Koinichiwa Steve,
Nice one. But may I suggest a few minor changes?
First, make the Input Transformer a Stepdown, say 4:1. This adds actual interesting functionality to the whole schebang. First we get a 16:1 stepdown in source impedance, which is nearly as good (or should I dare say "even better"?) than an active Buffer.
If we had a source impedance of 12k5 (50k passive preamp in -6db position) we would have < 1k source impedance to our Gainclone.
Then change the Feedback Network to compensate the loss of gain. Using a 2k2 input resistor and a 220K feedback resistor with a 4:1 stepdown transformer will result in an Input sensitivity for full output with 38V Rails (25V transformer Windings at nominal mains) of around 1V, but at that level we are already in the dynamic headroom. The reflected input impedance will be around 35KOhm.
The last thing to do is to make the Impedance on the positive input equal to that on the negative one. If we do not account for the source impedance but account for the DCR of the Transformer the resistor on the positive input to ground our resistor needs to be (Rdc+2k2) // 220k = app. 2k2....
And now we have an inverting gainclone with polarity (not phase) selection, the ability to use balanced or unbalanced inputs (provided a suitable transformer is used), a relatively high input impedance and galvanic insulation from the input.
Of course, if we do all this and add such a major cost part as a good quality input transformer, how about progressing to the "Bridgeclone" with (say) 6pcs LM3875 per channel (see Bridgeclone Thread), distortion cancellation (as in the Chipamp - TNG Thread) and we would a most excellent major bad mutha kick ***, high power, high end amp.
And if we go then for a really nice switching supply with choke input supplies seperated for each chip and using only Film Cap's (22uF/63V Mylars should do great) we may actually have something really outstanding.
Anyway, just some direct and leading thoughts.
Sayonara
Steve Eddy said:
Nice one. But may I suggest a few minor changes?
First, make the Input Transformer a Stepdown, say 4:1. This adds actual interesting functionality to the whole schebang. First we get a 16:1 stepdown in source impedance, which is nearly as good (or should I dare say "even better"?) than an active Buffer.
If we had a source impedance of 12k5 (50k passive preamp in -6db position) we would have < 1k source impedance to our Gainclone.
Then change the Feedback Network to compensate the loss of gain. Using a 2k2 input resistor and a 220K feedback resistor with a 4:1 stepdown transformer will result in an Input sensitivity for full output with 38V Rails (25V transformer Windings at nominal mains) of around 1V, but at that level we are already in the dynamic headroom. The reflected input impedance will be around 35KOhm.
The last thing to do is to make the Impedance on the positive input equal to that on the negative one. If we do not account for the source impedance but account for the DCR of the Transformer the resistor on the positive input to ground our resistor needs to be (Rdc+2k2) // 220k = app. 2k2....
And now we have an inverting gainclone with polarity (not phase) selection, the ability to use balanced or unbalanced inputs (provided a suitable transformer is used), a relatively high input impedance and galvanic insulation from the input.
Of course, if we do all this and add such a major cost part as a good quality input transformer, how about progressing to the "Bridgeclone" with (say) 6pcs LM3875 per channel (see Bridgeclone Thread), distortion cancellation (as in the Chipamp - TNG Thread) and we would a most excellent major bad mutha kick ***, high power, high end amp.
And if we go then for a really nice switching supply with choke input supplies seperated for each chip and using only Film Cap's (22uF/63V Mylars should do great) we may actually have something really outstanding.
Anyway, just some direct and leading thoughts.
Sayonara
AudioFreak said:
Quite true!
In fact if becomes a 'buffer' with a defined input impedance and much lower output impedance (and sufficient current etc) to drive the loudspeaker load.
Another way of describing what KYW's circuit is doing (beyond being a buffer), is that it has the same signal/noise ratio AS IF it had a gain of +20dB. As a buffer though it is unity gain (in theory could be configured to have even less?). If the S/N was not maintained it would be unstable.
Incidentally, while I'm at it, the 'inbuilt' capacitance between output and (-) must be quite low - using 1M feedback resistor and a couple of pF in parallel brings the response, otherwise flat, several dB down @ 20KHz. I found this out while developing the JLTi amp (http://members.ozemail.com.au/~joeras).
So where do we get the gain? My SACD player with VSE module (www.vacuumstate.com) has only 700mV out.
Joe
Kuei Yang Wang said:
This is your idea of 'minor' changes?????
Since an inverted circuit is essentially an I/V converter, where the 'normal' IGC converts voltage to current then back to voltage... and now we put a tx in front of that... another current exchange... didn't realise till lately how much fun/intrigueing inverted circuits are. Get the tx to reverse phase and we have...
A non-inverting inverting Gainclone.
Add switchmode PS and L input supplies and while we are cooking, can we think of anything else to chuck into this pot-pouri?
Now all we need is find someone to build it?
