john curl said:
Any special techniques you recommend?
2-switch contacts?
wide physical channel spacing?
1/4" aluminum between channels?
separate regulators for power supply?
etc?
Wire spacing is most important. Then shielding, but distance makes it a lot easier. IF you do have xtalk, and some cannot be avoided, then make sure that the xtalk is SYMMETRICAL. In other words that R adds to L, just like L adds to R. This was a true problem with the Levinson JC-2 preamp, and it was never completely fixed as it was caused by circuit board layout. It was one of the final arguments between Mark Levinson and me, before I was forced out. (He laid out the motherboard). Van Alstine 'cleaned our clock' at the time in imaging, and this is how I discovered the problem.
It would seem that xtalk is not what people want to argue about. Any other interesting topics?
john curl said:
John, how about a thermal modulation in FETs, particularly in small-signal JFETs in a situation when power dissipation could change by a noticable amount with the signal?
Cheers
Alex
Demian and I were just discussing thermal modulation in jfets, this last week, privately. It is easily seen with a Quantek Noise Analyzer, or in a sensitive jfet front end. All you have to do is the wave your hand over the input stage to move some air, or blow on it, and you will get a real, if not big, response. The air is just another input to the circuit.
However, a class A input stage is inherently pretty stable, especially a thermally coupled differential pair, as the total heat generated is the same within the same package. Usually, we like to see a heatsink, either through the leads or even better, on the outside case, in order to keep the thremal capacitance high, and the device relatively immune from changes from the air. You might also notice that both the Vendetta Research Phono Stage and the CTC Blowtorch cases are completely sealed, to keep the interior air currents to a minimum. This is absolutely necessary with the phono stage, because I have a gain of about 10,000 between 1-50 Hz. If I just remove the cover to adjust something, the 1/f noise jumps off the screen. Amazingly, it is not too audible, because it is of such low frequency, unless you do something stupid like blowing on the input stage. Then you can completely overload the preamp, and even blow your speakers.
With the Vendetta we have chosen to let the heat producing components just contact the air through a large surface area, such as a T0-220 case and/ or a clip-on heatsink. However, with the CTC, we took it to another extreme, where we mounted all the power devices, both fets and resistors on the bottom of the board and then thermally mounted the devices to the outside case of the CTC so that the top of the CTC preamp acts like a large heatsink. It gets warm, too! This way, we could keep as much heat away from the other components as possible, without adding vent holes that would disturb the air. It might not be obvious from the early pictures of the Blowtorch, but the circuit board is mounted at the top of the case, not the bottom, for this very reason.
However, a class A input stage is inherently pretty stable, especially a thermally coupled differential pair, as the total heat generated is the same within the same package. Usually, we like to see a heatsink, either through the leads or even better, on the outside case, in order to keep the thremal capacitance high, and the device relatively immune from changes from the air. You might also notice that both the Vendetta Research Phono Stage and the CTC Blowtorch cases are completely sealed, to keep the interior air currents to a minimum. This is absolutely necessary with the phono stage, because I have a gain of about 10,000 between 1-50 Hz. If I just remove the cover to adjust something, the 1/f noise jumps off the screen. Amazingly, it is not too audible, because it is of such low frequency, unless you do something stupid like blowing on the input stage. Then you can completely overload the preamp, and even blow your speakers.
With the Vendetta we have chosen to let the heat producing components just contact the air through a large surface area, such as a T0-220 case and/ or a clip-on heatsink. However, with the CTC, we took it to another extreme, where we mounted all the power devices, both fets and resistors on the bottom of the board and then thermally mounted the devices to the outside case of the CTC so that the top of the CTC preamp acts like a large heatsink. It gets warm, too! This way, we could keep as much heat away from the other components as possible, without adding vent holes that would disturb the air. It might not be obvious from the early pictures of the Blowtorch, but the circuit board is mounted at the top of the case, not the bottom, for this very reason.
Similar to John's work I try to isolate the inputs from thermal effects. I however try to keep them in still air isolated from the case so a cold breeze doesn't cause a problem. They can be good air velocity meters if you aren't careful.
I would put them in a thermally stabilized housing but its not realistic except for the most exotic applications. I'm working on a new phono stage with the unusual feature of flat response to DC. No LF rolloff. Normally it would be very feedback prone but it doesn't seem to be so far. However the very high gain makes it sensitive to external thermals- every connection can be a thermocouple and generate a voltage. I should have the first pilot run shortly and can see if its really practical. It does seem to make a sonic difference.
As for FET parameter changes with temperature; its a big problem with IC's and a lot of effort goes into layout to reduce the effect. Second, on a high bias fet input (I use 5 mA ea.) the change in operating point is very small even when swinging at maximum so the internal temps don't change much. I'll leave the discussions about fet outputs to John.
