Re: IC's and Transformers
We apparently are of similar vintage. Aged cheeses, wines, furniture, and all that.
The most unique "brain fart" amplifier I ever produced was based on the extraordinary linearity of many-transistor high (but not "power") output op-amps working in unity-gain mode, and a bunch of hand-wound (me, when I could still see the filamentary 36 GA wire and quarter millimeter ( 10 mil... for old-timers ) thick itty-bitty EI core sheets.) interstage step-up transformers. Only 3.1 to 1 (square-root of 10), so that each stage was 0.1% within 10 dB gain, and all the gain was passive.
The listening result was exactly as predicted: glorious. Clean, crisp, sweet without harshness. Precisely NO feedback at any stage (except if technically you want to cite the IC OpAmps in unity-gain as NFB-ultimate devices; i just think of them as 'near-perfect' current-gain elements), except for that derived from ratios of transformer windings. No capacitors interstage, no resistors in signal path. Only a bundle of 0.22 ohm load-sharing/leveling resistors on the final discrete bank of complimentary PNP/NPN transistors cobbled into an H-bridge orientation (with their complimentary signal provided by the last interstage transformer, of course!). 400 w/channel, pure bliss. Maybe not a "beast", but boy, did it have dynamic range. As quiet as a mouse, too... absolutely NO hum, hiss, or anything, even wide open. A "soft turn on" 3-stage power supply fixed the turn-on/turn-off thumps.
But that is an amplifier of 40 years gone. Today, even I would do it somewhat different. Maybe capitalize on the self-leveling (current sharing) of MOSFET transistors. Hard to say.
GoatGuy
We apparently are of similar vintage. Aged cheeses, wines, furniture, and all that.
The most unique "brain fart" amplifier I ever produced was based on the extraordinary linearity of many-transistor high (but not "power") output op-amps working in unity-gain mode, and a bunch of hand-wound (me, when I could still see the filamentary 36 GA wire and quarter millimeter ( 10 mil... for old-timers ) thick itty-bitty EI core sheets.) interstage step-up transformers. Only 3.1 to 1 (square-root of 10), so that each stage was 0.1% within 10 dB gain, and all the gain was passive.
The listening result was exactly as predicted: glorious. Clean, crisp, sweet without harshness. Precisely NO feedback at any stage (except if technically you want to cite the IC OpAmps in unity-gain as NFB-ultimate devices; i just think of them as 'near-perfect' current-gain elements), except for that derived from ratios of transformer windings. No capacitors interstage, no resistors in signal path. Only a bundle of 0.22 ohm load-sharing/leveling resistors on the final discrete bank of complimentary PNP/NPN transistors cobbled into an H-bridge orientation (with their complimentary signal provided by the last interstage transformer, of course!). 400 w/channel, pure bliss. Maybe not a "beast", but boy, did it have dynamic range. As quiet as a mouse, too... absolutely NO hum, hiss, or anything, even wide open. A "soft turn on" 3-stage power supply fixed the turn-on/turn-off thumps.
But that is an amplifier of 40 years gone. Today, even I would do it somewhat different. Maybe capitalize on the self-leveling (current sharing) of MOSFET transistors. Hard to say.
GoatGuy
What was the power bandwidth of said amplifier?
How many stages to get the requisite voltage gain?
Did you look at a square wave? How did it look?
Presumably this was before it was easy to do distortion spectra...
On another more philosophical note, it is an interesting mental exercise to consider that one *can* make an amplifier of "n" non-inverting stages, and maintain the same phase polarity throughout. However (and of course) doing the same with "n" inverting amplifiers will not maintain the same phase polarity throughout. Silly, eh?
_-_-
How many stages to get the requisite voltage gain?
Did you look at a square wave? How did it look?
Presumably this was before it was easy to do distortion spectra...
On another more philosophical note, it is an interesting mental exercise to consider that one *can* make an amplifier of "n" non-inverting stages, and maintain the same phase polarity throughout. However (and of course) doing the same with "n" inverting amplifiers will not maintain the same phase polarity throughout. Silly, eh?
_-_-
Member
Joined 2009
Paid Member
Hi Bear, GoatGuy, Sakis,
Thank you for posts.
I'm just asking about this amp; is it possible...... but it would be an audio, with the usual parameters required, 20-20k, 0.02% max, 80W or so, and less than -100dB noise.