Joe
Kuei Yang Wang said:
Sure, you could do that. But I don't see any real advantage. The output of the transformer's going to be going straight to the input of the amp so you're not driving any lines with it so I don't see any particular advantage there. And the 12k input impedance shouldn't be problematic even for tube sources.
I guess I've just never cottoned to the idea of attenuating the signal only to jack it back up again. It's just never worked well for me.
se
Non-unity gain buffer and near-unity gain amp chips?
Hi, folks!
I'm a newbie, waking up this dormant thread, heh!
OK, I was thinking (dreaming??) about building a powerful
(possibly multichannel) amp out of these LM3875/3886 things.
Thus, it will naturally go to one of BP(bridged/parallel) configurations.
There were good discussions on BP in other threads, especially about phase-splitting methods. It's claimed that using (good) transformer is the best, but cost of good transformers seems a bit prohibitive. (How much will Jensen transformers cost you if you need six, huh?) So, I'd prefer ways of using not-so-expensive op amps, being inverters and buffers.
Here goes my first question. If you need buffers anyway, is there any definite reason that they should be of unity gain? In other words, if buffers are allowed to have gain of, say, 5, the following LM3875 stage will need gain of 4 to 6, depending on your power need. Of course, you need the "magic" trick discussed in this thread to prevent oscillation.
Another thing I¡¯m not sure is, if you have a buffer stage of gain 5 and max 1V(rms) input signal (from a source or preamp), you will have 5V(rms) input to LM3875 chips. The question is, will LM3875 handle this large input signal nicely?
OK, my final question. Does this overall scheme make a sense? In other words, will it sound better than the conventional, unity gain buffer followed by normal gain (20 to 30) LM3875 configuration?
Anyway, I suppose that fine video op amps such as LM6171/2, LM6181/2, LM7372, and AD8021/2 would be good candidates for the buffer/inverter. And I guess I would need a LPF in front of the buffer, otherwise it will amplify some AM radio signal as well.
Take care,
Hi, folks!
I'm a newbie, waking up this dormant thread, heh!
OK, I was thinking (dreaming??) about building a powerful
(possibly multichannel) amp out of these LM3875/3886 things.
Thus, it will naturally go to one of BP(bridged/parallel) configurations.
There were good discussions on BP in other threads, especially about phase-splitting methods. It's claimed that using (good) transformer is the best, but cost of good transformers seems a bit prohibitive. (How much will Jensen transformers cost you if you need six, huh?) So, I'd prefer ways of using not-so-expensive op amps, being inverters and buffers.
Here goes my first question. If you need buffers anyway, is there any definite reason that they should be of unity gain? In other words, if buffers are allowed to have gain of, say, 5, the following LM3875 stage will need gain of 4 to 6, depending on your power need. Of course, you need the "magic" trick discussed in this thread to prevent oscillation.
Another thing I¡¯m not sure is, if you have a buffer stage of gain 5 and max 1V(rms) input signal (from a source or preamp), you will have 5V(rms) input to LM3875 chips. The question is, will LM3875 handle this large input signal nicely?
OK, my final question. Does this overall scheme make a sense? In other words, will it sound better than the conventional, unity gain buffer followed by normal gain (20 to 30) LM3875 configuration?
Anyway, I suppose that fine video op amps such as LM6171/2, LM6181/2, LM7372, and AD8021/2 would be good candidates for the buffer/inverter. And I guess I would need a LPF in front of the buffer, otherwise it will amplify some AM radio signal as well.
Take care,
the 'trick' for I/V
Hi Sayonara,
I bump into this thread, because I am for some time looking for a solution to use opa637 as I/V converter. I currently use opa627BP's and have tried the 637 and the sound much better however they oscillate, which masks the inner details and is specially annoying with music parts with low signal.
I therefor am very curious how your trick works out. I understand the trick as described to solve the instability problems of unity-gain unstable opamps. However in an I/V design, a current is converted to a voltage (which is special kind of amplification) and I don’t know what this implies to your trick. Could you elaborate on this.
I have a TDA1545 with its output connectted to the negative input of the opamp, the feedback consists of a 3k9 resistor paralleled with a 1nF Cap. The postive side of the opamp is at Vref (of the TDA1545), which is not Grnd (but 2,1 V). What is the implication of that.
Thanks Henk
Hi Sayonara,
I bump into this thread, because I am for some time looking for a solution to use opa637 as I/V converter. I currently use opa627BP's and have tried the 637 and the sound much better however they oscillate, which masks the inner details and is specially annoying with music parts with low signal.
I therefor am very curious how your trick works out. I understand the trick as described to solve the instability problems of unity-gain unstable opamps. However in an I/V design, a current is converted to a voltage (which is special kind of amplification) and I don’t know what this implies to your trick. Could you elaborate on this.