I would put them in a thermally stabilized housing but its not realistic except for the most exotic applications. I'm working on a new phono stage with the unusual feature of flat response to DC. No LF rolloff. Normally it would be very feedback prone but it doesn't seem to be so far. However the very high gain makes it sensitive to external thermals- every connection can be a thermocouple and generate a voltage. I should have the first pilot run shortly and can see if its really practical. It does seem to make a sonic difference.
As for FET parameter changes with temperature; its a big problem with IC's and a lot of effort goes into layout to reduce the effect. Second, on a high bias fet input (I use 5 mA ea.) the change in operating point is very small even when swinging at maximum so the internal temps don't change much. I'll leave the discussions about fet outputs to John.
any chance of pros and cons of current mirrors in gain stages?
🙂
pma's simulations/analysis on the jfet gain stages a while back was great input to the earlier discussion. maybe he could be enticed to come back for more as time permits ...
🙂
🙂
pma's simulations/analysis on the jfet gain stages a while back was great input to the earlier discussion. maybe he could be enticed to come back for more as time permits ...
john curl said:
🙂
Crosstalk
I, for one, am interested in understanding the mechanism by which this comes about. Asymmetric crosstalk, that is.
Mr Curl, would you know if the ML-1 motherboard was the same as the JC-2? If so, I would be tempted to pull my ML-1 apart and inspect the layout.
I, for one, am interested in understanding the mechanism by which this comes about. Asymmetric crosstalk, that is.
Mr Curl, would you know if the ML-1 motherboard was the same as the JC-2? If so, I would be tempted to pull my ML-1 apart and inspect the layout.
Still the problem of high capacitance exists with these shielded xformers, since the shield must be connected to somewhere (earth, that is). If one wants true low coupling (low leakage/"GND" currents) there is no other way than true low capacitance xformer design like R-Core, UI and specialist toroids with sectoral (non-concentric) windings (Avel-Lindberg has some small wattage models).KSTR said:
- Klaus [/QUOTE]
If the goal is low line noise then isolation the secondary side from capacitance is part of the problem, the magnetic coupling is also a part of the problem. And the fields from the transformers are more insidious than one might think.1audio said:
A true double shielded transformer (ultra-isolation) when implemented properly should reduce the line induced noise to a minimum, but its important to tie the shields to the right places or you won't get anywhere with one. [/B][/QUOTE]
Hi Guys,
Iron is iron! You can wind the copper in many ways, each having advantages/disadvantages.
I've used sectoral windings with toroids to get good HF isolation, but this requires specialized toroidal winders that are limit-stop reversible to avoid overlapping the windings, and it's difficult to do with small ID cores. You can wind bifilar to get the bet matching between windings, but this may require dual shuttles since the yield is not so good running two conductors in one shuttle simultaneously.
The E-I core has lots of fringing flux and is less efficient than the R or toroid cores, but it is cheap. You can use thinner (1-mil) laminations to minimize eddy-current losses (no relation to Steve, I assume), but you sacrifice stacking factor, resulting in a larger transformer. We've had custom short-E/long-E laminations punched so the gaps end up under the coils, and this helps somewhat, but then it might make sense to go with C-cores which have the lapped faces to minimize the gap.
The toroid and R-core have no fringing flux, but are more expensive. The toroid has a square profile since the iron tape has a fixed width. The best thing about the R-core is it's circular cross-section, which requires punching oval laminations of many different lengths/widths to approximate the round cross-section. But this cross-section results in the shortest length of copper per turn, optimizing the L/R ratio (a circle has 1/2 half the circumference-to-area ratio of a square.
You can take the best R-core and still wind it so it's a lousy transformer for audio applications. It's all depends on the parameters you decide to compromise.
Best, Chuck Hansen
Great input, Chas. Perhaps it will answer some questions in advance from members who have little or no experience with transformers. You also taught me a thing or two as well.
We used to think that toroid was a real breakthrough, until the problem of cap coupling became obvious. We still tend to use toroid transformers for 100W and up, but we would prefer R core, if we would easily get it. For preamps, however, R-core, or even an E-I high isolation transformer (if the transformer is located away from the signal electronics) is cheap and effective. We just used E-I high isolation tranformers for the CTC Blowtorch, but the Blowtorch is a two chassis design, so this is practical. We used an R-core for the Parasound JC-2 preamp, because we could get it, and it is located in the preamp chassis, itself. The huge JC-1 power amp transformer is toroid, because we could not afford to put an R-core there, but some very expensive power amps have used an R-core for high power. It is really the best solution.
We used to think that toroid was a real breakthrough, until the problem of cap coupling became obvious. We still tend to use toroid transformers for 100W and up, but we would prefer R core, if we would easily get it. For preamps, however, R-core, or even an E-I high isolation transformer (if the transformer is located away from the signal electronics) is cheap and effective. We just used E-I high isolation tranformers for the CTC Blowtorch, but the Blowtorch is a two chassis design, so this is practical. We used an R-core for the Parasound JC-2 preamp, because we could get it, and it is located in the preamp chassis, itself. The huge JC-1 power amp transformer is toroid, because we could not afford to put an R-core there, but some very expensive power amps have used an R-core for high power. It is really the best solution.
