Hi atmasphere,
Thank you for the link sir, very kind of you.
A motorized variac control is neat for sure, and low noise. We also use variacs to dim lights in recording studios. Bloody big variac though! These days I think you could find a more elegant way to do this.
A slew rate of 600V/uS is a feat, however having bandwidth well outside your signal passband can cause way more problems than it solves. I also have a test and measurement and telecom background. I would limit the bandwidth to some reasonable value. It may also allow you to lower some currents in areas. You are not going to slew rate limit ever, the super high cut off frequency honestly doesn't buy you anything. Maybe ad copy - but that's it. I also don't believe you need anywhere near that much capacitance in the supply. On the input filter section it merely reduces the conduction angle of your rectifiers and runs the peak currents sky high - for nothing. You say it is also regulated and LF ripple is much easier to get rid of than the higher frequency components very high capacitance gives you. Vregs are merely amplifiers with a steady state reference voltage and error amp. As frequency goes up, rejection drops due to the reduced OL gain available. That and capacitance across the pass element and into the circuit (depending on many things).
Output protection. Must have in my book (not a crowbar, although I have seen that before too!). There should always be an effective method to remove energy from the load in a fault condition. Fuses are generally slow acting for overloads unless they are very high (shorted cap or rectifier for example). Not fast enough for a speaker voice coil. They can loosen the glue in the VC, causing a buzz at certain frequencies that may not be noticed for quite some time. You're talking total, catastrophic failure. Several manufacturers of amplifiers have said just as you have, but damage isn't always apparent soon enough "to count". I did speaker warranty for several brands as well, and studied the failure modes.
You do make one of the best OTL amps on the market. That still isn't the direction I would go knowing what I do know.
Thanks for getting back to the Croft.
Thank you for the link sir, very kind of you.
A motorized variac control is neat for sure, and low noise. We also use variacs to dim lights in recording studios. Bloody big variac though! These days I think you could find a more elegant way to do this.
A slew rate of 600V/uS is a feat, however having bandwidth well outside your signal passband can cause way more problems than it solves. I also have a test and measurement and telecom background. I would limit the bandwidth to some reasonable value. It may also allow you to lower some currents in areas. You are not going to slew rate limit ever, the super high cut off frequency honestly doesn't buy you anything. Maybe ad copy - but that's it. I also don't believe you need anywhere near that much capacitance in the supply. On the input filter section it merely reduces the conduction angle of your rectifiers and runs the peak currents sky high - for nothing. You say it is also regulated and LF ripple is much easier to get rid of than the higher frequency components very high capacitance gives you. Vregs are merely amplifiers with a steady state reference voltage and error amp. As frequency goes up, rejection drops due to the reduced OL gain available. That and capacitance across the pass element and into the circuit (depending on many things).
Output protection. Must have in my book (not a crowbar, although I have seen that before too!). There should always be an effective method to remove energy from the load in a fault condition. Fuses are generally slow acting for overloads unless they are very high (shorted cap or rectifier for example). Not fast enough for a speaker voice coil. They can loosen the glue in the VC, causing a buzz at certain frequencies that may not be noticed for quite some time. You're talking total, catastrophic failure. Several manufacturers of amplifiers have said just as you have, but damage isn't always apparent soon enough "to count". I did speaker warranty for several brands as well, and studied the failure modes.
You do make one of the best OTL amps on the market. That still isn't the direction I would go knowing what I do know.
Thanks for getting back to the Croft.
That's OK about the lifting- I broke my back so I'm not a fan of heavy amps either.
The reason you want an amplifier with that kind of speed is so it will follow the incoming waveform with greater accuracy.
When you have lots of feedback the differences between tube and solid state go away. Back in the early 70s though, no amplifier made had feedback applied correctly so it was adding higher ordered harmonics of its own due to non-linearities at the feedback node (Norman Crowhurst was writing about this in the late 1950s). The result is such amplifiers tend to sound brighter and harsher compared to real life because the ear interprets the added higher ordered harmonics that way.
All of our OTLs are bandwidth limited by the voltage amplifier. The reason to use high capacitance in the output is because it reduces IMD, especially at higher power levels, if the amplifier is zero feedback. If the amplifier employs enough feedback, its able to correct and so not so much power supply capacitance is needed.
Hi atmasphere,
Okay, so that is the slew rate of the output stage. You certainly do want to limit the frequency response at the start!
You said you want the high slew rate so it can follow the input signal more accurately. You do know this is not factually true. The only signal details outside of the passband will be noise. You only need a touch higher slew rate to ensure the following stage(s) can accurately follow the input signal. That is a fact.
Feedback reduces distortion given a design without serious defects (impairments). As you approach higher fidelity, the technology used becomes irrelevant. That's why tube or solid state designs sound more alike as they get better. This is a very good thing. They also sound more "musical", anything else is a form of distortion (by definition). Like a tone control or effects pedal or unit.
