Actually, Wavebourn, you are showing your ignorance. CCIF two tone distortion is very important in this case, because the IM difference tone is going to be amplified (effectively) by the RIAA EQ. that can be as much as 40dB. Therefore, a 50Hz difference tone, for example, might be multiplied from .01% to almost 1%. This effect was noted by AMPEX many decades ago, even before my time, with analog tape reproduction. My reference is: J.G.(Jay) McKnight, AES member, and available in the bay area to clue you in.
For those who have some academic background and interest, this problem is why I switched over to complementary jfets, back in 1981, after successfully getting the same noise performance from bipolar transistors for about 8 years, with a patented and proven topology. The open loop performance was 10-100 times better with distortion cancellation due to complementary design, than my earlier effort.
While a paralleled row of fets will give low noise, the inherent distortion is very high, and it will be amplified out of proportion by the RIAA network with some inputs. This is what makes a Vendetta design exceptional, not its nominal noise, which is noteworthy, but not exceptional.
Making negative references to Jack Bybee in this context, just confuses people, and is completely just a random attack on someone, whom you have never met, but I might introduce you to his cousin ( also named Jack Bybee) who is his lawyer, and who lives in Walnut Creek. I have met him and had lunch with him, and he is the guy who will conduct the lawsuit, if it becomes necessary. It has happened before, and the effects were serious.
For those who have some academic background and interest, this problem is why I switched over to complementary jfets, back in 1981, after successfully getting the same noise performance from bipolar transistors for about 8 years, with a patented and proven topology. The open loop performance was 10-100 times better with distortion cancellation due to complementary design, than my earlier effort.
While a paralleled row of fets will give low noise, the inherent distortion is very high, and it will be amplified out of proportion by the RIAA network with some inputs. This is what makes a Vendetta design exceptional, not its nominal noise, which is noteworthy, but not exceptional.
Making negative references to Jack Bybee in this context, just confuses people, and is completely just a random attack on someone, whom you have never met, but I might introduce you to his cousin ( also named Jack Bybee) who is his lawyer, and who lives in Walnut Creek. I have met him and had lunch with him, and he is the guy who will conduct the lawsuit, if it becomes necessary. It has happened before, and the effects were serious.
No. It is another design problem that keeps repeating itself here. However, it is easy to be less ignorant about 7th harmonic distortion. Just Google: '7th harmonic music' and see what you get. Many here, without engineering degrees, can learn something important, just by doing a little 'research' on the internet.
The second reason for using fets is that their inherent transfer function does not easily allow much generation of 7th harmonic distortion. Not too much 5th either, but they DO put out a lot of 2'nd harmonic and some 3rd.
If you cancel the 2'd harmonic by differential (pair) input, you reduce the distortion considerably, BUT you add 3dB of input noise, and you have to use twice as many jfets.
In fact, to get back to the basic paralleled fet array, you have to use 4 times as many jfets. Think about it!
When you use a complementary jfet array, you can get low noise, and 2'nd harmonic cancellation with only the same number of fets as you might use in a single array. The downside is MORE non-linear input capacitance, because the P channel jfets have more non-linear input capacitance for the same transconductance, and usually a slightly lower peak operating voltage, due to the fact than many p channel jfets made for low noise operation have a lower rated max voltage.
The second reason for using fets is that their inherent transfer function does not easily allow much generation of 7th harmonic distortion. Not too much 5th either, but they DO put out a lot of 2'nd harmonic and some 3rd.
If you cancel the 2'd harmonic by differential (pair) input, you reduce the distortion considerably, BUT you add 3dB of input noise, and you have to use twice as many jfets.
In fact, to get back to the basic paralleled fet array, you have to use 4 times as many jfets. Think about it!
When you use a complementary jfet array, you can get low noise, and 2'nd harmonic cancellation with only the same number of fets as you might use in a single array. The downside is MORE non-linear input capacitance, because the P channel jfets have more non-linear input capacitance for the same transconductance, and usually a slightly lower peak operating voltage, due to the fact than many p channel jfets made for low noise operation have a lower rated max voltage.
John;
Not being afraid of a lawsuit, I can tell you my opinion that cancellation of the 2'nd order harmonic by diffstages and symmetrical topology is worthless. Been there, done that, discarded as a dead end idea that have no advantages except economy of an energy consumption.
Better results I got by:
1. Approximation: several devices in parallel working on different parts of their transfer functions. One example you've seen: degraded BJT in parallel with MOSFET. However, I used that in output stages, did not try on inputs.