Somehow the nested quoting got wrong in Charles' post... but in the end it doesn't matter much who said what, since this is technical discussion😉
It would appear that this forum has a limit on how many nested quotes there can be. Here is a recap of what has been said about transformers:
Cauhtemoc said:
Wow, you have really mastered the nested quotes - it was perfect.
Thanks, Chuck Hansen
Well, now that we have apparently summed up transformers, I might bring up the problem with rectifier diodes. I might mention that most manufacturers and amateurs alike will simply chose a 'diode bridge, composed of 4 rectifiers to do the job of converting AC to DC. It is cheap, clean, easy to solder, and reliable. Who could want anything more? Well... We uptown audio designers have found problems with common diode bridges, and common diodes, themselves. This is because the typical (and rugged) rectifiers of common manufacture do not turn off quickly enough to not cause some RFI to be generated.
For many years, we tried to replace tube rectifiers with solid state, thinking that we would get better regulation, less voltage drop, and higher peak currents, compared to a tube rectifier. I modded a Dyna Stereo 70 this way, BUT with time, I realized that it did not sound right, and I put the tube rectifier (GZ34) back in operation and it sounded better.
Later, back more than 20 years ago, my tech came in with a diode bridge that was made of faster turn-off diodes, for only a few dollars more. I, at first, scoffed at the idea of improving audio sound when all I was only rectifying 60Hz, and my diode bridge was located in another enclosure, separate from the audio electronics. I ignored it at the time.
However, an associate of mine, who owned one of my Vendetta Research preamps decided to try to replace my standard 1A diode bridge with faster diodes. No bridge was available, so he had to make them himself from 4 separate diodes. This was not easy, but he did it about 15 years ago, and he insisted that it made a real difference.
I started to research the problem at this time, especially encouraged by a friend of mine, Rick Miller, who tested different diodes with an RF spectrum analyzer and published it in TAA back in the early 90's. I found that with a CURRENT PROBE that I could capture the current coming from the rectifier bridge, and it was much easier to see the problem. I had not owned a current probe previously, so I was not easily able to make this measurement, but it was obvious, once I did.
Today, you will find the rectifier bridges of the Blowtorch, Parasound JC-1, and JC-2 preamps, Vendetta Research upgrades, etc, all composed of individual diodes of various sizes and even difficult to purchase. It is possible that there are quality diode bridges that are composed of high speed, soft recovery diodes, with the voltage-current ratings that we need, but I have not found any easily available at this time.
Some think that a small RC network across the individual diodes works much the same, but I prefer to eliminate the problem from the source, rather than attempt to fix it after I know that it has occurred, so I will stick to high speed, soft recovery diodes for my best efforts.
For many years, we tried to replace tube rectifiers with solid state, thinking that we would get better regulation, less voltage drop, and higher peak currents, compared to a tube rectifier. I modded a Dyna Stereo 70 this way, BUT with time, I realized that it did not sound right, and I put the tube rectifier (GZ34) back in operation and it sounded better.
Later, back more than 20 years ago, my tech came in with a diode bridge that was made of faster turn-off diodes, for only a few dollars more. I, at first, scoffed at the idea of improving audio sound when all I was only rectifying 60Hz, and my diode bridge was located in another enclosure, separate from the audio electronics. I ignored it at the time.
However, an associate of mine, who owned one of my Vendetta Research preamps decided to try to replace my standard 1A diode bridge with faster diodes. No bridge was available, so he had to make them himself from 4 separate diodes. This was not easy, but he did it about 15 years ago, and he insisted that it made a real difference.
I started to research the problem at this time, especially encouraged by a friend of mine, Rick Miller, who tested different diodes with an RF spectrum analyzer and published it in TAA back in the early 90's. I found that with a CURRENT PROBE that I could capture the current coming from the rectifier bridge, and it was much easier to see the problem. I had not owned a current probe previously, so I was not easily able to make this measurement, but it was obvious, once I did.
Today, you will find the rectifier bridges of the Blowtorch, Parasound JC-1, and JC-2 preamps, Vendetta Research upgrades, etc, all composed of individual diodes of various sizes and even difficult to purchase. It is possible that there are quality diode bridges that are composed of high speed, soft recovery diodes, with the voltage-current ratings that we need, but I have not found any easily available at this time.
Some think that a small RC network across the individual diodes works much the same, but I prefer to eliminate the problem from the source, rather than attempt to fix it after I know that it has occurred, so I will stick to high speed, soft recovery diodes for my best efforts.
- Status
- Not open for further replies.