I wonder if the input offset voltages are being misunderstood in the previous posts.
Let's start with two identical input transistors in the LTP.
Arrange for the currents and voltages to these two identical transistors to be the same.
Now measure the output offset of the amplifier.
Any small error is due to the input offset currents not being identical.
One can unbalance the LTP to create an opposite offset effect to reduce the output offset. This in my opinion increases the drift with regard to different operating temperatures.
I believe it is better for reduced drift with temperature to adjust the two input bias resistors.
This can be done by reducing one, or other bias resistor to bring the input offset down to reduce the output offset.
This small reduction in input bias resistor value does not increase noise at the input.
The alternative DC servo method injects a small current into one of the input pair's bias resistor to reduce the input offset voltage difference. It is an automatic method that is achieving an almost identical correction for output offset as I have described above.
This automatic method has advantages.
Let's start with two identical input transistors in the LTP.
Arrange for the currents and voltages to these two identical transistors to be the same.
Now measure the output offset of the amplifier.
Any small error is due to the input offset currents not being identical.
One can unbalance the LTP to create an opposite offset effect to reduce the output offset. This in my opinion increases the drift with regard to different operating temperatures.
I believe it is better for reduced drift with temperature to adjust the two input bias resistors.
This can be done by reducing one, or other bias resistor to bring the input offset down to reduce the output offset.
This small reduction in input bias resistor value does not increase noise at the input.
The alternative DC servo method injects a small current into one of the input pair's bias resistor to reduce the input offset voltage difference. It is an automatic method that is achieving an almost identical correction for output offset as I have described above.
This automatic method has advantages.
Last edited:
Andrew, no misunderstanding on Tom's or my part. Both of us are painfully aware of the foibles of op-amp circuits. Yes, you are entirely correct that correcting the residual Vos of a well-matched input stage by trimming out the offset (almost always by trimming various resistors in the circuit) will cause the drift to increase. No free lunch, etc., and something to be dealt with in any op-amp design where one intends to trim for highest performance.
The great benefit of the servo is that the application of a servo containing a high-precision op-amp (designed for low offset and drift) will impart most of this DC precision to the amp (e.g. LM3886) being corrected.
The great benefit of the servo is that the application of a servo containing a high-precision op-amp (designed for low offset and drift) will impart most of this DC precision to the amp (e.g. LM3886) being corrected.
Tom and others,
A servo is a good method to regulate the output bias voltage to zero volt and thus avoiding a elektrolytic capacitor in the feedback signal.
Of course I know that the extra resistor to the input will increase the noise and maybe lower the bandwith and the effectiveness of Cc. That's why I put a very good capacitor of 1,0 microF over this resistor.
Indeed there maybe is a chance of temperature drift. I built two LM3886 amplifiers and measured several times the temperature drift at almost full power while the amplifiers became hot. The drift was only a few mV and after that the bias output voltage returned to the original value. (This is no guarantee for other opamps)
As we say in Holland: measuring is knowing. But I'll read chapter 5 if I can find the book.
Marc.
A servo is a good method to regulate the output bias voltage to zero volt and thus avoiding a elektrolytic capacitor in the feedback signal.
Of course I know that the extra resistor to the input will increase the noise and maybe lower the bandwith and the effectiveness of Cc. That's why I put a very good capacitor of 1,0 microF over this resistor.
Indeed there maybe is a chance of temperature drift. I built two LM3886 amplifiers and measured several times the temperature drift at almost full power while the amplifiers became hot. The drift was only a few mV and after that the bias output voltage returned to the original value. (This is no guarantee for other opamps)
As we say in Holland: measuring is knowing. But I'll read chapter 5 if I can find the book.
Marc.
So to avoid a capacitor in the feedback path where it does no harm, you would rather put a capacitor directly in series with the input to make sure that it does as much harm as possible. You're obviously free to do whatever you'd like, but I do question your logic there. 🙂
You can see the full performance measurements of the LM3886DR on my website. I measured the performance using an Audio Precision APx525 audio analyzer.
Tom
It is not to avoid a capacitor in the feedback path but to avoid a elektrolytic capacitor in the feedback path.
I just wanted to react on TioFrancotirador who stated "Since I do not want to use feedback cap I wanted to minimize DC with other means". I only suggest such an other - very simple - means.
I don't want to - and I can't - compete with you about the measurement equipement. But see for my measurements the thread: LM3886: the effect of the compensation network Cc, Rf2 and Cf.
Marc.
I just wanted to react on TioFrancotirador who stated "Since I do not want to use feedback cap I wanted to minimize DC with other means". I only suggest such an other - very simple - means.