What I tried on input, was:
2. Transistors with bigger emitter areas than usually used there: smaller Rbb, flatter H21 on working currents.
Then, 2 ways to make a single device more happy:
3. Bootstrapped device: more linear transfer function due to smaller variations of a voltage across it's terminals,
4. Output current cancellation by a counter-modulated current source: less distortions due to smaller current variations through the device.
And, of course,
5. Using of more linear devices in combination with transistors, even though they are considered as obsolete and less linear. Yes, I mean vacuum tubes.
Not being afraid of a lawsuit, I can tell you my opinion that cancellation of the 2'nd order harmonic by diffstages and symmetrical topology is worthless. Been there, done that, discarded as a dead end idea that have no advantages except economy of an energy consumption.
Better results I got by:
1. Approximation: several devices in parallel working on different parts of their transfer functions. One example you've seen: degraded BJT in parallel with MOSFET. However, I used that in output stages, did not try on inputs.
What I tried on input, was:
2. Transistors with bigger emitter areas than usually used there: smaller Rbb, flatter H21 on working currents.
Then, 2 ways to make a single device more happy:
3. Bootstrapped device: more linear transfer function due to smaller variations of a voltage across it's terminals,
4. Output current cancellation by a counter-modulated current source: less distortions due to smaller current variations through the device.
And, of course,
5. Using of more linear devices in combination with transistors, even though they are considered as obsolete and less linear. Yes, I mean vacuum tubes.
Single ended design can sound pretty good, but with fets, it gets a bit 'mushy'. It is better to do single ended with tubes, for the most part. They have lower overall distortion, and what they do have is mostly 2'nd harmonic.
It is possible, but difficult to cancel the second harmonic distortion of a fet by using a certain combination of idle current and a specific value of load resistor. This is because the output conductance changes with idle current and can be made to track with the change in transconductance of the fet(s). It is fussy, but it will work. We, once made a microphone input stage this way (lowZ, not hiZ). Worked OK, but each channel had to be individually adjusted.
It is possible, but difficult to cancel the second harmonic distortion of a fet by using a certain combination of idle current and a specific value of load resistor. This is because the output conductance changes with idle current and can be made to track with the change in transconductance of the fet(s). It is fussy, but it will work. We, once made a microphone input stage this way (lowZ, not hiZ). Worked OK, but each channel had to be individually adjusted.
Mike, batteries work, they just are not reliable. However, batteries have something special: They are NOT connected to the power line. Think about it! What could connection to the power line, even with unmeasurable garbage or noise coming from the regulator outputs, do to effect the sound? It has to be the ground, and parasitic caps, etc connecting the power line to it.
john curl said:
The Vds characteristic is also a part of that equation, and as Scott has
pointed out (and I enjoy exploiting) there is also the opportunity for
the "fortuitous null" through selection of the Vds operating point against
the Ids operating point, load value and projected voltage swing.
This is old news to the brighter members of the triode crowd, but
seems unfamiliar to most people discussing fets.
Nelson Pass said:
the "fortuitous null" through selection of the Vds operating point
Yes, -90dB deep on a simple source follower. My experiments also seem to show that it is possible to always null the seconds when the devices are not exact compliments in complimentary differential circuits. The trick is finding the point where offset and seconds are simultaneously nulled, this means two interactive trims.
john curl said:
With my preamp power supply I noticed that the noise across +/- rail voltages was very low. But, when measuring from one rail to ground the noise was about 10 times worse. I guess this is what you mean by ground contamination? Anyway, I got rid of most of this high frequency noise using a single phase supply - used just one zero recovery diode on each rail for half wave rectification. Got more 60hz noise but that can be reduced too.
I plan on using batteries in a sub crossover/tone control I'm building. Just got to filter the high frequency battery noise.
john curl said:Oh great Wavebourn, let us see your equations for load line even harmonic cancellation!
Equations are simple, John, for minimization of all harmonics: we long time ago laser trimmed relatively cheap thick film ICs, so why modern Taiwanese manufacturers can't employ some technicians to use screwdrivers turning trimpots in audiophile devices that cost enough to return investments in corresponding measurement sets and technological instructions in weeks? It is needed only once, while feedbacks and other attempts to under-over-compensate have to work constantly with speeds greater than is needed for reproduction of the highest frequencies of sounds, producing worse results...
However, design of such a gear requires an art similar to music composing and poetry, instead of straight employment of standard topologies millions of times in last 40 years discussed in magazines and on online forums.
Also, there is such an input test that is needed to reject devices that can't be used. Cheap enough, if to keep in mind prices for gear and compare to costs of selecting of 4 matched FETs for complementary inputs. I am designer and technologist according to diploma, so understand how equilibrium between design and technology may be used to get more optimal results.
