well thats interesting... long story short i have always been plagued with mains hum so i have earthed this and that etc.etc to get rid of it or reduce it to a very very faint hum but still hear master tape hiss when present on recordings. well tonight i decided to try something different and i disconnected all the COM from the amp case and would you credit it there is no hum and i was greeted with an inky black background and a lovely smooth clear sound. there is no com either from the headphone socket or boards connected to the amp case. my amp is a two box with one being the PS and the other being the amp with an umbilical of only DC between the two. the PS case is still PE connected for safety.
Remember that COM GND IN- and OUT- are all the same thing. So none of these are connected to chassis? (easy to check: infinite resistance between metal chassis and the GND pad)
If so, its interesting. There is no reason you can't have the chassis floating if the power supply is external. It is just standard practice to connect it to the circuit common, usually at the input, so that the chassis can act as an electrostatic shield.
If there is less hum with the chassis floating, it probably means that the point where the chassis was previously connected to the circuit common was not the right one.
I don't know what the right one is, formally. GND seems to work well, but you can also try connecting the chassis at the input RCA jack, the output jack, or IN- to see if it makes any difference. I'm afraid it is trial and error at the moment.
If so, its interesting. There is no reason you can't have the chassis floating if the power supply is external. It is just standard practice to connect it to the circuit common, usually at the input, so that the chassis can act as an electrostatic shield.
If there is less hum with the chassis floating, it probably means that the point where the chassis was previously connected to the circuit common was not the right one.
I don't know what the right one is, formally. GND seems to work well, but you can also try connecting the chassis at the input RCA jack, the output jack, or IN- to see if it makes any difference. I'm afraid it is trial and error at the moment.
tried both ways.
first attempt was headphone socket connected to case, this was ok.
second attempt was no connection via headphone socket and both COM from board connected to a single point on case bottom with anodising scraped off, this was a bit better.
third is floating (nothing connected to case) this has resulted in zero hum from what i can tell, it has also gotten rid of a lot of harshness that i never knew was there until now. from top to bottom everything is tight and clean (OPA227) its also tamed a lot of the S&T sibilance i was getting which i put down to the Raspberry Pi with DAC (which is in the same case) due to the naff PS supplying it. RPI R/L out's is connected directly to the R/L Sapphire in's.
the RPi is being run off a cheep 7805 linear regulator just now with a DC wall wart as the main power to the reg board.
one other thing i noticed before i lifted the GND was that if i put my hand near the Sapphire boards (lid off) it would hum like a good one the closer my hand got but if i touched the case at the same time it would disappear. now i just get a very very faint (almost inaudible) hum with my hand almost touching the boards (MM above) but makes no difference if i touch the case as well.
first attempt was headphone socket connected to case, this was ok.
second attempt was no connection via headphone socket and both COM from board connected to a single point on case bottom with anodising scraped off, this was a bit better.
third is floating (nothing connected to case) this has resulted in zero hum from what i can tell, it has also gotten rid of a lot of harshness that i never knew was there until now. from top to bottom everything is tight and clean (OPA227) its also tamed a lot of the S&T sibilance i was getting which i put down to the Raspberry Pi with DAC (which is in the same case) due to the naff PS supplying it. RPI R/L out's is connected directly to the R/L Sapphire in's.
the RPi is being run off a cheep 7805 linear regulator just now with a DC wall wart as the main power to the reg board.
one other thing i noticed before i lifted the GND was that if i put my hand near the Sapphire boards (lid off) it would hum like a good one the closer my hand got but if i touched the case at the same time it would disappear. now i just get a very very faint (almost inaudible) hum with my hand almost touching the boards (MM above) but makes no difference if i touch the case as well.
Connect a 10 ohm resistor that's paralleled with a 0.01 uF (or 0.1uF) cap to chassis ground. Then connect both pwb GNDs to the other side of the resistor/cap combo.
Or do the whole bridge, resistor, cap combo thing. See Rod Elliot's page on Earthing (aka Grounding) or get a House Ground kit from John Boskie's Glass-ware store. I use Boskie's kit because it was cheap, came with mounting parts, thus easy to install.
I'm guessing the hum is due to the RasberryPi being not grounded as the power supply is 2-pin. Your hand is acting as a capacitive coupled antenna.
Or do the whole bridge, resistor, cap combo thing. See Rod Elliot's page on Earthing (aka Grounding) or get a House Ground kit from John Boskie's Glass-ware store. I use Boskie's kit because it was cheap, came with mounting parts, thus easy to install.
