Chip manufacturers usually recommended to use 2% tolerance caps if possible (5% max), because anything less will degrade the CMR. This really is more critical when amplifying very low level signals than it is for higher voltage audio signals, but it's still good to match them as closely as possible. The whole purpose of the input filtering is to prevent any RF from being amplified. Say there were no filtering. You'd be relying on the RF to be radiated equally into both signal wires so that it is amplified equally by the input opamp amp and then rejected by the diff amp. Filtering makes any RF appear as a small DC offset, which translates to a slightly higher DC offset at the output instead of noise.
If your circuit will be housed in it's own enclosure then it might not even be needed. I have my test circuit (no decoupling on either the input opamp or LM3875 and no filtering on the input) on my desk next to my computer, halogen light, and cd player and the noise really is very low, all things considered. If I turn the halogen light up I can hear some hum caused by noise being radiated into the wiring.
If your circuit will be housed in it's own enclosure then it might not even be needed. I have my test circuit (no decoupling on either the input opamp or LM3875 and no filtering on the input) on my desk next to my computer, halogen light, and cd player and the noise really is very low, all things considered. If I turn the halogen light up I can hear some hum caused by noise being radiated into the wiring.
The parts came in and I made some time to build up a board which you can see in the photo below. The fake credit card is under the amp just to give you an idea of the actual size. Bulk caps are 680uF Panasonic FM, op amp caps are Elna SilmicII, op amp is OPA2134 (but I'll eventually give the LM4562 a try too). Supply rails are regulated +/-20V for the LM3875, which is regulated down to +/-17.9 for the op amp using 2N3904/6 transistors and 18V zeners. The 3875 rails can obviously go higher but I don't want to push it with the small heatsink you see in the picture. It doesn't get too hot to touch but it is nice and toasty playing into a 4 ohm nominal load (my two 8ohm speakers in parallel).
In it's current form I just routed the source ground to the inverting input and the source signal to the non-inverting input (both inputs have a nominal 100kohm input impedance set by the 100kohm resistors to ground). It sounds very nice and I can't wait to get another channel up and running. Using balanced inputs should be another step up in terms of noise performance but I don't have a balanced source/preamp.
It's worth noting that neither a zobel or parallel resistor/inductor are used because they don't seem to be needed with my setup. I'll be putting about 20 boards up for sale in the trading post if anyone is interested (details will be included in that thread).
In it's current form I just routed the source ground to the inverting input and the source signal to the non-inverting input (both inputs have a nominal 100kohm input impedance set by the 100kohm resistors to ground). It sounds very nice and I can't wait to get another channel up and running. Using balanced inputs should be another step up in terms of noise performance but I don't have a balanced source/preamp.
It's worth noting that neither a zobel or parallel resistor/inductor are used because they don't seem to be needed with my setup. I'll be putting about 20 boards up for sale in the trading post if anyone is interested (details will be included in that thread).
Attachments
Hi,
what size is that heatsink?
Looks about 10 to 14C/W to me.
You must be close to shut down driving a 100r resistive load and you go and connect a parallel pair of 8ohm speakers giving a reactive 4ohm load.
I thinks you're bonkers to rely on the chipamps self protection circuits to save your amp from destruction.
what size is that heatsink?
Looks about 10 to 14C/W to me.
You must be close to shut down driving a 100r resistive load and you go and connect a parallel pair of 8ohm speakers giving a reactive 4ohm load.
I thinks you're bonkers to rely on the chipamps self protection circuits to save your amp from destruction.
Honestly, the heat sink is only toasty, I can hold my hand on it indefinitely no problem. I've felt the heatslug on some of my Tripath based class d amps get warmer! Besides, I want to test this thing out, make sure it works and all.
My speakers are 90dB @ 1W/1m and I'm sitting 3ft away so the amp really isn't being pushed that hard. The case they are going in will provide more adequate heatsinking
If I had to guess I'd say the heat sink shown in the photo is probably around 10 degC/W (maybe a little less).
My speakers are 90dB @ 1W/1m and I'm sitting 3ft away so the amp really isn't being pushed that hard. The case they are going in will provide more adequate heatsinking
If I had to guess I'd say the heat sink shown in the photo is probably around 10 degC/W (maybe a little less).
Hi Brian,
congratulations on the good work and the beautiful PCB.
But one question doesnt't leave me alone: What's the actual use? I mean I of course understand that it has
but why an instrumentation configuration for a line level circuit? Wouldn't hou have had nearly all of it by simply wiring a single opamp buffer in a differential fashion? Wouldn't you have gained all the benefits you need and only missed out the two least needed: extreme impedance and open-loop gain? Wouldn't you achieve the same CMMR with a single amp with a lot less tight-toleranced passive components? But it looks like a lot of fun, though!
