New chipamp circuit

[vossie]

Below is a circuit idea I've been playing around with on the simulator. Tuesday night I build it with LM3875T chipamps and it does not sound bad for a first try.

I used OPA541/BB opamps for simulation purposes.

The circuit works well with single or balanced inputs. If I inject a "noise" signal at either one of the output terminals it gets duplicated in similar phase at the other output. Matching similar component values are very important to maximize CMR.

This idea started from Nelson's SUSY - I wanted to reproduce noise and distortion from one side in common phase in the other to cancel over the load. Although this is not SUSY by a far cry, it does seem to work to a degree when the impedance seen at it's inputs are high (with these component values the simulation indicates above 1k). Amplification seems to be OK as the resistance at the input affects only the symmetrical re-creation of the error on the other side.

After a period of excitement I discovered that similar circuits are being used in existing line driver circuits like the SSM2142, the MAX435 and others. Rod Elliott also has a line driver project on these which is probably a good read. I have not seen these circuits used for power amplifiers (probably because heat is involved which could cause havoc with component tolerances). Mine is not exactly the same (a bit simpler).

Because of the loop gain sensitivity I compensated by making R5 & R7 slightly smaller than R10 & R12. This allows for larger tolerances in component values and chipamp DC offset. Making R5 & R7 equal to R10 & R12 would almost surely pull one side to a supply rail and toast the speakers. I see similar circuits compensate for this by placing resistors in the feedback loop and at the outputs.

Speaking of toasting speakers - when driving this circuit single-ended it is not a good idea to reference the non-driven input directly to the circuit GND. The chance is good that the input pin that is driven is floating (either not plugged in or driven through a cap). This leaves a voltage over the load... I referenced both inputs to circuit GND through 10k resistors and simply drive it with the CD player inputs floating. This way the circuit sees it as a differential source.

One should be able to couple two discrete amplifiers together in this way - as long as the coupling resistors are matched well and the system is compensated for the small mismatches...

[vossie]

Here is the same circuit with more symmetrical correct values but more resistors. this performs slightly better with square waves because of the theoretical (simulated) precisly matched parts...

(I've omitted the RC filter network on both outputs.)

This is nifty...

[wensan]

This two circuits below are totally symmetrical.
They are equivalent.


But such circuits have serious common-mode DC drift problem.
Because their common-mode equivalents will like this:

or this:

If the OP's common-mode gain is positive will cause positive feedback loop.
I have had the experience of experimenting on such circuit,
But the circuit that I made become a flip-flop.
The two outputs drift to Vcc or Vee together.
And input signal can trigger the outputs change status.
But I can not simulate such situation by PSPICE.


This is a method to restrain the positive feedback loop.

It deduce a totally balanced circuit with very high CMRR like this:

Because it has very high CMRR.
So it can become a singal-end-to-differential-converter by short one input to GND.

[wensan]

This is a balanced amp with DC servo to eliminate DC ingredients of input signal that I made.
http://home.pchome.com.tw/shop/cherry1chiu/BAL604.jpg

This is the balanced circuit with OPA604.
http://home.pchome.com.tw/shop/cherr...OPA604PC_2.jpg
This is the balanced outputs with 1KHz square wave singal-end input.
http://home.pchome.com.tw/shop/cherr...OPA604S3_2.jpg
http://home.pchome.com.tw/shop/cherr...OPA604S4_2.jpg

This is the balanced circuit with NE5534.
http://home.pchome.com.tw/shop/cherr.../NE5534P_2.jpg
This is the balanced outputs with 1KHz square wave singal-end input.
http://home.pchome.com.tw/shop/cherr...NE5534P1_2.jpg
http://home.pchome.com.tw/shop/cherr...NE5534P2_2.jpg

This is the balanced circuit with a OP module designed by myself.
http://home.pchome.com.tw/shop/cherry1chiu/OPM_2.jpg
This is the balanced outputs with 1KHz square wave singal-end input.
http://home.pchome.com.tw/shop/cherry1chiu/OPM1_2.jpg
http://home.pchome.com.tw/shop/cherry1chiu/OPM2_2.jpg

If you can read Chinese, you may refer to this web page.

[wensan]

Such modification will reduce output "Signal balance ratio(SBR)".
Because the original circuit, Va = Vd, Vb = Vc.
But your modification will let Va >< Vd, Vb >< Vc.
It will cause two OP's differential inputs not equal.

[wensan]

Such circuit structure has been put in use for long long ago.

[realmadrid]

wensam:
Can you please send the schematic of your balanced circuit with OPA604?
Is that valid for the OPA627 too?

Tia
Juan

[wensan]

Originally posted by realmadrid
wensam:
Can you please send the schematic of your balanced circuit with OPA604?
Is that valid for the OPA627 too?

Tia
Juan

Yes, it is valid for OPA627.

[realmadrid]

I mean the R values of this:
http://home.pchome.com.tw/shop/cherry1chiu/BALtc.jpg

Thanks in advance
Juan

[wensan]

Originally posted by realmadrid
I mean the R values of this:
http://home.pchome.com.tw/shop/cherry1chiu/BALtc.jpg

Thanks in advance
Juan

The R values depend on circuit's gain and input impedance by your design.

You can experiment on it.