Split thread??
Maybe it was I who first brought the topic up, but anyway there
has been quite many posts on protection schemes and issues
for amplifiers, which seems a bit off-topic for this particular thread.
So, unless we're done with that subdiscussion I suggest it is
split off as a separate thread.
Maybe it was I who first brought the topic up, but anyway there
has been quite many posts on protection schemes and issues
for amplifiers, which seems a bit off-topic for this particular thread.
So, unless we're done with that subdiscussion I suggest it is
split off as a separate thread.
HI subwo1!What simulate sofware are you using?I use ORCAD,but it can't simulate THD to frequency
Jam,
The schematic you posted wih the 2SK389 and 2SJ109 (supermatched pairs). Have you built one of these? What are its strenghts and its weaknesses? The topology has nice symetry.
The schematic you posted wih the 2SK389 and 2SJ109 (supermatched pairs). Have you built one of these? What are its strenghts and its weaknesses? The topology has nice symetry.
"Series inductors delay NFB node sensing of loudspeaker back emfs. Would need to be no more than 250nH to maintain 0.01% distortion specifications."
Well, it sounded too easy. TINSTAAFL.
Well, it sounded too easy. TINSTAAFL.
Nelson,
That's encouraging. I like the input stage of this as a building block for a more complete design.
That's encouraging. I like the input stage of this as a building block for a more complete design.
Hi Jonathan,
Too busy here at moment to seek out circuits.
Hi subwo1.
Surge gone but still symmetrical. Good.
There is however another potential problem with shunt feedback, in that the user's pre-amp becomes an integral part of the amplifier's NFB circuit. Unlike your virtual signal generator most pre-amps do not have perfectly zero output impedance at all frequencies. Actually few pre-amps are lower than 100R, and have variable Z. Many are 600 ohm, some 10k !
As is quite normal the capacitors in your circuit are obviously affecting low frequency phase, which is causing level shifting over the first two 10Hz cycles. Your simulator is correctly calculating the first cycle waveform distortion. It is though essential that you complete this simulation with an equivalent bass loudspeaker load instead of 8R, just to checK that there is not 'input C' interaction, say around 55Hz with the Ariel design I included.
Compare the bass driver waveform with computer generated loudspeaker input, to the waveform with your amplifier as the source.
The 1kHz simulation is fine. 0.015% amplitude distortion should be acheivable at other frequencies too if you start distortion measurement after the first few cycles of exponential charge settlement.
However we do hear every first cycle as it arises in real time !!!!!!!
The 100kHz trace reveals an initial propagation delay which is very hard to estimate from post 117. To show that this is not significant you need several first cycle distortion measurements with say the Ariel load at maybe 2k, 5k and 10kHz.
A good bipolar amp leading edge will intercept the axis no later than 50nS.
Again check tweeter waveform for computer / amplifier.
These checks show whether the amplifier interacts with the loudspeaker and thereby modifies reproduced tonality.
Don't be shocked by the first cycle crossover induced distortion, (this is why full rangers sound cleaner !) just make sure that your circuit does not introduce changes.
Watch the first two cycles at bass and first 90 degrees at treble, for lagging - leading problems which are the NFB inversions of leading capacitor current in bass and delayed currents at hf.
Cheers for now ........... Graham.
Too busy here at moment to seek out circuits.
Hi subwo1.
Surge gone but still symmetrical. Good.
There is however another potential problem with shunt feedback, in that the user's pre-amp becomes an integral part of the amplifier's NFB circuit. Unlike your virtual signal generator most pre-amps do not have perfectly zero output impedance at all frequencies. Actually few pre-amps are lower than 100R, and have variable Z. Many are 600 ohm, some 10k !
As is quite normal the capacitors in your circuit are obviously affecting low frequency phase, which is causing level shifting over the first two 10Hz cycles. Your simulator is correctly calculating the first cycle waveform distortion. It is though essential that you complete this simulation with an equivalent bass loudspeaker load instead of 8R, just to checK that there is not 'input C' interaction, say around 55Hz with the Ariel design I included.
Compare the bass driver waveform with computer generated loudspeaker input, to the waveform with your amplifier as the source.
The 1kHz simulation is fine. 0.015% amplitude distortion should be acheivable at other frequencies too if you start distortion measurement after the first few cycles of exponential charge settlement.
However we do hear every first cycle as it arises in real time !!!!!!!
The 100kHz trace reveals an initial propagation delay which is very hard to estimate from post 117. To show that this is not significant you need several first cycle distortion measurements with say the Ariel load at maybe 2k, 5k and 10kHz.
A good bipolar amp leading edge will intercept the axis no later than 50nS.
Again check tweeter waveform for computer / amplifier.
These checks show whether the amplifier interacts with the loudspeaker and thereby modifies reproduced tonality.