Sebastian.
congratulations on the good work and the beautiful PCB.
But one question doesnt't leave me alone: What's the actual use? I mean I of course understand that it has
but why an instrumentation configuration for a line level circuit? Wouldn't hou have had nearly all of it by simply wiring a single opamp buffer in a differential fashion? Wouldn't you have gained all the benefits you need and only missed out the two least needed: extreme impedance and open-loop gain? Wouldn't you achieve the same CMMR with a single amp with a lot less tight-toleranced passive components? But it looks like a lot of fun, though!
Sebastian.
Hi Sebastian. Thank you for the compliments and questions.
I should preface the following by saying this is the first time I built an amp based on a high power op amp, so I don't know how a minimalist implementation sounds. Therefore, my reason for choosing to try an instrumentation configuration was solely due to it's attractiveness in theory Having listened to the end result for a few hours now (dual mono) I can say that it does in fact work quite well in practice.
A differential reciever using one opamp in front of the LM3875 would work fine, but would be more sensitive to impedance mismatch from the source due to the lower input impedance of the differential stage. It would also still require precise resistor matching but you could just as easily get chips with laser trimmed resistors already on the die.
I guess there are three main reasons I liked the two op amp buffer better. The first reason is, as you mentioned, the very high input impedance which is less sensitive to impedance mismatch (the OPA2134 has input bias currents about 2 orders of magnitude lower than the LM3875; 5pA typical vs 0.2uA typical). While it's not optimal to use a single ended source you can do so and still get very good results. This is what I'm doing and it sounds very good. The second reason is that two op amps provide a nice low output impedance to drive each input of the high power op amp. Many people have used a buffer in front of just one input and obtained good results so why not buffer both inputs? Put that second op amp in the dual package to good use The third reason has to do with DC output offset voltage. There's a lot of talk about high DC output offsets with the standard Gainclone configurations and ways to cure it, but the instrumentation configuration will have low DC output offset simply due to the way it works. The first two op amps buffer (or amplify if desired) the input signal while the third op amp rejects the common mode signal and amplifies the differential mode signal with respect to a reference voltage that you specify. For the application of an audio amplifier you simply use the speaker ground as the reference. Viola, the output voltage of the high power op amp is now referenced directly to the speaker ground instead of an input ground. DC offset now becomes more a function of the chip's specs and the quality of the PCB layout.
Other nice features of the instrumentation configuration include the option to use either balanced or unbalanced sources and less susceptibility to noise (radiated EMI, ground loop hum, etc.). Multiple amplifiers on one power supply shouldn't be a problem since both inputs are high impedance which should break any ground loops that would normally form with the standard implementations.
I should preface the following by saying this is the first time I built an amp based on a high power op amp, so I don't know how a minimalist implementation sounds. Therefore, my reason for choosing to try an instrumentation configuration was solely due to it's attractiveness in theory
Having listened to the end result for a few hours now (dual mono) I can say that it does in fact work quite well in practice.A differential reciever using one opamp in front of the LM3875 would work fine, but would be more sensitive to impedance mismatch from the source due to the lower input impedance of the differential stage. It would also still require precise resistor matching but you could just as easily get chips with laser trimmed resistors already on the die.
I guess there are three main reasons I liked the two op amp buffer better. The first reason is, as you mentioned, the very high input impedance which is less sensitive to impedance mismatch (the OPA2134 has input bias currents about 2 orders of magnitude lower than the LM3875; 5pA typical vs 0.2uA typical). While it's not optimal to use a single ended source you can do so and still get very good results. This is what I'm doing and it sounds very good. The second reason is that two op amps provide a nice low output impedance to drive each input of the high power op amp. Many people have used a buffer in front of just one input and obtained good results so why not buffer both inputs? Put that second op amp in the dual package to good use
The third reason has to do with DC output offset voltage. There's a lot of talk about high DC output offsets with the standard Gainclone configurations and ways to cure it, but the instrumentation configuration will have low DC output offset simply due to the way it works. The first two op amps buffer (or amplify if desired) the input signal while the third op amp rejects the common mode signal and amplifies the differential mode signal with respect to a reference voltage that you specify. For the application of an audio amplifier you simply use the speaker ground as the reference. Viola, the output voltage of the high power op amp is now referenced directly to the speaker ground instead of an input ground. DC offset now becomes more a function of the chip's specs and the quality of the PCB layout.Other nice features of the instrumentation configuration include the option to use either balanced or unbalanced sources and less susceptibility to noise (radiated EMI, ground loop hum, etc.). Multiple amplifiers on one power supply shouldn't be a problem since both inputs are high impedance which should break any ground loops that would normally form with the standard implementations.
