How to simplify this interstaging circuit?

[HiFiNutNut]

I want to simplify this circuit as much as possible. i.e. less caps and less opamps in the signal path.





1. Because I can't measure any DC, C2 and R5 in the power amp are removed. The imput impedance of the power amp is 10k and can't be changed easily.

2. I am not running long interconnect cables so normally Stage 2 buffer / isolation is unnecessary and can be removed. However, I am worried that P1 in Stage 3 may alter the filter response of Stage 1 because P1 effectively adds a resistor in parallel with R1 and R3 to ground. Will this be a problem? With stage 2 put in this is definitely not a problem but of course, there is an additional opamp in the signal path.

3. This is the low pass filter part of the XO/EQ and I want some adjustable attenuation to the woofer hence I had Stage 3. Once C2 and R5 are ditched I guess U3 will be necessary otherwise with the low input impedance of the power amplifier the output impedance of the trim pot will effectively shunt the input of the power amp. The power amp has a differential input and the impedance of the input and feedback should be matched otherwise I guess it would have DC offset and all other different sorts of problems.

I just found the circuit too clumpsy to have so many opamps and if you have any better ideas of simplifying the circuit they will be greatly appreciated.

Regards,
Bill

[sam9]

Look up the datasheet for BUF634. Note the input impedance and the output currenct capability. Note the noise and distortion figures as well as generally not requiring bypass caps.

Maybe this is a bad idea, but consider replacing U2 with BUF634, then dropping both U3 and R3.

[HiFiNutNut]

Thanks Sam. The opamps I use are all OPA2134a. I will look at BUF634.

Any other suggestions?

[adrome00]

You should be able to delete U2, U3, and P1 by putting a pot used as variable resistor in series with R3. It won't attenuate the signal to zero or give you linear volume changes, but will work if you just need some control.

The value would depend on the actual value of R3, but crudely speaking, taking ten times the largest value (of either 10k or R3) will give you up to 20 dB of volume attenuation.

A more extreme idea, depending on the impedance of the source signal, would involve elimination of stage 1, stage 2, U3, and possibly R3 by placing an appropriate capacitance in parallel with the feedback resistor in the power section. I'm not really suggesting this without knowing more detail, but it may well be possible.

Good luck!

[sajti]

Just simply remove U2 and U3. Put the input of the P1 directly to the output of U1, and put the output of P1 to the input of U4. There will be not problem, due the output impedance of the U1 is low, and it can drive P1. The input impedance of U4 is high enough for P1, especially if the highlighted parts are removed, and R4 increased to 100k.
The original circuit is overkill...

sajti

[ilimzn]

Sajti, thanks again for being the voice of reason. i was just looking at the schematic and thinking "they must be kidding, right?". It pays to look up the definition of the term 'operational amplifier'...

[HiFiNutNut]

Hi,

I have now eliminated the original stage 2:



Is there still any way to simplify it (without altering the power amp)?

I guess the buffer is still needed because the next stage (power amplifier) has an input impedance of only 10k. To delete the buffer opamp I guess the power amp input impedance must be changed to something like 100k but that requires changes to the feedback network. The power amp sounds fantastic so I don't want to change it.

I'd love to get rid of that opamp but am now running out of tricks.

Regards,
Bill

[HiFiNutNut]

adrome00,

Thank you for your post. Your ideas have been much appreciated and carefully studied. Here is your suggested schematic:



I have one worry here. Let's assume we want an attenuation of 2 or 6dB so we adjust R3 to be 10k. From the power amp's perspective the input impedance would become R3 * R5 / (R3 + R5) = 5k.

Now because the amplifier has a designed 10k input impedance which matches to its global negative feedback impedance to the differential input pair, any alternations to the input impedance must be matched to the impedance of the feedback network, or DC offset or other problems can occur. There are also other risks involved. Therefore, I would rather not to alter the circuit in the power amp except that I took out the input cap.

I guess your circuit will work perfectly well if the input cap stays on. But if I have to choose to eliminate either the input cap or an unity gain buffer then I would get rid of the cap instead of the opamp because even the most expensive cap will introduce much more distortion than the opamp buffer.

Regards,
Bill

[AndrewT]

Hi,
is your filter stage around U1 just a simplified indication of the schematic? or is this the full filter?

[adrome00]

Hi Bill.
I think I see what you're talking about with the impedance issue. Is R5 an actual resistor or a modeled impedance? Also I'm not sure what 100R is for R3 on your original diagram.

But with this issue in mind, you have limited choices:

1. I think Sajti has a good point - you can eliminate U3. But then P1 must be optimized between two limits. By keeping P1 much lower than the input impedance of the power section (keeping R3 in) then the power amp will not see much variation in impedance. But P1 must be high enough (guess at least ~15v/~50mA = 300 ohms) for the current capability of U1.
However, low load impedance can likely affect performance - but it's possible that, say, 1k might not cause problems and perform wonderfully.

2. You could eliminate U3 and move P1 to the + input of stage 1. Whatever the signal impedance is, you can use something higher than that for P1 (maybe 500k), U1 won't care, and you won't have to worry about loss of upper bandwidth. LF noise could possibly be an issue if stage 1 has a lot of gain. In that case, a high quality pot should minimize this. Also, P1 could affect the U1 offset, which would be passed on to the power amp.
If you're working on a personal system, a few custom bias-tweaking components might be worth the advantage of minimal signal-path.

Best of luck - Richard