Balanced CFA IPS -
2007? No. I've only been on this site for about 2 years. Same idea though.
How to integrate the input of a CFA for balanced input with good cmr ?
[BTW -- This bal ips topology - I first heard of the idea was from John Curl decades before 2007. But who is counting?]
Its the concept I am after here for a balanced CFA input -- what ideas are there on that?
Thx-RNMarsh
2007? No. I've only been on this site for about 2 years. Same idea though.
How to integrate the input of a CFA for balanced input with good cmr ?
[BTW -- This bal ips topology - I first heard of the idea was from John Curl decades before 2007. But who is counting?]
Its the concept I am after here for a balanced CFA input -- what ideas are there on that?
Thx-RNMarsh
Last edited:
Richard,
Would a balance ips section for a CFA require two feedback paths, or would one central feedback path supply both mirrored halves of the circuit?
Would a balance ips section for a CFA require two feedback paths, or would one central feedback path supply both mirrored halves of the circuit?
Could be either way.... but simpler would be nice.... maybe easier to make stable, also.
Besides higher power ... sooner or later, one would be asked for a balanced input feature.
[And not a kludged attachment.]
-RNM
Besides higher power ... sooner or later, one would be asked for a balanced input feature.
[And not a kludged attachment.]
-RNM
Last edited:
Nothing wrong with balanced inputs at all, and if the topology allows for it then do it. But, I don't think they are strictly necessary for good sound. In a dirty electrical/signal environment, then of course you cannot do without them.
It's quite easy to design a balanced CFA - this would essentially be the H bridge IPS config driving two OPS taken from the opposite level shifter outputs. Two separate feedback paths.
But, this is a lot of extra complexity IMV . . .
But, this is a lot of extra complexity IMV . . .
I feel the same way -- especially with the short distances between home gear - usually. But then, there are other situations and apps in the world for amps. Like CI and studio etal. And, VFA have them... would not want to be forced into VFA just for lack of this feature.
Just for 'do-diligence' would you pls put up SIM data and schematic of your description in # 4789. So, if anyone wants to have bal CFA input, it can be referred to as an example to use.
If it is a time issue for you, then the priority might be for the higher power amp, first. Because it is not related to the issue of bal/unbal but of clipping and the real need for adequate power output. Or, others might pick up the Bal ball and run with it.
Thx-RNMarsh
Last edited:
That's a feature of standard symmetrical amplifiers like the ones OS, Damir and me have been posting up. You don't need to go to the full bridged balanced design to get that. The advantage of bridged amps is the lower ( or zero) power signal ground - all the signal current flows between the supply rails and the load. NP's super symmetrical exploits this approach.
It's possible, if You use balanced feedback as well. It works great, with no problem. However it has one backwards:The balanced amplifiers sound worse if You drive them from single ended source. Of course this is my experience only..
Sajti
Here is the concept 'fx-Amp' mentioned earlier. There is a transient file, a slew rate file and a loop gain file.
The bjt models file extension will need to be changed to .bjt if you use it.
Here is a quick summary of the simmed results:-
11ppm at 20 k at ~420W into 8 Ohms; ~3ppm with AFEC
Into 2 Ohms with AFEC the distortion is ~3ppm as well
Slew Rate is ~150V/us with filters in situ
The OPS is biased up for c. 75mA per pair. This will give around 6.5W class A. You can bias up higher, but will need decent heatsinking.
-3 dB bandwidth with comp as shown is ~1 MHz - probably too wide for a practical amp, but easily reduced
PSRR (positive side only measured) is -80dB at 1 Hz, 120 dB at 1 kHz and better than that at 10 kHz. These results will not be attainable in practice (capacitor L and esr), but nevertheless will be as good as VFA.
The bjt models file extension will need to be changed to .bjt if you use it.
Here is a quick summary of the simmed results:-
11ppm at 20 k at ~420W into 8 Ohms; ~3ppm with AFEC
Into 2 Ohms with AFEC the distortion is ~3ppm as well
Slew Rate is ~150V/us with filters in situ
The OPS is biased up for c. 75mA per pair. This will give around 6.5W class A. You can bias up higher, but will need decent heatsinking.
-3 dB bandwidth with comp as shown is ~1 MHz - probably too wide for a practical amp, but easily reduced
PSRR (positive side only measured) is -80dB at 1 Hz, 120 dB at 1 kHz and better than that at 10 kHz. These results will not be attainable in practice (capacitor L and esr), but nevertheless will be as good as VFA.
Last edited:
Something about AFEC bother me. It is very similar to ODNF(Only Distortion Negative feedback). Main difference is that AFEC decrease distortion of the GNFB amp, and ODNF is to be applied with on NGNFB amp. Both have one drawback and that was the reason I dropped ODNF. If you want to get the best of both type you need to do fine ratio adjustment of the output level input in the OPA. As this here is used in CFA the gain is not just 1+Rf/Rg (in your AFEC explanation, 1+R2/R3) as there is low impedance - input to take in the calculation. Only way, as I see it, is to use distortion analyzer and fine adjust AFEC or ODNF.
Tell me if I got it all wrong.
BR Damir
ps. in my ODNF simulation if that ratio was not exact you got higher distortion then with no ODNF. I think, with AFEC you could lose all GNFG distortion decrease.
Last edited:
Hi,
I will do it, but now I'm in the office, and my schemas are at home on my PC.
Sajti
Damir,
If you have mismatches in the gain between the AFEC stage and the main amp you will lose the distortion cancellation. When I tried this, IIRC, I did not see increases in distortion with small mismatches - eg like 1-2%, just a reduction in distortion cancellation. If the mismatch is too big, you will see changes in gain - either higher or lower.
You are correct that the feedback networks have to be accurately matched to null the distortion. One way to do this is to connect a pot in parallel with the feedback network to the AFEC and then trim it for a null. You could use a computer sound card for this and trim for a null in the harmonics.
Let me add here that for AFEC to work, I think a lot of care in layout, opamp choice and
nulling is required. I have not built an AFEC amp yet - but I will start on this soon.
If you have mismatches in the gain between the AFEC stage and the main amp you will lose the distortion cancellation. When I tried this, IIRC, I did not see increases in distortion with small mismatches - eg like 1-2%, just a reduction in distortion cancellation. If the mismatch is too big, you will see changes in gain - either higher or lower.
You are correct that the feedback networks have to be accurately matched to null the distortion. One way to do this is to connect a pot in parallel with the feedback network to the AFEC and then trim it for a null. You could use a computer sound card for this and trim for a null in the harmonics.
Let me add here that for AFEC to work, I think a lot of care in layout, opamp choice and
nulling is required. I have not built an AFEC amp yet - but I will start on this soon.
Last edited:
Doesn't the slew rate of an amp double when it is strapped/bridged/balanced? Why would a push pull balanced config cancel even harmonics? We want a config that cancels 3,5,7th. Thanks, GRH
Bonsai,
Good lack with AFEC, I think that AFEC is less sensitive to matching then ODNF.
Even in simulation it was not easy to get good cancelation in ODNF, but ODNF is doing all cancelation, no negative feedback is used.