I have done some of my best simulations in my head... works best with some good weed though... draw a blank otherwise
Simulations don't always work . In trying to find something simple for students to build there is a network that uses transistors connected in the power supply rails of an operational amplifier. The problem was that the operational amplifier model in the simulator didn't include AC characteristics in the power supply rails hence couldn't be simulated.
Below is a link for an amplifier design 50 years ago that was used in numerous motional feedback subs. It still looks good by todays standards, except for that electrolytic capacitor in the feedback path looking highly suspect as perhaps the limiter in its performance. The question is if things have really improved that much?
http://espacenomeutente.altervista.org/wp-content/uploads/2020/04/RE_Mar-Apr1973.pdf
Simulations don't always work . In trying to find something simple for students to build there is a network that uses transistors connected in the power supply rails of an operational amplifier. The problem was that the operational amplifier model in the simulator didn't include AC characteristics in the power supply rails hence couldn't be simulated.
Below is a link for an amplifier design 50 years ago that was used in numerous motional feedback subs. It still looks good by todays standards, except for that electrolytic capacitor in the feedback path looking highly suspect as perhaps the limiter in its performance. The question is if things have really improved that much?
http://espacenomeutente.altervista.org/wp-content/uploads/2020/04/RE_Mar-Apr1973.pdf
Last edited:
You mean have things improved from this?
Yes. Yes they have. See my Modulus amps or the Wolverine for examples.
Tom
Yes...from DBT only ....?
Haven't looked at it for years, though still an interesting design. The output "triple" creates the opportunity to regulate the input network. Not sure why he didn't split D1 into 2 zeners at 18V each split referenced to ground. This would isolate the power supply caps from affecting the input network. You know... kill all that DA...
Haven't looked at it for years, though still an interesting design. The output "triple" creates the opportunity to regulate the input network. Not sure why he didn't split D1 into 2 zeners at 18V each split referenced to ground. This would isolate the power supply caps from affecting the input network. You know... kill all that DA...
Last edited:
I'd love to have an EM field solver (and person to run it) for my work. All I can do without spending thousands on field solver software that I wouldn't have time to learn is to visualize current densities in my head. It sure would be nice to be able to check my hypotheses.
True that. LTspice and others also assume that a BC547B is a BC547B never mind that a BC547B may be subtly (but quantifiably) different from manufacturer to manufacturer. Then add behavioural models like the LM3886 and many (but not all) bets are off.
All models are wrong but some are useful.
Tom
There are free EM simulators eg Quickfield. They have ropey user interfaces but generate accurate results. High end tools are very expensive.
OT posts have been moved to a new thread:
https://www.diyaudio.com/community/threads/measuring-the-imaginary.409662/
Hugo
Moderator team
https://www.diyaudio.com/community/threads/measuring-the-imaginary.409662/
HugoModerator team
The other one you might want to look at is FEMM by David Meeker. I forgot to mention, Quickfield and FEMM produce 2D sim results, so its a slice through an EM field.
Would be very nice if KiCAD dropped the attempt at integrating a spice simulator (there are plnety of good simulators out there - LTspice is not to be sniffed at since it comes with a very good library and support - it is not a hobbyist toy as someone alluded to earlier) and instead hooked to someone who could help them develop a EM simulator to run on a PCB layout. That would be an absolute breakthrough product for individuals and small organisations who can't afford professional packages that can do this stuff. It's computationally very complex because all of these EM simulators use finite element analysis for 2D simulations but if you want a 3D sim it will involve tensor calculus on top of that. It needs someone with a very big brain.
When the Dan Meyer amp was published in 1973 it was still the dark ages of solid-state amp design. Practitioners didn't start getting a grip on things until the early 1990's, although Cordell was light years ahead of most. Each half of the LTPs are being run at 425uA (too low in my opinion) and primary compensation is very heavy-handed VAS shunt loading via R15 and C3 which both Self and Cordell talk about avoiding.
Keep in mind as well, that there have been huge improvements in components, especially output and driver transistors, so you get high fT (30 MHz) and superb SOA together which the guys in 1973 did not have. Nowadays, PSU components are far cheaper (and better) as are OP devices, so CFP OPSs aren't needed.