I'm guessing the hum is due to the RasberryPi being not grounded as the power supply is 2-pin. Your hand is acting as a capacitive coupled antenna.
But wouldn't loop breakers make sense only if you had rca interconnects between gear boxes?
In Bibio case everything is in a single chassis.
I have similar setup - lifted ground everywhere, and have no problems with jfet-based input opamps (input impedance ~ 10**13Ohm). But when replaced with lt1357 I got audible hum (input impedance ~ 10**6Ohm).
Like I stated, I'm guessing here. Single case or not, there needs to be a single ground point for incoming power. And, for some, they like to keep the audio 10 ohms off power common and I'm one of those.
Richard will correct me if I get the following wrong! The Sapphire board floats off the transformer/diode path. The input ground, output ground, and power supply ground all star connect to the GND point. Taking that point to the chassis via 10 ohms with an RF bypass cap, should be the standard path.
Whatever connects to the Sapphire's input should have it's ground taken to the chassis at the same point as the Sapphire, especially if it's in the same chassis. Whether it's done via a 10 Ohm lift is something to be experimented with.
Richard will correct me if I get the following wrong! The Sapphire board floats off the transformer/diode path. The input ground, output ground, and power supply ground all star connect to the GND point. Taking that point to the chassis via 10 ohms with an RF bypass cap, should be the standard path.
Whatever connects to the Sapphire's input should have it's ground taken to the chassis at the same point as the Sapphire, especially if it's in the same chassis. Whether it's done via a 10 Ohm lift is something to be experimented with.
let me make it a little more complicated... and maybe one of you bright sparks can understand why.
GND lifted:
turn on headphone amp.. hummmmmmmmmm
turn on RPi.. hummmmmmmmm
start playing music.. inky black silence
GND connected to case:
turn on headphone amp.. very very slight hummm
turn on RPI.. very very slight hummm
start playing music.. very slight hum continues but a lot more faint.
it seams that once the RPi engages the DAC in handshaking that the audio path is acting as a GND. the silence continues all the time as long as both units are not turned off. changing albums, pausing configuring the RPi, reboot has no effect on hum till both units are turned off then started from cold.
could this also be the case for people using the RPi as a stand-alone plugged into separate amp who say that there is a lot of noise and HF hash.
GND lifted:
turn on headphone amp.. hummmmmmmmmm
turn on RPi.. hummmmmmmmm
start playing music.. inky black silence
GND connected to case:
turn on headphone amp.. very very slight hummm
turn on RPI.. very very slight hummm
start playing music.. very slight hum continues but a lot more faint.
it seams that once the RPi engages the DAC in handshaking that the audio path is acting as a GND. the silence continues all the time as long as both units are not turned off. changing albums, pausing configuring the RPi, reboot has no effect on hum till both units are turned off then started from cold.
could this also be the case for people using the RPi as a stand-alone plugged into separate amp who say that there is a lot of noise and HF hash.
It could be that the DAC is muted on the digital side until music plays. The analog side would still be active. Or it could be the DAC mutes by shunting the analog to ground.
On the DAC side, there are two (actually 3) grounds. There is the digital ground (and the power supply ground) and there is the analog ground. The analog ground should attach to the digital ground on one place only. The data sheet for the DAC and/or DAC board should spell that out somewhere.
If it were me, I'd try removing the ground input on the Sapphire from the RCA jack And have the RPi DACs out ground tied to the power supply common point or chassis ground.
As to the hash, that's most likely due to a not so great digital filter. It could be the DAC implementation as well - non-stable oscillators, no filter at the DAC outputs (or ineffective filter).
Typically, one would want something like -
I2C -> DAC w/temp stable osc -> I to V convert -> 22 KHz analog filter -> Buffer
And with no intention of trashing anyone's hardware, I looked at the HiFiBerry DAC+ Pro and I don't see enough circuitry on that board to implement a proper filter - the coil alone would cost just under $4 US.
On the DAC side, there are two (actually 3) grounds. There is the digital ground (and the power supply ground) and there is the analog ground. The analog ground should attach to the digital ground on one place only. The data sheet for the DAC and/or DAC board should spell that out somewhere.
If it were me, I'd try removing the ground input on the Sapphire from the RCA jack And have the RPi DACs out ground tied to the power supply common point or chassis ground.
As to the hash, that's most likely due to a not so great digital filter. It could be the DAC implementation as well - non-stable oscillators, no filter at the DAC outputs (or ineffective filter).