Hi John,
Batteries are noisy as well. You still need a power supply to kill that noise. The amount of noise is dependent on chemistry. Mains connected supplies can easily have lower noise than a simple battery connection. If you really want to use batteries, consider the "Cyclone" brand. They are very common in process control instruments and are reliable. Of course they will require replacement every 4 ~ 7 years. Using industry standard parts both reduces costs and makes replacement availability in the future more certain. Here is a link to an applications manual for the Cyclon batteries. It's a large file. Note that these are a lead acid type. The equipment uses include some working at low noise levels.
If your power supplies are designed properly, the power cord will not be heard. However, for a low level device, series resistance in the transformer primary circuit would be beneficial in reducing HF noise from the rectifiers. Think in terms of even 0.1 ohms or there abouts. This would probably be close (hopefully) to the sum of resistances to your electrical box in the average home. Adding resistance will reduce the peak charging currents in your filter capacitors. If you can hear the effect of a power cord on your preamp, adding resistance should really make stunning changes.
John, I'm not poking fun at you here. From past experience with equipment where changing power cords did make an audible difference, I was able to improve the power supplies to a point where no one could hear the effects of a power cord any longer. This should be something worth investigating from your point of view as it would allow you to manufacture and sell the Blowtorch again.
I may not be an expert or certified engineer, but I do know the odd thing or two, as many others here do.
Now, to get back to the current discussion. A new Corvette would be the way to go for a fun, sporty ride. They are good on gas, handle very well and are comfortable. They are not twitchy like some other "roadsters". The older Jags really are better off with the Chev. upgrade, in every aspect of handling, power and cost of ownership. A stiffer suspension also does wonders for a Jag.
I can't remember whether the 928 or the 944 was the more comfortable car. Most are no fun at all on the average North American road. You won't be feeling too frisky after 45 minutes in a Porshe hitting potholes. That and you have to get use to the higher revving engine.
Ever drive a Jensen Interceptor? That's insane. We had one of these at our disposal in black. It's got a Chrysler 440 interceptor engine under the hood! The tranny tunnel is pretty high, which means your arm is almost horizontal while manipulating a very short shifter. Far too much horsepower in a tiny package.
-Chris
Edit: Here's one.
Batteries are noisy as well. You still need a power supply to kill that noise. The amount of noise is dependent on chemistry. Mains connected supplies can easily have lower noise than a simple battery connection. If you really want to use batteries, consider the "Cyclone" brand. They are very common in process control instruments and are reliable. Of course they will require replacement every 4 ~ 7 years. Using industry standard parts both reduces costs and makes replacement availability in the future more certain. Here is a link to an applications manual for the Cyclon batteries. It's a large file. Note that these are a lead acid type. The equipment uses include some working at low noise levels.
If your power supplies are designed properly, the power cord will not be heard. However, for a low level device, series resistance in the transformer primary circuit would be beneficial in reducing HF noise from the rectifiers. Think in terms of even 0.1 ohms or there abouts. This would probably be close (hopefully) to the sum of resistances to your electrical box in the average home. Adding resistance will reduce the peak charging currents in your filter capacitors. If you can hear the effect of a power cord on your preamp, adding resistance should really make stunning changes.
John, I'm not poking fun at you here. From past experience with equipment where changing power cords did make an audible difference, I was able to improve the power supplies to a point where no one could hear the effects of a power cord any longer. This should be something worth investigating from your point of view as it would allow you to manufacture and sell the Blowtorch again.
I may not be an expert or certified engineer, but I do know the odd thing or two, as many others here do.
Now, to get back to the current discussion. A new Corvette would be the way to go for a fun, sporty ride. They are good on gas, handle very well and are comfortable. They are not twitchy like some other "roadsters". The older Jags really are better off with the Chev. upgrade, in every aspect of handling, power and cost of ownership. A stiffer suspension also does wonders for a Jag.
I can't remember whether the 928 or the 944 was the more comfortable car. Most are no fun at all on the average North American road. You won't be feeling too frisky after 45 minutes in a Porshe hitting potholes. That and you have to get use to the higher revving engine.
Ever drive a Jensen Interceptor? That's insane. We had one of these at our disposal in black. It's got a Chrysler 440 interceptor engine under the hood! The tranny tunnel is pretty high, which means your arm is almost horizontal while manipulating a very short shifter. Far too much horsepower in a tiny package.
-Chris
Edit: Here's one.
- Status
- Not open for further replies.