Hi Brian,
thanks for the detailed response.
Which I perfectly understand!
I'd say, e.g., an INA134/137 provides most of it with a single 'device', but it:
- has been tried before,
- costs more,
- is less DIY,
- depends on availability of a proprietary TI device (or it's proprietary AD or THAT counterparts),
- probably needs a servo for DC compensation,
- is less scientific,
- doesn't feature the high input impedance, large common mode voltage range, low output impedance, etc.
- ...
And - you can really use it for instrumentation! It would work as a measurement amplifier for in-circuit measurements and could directly drive a low-impedance measurement instrument (channel), such as driving:
- a speaker without influencing the measured signal (like a detector amplifier),
- a CRT deflection coil to visualize signals (how about a 30" correlation meter),
- a motor that needs current to work precisely (like a servo in a control loop with a sensor),
- ...
And i'd say it's a proof of concept! Just a little unfortunate that the board doesn't cope with LM3876/86 (more current, lower impedance drive, more headroom/dynamic range), but hey!
My opinion: good that someone has tried it and that a board could be available for DIY.
Cheers.
thanks for the detailed response.
Which I perfectly understand!
I'd say, e.g., an INA134/137 provides most of it with a single 'device', but it:
- has been tried before,
- costs more,
- is less DIY,
- depends on availability of a proprietary TI device (or it's proprietary AD or THAT counterparts),
- probably needs a servo for DC compensation,
- is less scientific,
- doesn't feature the high input impedance, large common mode voltage range, low output impedance, etc.
- ...
And - you can really use it for instrumentation! It would work as a measurement amplifier for in-circuit measurements and could directly drive a low-impedance measurement instrument (channel), such as driving:
- a speaker without influencing the measured signal (like a detector amplifier),
- a CRT deflection coil to visualize signals (how about a 30" correlation meter
),- a motor that needs current to work precisely (like a servo in a control loop with a sensor),
- ...
And i'd say it's a proof of concept! Just a little unfortunate that the board doesn't cope with LM3876/86 (more current, lower impedance drive, more headroom/dynamic range), but hey!
My opinion: good that someone has tried it and that a board could be available for DIY.
Cheers.
sek said:
I originally wanted to use the 3886 but the pinout of the 3875 lended itself to a nicer layout. In reality the 3875 doesn't seem to have any problem driving a 4 ohm load, but I didn't push it hard at all. That will be something to try! It also doesn't seem to need a zobel or parallel resistor/inductor on the output like most people say the 3886 requires. It wouldn't be too hard to change the layout to work with the 3886 (it's unfortunate National made their pinouts so different) but an even better solution might be to use one of the TI power op amps that is stable at lower gains. That way you can use the buffers for most of the voltage gain and the high power chip for mainly just current gain.
I tried to stay away from the special chip solutions like the INA134 because they limit part choices and generally cost more.
Hi Owen. Did your schematic look like the one posted a few pages back? What op amp and power op amp did you use? And finally, you made it sound like you never got to actually hear it. If you did, how would you rate it against a standard configuration? I'm slowly working my towards a differential signal chain 
Upon request, here is the schematic with all part values. These are what I am using on my boards with very good results.
C14, C15, C16, and C17 are optional EMI/RFI filtering caps. I would recommend at least using C16 or C14, C15, and C16 if possible. C14 and C15 should be 5% tolerance or better.
C9 is an optional rail to rail bypass cap. 100V rating minimum.
R13, R14 and R15, R16 may need to be recalculated if you plan on changing the zener voltages.
R17, R18 are optional loading resistors. Be careful of the rated power dissipation if you decide to lower their values. I would recommend using these to help ensure a symmetrical supply voltage and to help discharge the supply caps when the power is turned off. This configuration only has a small turn on click and a small turn off click accompanied by some other rather quiet noises due to the op amp supply rails discharging.
Q1, Q2 can be changed to any suitable NPN, PNP devices. 2N3904 and 2N3906 are cheap and widely available.
IC1 is the dual op amp. I'm using the OPA2134 but other devices may be used. I plan on trying the LM4562 when I get the itch to swap out the 2134.
IC2 is the chip amp. I'm using the LM3875 but other devices may be used as well. If you use a different chip you may change the feedback resistors to change the gain of the chip amp. Be careful with stability issues and lower gain settings.
