Building a Gainclone with LM3886 General Direction and PSU

[pacificblue]

Originally posted by bmwpowere36m3
It says that the feedback resistance should be matched with the input resistance, i.e. the same and that the ideal input resistance is 100k minimum. By setting the feedback resistance equal to 100k it minimizing DC offsets at the output. Is this correct?

The impedances at both inputs should be the same for best common mode rejection. That means you have to put yourself in the place of each input pin and find out the total impedance of what you see. In that case you also need to take into account, what the preamp contributes to that impedance. If you do that, you will find that the chipamp schematic is not a good match for the Millett. You have to take into account the 2k resistor to ground in the Millett's output and match the feedback loop accordingly.
The ideal value is always, what you think is the best compromise between low offset voltage at the output, high offset voltage at the input, low noise, losses, etc. It is also a question of what common mode rejection is acceptable for you and the amplifier.

Originally posted by bmwpowere36m3
Finally as I understand it the purpose of Rin is to keep the chipamp stable when there is no "preamp" or source connected.

The purpose of Rin is to provide a path to ground for the offset current and to provide a load with which the preamp can work stably. In some setups it also serves as a terminating resistance for the cables impedance, as snubber for the input capacitor and as attenuator.

Originally posted by bmwpowere36m3
For one, I assume the output impedance of the millet is 2k, right as formed by the RC.

The output impedance is in line with the signal, so basically the IRF510's 0,54 Ohm drain-to-source resistance plus C3 plus the lead and trace impedances in the path.

Originally posted by bmwpowere36m3
Second I know the output capacitors are large in efforts to maintain bass when low-impedance headphones where plugged in. With the high impedance of the LM3386, they could be smaller yet preserve "bass" or bandwidth, correct?

Yes. The limiting factors in that case are R6 and R12, which will maintain the load for the capacitor below 2k.

Originally posted by bmwpowere36m3
What is the best way to couple these two sections, noting that the headphone amp will ONLY be used with the chipamp. Finally I read some suggestions as to various changes to the selected resistances and capacitors used in chipamp's PCB....

If you are absolutely sure that you will never ever connect anything else than this preamp to the LM3886, skip Rin, R1 and R2 in the chipamp schmematic and connect the preamp out directly to the LM3886's pin 10. Ci should then be in the range of 1000-1500 µF to create a meaningful filter in conjunction with the actual values of C3 and R6 or C5 and R12 in the Millett.
If you have the slightest inkling that you may in the future decide to try different preamps or none at all, you should prepare the LM3886 to be compatible. Then you should have a look at Figure 5 on page 7 or at Figure 7 on page 21 in the LM4780's datasheet.
You should consider not to use the Millett as preamp. It provides a too high voltage level of ~19 V. Its output transistors are power transistors and are not likely to work well into high impedances. The 2k to ground in the output compels the use of too unpractical values for the LM3886's circuit to make it compatible with any thing else afterwards.

[bmwpowere36m3]

Hmmm, it's looking like the MH isn't going to be a good preamp... what about setting up the chipamp for "general" use with any preamp or source that has its own volume control?

[bmwpowere36m3]

Back again, I've decided to go the simple route with the gainclone build, I have a quick question regarding the amp circuit. I'm going with the simple build like this:



My question is regarding Rg (input resistor)... I'll be using a potentiometer with this amp circuit, so I'll be feeding it from a source (Gamma1 DAC). So do I use Rg along with (in parallel) a 10k pot or just use the pot (as the input impedance)? The rest of the circuit remains the same. Thanks.

[danielwritesbac]

Some of the data sheets show 20k potentiometer // 20k resistor.

However you do it, that schematic requires at least a 10k load at the input of that amplifier in order to keep DC offset low.

A 5k load, which is illustrated within your question above, results in even lower DC offset. Low DC offset is a good thing.

Point of reference:
Inappropriate DC offset occurs at 18k and higher figures for the amplifier's input load resistor, for the amplifier that's pictured in the schematic.
The fix is either to change the resistor value (of course) or to insert a capacitor in series with "RI" and this capacitor is known as "NFB cap" for its common name.