I'm throwing an idea to discuss, and I have some ideas but have not resolved it all and I'm not giving detail of course, but I'm trying to find out if this has been in the public domain over the last thirty years. I'm reasonably au fait with the technology, but I may have missed this..... so I ask here.
It's a long shot, and indeed may not be possible, but 110 years ago airplanes were impossible too........
Bear, you are very close. A non-inverting stage can be configured easily with a feedback node - rush cascode. As you say, an interesting philosophical exercise.
Hugh
Thank you for posts.
I'm just asking about this amp; is it possible...... but it would be an audio, with the usual parameters required, 20-20k, 0.02% max, 80W or so, and less than -100dB noise.
I'm throwing an idea to discuss, and I have some ideas but have not resolved it all and I'm not giving detail of course, but I'm trying to find out if this has been in the public domain over the last thirty years. I'm reasonably au fait with the technology, but I may have missed this..... so I ask here.
It's a long shot, and indeed may not be possible, but 110 years ago airplanes were impossible too........
Bear, you are very close. A non-inverting stage can be configured easily with a feedback node - rush cascode. As you say, an interesting philosophical exercise.
Hugh
Last edited:
re: Transformers and Fuzzy Wires
* POWER bandwidth was about 35 Hz to 18 kHz. The limit was entirely due to my hand-wound interstage transformers. Better transformers, more attention to multifilar windings... could have gotten it 20 Hz to 25 kHz, no problem.
* STAGES, I ended up using 4 stages, each at 3.1x per stage, except the interstage transformer leading to the final stage was 2:1 wound. Was trying to get rated power at 0.775V dBm input level, and 3.1x was a bit overdriven.
* SQUARE WAVE looked "as expected". Almost no ringing (a wee bit, very small), and the usual inductor/transformer soft-rise, soft-fall characteristics. Same for low frequency square wave, except slanted tops, due to core-drive saturation. Hence "need better interstage" cores.
* HOW DID IT LOOK - the square wave? Pretty as a button. The unit? Fuzzy, and full of decidedly retro cloth-covered wires. Nasty stuff.
* DISTORTION SPECTRUM - who knows. It was in the era of hand-plots from RMS output measurements using an HP "true RMS" VTVM and a correspondingly irascible calibrated signal generator. My lust for a Tektronix spectrum analyzer and a *******' purpose-built Polaroid camera to capture the spectra ... was palpable. I grew to like green-grid engineering paper though, and Stadler pencils.
GoatGuy
* POWER bandwidth was about 35 Hz to 18 kHz. The limit was entirely due to my hand-wound interstage transformers. Better transformers, more attention to multifilar windings... could have gotten it 20 Hz to 25 kHz, no problem.
* STAGES, I ended up using 4 stages, each at 3.1x per stage, except the interstage transformer leading to the final stage was 2:1 wound. Was trying to get rated power at 0.775V dBm input level, and 3.1x was a bit overdriven.
* SQUARE WAVE looked "as expected". Almost no ringing (a wee bit, very small), and the usual inductor/transformer soft-rise, soft-fall characteristics. Same for low frequency square wave, except slanted tops, due to core-drive saturation. Hence "need better interstage" cores.
* HOW DID IT LOOK - the square wave? Pretty as a button. The unit? Fuzzy, and full of decidedly retro cloth-covered wires. Nasty stuff.
* DISTORTION SPECTRUM - who knows. It was in the era of hand-plots from RMS output measurements using an HP "true RMS" VTVM and a correspondingly irascible calibrated signal generator. My lust for a Tektronix spectrum analyzer and a *******' purpose-built Polaroid camera to capture the spectra ... was palpable. I grew to like green-grid engineering paper though, and Stadler pencils.
GoatGuy
Member
Joined 2009
Paid Member
Hugh,
You are to a degree confusing me here.
Do you want to design an amp that has
a) minimal phase shift from IP to OP or are you
b) trying to design an amp comprised of individual stages that are non
inverting themselves to have some kind of positive effect WRT when the
feedback loop is closed?
Either way, I would suggest, looking at your criterion, just build an amp
that is open loop, right from IP to OP. As such you can make the amp
blindingly fast without any requirement for compensation and as such have
extremely low phase shift - if that is your design criterion.