Don't be shocked by the first cycle crossover induced distortion, (this is why full rangers sound cleaner !) just make sure that your circuit does not introduce changes.
Watch the first two cycles at bass and first 90 degrees at treble, for lagging - leading problems which are the NFB inversions of leading capacitor current in bass and delayed currents at hf.
Cheers for now ........... Graham.
Hi thanh,
I have the MicroCap demo version. I also have LTspice, but don't know how to do distortion measurements with it. My present understanding is that it doesn't do distortion vs. frequency, just fast Furrier. But, LT spice has no limitations on circuit size or node numbers.
I have the MicroCap demo version. I also have LTspice, but don't know how to do distortion measurements with it. My present understanding is that it doesn't do distortion vs. frequency, just fast Furrier. But, LT spice has no limitations on circuit size or node numbers.
Hi Graham
I never suspected a preamplifier could have an output impedance over a few hundred ohms.:-0 I consider a preamp's job to amplify current as well as voltage. Glad to get your thought there.
You touched on a very important point with the computer generated loudspeaker input compared to amplifier input. I really am unfamiliar with this kind of test but get an impression of what you mean.
When you talk about that first cycle, isn't that involving the simple total phase shift of the amplifier? Is it related to absolute phase? I am thinking absolute phase is important because depending on the configuration of the loudspeakers and the use or lack of multi-amping, the fidelity of the sound in an area may be altered. I input a 10mhz 0.1v sine wave and saw the mentioned delay of 50ns for the output signal,
I input a 10mhz 0.1v sine wave and saw the mentioned delay of 50ns for the output signal. Your thought on the forward effect on bass output and the lagging effect on treble output of the feedback dc blocking capacitor makes a lot of sense. I see what you mean about the amp loading effects of multi-way speakers vs. single a driver.
Thanks.
I never suspected a preamplifier could have an output impedance over a few hundred ohms.:-0 I consider a preamp's job to amplify current as well as voltage. Glad to get your thought there.
You touched on a very important point with the computer generated loudspeaker input compared to amplifier input. I really am unfamiliar with this kind of test but get an impression of what you mean.
When you talk about that first cycle, isn't that involving the simple total phase shift of the amplifier? Is it related to absolute phase? I am thinking absolute phase is important because depending on the configuration of the loudspeakers and the use or lack of multi-amping, the fidelity of the sound in an area may be altered. I input a 10mhz 0.1v sine wave and saw the mentioned delay of 50ns for the output signal,
I input a 10mhz 0.1v sine wave and saw the mentioned delay of 50ns for the output signal. Your thought on the forward effect on bass output and the lagging effect on treble output of the feedback dc blocking capacitor makes a lot of sense. I see what you mean about the amp loading effects of multi-way speakers vs. single a driver.
Thanks.
subwo
I like inverting circuits like that one, basically because there are no SS components in the feedback path and also the summing of feedback and input signal is totally passive.
I am aware although that it is more demanding (to the circuit driving it) than conventional topologies.
Regards
Charles
I like inverting circuits like that one, basically because there are no SS components in the feedback path and also the summing of feedback and input signal is totally passive.
I am aware although that it is more demanding (to the circuit driving it) than conventional topologies.
Regards
Charles
Hi subwo1.
If 50nS is all you have for a delay, then that sounds just fine.
If you had a differential input stage with a series input capacitor, then its low frequency amplitude phase shift would be predictable.
However when you have shunt feedback, the NFB is modified not only by pre-amp and interconnecting cable impedance but internal power amplifier gain limitations which determine the effective output impedance and the amplifier's ability to respond. I do not know the gain and NFB degeneration of your circuit.
The series input capacitor will control the small signal input response, but I have suggested more realistic loading to check output capabilities when might drop off unpredictably with current loading that is out of phase with simplistic resistor-voltage observation. This is where you can tell the difference between a hi-fi audio amplifier and a plain ordinary audio frequency amplifier.
A circuit can measure flat from 20-20k and have less than 0.01% distortion with an 8R load, as with the one I posted, but it can still sound worse than a 1% chassis, as illustrated by some unpredictable internal reaction caused by leading or lagging back emf.
You had 0.015% at 1kHz, but over 3% at 10Hz. Simulate with a 1,000uF input capacitor; does the distortion drop by a factor of 100. If yes then it is the finite gain and thus output impedance of the amplifier circuitry that is interacting with the series input capacitor, and loudspeaker loading might make it worse, as with just below resonance with the Ariel. Circa 45Hz.
Maybe you need to go back to the differential input circuitry. They do introduce an extra stage of hf phase change, but if they are good transistors running at reasonable input current, it is the following stages that dominate the stability response.
Cheers for now ............ Graham.
If 50nS is all you have for a delay, then that sounds just fine.