I don't want to - and I can't - compete with you about the measurement equipement. But see for my measurements the thread: LM3886: the effect of the compensation network Cc, Rf2 and Cf.
Marc.
Not all electrolytic caps are created equal. The Nichicon Muze ES and UES series are sonically transparent. They contribute no THD (or anything else) that I can measure and I can't hear any effect of them either. If you value measurements as much as you say you do, you should really reconsider whether you're barking up the right tree here.
In my opinion, you are trying to solve a problem that does not exist. In your solution you are creating a problem that definitely does exist. But hey... It's your time, your money, so it is really up to you. However, I am not going to support a "solution" to a problem that doesn't exist - in particular when the "solution" is documented to degrade performance.
If you would like me to look at your threads, please link to them. Listing the name does not make them easier to find.
Tom
In my opinion, you are trying to solve a problem that does not exist. In your solution you are creating a problem that definitely does exist. But hey... It's your time, your money, so it is really up to you. However, I am not going to support a "solution" to a problem that doesn't exist - in particular when the "solution" is documented to degrade performance.
If you would like me to look at your threads, please link to them. Listing the name does not make them easier to find.
Tom
Last edited:
Of course if the electrolytic caps do no sonic harm, than there is no problem. But why are there so many discussions about "audio grade" electrolytic caps and the differences between these caps? And the audiophile caps are mostly rather expensive. Thus maybe it is good to avoid them. For that I proposed an alternative solution. Where is documented that this solution degrades the performance?
See for my other measurements: http://www.diyaudio.com/forums/chip-amps/217790-lm3886-effect-compensation-network-cc-rf2-cf.html .
Marc.
See for my other measurements: http://www.diyaudio.com/forums/chip-amps/217790-lm3886-effect-compensation-network-cc-rf2-cf.html .
Marc.
Because people need something to geek out about. 😉 What the expensive audiophile caps tend to have in common is that they're grossly under-specified (i.e. have no data sheet or a data sheet that contains hardly any information). Those I have measured have measured considerably worse (higher ESR, higher ESL -> further from an ideal cap) than commonly available parts that you can get from Digikey/Mouser/et al.
Yep. And many people buy them for that reason. "They're really expensive so they must be really good." Humans are funny critters.
You can measure the THD of capacitors with a distortion analyzer (or good sound card + software). It's not hard.
If you put a capacitor in series with the signal, I suggest choosing a capacitor with polypropylene dielectric or one of the audio grade electrolytics (Nichicon MUZE ES or UES for example).
The LM3886DR provides stellar THD (slightly better than the measurements shown in the LM3886 data sheet) with readily available parts. Its output offset is around ±1-2 mV in most builds (±10 mV max).
Tom
Last edited:
Tom, did you ever measure or look at the Elna Silmic IIs ? I believe these were recommended by Nelson Pass as coupling capacitors in power amplifiers previously.
Silmic II capacitors are pretty ordinary for ESR, not especially low. I do use them as coupling capacitors but not in power supplies. The attachment shows measurements I made six years ago.
Thanks! I appreciate the measurement. Any other measurements on other capacitors that we can compare with ?
Here's the Elna SILMIC II data sheet: http://www.elna.co.jp/en/capacitor/alumi/catalog/pdf/rfs_e.pdf
You can compare against similar components from Nichicon, Panasonic, etc. Mouser/Digikey/et al have the data sheets if you don't feel like going to the manufacturer's website directly.
The SILMIC appears to be a regular electrolytic can. Tan(delta) for a 25-35 V type is 0.1. That's pretty standard fare for an electrolytic cap.
Tom
You can compare against similar components from Nichicon, Panasonic, etc. Mouser/Digikey/et al have the data sheets if you don't feel like going to the manufacturer's website directly.
The SILMIC appears to be a regular electrolytic can. Tan(delta) for a 25-35 V type is 0.1. That's pretty standard fare for an electrolytic cap.
Tom
Thanks Tom. I looked at the Muses ES series of capacitors that you suggested. Those are about 0.14 to 0.16 of Tan(delta) at 25-35V. Does this mean the Silmics are somewhat better for audio usage ?
It was found on DigiKey's site back in 2009. Here is some Panasonic FM series curves.
Thanks again Bill! That is my favorite capacitor for low ESR applications. Although have you tried these for audio-frequency coupling ?
No, really low ESR isn't required for coupling so I have not used FM caps. Here is my final set of curves for Nichicon ES.
Very nice, thanks Bill. I am going to compare these and see what practical conclusions I can draw.
The Nichicon ES and Elna RFS series seem very similar at similar sizes, but the ES series consistently measure lower. As the ES series are bipolar, I am not sure how this affects their intended use as signal coupling caps...