The gain of this configuration can be changed with only one resistor - R11. 20k gives an overall gain of 20V/V. 10k gives an overall gain of 30V/V. Set your gain based on your source and power supply voltages. The gain formula is on the schematic.
C14, C15, C16, and C17 are optional EMI/RFI filtering caps. I would recommend at least using C16 or C14, C15, and C16 if possible. C14 and C15 should be 5% tolerance or better.
C9 is an optional rail to rail bypass cap. 100V rating minimum.
R13, R14 and R15, R16 may need to be recalculated if you plan on changing the zener voltages.
R17, R18 are optional loading resistors. Be careful of the rated power dissipation if you decide to lower their values. I would recommend using these to help ensure a symmetrical supply voltage and to help discharge the supply caps when the power is turned off. This configuration only has a small turn on click and a small turn off click accompanied by some other rather quiet noises due to the op amp supply rails discharging.
Q1, Q2 can be changed to any suitable NPN, PNP devices. 2N3904 and 2N3906 are cheap and widely available.
IC1 is the dual op amp. I'm using the OPA2134 but other devices may be used. I plan on trying the LM4562 when I get the itch to swap out the 2134.
IC2 is the chip amp. I'm using the LM3875 but other devices may be used as well. If you use a different chip you may change the feedback resistors to change the gain of the chip amp. Be careful with stability issues and lower gain settings.
The gain of this configuration can be changed with only one resistor - R11. 20k gives an overall gain of 20V/V. 10k gives an overall gain of 30V/V. Set your gain based on your source and power supply voltages. The gain formula is on the schematic.
Attachments
LM3875, in differential input mode - it reduced the parts count alot (no need for the additional buffers, as it was being driven differentially in the first place, from the DCX2496.
Sound quality reported was very good - I'm slowly getting the parts together to do a transformer input gainclone... story of my life really, no time to build much at all
Owen
Sound quality reported was very good - I'm slowly getting the parts together to do a transformer input gainclone... story of my life really, no time to build much at all
Owen
Mr Squeegs built it
http://www.diyaudio.com/forums/showthread.php?threadid=56243
Lets see if he shows up....
Owen
http://www.diyaudio.com/forums/showthread.php?threadid=56243
Lets see if he shows up....
Owen
A transformer coupled input would be ideal, but those little things are expensive for most of us. The active instrumentation method really doesn't add too many parts and can be cheaper depending on what parts you use.
Owen answered that nicely in his post above. If you want lower gains you'll need to use a chip that is stable at those gains. I believe the OPAs (the high power ones from TI) are unity gain stable. The 3875's specified minimum gain is 10, but it might be able to go a bit lower without any issues. So far the 3875 has shown no problems with oscillation when used as a differential amplifier with a gain of 10 with no parallel resistor-inductor or zobel on the output.
AndrewT said:
Owen answered that nicely in his post above. If you want lower gains you'll need to use a chip that is stable at those gains. I believe the OPAs (the high power ones from TI) are unity gain stable. The 3875's specified minimum gain is 10, but it might be able to go a bit lower without any issues. So far the 3875 has shown no problems with oscillation when used as a differential amplifier with a gain of 10 with no parallel resistor-inductor or zobel on the output.
The transformer input does not have to be too expensive... If the stage before has little dc offset ( <10 mV), you can comfortably use small VA mains torroids - high inductance, extremely good coupling, low core artifacts and great bandwidth - or use a parafeed style input (it would still then have the capacitor input) if DC offset of the previous stage was higher.
From Susan Parkers site
15VA Input Transformer Update - 27th Oct 04
Low level bandwidth is 10Hz to 25.5 kHz (-3 dB), 31.5 kHz (-6dB).
And you could use:
http://www.oep.co.uk/audio_transformers_high_performance/audio_transformers_high_performance.html
These are available for not a huge outlay. The key to remember is that this is a low signal level, low current transformer, and doesnt need to be a huge lump of iron unless it is a classic value pre with output transformer needed to handle large DC Voltages at moderate currents.... then you do need an airgap...
Have fun
Owen
From Susan Parkers site
15VA Input Transformer Update - 27th Oct 04
Low level bandwidth is 10Hz to 25.5 kHz (-3 dB), 31.5 kHz (-6dB).
And you could use:
http://www.oep.co.uk/audio_transformers_high_performance/audio_transformers_high_performance.html
These are available for not a huge outlay. The key to remember is that this is a low signal level, low current transformer, and doesnt need to be a huge lump of iron unless it is a classic value pre with output transformer needed to handle large DC Voltages at moderate currents.... then you do need an airgap...
Have fun
Owen
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.