A few years ago I built a stereo power amplifier with the LM3886 chip amp and have been using it daily with a single iPhone source as my primary sound system. It sounds great, it's a HiFi design with high end components, but it's also a bit too minimal, with only one stereo input and one stereo output. I am planning to build a HiFi preamplifier with an input select switch, volume, tone, and balance controls. My audio sources will be an iPhone, CD player, and TV. I read that these have a relatively high output voltage, so I decided on a moderate maximum gain of ~13 db, which is set by a Baxandall active volume control circuit.
It seems like the best performing audio op-amp that still comes in a DIP package is the LM4562, so that is what I plan on using for the op amps. The preamplifier uses a high pass and low pass filter at the input to set the bandwidth, then an input buffer stage with parallel LM4562 op-amps. The input impedance is set to 47.1K. The input buffer stage is followed by a Baxandall active volume control circuit. After the volume control circuit is another parallel op-amp buffer (not sure if this is necessary), then a Baxandall tone control, again followed by a parallel op-amp buffer. Then a balance control circuit, followed by an output buffer consisting of two op-amps in series.
I am trying to optimize this circuit for low noise, good bass response, and clear mids and highs. I have designed and built amplifiers with chip amps like the LM3886, LM386, TDA2050, and TDA2003 before. But I don't have a lot of experience with op-amp based amplifiers, so I was wondering if anyone could check out my circuit to make sure I'm not missing anything?
I understand the importance of having a high impedance at the preamplifier's input, but is it necessary to set the input impedance at each set of op-amp buffers? For example are the 47K input impedance setting resistors R5, R8, R17, and R21 necessary?
If they are, would it be better to set them to a lower resistance for less Johnson noise? The datasheet says that the LM4562 can drive 600R loads, so can I set the input impedance of all buffer stages after the input to 600R instead of 47K?
The input impedance of the preamplifier is 47.1K, should I balance the input bias currents of the op-amps (U1 and U2) by placing a 47.1K resistor in the feedback loop?
Can I balance the input bias currents of U4 (in the active volume control section) by placing a ~500R resistor from the positive input to ground?
The values of the 2 uF DC blocking caps around the volume, tone, and balance pots were chosen arbitrarily based on the value of C1. Is there a better value for those?
Thanks for any input or suggestions!
It seems like the best performing audio op-amp that still comes in a DIP package is the LM4562, so that is what I plan on using for the op amps. The preamplifier uses a high pass and low pass filter at the input to set the bandwidth, then an input buffer stage with parallel LM4562 op-amps. The input impedance is set to 47.1K. The input buffer stage is followed by a Baxandall active volume control circuit. After the volume control circuit is another parallel op-amp buffer (not sure if this is necessary), then a Baxandall tone control, again followed by a parallel op-amp buffer. Then a balance control circuit, followed by an output buffer consisting of two op-amps in series.
I am trying to optimize this circuit for low noise, good bass response, and clear mids and highs. I have designed and built amplifiers with chip amps like the LM3886, LM386, TDA2050, and TDA2003 before. But I don't have a lot of experience with op-amp based amplifiers, so I was wondering if anyone could check out my circuit to make sure I'm not missing anything?
I understand the importance of having a high impedance at the preamplifier's input, but is it necessary to set the input impedance at each set of op-amp buffers? For example are the 47K input impedance setting resistors R5, R8, R17, and R21 necessary?
If they are, would it be better to set them to a lower resistance for less Johnson noise? The datasheet says that the LM4562 can drive 600R loads, so can I set the input impedance of all buffer stages after the input to 600R instead of 47K?
The input impedance of the preamplifier is 47.1K, should I balance the input bias currents of the op-amps (U1 and U2) by placing a 47.1K resistor in the feedback loop?
Can I balance the input bias currents of U4 (in the active volume control section) by placing a ~500R resistor from the positive input to ground?
The values of the 2 uF DC blocking caps around the volume, tone, and balance pots were chosen arbitrarily based on the value of C1. Is there a better value for those?
Thanks for any input or suggestions!
Too complicated for my taste. Why so many (8) stages? Why do you need to parallel opamps?
You can bypass all 2uF coupling caps extept the first one on the input, but possibly add one on the output. With bypassed caps you don't need R5, R8, R17 and R21. C2 is too high. 1nF should be enough.
You can bypass all 2uF coupling caps extept the first one on the input, but possibly add one on the output. With bypassed caps you don't need R5, R8, R17 and R21. C2 is too high. 1nF should be enough.
The shorted 33 pF capacitors are quite unnecessary. The same holds for R8 and R17, as there is already a DC path through the output of the previous stage. R15 is also unnecessary, it can best be replaced with a direct connection to ground.
The Zobel network at the output would make perfect sense for a power amplifier, but not for a line output. It will only load the op-amps too much.
I also don't understand what the paralleled buffers are for. In fact, I think you can leave out U2, U5, U6, U8, U9 and U11.
