believe that I do not understand anything of basic physics rather than explain what happens in a ground loop.
Fortunately, we have an explanation here :
As nobody disputes this explanation, I can assume it is accepted by "isolated connectors" supporters.
We can have a look on this explantion and try to see who understand basic physic.
This explanation is wrong.
Magnetic fields don't induce currrent in the loop.
Magnetic fields induce VOLTAGE in the loops.
As open loops don't exist, because there is always a parasitic impedance to close it - as the capacitor between "isolated" connector and chassis- this voltage exist even if you intentionaly close the loop or if you belived to have an open one.
The explanation is wrong and forget half of the law.
The loop voltage creates a current I= U/Z where Z is the loop impedance. This current creates a magnetic field in opposition to the perturbation.
If the loop impedance is very low, it can dramatically reduce the perturabation.
If you intentionnnaly create many low impedance loops, reduce the loop area by clamping the cables to chassis, you can have perturbation under the noise.
This has been done for the measurements I published where results are better than the "isolated" way, but of course this is only the wrong opinion of someone who doesn't understand basic physics.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.