I can give you details of a previous amp I worked on which is SS / tube
hybrid, which will have very low phase shift and distortion within the
constraints of being open loop and maybe that will meet your criteria.
It all depends what you are trying to achieve. Those distortion numbers you
have quoted are quite achievable with open loop design, the only gotcha is
for decent sound and open loop OP stage you will have to run more than usual
bias. I'm not talking crazy class A bias currents but certainly around 0.5 to 1
amp.
Give me a call and we can discuss.
PS yes I know I have been hard to get hold of, lots going on ATM as you are
well aware
cheers
Terry
Is a common emitter stage _really_ an inverter?
While I think the premise of this thread is interesting and has the potential to provoke some good ideas, I think there is another way to look at the inverting stage(s) in an amplifier.
They are only inverting if you consider the output signal to be a voltage. But at a more fundamental level, the transistor or tube or whatever it is is acting as a transconductor. Voltage in, current out. You can assign a polarity to the output current in either way you like. You can prevent an inverted phase voltage from appearing directly on the collector or the plate with a cascode.
So, in that sense, I don't really consider that the inherent phase inversion really counts as a phase _shift_, because it's really an ideal sign change: a topological artefact but not a nonideality.
That said, it is true that a non-cascoded CE stage has some inherent phasey negative feedback from collector to base. But again, if you load it with that low impedance cascode, you can consider that you are just isolating that current output whose sign you can assign either way.
Finally, in fact, you get some linearity benefit from the inherent negative feedback from C to E, especially if you use it for a big Miller cap.
Still, I like the idea from a design challenge POV, constraints make things interesting.
While I think the premise of this thread is interesting and has the potential to provoke some good ideas, I think there is another way to look at the inverting stage(s) in an amplifier.
They are only inverting if you consider the output signal to be a voltage. But at a more fundamental level, the transistor or tube or whatever it is is acting as a transconductor. Voltage in, current out. You can assign a polarity to the output current in either way you like. You can prevent an inverted phase voltage from appearing directly on the collector or the plate with a cascode.
So, in that sense, I don't really consider that the inherent phase inversion really counts as a phase _shift_, because it's really an ideal sign change: a topological artefact but not a nonideality.
That said, it is true that a non-cascoded CE stage has some inherent phasey negative feedback from collector to base. But again, if you load it with that low impedance cascode, you can consider that you are just isolating that current output whose sign you can assign either way.
Finally, in fact, you get some linearity benefit from the inherent negative feedback from C to E, especially if you use it for a big Miller cap.
Still, I like the idea from a design challenge POV, constraints make things interesting.
Zen (Terry),
Yes, an amp with no stages with any inversion, AND very low phase shift input to output, and of course, with global feedback.
Question: is there a connection between zero phase shift and good sonics? I have some doubts and that anyone can definitely answer that........
However, since minimal phase shift around the crossover of a speaker makes a difference to imaging, perhaps there is a good reason to try this.
Next, if a non-inverting series of stages can be created, we would clearly be needing a grounded base VAS. Such topologies have blindingly speed.
Last, since fb necessarily needs a lag compensation system to pull the loop gain to unity by the Bode pole, how could we easily stabilise it? Would it be so fast in the early stages that the slowest outputs would keep it stable without compensation? And would shunt compensation be the best?
Cheers,
Hugh
Yes, an amp with no stages with any inversion, AND very low phase shift input to output, and of course, with global feedback.
Question: is there a connection between zero phase shift and good sonics? I have some doubts and that anyone can definitely answer that........
However, since minimal phase shift around the crossover of a speaker makes a difference to imaging, perhaps there is a good reason to try this.
Next, if a non-inverting series of stages can be created, we would clearly be needing a grounded base VAS. Such topologies have blindingly speed.
Last, since fb necessarily needs a lag compensation system to pull the loop gain to unity by the Bode pole, how could we easily stabilise it? Would it be so fast in the early stages that the slowest outputs would keep it stable without compensation? And would shunt compensation be the best?
Cheers,
Hugh
A polarity inversion is not in any sense a phase shift; there's no frequency-dependent behavior. Within an overall feedback loop, a polarity inversion does indeed contribute to the overall phase shift by adding 180 degrees at all frequencies.
We want linear phase behavior (equivalent to a pure time delay) to preserve waveform shape, and zero phase is just a special case.
Member
Joined 2009
Paid Member
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.