Reducing R1 would reduce the noise with open input, but would only make things worse when there is a signal source connected. When driven from a signal source with nonzero impedance, it would attenuate the signal more than it reduces the noise, so you would end up with a worse signal-to-noise ratio. (I could also explain this in terms of increased equivalent input noise current.)
The AC coupling capacitors around potmeters cause some loss of bass and low-frequency phase shift, but they may also prevent anodization of the wiper if you use carbon potmeters and reduce issues with scratching sounds when the wiper temporarily loses contact.
The Zobel network at the output would make perfect sense for a power amplifier, but not for a line output. It will only load the op-amps too much.
I also don't understand what the paralleled buffers are for. In fact, I think you can leave out U2, U5, U6, U8, U9 and U11.
Reducing R1 would reduce the noise with open input, but would only make things worse when there is a signal source connected. When driven from a signal source with nonzero impedance, it would attenuate the signal more than it reduces the noise, so you would end up with a worse signal-to-noise ratio. (I could also explain this in terms of increased equivalent input noise current.)
The AC coupling capacitors around potmeters cause some loss of bass and low-frequency phase shift, but they may also prevent anodization of the wiper if you use carbon potmeters and reduce issues with scratching sounds when the wiper temporarily loses contact.
Last edited:
Have a look at rod elliott's project:
http://sound-au.com/project88.htm
Or this one, with tone controls:
http://sound-au.com/project97.htm
http://sound-au.com/project88.htm
Or this one, with tone controls:
http://sound-au.com/project97.htm
Ditto about C3,4,7,8,9,10,15,16,19,20 - doing nothing.
Parallelled buffers on the input reduce noise which might be useful before the gain stage, but not needed thereafter.
C2 is too big and will act as low pass filter for a high impedance source - 1nF is plenty, 470pF is probably a good general value. Add another ceramic 220pF directly on the input connectors to stop RF getting into the case too, LM4562's are very sensitive to RF. Ceramic caps must be C0G/NP0
You don't need any buffers after the volume section or tone section as their outputs are direct from an opamp.
Parallelled buffers on the input reduce noise which might be useful before the gain stage, but not needed thereafter.
C2 is too big and will act as low pass filter for a high impedance source - 1nF is plenty, 470pF is probably a good general value. Add another ceramic 220pF directly on the input connectors to stop RF getting into the case too, LM4562's are very sensitive to RF. Ceramic caps must be C0G/NP0
You don't need any buffers after the volume section or tone section as their outputs are direct from an opamp.
I am not sure. Baxandal circuits can have very low Z inputs in max pot position. The op amp feed back path is then under heavy load.
C5 and C11 should be 100uF .. 470uF to avoid THD and bass loss.
Last edited:
The volume control will not get you to zero output.
None of the stages have any gain.
It is mono.
The bandwidths listed seem excessive - less BW = less noise..
Are these are design choices?
A couple thoughts:
None of the stages have any gain.
It is mono.
The bandwidths listed seem excessive - less BW = less noise..
Are these are design choices?
A couple thoughts:
- You might consider starting with an app note circuit from the op amp data sheet.
- And possibly think about a multichannel digital volume control (and something similar for tone controls) so it can be expanded into a stereo or surround amp.
- You haven't really provided any input source switching. Perhaps a bluetooth receiver IC for your iphone might be worth considering too.
- - The TI version spec sheet recommends an updated IC for new designs: https://www.ti.com/document-viewer/opa891/datasheet
Last edited:
I agree, it ended up more complicated than I would have wanted too. I got the idea of parallel op-amp buffer stages from Doug Self's Small Signal Audio Design. In that book, he explains that op-amps in parallel have some noise cancelling effect. Also from what I understand, he was also recommending to put buffers before and after the stages with potentiometers so that the pot setting doesn't change the input and output impedances of the section. But if that is not really necessary then I would love to get rid of some of the buffer stages to reduce the parts count.
Could you explain more about what you mean by this? Do you mean that if I remove the 2 uF caps (except at the input), that I can also remove R5, R8, R17, and R21?
Ok good to know. I read somewhere online that the 33 pF caps can help stabilize the op-amps, but if you think they aren't needed I will remove them. Nothing wrong with less components! I'm guessing the PCB layout will have the most effect on stability and oscillations?
This came from Doug Self's Small Signal Audio Design book. He talks about how parallel op-amps have some noise cancelling effect. But is the additional cost and complexity of the circuit worth a few db's of noise reduction? I don't know, maybe not. I will consider removing the extra op-amps. Would you still recommend single op-amp buffer stages between the volume, tone, and balance circuits though?
Ok great to know caps and the buffers after the volume and tone controls are not needed, I definitely want to get rid of anything that's not needed. I'll incorporate all of these suggestions in the next version of the schematic.
