I realize the OP is looking for a simple solution. Rules out the instrumentation input (which has excellent CMRR, especially when you operate it with gain).
I think the next best solution is to use one opamp and balance the impedances at the source. You can measure the output impedance and add the same resistance for the "cold" connection. Use microphone cable of course. As shown it has a gain of -1dB and a cut frequency of 1.5MHz. (This universal input can be adapted for any type of source.)
I think the next best solution is to use one opamp and balance the impedances at the source. You can measure the output impedance and add the same resistance for the "cold" connection. Use microphone cable of course. As shown it has a gain of -1dB and a cut frequency of 1.5MHz. (This universal input can be adapted for any type of source.)
Surely the way to look at this is to short the + and - input legs together and to then apply a noise stimulus between the junction and ground. If you do this with a standard single opamp balanced input circuit and 0.1% resistors the CMRR is >50 dB. Quite a respectable number for little cost.
This one is "less universal", it does not present the same impedances to the source, but it is a lot better in
CMRR with a balanced impedance source. Below is the simulated CMRR.
CMRR with a balanced impedance source. Below is the simulated CMRR.
Reply to post #22: R3 should be equal to R4 and R6 to R5. You can immediately see why for frequencies where C1 and C2 have a negligible impact: just replace R3 and the resistors to the left of it with one resistor with their series value and do the same with R4 and the resistors to the left of it.
Hi, as pictured, circuit will still need very careful resistor matching which is normal, but in addition it requires very low current noise opamp (eg jfet input device).
What you said is correct, source needs balanced impedances which is actually super easy to make; just terminate source output (assumed it is not tube and it's output Z is close to 0) and return wire (this mostly being analog ground) with matched resistors in range 50 to 200 ohm.
Usual differential receivers (ssm2141 THAT124x) can be used at other end then and they already have precisely matched resistors on board... Hum gone 😊
What you said is correct, source needs balanced impedances which is actually super easy to make; just terminate source output (assumed it is not tube and it's output Z is close to 0) and return wire (this mostly being analog ground) with matched resistors in range 50 to 200 ohm.
Usual differential receivers (ssm2141 THAT124x) can be used at other end then and they already have precisely matched resistors on board... Hum gone 😊
That sim assumes perfectly-matched resistors. Try the simulation with a 1% mismatch in the 15K resistors and see what you get. R1 & R2 also will affect the CMRR but to a lesser degree (percentagewise).
If you want to take that route, an instrumentation amplifier would perform better. Finding a resistor array with closely-matched resistors would work, too, but I suspect the combination would not as cost-effective. I found some .1% matched 4-resistor arrays on DigiKey in the 10K range. They will cost you somewhere between $4 and $10 apiece. For comparison, you can get a single instrumentation amplifier in the TI INAxxx product line for something like $2.77. That's for models with decent bandwidth.
If you want to take that route, an instrumentation amplifier would perform better. Finding a resistor array with closely-matched resistors would work, too, but I suspect the combination would not as cost-effective. I found some .1% matched 4-resistor arrays on DigiKey in the 10K range. They will cost you somewhere between $4 and $10 apiece. For comparison, you can get a single instrumentation amplifier in the TI INAxxx product line for something like $2.77. That's for models with decent bandwidth.
If you need MATACHED resistors use Dip or Sip resistor packs,they are usally 0.01%. The absolute omic value maybe 1-2%.
Exactly to the point, nobody will match resistors as they are matched at such chips, this just works.
Haven't done that yet, but mismatching one resistor (1% off) will degrade CMRR to -40dB.
Yes, you need good matching or you should select. That said, modern metal film resistors on a reel (SMD) are remarkably close in tolerance. I used this circuit on the front end of the ax-Amplifier and measured 45 dB without matching using 1% tolerance resistors. For a few pennies more, you get 0.1% tolerance and 54dB worst case. If you couple 45 dB CMRR and have good layout and hum mitigation in the amp overall, you are ‘off to the races’ in other words you will have a very good measured result.
With corrected C2, the CMRR of the gain-of-20 version of post #1 is much higher at a given percentage of resistor mismatch than that of a unity-gain version. You can see that by taking the limit for closed-loop gain approaching infinity.
I am dealing with RCA interface. I have no interest in implementing a full fledged balanced interface. Even an extra opamp is too much hassle for me. Also the impedance matching is not guaranteed with a volume pot at the front.
Thus, the solution in Post #11 is my preferred solution.
I modified my Quad 405 yesterday. It is rather simple as it has the separated signal ground already. The only thing I have to do is to make sure the signal ground is insulated from the chassis at the RCA socket. It basically resolved the ground loop that is caused by RCA cables.
I also did my Hafler 9290. This was a little bit tricky. I had to cut a PCB trace to create a separated signal ground, and bridge them with a 10 Ohm resistor. The RCA jacks are insulated from the chassis already. I don’t have to do anything about jacks.
These 2 amps have 3-prong plug. They can introduce ground loop very easily with other audio components. The mod is necessity for me.
Thus, the solution in Post #11 is my preferred solution.