Typically, one would want something like -
I2C -> DAC w/temp stable osc -> I to V convert -> 22 KHz analog filter -> Buffer
And with no intention of trashing anyone's hardware, I looked at the HiFiBerry DAC+ Pro and I don't see enough circuitry on that board to implement a proper filter - the coil alone would cost just under $4 US.
i dont have any RCA sockets (RPi DAC out and Sapphire in are wired direct) but i think i know what you mean. remove cold from the Sapphire in and send the cold from the RPI DAC out to the case GND and use the Sapphire COM as the cold in a sort of star GND e.g. connect Sapphire COM and RPi DAC cold to case at the same point?
FWIW - If JG's filter on the front of his buffer is implemented on the Sapphire - adding it in front of C1, the hash/noise from the DAC should go away. Note if the opamp has less than 1uA of input bias current, C1, R1, and R2 could removed if the RPi has no DC on the output - test to be sure!!!
The filter is -
In -> 240R -> 4n7 film to GND -> 18 mH -> 18K9 to GND -> 470R -> 150pF (or 100) to GND -> Op amp +IN
The coil is a Fastron 07MFG-183J-50, Mouser PN 434-02-183J.
The filter is -
In -> 240R -> 4n7 film to GND -> 18 mH -> 18K9 to GND -> 470R -> 150pF (or 100) to GND -> Op amp +IN
The coil is a Fastron 07MFG-183J-50, Mouser PN 434-02-183J.
Last edited:
My understanding is that the main task here is to provide equipotential plane to every board (which is somewhat naive concept anyway), conveniently labeled GND/COMM/whatever, and that every conductor has some impedance. Then any current from point a to point b develops voltage difference. To reduce that effect you can: 1) reduce current 2) reduce conductor impedance.
1) can be achieved using opamp with very high input impedance (this would explain jfet vs bjt anomaly in my case).
I presume your advice relates to 2), that somehow using star ground scheme with direct 10R connection to chassis would do the trick. but chassis is/must/ought be connected to PE. Using PE for return currents is not elegant at best. Besides increases overall inductance, which is not a good idea, I guess, when you have Rpi board with its switching regulators nearby. so this concept is unclear to me.
That's a typical trick done by delta-sigma DAC ICs.
PCM5122 has an internal I/V conversion circuit. And according to the datasheet all you need to do is to put external LPFs on its outputs.
Last edited:
output DC from RPi to input of Sapphire boards is 0.1ma but my DMM is seriously shizz.
output offset at headphone socket with opa134 is around 10mv, OPA227 is around 1mv OPA228 is around 0.2mv with a gain of 20?
all i do is watch youtube videos (EEVblogs are great) to try and get an idea of electronics especially opamps but it goes right over my head. it will sink in eventually but its going to take a while... lol
i'll read the JG thread and see if i can make sense of it.
output offset at headphone socket with opa134 is around 10mv, OPA227 is around 1mv OPA228 is around 0.2mv with a gain of 20?
all i do is watch youtube videos (EEVblogs are great) to try and get an idea of electronics especially opamps but it goes right over my head. it will sink in eventually but its going to take a while... lol
i'll read the JG thread and see if i can make sense of it.
I don't want to interrupt as it looks like you guys are making some great progress already.
Let me just say that the key point to me would be what the DAC does electrically when it "plays music", and that the particulars here are probably more to do with the RPi than the Sapphire per se., but that's not to say the Sapphire is entirely blameless.
Just to state the obvious: if you have a DAC and the Sapphire in a shared chassis, the chassis should only attach to the circuit common at a single point. If the connection is made at the Sapphire GND pads, the DAC should not have any direct connection to the chassis, or vice versa.
Lance, could we get a photo of your chassis internals, just for reference?
Let me just say that the key point to me would be what the DAC does electrically when it "plays music", and that the particulars here are probably more to do with the RPi than the Sapphire per se., but that's not to say the Sapphire is entirely blameless.
Just to state the obvious: if you have a DAC and the Sapphire in a shared chassis, the chassis should only attach to the circuit common at a single point. If the connection is made at the Sapphire GND pads, the DAC should not have any direct connection to the chassis, or vice versa.
Lance, could we get a photo of your chassis internals, just for reference?
Last edited:
i know its a mess but here goes.. new case will be getting done sometime in the new year once i have got all the bugs sorted. ignore the stepped attenuator its not connected
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