If you had a differential input stage with a series input capacitor, then its low frequency amplitude phase shift would be predictable.
However when you have shunt feedback, the NFB is modified not only by pre-amp and interconnecting cable impedance but internal power amplifier gain limitations which determine the effective output impedance and the amplifier's ability to respond. I do not know the gain and NFB degeneration of your circuit.
The series input capacitor will control the small signal input response, but I have suggested more realistic loading to check output capabilities when might drop off unpredictably with current loading that is out of phase with simplistic resistor-voltage observation. This is where you can tell the difference between a hi-fi audio amplifier and a plain ordinary audio frequency amplifier.
A circuit can measure flat from 20-20k and have less than 0.01% distortion with an 8R load, as with the one I posted, but it can still sound worse than a 1% chassis, as illustrated by some unpredictable internal reaction caused by leading or lagging back emf.
You had 0.015% at 1kHz, but over 3% at 10Hz. Simulate with a 1,000uF input capacitor; does the distortion drop by a factor of 100. If yes then it is the finite gain and thus output impedance of the amplifier circuitry that is interacting with the series input capacitor, and loudspeaker loading might make it worse, as with just below resonance with the Ariel. Circa 45Hz.
Maybe you need to go back to the differential input circuitry. They do introduce an extra stage of hf phase change, but if they are good transistors running at reasonable input current, it is the following stages that dominate the stability response.
Cheers for now ............ Graham.
Chris Wood,
I have built one like Mr.Pass and agree with that it is a very nice sounding unit. I am working on a more complete unit with mosfet cascodes which I will present later, hopefully we will have circuit boards available as well.
Reagrds,
Jam
I have built one like Mr.Pass and agree with that it is a very nice sounding unit. I am working on a more complete unit with mosfet cascodes which I will present later, hopefully we will have circuit boards available as well.
Reagrds,
Jam
Jam, re post #85 : You are on the right track, but I would choose a 50 ohm resistor and use V spec 2SK389-J109, if possible. This input capacitance of the second stage is pretty large, so higher current drive from the source devices is necessary.
Thanks for the hints MrCurl.
I had built a version with doubled up fets. I will try to get some V grade fets. I also had cascoded the fets with bipolars and I will try mosfets next.
Regards,
Jam
I had built a version with doubled up fets. I will try to get some V grade fets. I also had cascoded the fets with bipolars and I will try mosfets next.
Regards,
Jam
I will try to get some V grade fets.
Good Luck............. oh you'll find out. Get me some too if you can. I am not holding my breath though since they are very hard to find. I had to trade some pretty good stuff for my small handful of V J109 and K389s.......
Good Luck............. oh you'll find out. Get me some too if you can. I am not holding my breath though since they are very hard to find. I had to trade some pretty good stuff for my small handful of V J109 and K389s.......
john curl said:
Those output devices are Toshiba's audio MOSFETs. They are designed to have good complementarity (is that a real word?), but otherwise are fairly ordinary vertical devices (similar to the IR parts).
In a source follower configuration, the lateral parts have a much lower input capacitance. I've never looked at this from the standpoint of a common source output stage, but it may be worth exploring.
On a separate note, why not get rid of the differential pair, leaving just the input transistors? The second set of transistors is simply buffering the feedback signal. If this were a preamp or other design where the output stage didn't have the capability of driving the sources of the input stage, that might be a good idea. But presumably if the output stage can drive loudspeaker loads, it should also be able to easily drive the sources of the input pair without the need for another pair of buffer transistors.
Disclaimer: I don't use feedback, so I've never tried this. It just seems like if you are going for simplicity that it would be a good idea to cut your transistor count by 33%.
Re: I will try to get some V grade fets.
Hello Fred,
If you need just 2 of them, send me a PM.
Regards,
Audiofanatic 😉
Fred Dieckmann said:Good Luck............. oh you'll find out. Get me some too if you can. I am not holding my breath though since they are very hard to find. I had to trade some pretty good stuff for my small handful of V J109 and K389s.......
Hello Fred,
If you need just 2 of them, send me a PM.
Regards,
Audiofanatic 😉
There is no such thing as a free lunch
"In a source follower configuration, the lateral parts have a much lower input capacitance. "
With a corresponding decrease in transconductance. To get the same transconductance you would have to parallel several and then your right back to having high input capacitance. I wonder if a comparison, the hexfets might have greater transconductance per unit input capacitance. I'll bet NP would know............ and by how much.
"In a source follower configuration, the lateral parts have a much lower input capacitance. "
With a corresponding decrease in transconductance. To get the same transconductance you would have to parallel several and then your right back to having high input capacitance. I wonder if a comparison, the hexfets might have greater transconductance per unit input capacitance. I'll bet NP would know............ and by how much.
- Status
- Not open for further replies.