[bmwpowere36m3]

Originally Posted by danielwritesbac

Some of the data sheets show 20k potentiometer // 20k resistor.

However you do it, that schematic requires at least a 10k load at the input of that amplifier in order to keep DC offset low.

A 5k load, which is illustrated within your question above, results in even lower DC offset. Low DC offset is a good thing.

Point of reference:
Inappropriate DC offset occurs at 18k and higher figures for the amplifier's input load resistor, for the amplifier that's pictured in the schematic.
The fix is either to change the resistor value (of course) or to insert a capacitor in series with "RI" and this capacitor is known as "NFB cap" for its common name.

That's opposite of what the National datasheet specs, " The desired input impedance is set by RIN. Very high values can cause board layout problems and DC offsets at the output. The value for the feedback resistance, Rf1, should be chosen to be a relatively large value (10 kΩ– 100 kΩ), and the other feedback resistance, Ri, is calculated using standard op amp configuration gain equations."

And since setting RIN equal to Rf1 should reduce output offset (or so it's said), that leads me to believe that RIN also should be in the range of 10-100k. I know that higher input impedances are more susceptible to noise, but make it easier for the source to drive them.

[danielwritesbac]

Originally Posted by bmwpowere36m3

That's opposite of what the National datasheet specs, " The desired input impedance is set by RIN. Very high values can cause board layout problems and DC offsets at the output. The value for the feedback resistance, Rf1, should be chosen to be a relatively large value (10 kΩ– 100 kΩ), and the other feedback resistance, Ri, is calculated using standard op amp configuration gain equations."

And since setting RIN equal to Rf1 should reduce output offset (or so it's said), that leads me to believe that RIN also should be in the range of 10-100k. I know that higher input impedances are more susceptible to noise, but make it easier for the source to drive them.

To quote Peter of AudioSector: "Use 10k"
He was talking about the input load resistor in the context of DC offset.

It is known that the source is a headphone level device owned by the creator of this thread. We don't want it to start clipping well before the potential of the power amp is reached. That source might be helpless if its load were 100k; however, if its load is 10k, then its got a chance to work right.

Ooh! That gives me an idea! Thanks!!!!

[bmwpowere36m3]

Originally Posted by danielwritesbac

To quote Peter of AudioSector: "Use 10k"
He was talking about the input load resistor in the context of DC offset.

It is known that the source is a headphone level device owned by the creator of this thread. We don't want it to start clipping well before the potential of the power amp is reached. That source might be helpless if its load were 100k; however, if its load is 10k, then its got a chance to work right.

Ooh! That gives me an idea! Thanks!!!!

I am the creator of "that" thread. My original plans were to use a headphone amp (Millet "SS" Hybrid) as a preamp for the gainclone to add some "tube warmth." But my plans have changed, for now the gainclone will be a standalone power amp (with potentiometer) that'll I'l feed from a source (AMB Labs ϒ1 DAC). I might still experiment with a headphone amp as a preamp.

I think I found my amp design, taken from the same site:



I'm also going to get a couple of pots (10k & 50k) to experiment with. Components include: Rf 22k, Rg 680R, Rm 10k, and Cs 2200uF.

[Pentode]

Trouble is with regulated supplies for PAs... won't they have some hysteresis, meaning will the be slow to regulate say... hf signals above a few hundred hertz, just thinking this would add distortion under certain musical conditions and power levels.... another reason why some don't use Fuses in speakers and HT lines to help keep resistances down to a bare minimum when passing 'peaky' amperes - just my thoughts.

Dave

[star882]

A good voltage regulator should be able to respond to transient loads very quickly. Test it with a load resistor and a MOSFET gated by a signal generator.

Power supply circuits for modern CPUs have to supply 10s of amps at a low voltage and respond very, very quickly to transient loads. Many even need to dynamically change the output voltage for energy saving features.

[danielwritesbac]

Oh yes. Decibel Dungeon has a briefing on how to use computer power supplies with a chip amp. The technology does seem to work well.