What do you think about the two op-amps in series at the output? Is there any real benefit to having a second op-amp there? This arrangement was described in Small Signal Audio Design as being better able to drive lower impedance loads than a single op-amp. I think the input impedance of my power amp is about 20K, and I also plan on having a second output in parallel for an active sub. Would a single op-amp at the output buffer be enough to handle that?
Ok sure, that makes sense. There is another DC blocking cap right before the balance pot (C17), should that also be 100-470uF? What about the caps between the pot wiper and the op-amp (C6 and C18) in the volume and balance circuits? Are those really necessary? I just want to make sure the pots don't make any scratching noises when turned.
Ok that's a problem then. I would like the volume to be silent at full attenuation. Is that because of the Baxandall design, the gain, or something else?
I actually had a hard time deciding on which frequencies to use for the upper and lower cutoffs. What is a good bandwidth for a HiFi preamplifier?
I didn't include the inputs and input select switch in the schematic, but I'm planning to have 5-6 RCA inputs controlled by a dual gang rotary switch. I'm also planning to add a preamp bypass switch and a tone control bypass switch.
My next project is going to be a Raspberry Pi audio player with a HiFi audio DAC, network attached storage, and bluetooth 🙂
I forgot the most important thing: the supply decoupling capacitors are missing! Without them, it will probably be an oscillator rather than an amplifier.
That's true but it would make sense for much weaker signals in sub mV range. Line level signals are already in V range. Paralleling would be overkill plus it could produce some side effects.
Yes, it makes sense to decouple a potentiometer from preceding high impedance source. U1 does that. But output of U4 is already low Z so U5 and U6 are redundant. Similarly, U8 & U9 are redundant because U7 output is low Z. Then U11 is arguably redundant to, except if you want to put some opamp with more potent output stage. You don't need Zobel R24 C21 there for sure. Series 100ohm R25 is sufficient to decouple output stage from cable capacitance and ensure robust stability (this is something that we cannot afford with power amps, hence Zobel networks come to help).
I don't get why do you need positive FB with R22. Is it some bootstrapping to increase input impedance of U11 (which is already intrinsically very high)? I don't find it useful.
The purpose of resistors like R5, R8, R17 and R21 is to provide DC path in order to define the DC voltage on the input nodes (+ or -). E.g. the + node of U3 is DC isolated because of C6. Therefore you need R5 to set 0V at the + node. However, since the DC output of U1 is essentially 0V you can bypass C5 and C6. Then R5 becomes redundant (BTW I think U3 is redundant to).
Going further, U4 output defines 0V DC for the next stage (R8 redundant). And so on.
There could be some justification (not worthy in my opinion) to keep C5, C11, C17 and C18 on. It's recommendable to isolate potentiometers from DC. Because DC can cause some scratchy noise when turning the pot, and it can possibly cause some metal deposition on the carbon tracks in the pot. However, DC offset of modern opamps is in 1mV range which is in my experience too low to cause mentioned problems. Ultimately it's designers choice.
Oh, and provide a bypass switch to "defeat" the Baxandall TC. Nice feature, I guess you will need it.
Forget opamps, take a look at Mark Johnson's Noir schematic.
Although aimed at headphone users, it make a fantastic preamp.
https://www.diyaudio.com/community/...class-a-single-ended-150ma-bias.343616/latest
Although aimed at headphone users, it make a fantastic preamp.
https://www.diyaudio.com/community/...class-a-single-ended-150ma-bias.343616/latest
https://123.physics.ucdavis.edu/week_2_files/Johnson_noise_intro.pdf for those 100k in bandaxal , 10k is let say ok but 100k ten times more noise.
This is part of what I have using
Attachments
OP's intention is to make own design using textbook approach. It includes tone controls and balance control.
Noire is completely different beast. It's a tested minimalistic project from the sea of good preamp solutions. Two transistor design, no tone controls. And probably most important, it's some other's solution.
Going to extreme, from a purely practical point, probably the best solution is to obtain a good secondhand preamp from a top tire manufacturer. But it's not DYI anymore.
How much of that noise will be audible is the question.
Rod elliott examples were more a suggestion to reduce number of (also potentially noisy) opamps.
The LM4562 has an equivalent input current noise of 1.6 pA/√Hz at 1 kHz, if you have a product from a good batch without popcorn noise and if the manufacturer didn't use spec tricks like specifying only the differential component of the noise current. That means that at resistive impedance levels above about 7 kohm, op-amp current noise will dominate over resistor thermal noise.
Yes it is completely different and that's the point. It requires a leap of faith away from the domination of opamps in our hobby,
Two would be needed - one before and one after the Baxandall per channel but of course it's more work to assemble them that a couple of op amps.
The Noir is completely silent using headphones, where any hiss or hum is easily audible.
Last edited:
Just didn't add them yet to keep the schematic simpler. But I was planning on a single 100 nF cap from each voltage pin (+V and -V) to ground, right next to each op-amp. Would adding a second cap with a higher capacitance, say around 10 uF -100 uF add any benefit?