If each piece in a system does its job properly, the entire system will perform as good as it can. You don't solve an issue in one part by introducing an impairment in another. You solve issues where they live. Right from the recording microphone on out to the loudspeaker.
Okay, so that is the slew rate of the output stage. You certainly do want to limit the frequency response at the start!
You said you want the high slew rate so it can follow the input signal more accurately. You do know this is not factually true. The only signal details outside of the passband will be noise. You only need a touch higher slew rate to ensure the following stage(s) can accurately follow the input signal. That is a fact.
Feedback reduces distortion given a design without serious defects (impairments). As you approach higher fidelity, the technology used becomes irrelevant. That's why tube or solid state designs sound more alike as they get better. This is a very good thing. They also sound more "musical", anything else is a form of distortion (by definition). Like a tone control or effects pedal or unit.
If each piece in a system does its job properly, the entire system will perform as good as it can. You don't solve an issue in one part by introducing an impairment in another. You solve issues where they live. Right from the recording microphone on out to the loudspeaker.
Actually its the slew rate of the driver and output section. Our first prototypes had full bandwidth in the voltage amplifier too, and we found by attaching an antenna to the speaker terminals that the amp made a competent power amp for a 27MHz radio transmitter. Obviously you'' burn up tweeters with that sort of bandwidth in an audio amplifier 😉
The high speed of the output isn't something we particularly tried for- the tubes are inherently wide bandwidth. But that sort of speed is helpful in reducing distortion in the output section. You certainly don't want it slower than the voltage amplifier!
I agree that as you get better fidelity, the kind of amp is less important. If you were to do a search on my posts on the web, you'll find that I've frequently said that class D has put tube power amps on borrowed time. Why bother with the heat and reliability of tubes when you can get better sound elsewhere? The only reason I can think of is doing it for the fun of it. That's why it could be fun getting something like this little Croft going, but if it needs more than just the right driver tubes and filter caps, I think it might have to be traced out. I've not found any schematics online.
The high speed of the output isn't something we particularly tried for- the tubes are inherently wide bandwidth. But that sort of speed is helpful in reducing distortion in the output section. You certainly don't want it slower than the voltage amplifier!
I agree that as you get better fidelity, the kind of amp is less important. If you were to do a search on my posts on the web, you'll find that I've frequently said that class D has put tube power amps on borrowed time. Why bother with the heat and reliability of tubes when you can get better sound elsewhere? The only reason I can think of is doing it for the fun of it. That's why it could be fun getting something like this little Croft going, but if it needs more than just the right driver tubes and filter caps, I think it might have to be traced out. I've not found any schematics online.
Yes, I agree that there is a lot of fun dealing with tubes. That's why I do it. My go-to system is solid state. The power required to have the same in tube amplification is very high, but for less than 80 wpc it isn't too bad. Mind you, I believe in one pair of output tubes per channel.
Once class D gets there, I may be interested. But not yet. It works for subwoofers and non-critical systems (TV's and such), but certainly not for full range listening.
One thing about class D. The driver chips keep being discontinued. That and they tend to be a secret, meaning they are throw away. I disagree strongly with that concept. We as a world community cannot afford it. This is not sustainable.
Once class D gets there, I may be interested. But not yet. It works for subwoofers and non-critical systems (TV's and such), but certainly not for full range listening.
One thing about class D. The driver chips keep being discontinued. That and they tend to be a secret, meaning they are throw away. I disagree strongly with that concept. We as a world community cannot afford it. This is not sustainable.
I've heard some class D amps that are perfectly fine running full range and sound as smooth and relaxed as tubes doing so- but are more transparent.
Driver chips do change- but they are often drop in replacements as improvements are made. We've had 3 types of driver chips used in our design so far and not had to change the board or design to accommodate them. We've seen the same thing with comparators. Tubes are throw-away when they fail and there's a lot more being tossed out in them. So I see class D as less problematic in that regard.
So I wonder after all of this prognostication how many of you have actually spent any time at all with an OTL amplifier with appropriate speakers?
There are likely many solutions to any given problem based on the attributes most important to the person making the decisions.
I have only worked a single Croft product, one of his pre-amplifiers - I was impressed by the build quality and what appeared to be a clever design. I don't know whether the Croft is a circlotron, inverted futterman or futterman amp. Is there an output coupling cap?
I would say that the amplifier is worth keeping if you are willing to pair it with something compatible.
High quality Autotransformers can be built to reflect a more reasonable load to an OTL amplifier. (And have comparatively few downsides if well designed.)
Decades ago I remember hearing a pair of Atmosphere 100W monos with a pair of the smaller Apogees (the Duetta?) - anyway they left a positive impression that has lasted to this day.
And yes I wiped out a woofer with my experimental 6C33 circlotron OTL 25 years ago, but when working it sounded better than any of my output transformer equipped amps at the time. No missing gears.
There are likely many solutions to any given problem based on the attributes most important to the person making the decisions.
I have only worked a single Croft product, one of his pre-amplifiers - I was impressed by the build quality and what appeared to be a clever design. I don't know whether the Croft is a circlotron, inverted futterman or futterman amp. Is there an output coupling cap?
I would say that the amplifier is worth keeping if you are willing to pair it with something compatible.
High quality Autotransformers can be built to reflect a more reasonable load to an OTL amplifier. (And have comparatively few downsides if well designed.)
Decades ago I remember hearing a pair of Atmosphere 100W monos with a pair of the smaller Apogees (the Duetta?) - anyway they left a positive impression that has lasted to this day.
And yes I wiped out a woofer with my experimental 6C33 circlotron OTL 25 years ago, but when working it sounded better than any of my output transformer equipped amps at the time. No missing gears.
Last edited:
Hi Kevin,
I have heard several OTL amps and worked on a few over the years.
They sounded okay on the right speakers, but not good enough for the costs.
I have heard several OTL amps and worked on a few over the years.
They sounded okay on the right speakers, but not good enough for the costs.
IME, OTLs are usually less expensive than the amps they beat. I think that's because a good output transformer that really has bandwidth is also really expensive.
It is. The last tube amp I designed had transformers designed by Hammond. Boy, they are great! However there is a price attached to quality as you've noted.
However all tube amplifiers are much more expensive to make than an equivalent transistor based amplifier. The AB class transistor amplifiers should have much lower maintenance costs.
However all tube amplifiers are much more expensive to make than an equivalent transistor based amplifier. The AB class transistor amplifiers should have much lower maintenance costs.
This tread is very interesting!
It looks like we need to throw out all the amplifiers that are from the early 70s, and all of them before that (50s, and 60s).
Note: Please send me all those amplifiers you are throwing out.
Note 2: If you want to hear lots of upper harmonic distortion, just put 50 Watts into the average loudspeaker that you have.
Please send me those bad loudspeakers too.
The above comments are just my plans of how to get good products for free.
It looks like we need to throw out all the amplifiers that are from the early 70s, and all of them before that (50s, and 60s).
Note: Please send me all those amplifiers you are throwing out.
Note 2: If you want to hear lots of upper harmonic distortion, just put 50 Watts into the average loudspeaker that you have.
Please send me those bad loudspeakers too.
The above comments are just my plans of how to get good products for free.
Yes: opamps and also certain class D amps of the self-oscillating variety. Also our smaller OTLs which employed a small amount of feedback.
If the feedback is arriving at a tube, transistor or MOSFET at the input of the amp, that device is distorting the feedback signal. So the feedback signal isn't able to do its job properly. Norman Crowhurst wrote about this problem in the late 1950s but very little was done about it...
If instead you mix the feedback signal using a resistive divider network, as seen in opamp circuits, the feedback signal is far less distorted! This is a good part of why opamps can be so neutral (also newer opamps have pretty impressive Gain Bandwidth Product values).
Hi atmasphere,
Yes, I have read a lot of stuff written by Norman Crowhurst and I think i have it all spinning around on my hard drive.
-Chris
That is not always true at all, and no I am not going to debate it. It depends on circuit topology and device type.
Yes, I have read a lot of stuff written by Norman Crowhurst and I think i have it all spinning around on my hard drive.
-Chris
😉 Then you must know of a device that is perfectly linear. I don't. Baxandall says you just add more feedback to get around this problem; I think he's full of beans on this one.
No actually Baxandall is correct, whether or not that is something you want to do in an audio circuit is another question. Automated precision voltage and current measurements rely on amplifier circuitry with very high open loop gains in order to provide sufficient feedback margin to make the amplifier gain and linearity errors irrelevant and to allow external passive components to determine the circuit gain accuracy and overall linearity. (Think of all of the ATE used to confirm the quality of semiconductors and the lab equipment used in the development of those parts and for confirming the accuracy and precision of the production test equipment used to test it.)
You are correct Kevin, completely and utterly correct.
In addition, the differential pair is inherently more linear than a single device, especially if those parts are tightly matched and at the same temperature. We are talking in general, not just audio as the requirements for audio are far less stringent than for some test equipment and other things.
In addition, the differential pair is inherently more linear than a single device, especially if those parts are tightly matched and at the same temperature. We are talking in general, not just audio as the requirements for audio are far less stringent than for some test equipment and other things.
I don't argue what you're saying here. That's how opamps work. But none of what you are saying addresses what I was talking about so this is a bit of a red herring.