Help! Parallel 2x 3875 per channel design - please advise.

[loong]

I might be able to find someone with hi-res multiter, but 0.1% Res is very difficult & too expensive.

As stated in the NS paper that '.. with 1% would give good result, but recommend 0.1%", so I don't think it will be serious/big concerns if I use 0.1% or 1% as long as it doesn't impact (noticable) the sound quality.

BTW,

Just confused if this one will work or not

[bigparsnip]

If you do use the 1% resistors, I would sujest using larger output resistors for each chip (0.33, or 0.47 perhaps), as this will decrease the likelyhood of the two chips drwaing or suppling current to each other if there is a voltage imbalance at their outputs.

[azira]

loong:
I would not recommend using that circuit you posted. You'll likely have a DC offset at each output and an imbalance between the outputs. When the opamps fight, you lose.
The app note uses a unity gain opamp buffer and a 1k resistor before all the inputs to the 3875s. The reason for this is for the power opamps to have equal base impedances at their inputs. In the schematic you are showing, each opamp has 23k on the + input and only 680 on the - input. If you're intent on not using the buffer then the only way to keep the input impedance high and match +/- input impedances is to make the feedback resistors x10 (so 220k and 6.8k) and make your +input resistors 6.8k and remove the 22k input to gnd or similar values. Those seem like pretty high values though..

[bigparsnip]

Originally posted by azira
loong:
The app note uses a unity gain opamp buffer and a 1k resistor before all the inputs to the 3875s. The reason for this is for the power opamps to have equal base impedances at their inputs. In the schematic you are showing, each opamp has 23k on the + input and only 680 on the - input. If you're intent on not using the buffer then the only way to keep the input impedance high and match +/- input impedances is to make the feedback resistors x10 (so 220k and 6.8k) and make your +input resistors 6.8k and remove the 22k input to gnd or similar values. Those seem like pretty high values though..

You are almost right here, but if you look again you will see that teh lower feedback resistor is capacitively coupled to ground, so the bias curent will flow through the 22k feedback resistor from the output instead, so if you match the input impedance on the + input to this you will reduce the offset in a more conventional fassion.

[azira]

Originally posted by bigparsnip



You are almost right here, but if you look again you will see that teh lower feedback resistor is capacitively coupled to ground, so the bias curent will flow through the 22k feedback resistor from the output instead, so if you match the input impedance on the + input to this you will reduce the offset in a more conventional fassion.

Hmmm... I think you're right. But he'll still need to change the Fb resistor to 23k or change the input resistor to 21k to match it up.

[Rudy]

as long as the resistors don't need to be the exact value but need to be closed matched, why not just buy 2 resistors on tape, some stores sales them by 10 or so but they or still taped, this way you get an accuratie off way better then 1% but stil have the price off the 1% resistors. Witch aint mutch compared to 0.1% and 0.01% resistors

[azira]

Originally posted by Rudy
as long as the resistors don't need to be the exact value but need to be closed matched, why not just buy 2 resistors on tape, some stores sales them by 10 or so but they or still taped, this way you get an accuratie off way better then 1% but stil have the price off the 1% resistors. Witch aint mutch compared to 0.1% and 0.01% resistors

Because unlike transistors, resistors are tested and "binned" before they are cut into tapes. They only produce one line of resistors and then test them to see what their actual value is. The ones that ended up getting made within .1% of the target value are put into the same bin and the rest that are 1% of the target value are put together, then 2% then 5%. So what you actually end up getting when you get 5% resistors is a set of resistors that were more than +/- 2% away from the actual value. Bimodally distributed. The only set of devices that should have a normal distribution is the highest tolerance bin.

[Rudy]

maybe so, but have you actually tried it, i think not ...

[Alcaid]

The output resistors should also be matched to within 1% or better.

[SimontY]

Originally posted by Rudy
maybe so, but have you actually tried it, i think not ...

I measured some resistors on a tape, infact, many values on many tapes, and whilst the ones next to each other were often the same as measured by my cheap DMM, there is always a point (maybe after 2, 5, or 8 in a row) where they will measure different. So if my meter measures consistently, it shows just taking them off a tape does not gurantee they'll be the same

(these were all 1% tol. metal film 1/4w)