I have a TDA1545 with its output connectted to the negative input of the opamp, the feedback consists of a 3k9 resistor paralleled with a 1nF Cap. The postive side of the opamp is at Vref (of the TDA1545), which is not Grnd (but 2,1 V). What is the implication of that.
Thanks Henk
I have thought about this for some time too, gainclone with gain=1, as my soundcard outputs almost 8,5v p-p and thats enough for me. Better to make use of total output signal than amplifying low signal with noise. What if a gainclone was used as a voltage follower (output and negative input shorted with audio signal on positive input)? Could anyone test it?
8.4Vpp is only just a little more than CD which will output between 6Vpp and 7.5Vpp.
With a gain of 1 (+0dB) your speaker level signal will be limited to a maximum of the 0dBfs of your source, meaning about 6.5Vpp to 8.5Vpp.
This is equivalent to ~2.5Vac or about 3/4 of a watt for the maximum signal and between 10mW and 80mW of average level signal.
A headphone amplifier is usually designed to put out more signal than this before clipping.
With a gain of 1 (+0dB) your speaker level signal will be limited to a maximum of the 0dBfs of your source, meaning about 6.5Vpp to 8.5Vpp.
This is equivalent to ~2.5Vac or about 3/4 of a watt for the maximum signal and between 10mW and 80mW of average level signal.
A headphone amplifier is usually designed to put out more signal than this before clipping.
Improving Performance with Decompensated Operational Amplifiers
A chip amp is essentially a decompensated power opamp - the circuits for getting unity gain are shown in figs 3 and 4 (non-inverting and inverting, respectively).
But it doesn't sound like the lower gain is a viable option, in his case, unless he can live with a fraction of a watt of output. Or am I missing something?
And wouldn't the source's noise also be attenuated by the divider? If so, then where does the greater noise come from?
Edit: Oh, from the better feedback situation. Sorry. I wasn't thinking. But he wouldn't be any worse off than a regular chipamp, at least, would he, with an input attenuator and then higher gain?
And wouldn't the source's noise also be attenuated by the divider? If so, then where does the greater noise come from?
Edit: Oh, from the better feedback situation. Sorry. I wasn't thinking. But he wouldn't be any worse off than a regular chipamp, at least, would he, with an input attenuator and then higher gain?
Last edited:
spindre17 is saying he can live with such low output levels. But then again, at such low levels the noise of your suggested approach is likely to be inaudible too
It appears a chipamp's noise is generally limited by its input. The schematics I've looked at show heavily degenerated LTPs (using 3k3 resistors in LM1875's case) as a first stage which tend to mean much higher noise than opamps. So higher signal levels at the input mean lower output noise in general. Attenuating then gaining up will amplify the input noise much more than running at lower gain.
Hi,
I have. My idea after all.
Later I have also made a unity gain stable non-inverting variant:
Fusion Power Multiplier Modules | Diy HiFi Supply
To compare a unity gain stage to one with gain is apples and pears. Normally there is no way to keep the rest or the system the same, as there is usually no way to compensate the 26dB (or more) gain difference
No, the buffer above (and also the inverting unity gain buffer) sounds essentially like what drives it, given a competently implemented powersupply.
Ciao T
I have. My idea after all.
Later I have also made a unity gain stable non-inverting variant:
Fusion Power Multiplier Modules | Diy HiFi Supply
To compare a unity gain stage to one with gain is apples and pears. Normally there is no way to keep the rest or the system the same, as there is usually no way to compensate the 26dB (or more) gain difference
No, the buffer above (and also the inverting unity gain buffer) sounds essentially like what drives it, given a competently implemented powersupply.
Ciao T
Nice, thanks:
From the Conclusion:
The most obvious application to use a decompensated amplifier is in higher gain circuits. In applications with gains of 10 or more, using the THS4021 instead of the THS4011 results in over ten times more bandwidth, noise reduction by a factor of 5, and 1.5 times more slew rate.
To consider unity-gain performance, five different circuits have been tested for distortion, transient response, and noise performance. By comparing the noninverting amplifiers (circuit a versus circuit d) to the inverting amplifiers (circuits b and c versus circuit e), the following conclusions are drawn about using an externally-compensated THS4021 versus using the internally-compensated THS4011.
A tabulated comparison is shown in Table 2. In the inverting amplifiers (circuits b and c versus e), significant improvement in THD performance is seen across the frequencies tested. There was no significant difference between circuits b and c. For the noninverting amplifiers (circuit a versus d), improvement in THD performance is also seen, but it diminishes with frequency, with no advantage seen at 16 MHz.
Transient performance shows mixed results. Slew rate and settling time are somewhat better when comparing the inverting topologies, but appears to be little change for the noninverting amplifier. The noninverting amplifier (circuit a) shows considerable overshoot, which may be undesirable.
If the noise is bandwidth limited, the THS4021 shows better noise performance; but little difference is expected if the full bandwidth of the parts is utilized.
Wow, inverted really seems like the winner!