I modified my Quad 405 yesterday. It is rather simple as it has the separated signal ground already. The only thing I have to do is to make sure the signal ground is insulated from the chassis at the RCA socket. It basically resolved the ground loop that is caused by RCA cables.
I also did my Hafler 9290. This was a little bit tricky. I had to cut a PCB trace to create a separated signal ground, and bridge them with a 10 Ohm resistor. The RCA jacks are insulated from the chassis already. I don’t have to do anything about jacks.
These 2 amps have 3-prong plug. They can introduce ground loop very easily with other audio components. The mod is necessity for me.
Originally, QUAD 405s have DIN rather than RCA sockets, I wonder how they connected pin 2 of those. It's an inverting amplifier, so none of the schematics in this thread are applicable.
Yep, from the DIN socket, the signal ground was shorted to the chassis at the socket from the factory. It was through a jumper wire at the socket. I just simply cut that wire. The RCA jacks were added at some point.
The idea of Post #11 is to use the “remote ground” for everything except for the speaker. You will find the gain of the common mode is always 1, even you have multiple stages like Quad 405. If you have 26dB gain in the chain, you get 26dB gain from differential component, but only 1 for common mode component. You get 26dB better noise ratio.
Last edited:
its a simple ground lift.
it has been covered many times in this forum.
did you miss it since 2009?
Hi,
Differential input can be done with RCA socket too, but not so elegant.
Not perfectly matched resistors will still give some CMR, grounded return signal will get zero CMR, but I understand, sometimes it is ok to built classic single ended input but care has to be taken a lot.
Here is what I do with them, as example last year I rebuilt 6 channel - dual power supply (cant get more complicated than this) from horrible hum to total silence. Once I will make small thread about it as many mechanical and wiring issues are shown there, still no time from my side.
This my raise some eyebrows as it is not how most people do it, sketch diagram:
All RCA connectors grouped together (not at each end of amp as some do) and its shield (ground) connected to the chassis as soon and as solid as possible.
Power input ground connected to the chassis also as soon as possible, at same place where RCA-s are grounded (if possible).
This will allow dirty ground signals to be shortened asap, without leading their ways into the electronics.
Input cable (coming from RCA) goes to amplifier; signal wire to + input, shield to SGND.
SGND is small point where Rin and Rg of feedback network connect . (this is simplified schematic without feedback condenser and other things)
Now we need to connect this SGND to system PS GND somehow, I take it from speaker negative terminal as this is where we want to measure feedback voltage, this reduces THD a bit too.
Speaker negative terminal is finally connected to properly built PS central ground. (by properly built I mean trafo- rectifier- capacitors and PS ground taken at the end of this chain)
Safety ground is connected to chassis directly, and to PS ground via input cable shield (I have 6 inputs , so plenty of cable surface for safety, cable with thicker shield can be used if anyone is concerned for safety) and than speaker - wire (this one is thick). There is no loop and chassis is extended shield.
Feedback is connected directly to speaker, where it belongs.
No ground lifters, no lifting resistors, ground insulators, and no hum.
Edit: forgot to mention; local PS capacitors at PCB: their ground is lead directly back to PS star ground, no connection to SGND.
Here is pic of input grounds grouped and tight together with safety ground to chassis:
Differential input can be done with RCA socket too, but not so elegant.
Not perfectly matched resistors will still give some CMR, grounded return signal will get zero CMR, but I understand, sometimes it is ok to built classic single ended input but care has to be taken a lot.
Here is what I do with them, as example last year I rebuilt 6 channel - dual power supply (cant get more complicated than this) from horrible hum to total silence. Once I will make small thread about it as many mechanical and wiring issues are shown there, still no time from my side.
This my raise some eyebrows as it is not how most people do it, sketch diagram:
All RCA connectors grouped together (not at each end of amp as some do) and its shield (ground) connected to the chassis as soon and as solid as possible.
Power input ground connected to the chassis also as soon as possible, at same place where RCA-s are grounded (if possible).
This will allow dirty ground signals to be shortened asap, without leading their ways into the electronics.
Input cable (coming from RCA) goes to amplifier; signal wire to + input, shield to SGND.
SGND is small point where Rin and Rg of feedback network connect . (this is simplified schematic without feedback condenser and other things)
Now we need to connect this SGND to system PS GND somehow, I take it from speaker negative terminal as this is where we want to measure feedback voltage, this reduces THD a bit too.
Speaker negative terminal is finally connected to properly built PS central ground. (by properly built I mean trafo- rectifier- capacitors and PS ground taken at the end of this chain)
Safety ground is connected to chassis directly, and to PS ground via input cable shield (I have 6 inputs , so plenty of cable surface for safety, cable with thicker shield can be used if anyone is concerned for safety) and than speaker - wire (this one is thick). There is no loop and chassis is extended shield.
Feedback is connected directly to speaker, where it belongs.
No ground lifters, no lifting resistors, ground insulators, and no hum.
Edit: forgot to mention; local PS capacitors at PCB: their ground is lead directly back to PS star ground, no connection to SGND.
Here is pic of input grounds grouped and tight together with safety ground to chassis: