Hello -
I found another detailed schematic for a zero feedback power amp. This was the Wingate that enjoyed a brief run in the late '80s. It was advertised as a class-A, all MOSFET, zero-feedback design. It had all of the buzzwords of the day and, if I recall correctly, received a very enthusiastic review from The Absolute Sound. However, it was only sold direct and the company failed after a short time (as all such companies tend to do).
Since the original thread was so cluttered, I chose to post this separately. It looks like a nice design. Certainly very original and not at all like the "app-note specials" you see from some companies. I don't think I'd agree with his use of the term "current mirrors" in the first stage, but I don't really want to start another argument about semantics!!
http://patft.uspto.gov/netacgi/nph-...t&s1=4688002.WKU.&OS=PN/4688002&RS=PN/4688002
Enjoy,
Charles Hansen
I found another detailed schematic for a zero feedback power amp. This was the Wingate that enjoyed a brief run in the late '80s. It was advertised as a class-A, all MOSFET, zero-feedback design. It had all of the buzzwords of the day and, if I recall correctly, received a very enthusiastic review from The Absolute Sound. However, it was only sold direct and the company failed after a short time (as all such companies tend to do).
Since the original thread was so cluttered, I chose to post this separately. It looks like a nice design. Certainly very original and not at all like the "app-note specials" you see from some companies. I don't think I'd agree with his use of the term "current mirrors" in the first stage, but I don't really want to start another argument about semantics!!
http://patft.uspto.gov/netacgi/nph-...t&s1=4688002.WKU.&OS=PN/4688002&RS=PN/4688002
Enjoy,
Charles Hansen
Can't say as I know any IC designers that would call that a current mirror... 
This configuration is not so different from the Hawk Audio A18 non loop/interstage feedback design, which has been cited in other posts at this forum, including a recent one in the DC Servo thread.
The block diagram labelling is curious, because what's called the voltage gain block is where it should appear, from the output of the OTA formed by the complementary input/level shift transistor pairs, but what isn't shown is a gain set resistor to ground. A little obfuscation, perhaps? Otherwise that block appears to merely be the bias stage, though this makes me wonder what FETs they were using on the output, as the thermal feedback mechanism necessary for VMOS appear to be lacking. But with lateral style FETs like Hitachi or Semelab, they'd be pretty much unnecessary.
BTW, the A18 is a pretty nice sounding little amp, if a bit low powered.
Best regards,
Jon
This configuration is not so different from the Hawk Audio A18 non loop/interstage feedback design, which has been cited in other posts at this forum, including a recent one in the DC Servo thread.
The block diagram labelling is curious, because what's called the voltage gain block is where it should appear, from the output of the OTA formed by the complementary input/level shift transistor pairs, but what isn't shown is a gain set resistor to ground. A little obfuscation, perhaps?
Otherwise that block appears to merely be the bias stage, though this makes me wonder what FETs they were using on the output, as the thermal feedback mechanism necessary for VMOS appear to be lacking. But with lateral style FETs like Hitachi or Semelab, they'd be pretty much unnecessary. BTW, the A18 is a pretty nice sounding little amp, if a bit low powered.
Best regards,
Jon
Steve, nobody wants to have stupid semantic arguments with you. You already have hijacked one thread with your silly arguments. If you want to argue about the definition of feedback, please go back to your other silly argument thread. You have already been warned about this once:
http://www.diyaudio.com/forums/showthread.php?postid=332964#post332964
Now kindly go away.
http://www.diyaudio.com/forums/showthread.php?postid=332964#post332964
Now kindly go away.
JonMarsh said:but what isn't shown is a gain set resistor to ground. A little obfuscation, perhaps?
Hello Jon,
As the mandrill says in "Lion King", loook haardeer....
R23 and R24 are the gain set resistors. I agree that the terminology is confusing. He uses resistors as current-to-voltage converters and calls it a "voltage gain" stage! Again, I wouldn't call it that, but I don't want to start another semantic argument.
He says in the text of the patent that all the MOSFETs are Hitachis. We can assume the outputs are the TO-3 package parts, both from the mechanical drawings of oval output devices and from the power requirements. He explicitly uses the smaller TO-220 parts on the voltage regulators, but we are left to guess which ones he uses on the input transistors. I would guess the smaller TO-220 parts there as well. His text also describes how the heat from the output devices will heat up the input devices, reducing current there and hence to the output to provide a very slow thermal feedback mechanism.
Ciao,
Chas
Ah, OK! I didn't notice the ground tap between R23 and R24, and I haven't read the patent text yet, just flipped through it and printed out the pretty pictures.
Mea culpa!
But it's a goofy way to draw the schematic!
Voltage gain, schmoltage gain, OTA's gotta have an current to voltage conversion somewhere...
This is the first time I've heard of the Wingate amplifier, and this patent. But then I've not had the interest in researching prior art that someone involved in commercial design should have. You might be right that Hawk Audio had a close look to this before developing the A18, but I suppose only Jacco Dekker can say for certain...
There was another guy who did a number of amplifier designs in the early 80's, reviewed frequently at Absolute Sound, probably Sterehpile, too, who's last design was supposedly NFB, but ran hotter than a pistol, Rappaport was his last name, I'm thinking Andy was his first, but that could be wrong. Any idea what that one was about? Also did the Amp-1 and Pre-1.
~Jon
Mea culpa!
But it's a goofy way to draw the schematic!
Voltage gain, schmoltage gain, OTA's gotta have an current to voltage conversion somewhere...
This is the first time I've heard of the Wingate amplifier, and this patent. But then I've not had the interest in researching prior art that someone involved in commercial design should have. You might be right that Hawk Audio had a close look to this before developing the A18, but I suppose only Jacco Dekker can say for certain...
There was another guy who did a number of amplifier designs in the early 80's, reviewed frequently at Absolute Sound, probably Sterehpile, too, who's last design was supposedly NFB, but ran hotter than a pistol, Rappaport was his last name, I'm thinking Andy was his first, but that could be wrong. Any idea what that one was about? Also did the Amp-1 and Pre-1.
~Jon
If I were to attempt a ZFB headphone amplifier, it would look kind of like this.
(I think I need more time to make this readable.)
I'll confess, this is Borbely's Super Buffer with the gain stages re-tuned to eliminate the global feedback. It should have a gain of 16. I don't plan to actually make it, but am toying with the idea of ZFB.
JF
(I think I need more time to make this readable.)
I'll confess, this is Borbely's Super Buffer with the gain stages re-tuned to eliminate the global feedback. It should have a gain of 16. I don't plan to actually make it, but am toying with the idea of ZFB.
JF
Attachments
Try it John, you might really like how it sounds. And it's not too hard to get fairly good "numbers" in a conventional sense at the voltages and currents they require.
There are less problems to deal with for preamps and headphone amps than a full blown power amp. Stability with capacitive loads, for instance.
But then the problems that occur in a power amp output stage and how they propagate through a feedback loop are a big reason, I suspect, for why the benefit can be quite high in that application. I certainly started developing some strong suspicions twenty years ago after listening to one of my late 70's solid state designs against a Conrad Johnson tube amp. The former benched quite well (0.002 THD and IM), and was certainly fast enough (100V/usec SR in a 100W/channel amp), but it did some characteristically solid state kinds of things that were not present in the same signal routed through the headphone outputs of my Stax preamp (FET input class A design). And not present in the Conrad Johnson amplifer. Nor in my current circuits.
Live and learn, I say. Or don't.
Have fun with the project!
~Jon
And it's not too hard to get fairly good "numbers" in a conventional sense at the voltages and currents they require. There are less problems to deal with for preamps and headphone amps than a full blown power amp. Stability with capacitive loads, for instance.
But then the problems that occur in a power amp output stage and how they propagate through a feedback loop are a big reason, I suspect, for why the benefit can be quite high in that application. I certainly started developing some strong suspicions twenty years ago after listening to one of my late 70's solid state designs against a Conrad Johnson tube amp. The former benched quite well (0.002 THD and IM), and was certainly fast enough (100V/usec SR in a 100W/channel amp), but it did some characteristically solid state kinds of things that were not present in the same signal routed through the headphone outputs of my Stax preamp (FET input class A design). And not present in the Conrad Johnson amplifer. Nor in my current circuits.
Live and learn, I say. Or don't.
Have fun with the project!
~Jon
John, I agree with Jon. "Try it, try it, you will see." (Yes, I have small children. )
In the interest of making the task less daunting (and therefore more likely to be attempted) I would suggest simplifying the circuit in the following ways:
1) Eliminate the paralleled input devices. This really only serves to lower the noise floor with small input signals, but is not needed with a line-level input. Besides the zener diodes in the second stage will contribute far more noise than you would get by using a single input pair.
2) Eliminate the paralleled output devices. I can't read your posted schematic, but assuming you are using 2SK170s and 2SJ74s, the output impedance of a single pair will be around 20 - 25 ohms, and they will be able to deliver 20 - 25 mA of current. This is probably sufficient for most headphones.
3) Eliminate the cascodes. The cascodes will extend the bandwidth up from several hundreds of kilohertz up to several megahertz, but I don't know that would necessarily be beneficial in this application.
At that point you have something that can be built fairly readily. I'd be interested in hearing your reaction to its sonic performance.
)In the interest of making the task less daunting (and therefore more likely to be attempted) I would suggest simplifying the circuit in the following ways:
1) Eliminate the paralleled input devices. This really only serves to lower the noise floor with small input signals, but is not needed with a line-level input. Besides the zener diodes in the second stage will contribute far more noise than you would get by using a single input pair.
2) Eliminate the paralleled output devices. I can't read your posted schematic, but assuming you are using 2SK170s and 2SJ74s, the output impedance of a single pair will be around 20 - 25 ohms, and they will be able to deliver 20 - 25 mA of current. This is probably sufficient for most headphones.
3) Eliminate the cascodes. The cascodes will extend the bandwidth up from several hundreds of kilohertz up to several megahertz, but I don't know that would necessarily be beneficial in this application.
At that point you have something that can be built fairly readily. I'd be interested in hearing your reaction to its sonic performance.
Currently, I've got most of my JFETs characterized. The next step is to build and adjust each stage on a proto board. Once it's DC balanced, I can try a couple of tests with global or no global feedback. I'll try to spend a little more time with spice to verify that I've got an idea of how the ZFB works.
Yes, I'm using mostly 2sk170/2sk74s. I should be good to go with parallel output devices. Cascodes are planned. Don't know if I have enough matches for parallel inputs, but was using it to properly bias the 2nd stage (I'll see what I can do).
I got the impression that jazz music is particularly good with ZFB. Any other recommendations?
JF
Yes, I'm using mostly 2sk170/2sk74s. I should be good to go with parallel output devices. Cascodes are planned. Don't know if I have enough matches for parallel inputs, but was using it to properly bias the 2nd stage (I'll see what I can do).
I got the impression that jazz music is particularly good with ZFB. Any other recommendations?
JF
Since, by definition, a typical feedback amp has scads of gain, I've always thought it likely that much of the improvement in sound quality in a "zero feedback" (semantics be damned...) amplifier was due to the fact that you weren't asking for as much gain. As such, the devices were more likely to stay in a relatively linear region.
Grey
Grey
Well, I have to admit I'm a little bit puzzled. Presumably you've run some sort of simulation on the circuit, and I would guess that it kind of works.
But the first stage won't have *any* gain to speak of. If we ignore the parallel transistors and the cascodes, consider that one transistor will be turning on as the other is turning off. Therefore the current through R3 and R4 will be essentially constant as the input signal changes.
What you need to do is run a resistor from the junction of R5 and R6 to ground. The gain for each half of the first stage will then be set by the formula:
Av = R3 / ((R5 + 1/gm) + (2 * Rnew))
In this case the transistors will be running at about 1 or 2 mA, so 1/gm will be around 100 ohms.
But the first stage won't have *any* gain to speak of. If we ignore the parallel transistors and the cascodes, consider that one transistor will be turning on as the other is turning off. Therefore the current through R3 and R4 will be essentially constant as the input signal changes.
What you need to do is run a resistor from the junction of R5 and R6 to ground. The gain for each half of the first stage will then be set by the formula:
Av = R3 / ((R5 + 1/gm) + (2 * Rnew))
In this case the transistors will be running at about 1 or 2 mA, so 1/gm will be around 100 ohms.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.