To gauge how much things have changed, consider the distortion vs frequency graph above. At 20kHz, a modern amp that follows the Self and Cordell guidelines will be about 8000x lower distortion, and at 1kHz, it will be about 6000x better; 19+20 kHz 1 kHz IMD residuals are typically -95 dB in a modern amp, with some designs 20 dB better than that. most of these improvements have come about because of a better understanding of the distortion mechanisms and a more thorough understanding of how to apply feedback.
Keep in mind as well, that there have been huge improvements in components, especially output and driver transistors, so you get high fT (30 MHz) and superb SOA together which the guys in 1973 did not have. Nowadays, PSU components are far cheaper (and better) as are OP devices, so CFP OPSs aren't needed.
To gauge how much things have changed, consider the distortion vs frequency graph above. At 20kHz, a modern amp that follows the Self and Cordell guidelines will be about 8000x lower distortion, and at 1kHz, it will be about 6000x better; 19+20 kHz 1 kHz IMD residuals are typically -95 dB in a modern amp, with some designs 20 dB better than that. most of these improvements have come about because of a better understanding of the distortion mechanisms and a more thorough understanding of how to apply feedback.
Oh no! You just told me the engineers who designed the wonderful wonderful midd eighties Luxman amps Ive heard didnt have a grip. Glad I didnt get around to buy one yet, but at a loss about what to do with other 80ies stuff I have. Sell as vintage I guess?
Cheers!
Btw I jumped from page 8 to 17 of this thred to see if it improved. Glad to see it did. Kudos to admins, if that is allowed.
ANP125
Oh...the thread !... I suggest to replace that awful electrolytic cap in the feedback path of the Tiger... perhaps a bigger non-polarized one with less DA.
ANP125 This thread link contains Figure 1: Auditory sensation area (DIN 45630) indicating the resolution required to achieve the performance limits necessary for all auditory perception. Advancements in general don't necessarily translate into improvement in auditory perception or enjoyment... except for the advancements in the tetrahydrocannabinol content of hemp capacitors. Unfortunately they don't blow as much as they used too.
Getting to the Tiger .01, as you indicated the gain/bandwidth and SOA of devices has improved dramatically over the years. Yet this is to a degree mitigated by nested feedback looping in his so called "triple" output stage, being realized as an internal CFA amplifier with a gain of 3 that is reduced to unity at high frequencies by C6 an C7 across R30 and R31, and returned back to ground through R28 and R29. Although the primary compensation R15 and C3 seems heavy handed this network is returned to ground, becoming a "low" source impedance for the input of the output CFA, as a critical condition for the high frequency performance of this network. It is actually brilliant, as likely responsible for the smooth highs as he characterized the amplifier, when connected to highly resolved electrostatic panels and headphones.
I don't know how Nelson Pass ever managed to pawn off those embarrassing sounding Thresholds back in the 70's? If only he had focused on the first watt or two instead of the rest his descent into obscurity into the annals of history might not have been so tragic...don't you think?
Oh...the thread !... I suggest to replace that awful electrolytic cap in the feedback path of the Tiger... perhaps a bigger non-polarized one with less DA.
ANP125 This thread link contains Figure 1: Auditory sensation area (DIN 45630) indicating the resolution required to achieve the performance limits necessary for all auditory perception. Advancements in general don't necessarily translate into improvement in auditory perception or enjoyment... except for the advancements in the tetrahydrocannabinol content of hemp capacitors. Unfortunately they don't blow as much as they used too.
Getting to the Tiger .01, as you indicated the gain/bandwidth and SOA of devices has improved dramatically over the years. Yet this is to a degree mitigated by nested feedback looping in his so called "triple" output stage, being realized as an internal CFA amplifier with a gain of 3 that is reduced to unity at high frequencies by C6 an C7 across R30 and R31, and returned back to ground through R28 and R29. Although the primary compensation R15 and C3 seems heavy handed this network is returned to ground, becoming a "low" source impedance for the input of the output CFA, as a critical condition for the high frequency performance of this network. It is actually brilliant, as likely responsible for the smooth highs as he characterized the amplifier, when connected to highly resolved electrostatic panels and headphones.
Last edited:
I remember hearing my first Threshold at the now defunct "Audio Centre" and thinking to myself "I wonder how much I can get for mom?" But digging her up wasn't worth the effort.
The Tiger amp is plagued by Miller capacitance in the VAS, driver, and output stages.
Ed
Those 1980’s and 1980’s amps looked good, but by modern standards were highly compromised. The problem as I remarked was to do both with available components and misunderstandings around the use of feedback.