Mac Unity coupled
-
- Posts: 2092
- Joined: Sat Jan 22, 2005 3:54 am
- Location: NYC
- Contact:
the whole output Z thing is confusing me. I am using the method where you drive a node with a 1A current source and then the plotted voltage gives the output Z. I confirmed the same behavior with math and the open circuit vs. short circuit voltages so whatever is going on it is consistent.
Here is the output Z of 36V & 37V bias with a 4K load resistor. The green looks like the cathode follower an the Blue is essentially the 4K load resistor.
Here is the output Z of 36V & 37V bias with a 4K load resistor. The green looks like the cathode follower an the Blue is essentially the 4K load resistor.
- Attachments
-
- mac output.asc
- (3.5 KiB) Downloaded 154 times
-
- Screen Shot 2021-01-19 at 8.35.00 AM.png (46.06 KiB) Viewed 7801 times
Last edited by dave slagle on Tue Jan 19, 2021 2:25 pm, edited 2 times in total.
Get Your Fix
www.hifiheroin.com
www.hifiheroin.com
-
- Posts: 2092
- Joined: Sat Jan 22, 2005 3:54 am
- Location: NYC
- Contact:
Now lets up the load resistor to 1 meg to take it out of the picture and it appears as if this has a CF Z-out at 36V bias and way more at 37V. It also appears to have both inductive and capacitive quantities with the SRF being around 3K where the 1 Meg load resistor dominates.
- Attachments
-
- Screen Shot 2021-01-19 at 8.37.22 AM.png (50.29 KiB) Viewed 7800 times
Get Your Fix
www.hifiheroin.com
www.hifiheroin.com
-
- Posts: 2092
- Joined: Sat Jan 22, 2005 3:54 am
- Location: NYC
- Contact:
Now looking at the current through one of the coils in the voltage driven transient realm and I see that the "squirrelly" has current clipping.
WTF? Is this just user error? spice error? or reality?
dave
WTF? Is this just user error? spice error? or reality?
dave
- Attachments
-
- Screen Shot 2021-01-19 at 8.39.52 AM.png (54.53 KiB) Viewed 7799 times
Get Your Fix
www.hifiheroin.com
www.hifiheroin.com
-
- Posts: 2092
- Joined: Sat Jan 22, 2005 3:54 am
- Location: NYC
- Contact:
I was going by the description found Here:
http://www.oestex.com/tubes/Mac/MC-3500.html
note how the plate of one tube is directly connected to the screen of the other!
The interstage is also bifilar and what appears to be a ax7 input / paraphase splitter.
http://www.oestex.com/tubes/Mac/MC-3500.html
note how the plate of one tube is directly connected to the screen of the other!
The interstage is also bifilar and what appears to be a ax7 input / paraphase splitter.
- Attachments
-
- Screen Shot 2021-01-19 at 4.27.30 PM.png (455.79 KiB) Viewed 7789 times
Get Your Fix
www.hifiheroin.com
www.hifiheroin.com
ah, i see it now... your drawing is crazy! i need to think about it. are you sure about your 6L6 model? i don't totally trust spice... i love it! but it's not entirely right. it's right to look at the current. the voltage swings can look like shit, but sum in the loads just perfectly. i will think about it some more.
my skin is thin, my skull is thick.
i redrew the circuit as i see it and made the generator impedance 0 to begin with. Mac had that bootstrap driver stage to get the crazy drive swing needed for the no gain output stage. i think 40H is way too high for the Mac tranny. but maybe you have better info than i do? i get 50 watts into 8 ohms at just under 1%, which is what mac claimed. with 400 volts... not 350. but again i do not know what the spice model is doing. and this is with an ideal driver, which mac definitely did not have. this is most definitely class B as you can see the cut off current in the windings.
- Attachments
-
- mac_50_watt_opt_113.png (178.38 KiB) Viewed 7743 times
my skin is thin, my skull is thick.
the screen current pulls at the plate supply when the plate is bottomed out at ground. so there is a big impedance difference there, but the cathode is helped out by the screen. there's your feedback mechanism. it error corrects the core magnetization for sure! but it does introduce an error term... and you can see it in the ripple of the opposed screen currents. that would be an odd order term... and most likely higher order. the trifilar winding scheme would eliminate that error.
my skin is thin, my skull is thick.
the advantages of not having to share B+ with the screen are manifold! having the screen supply isolated and at a lower voltage means the plate can be safely increased. with EL-34s you could have the screens at 250 and the plates at 700 and still get a long safe life, in class B. the driver stage would have to make 600 Vpp across the audio range, clean! but you could easily get a low distortion 70 watts out of a pair of tubes like that.
my skin is thin, my skull is thick.
you might be adding dead zone with that voltage difference. it should be at the grids for this circuit. there is a simple way to drive a class B stage from the cathodes with the grids grounded directly. but i can't imagine driving a splitload that way.... no idea off hand how that might look? there might be a way? i'll think about that too.
my skin is thin, my skull is thick.
i am sure you are throwing things off for the splitload feedback mechanism with your offset voltage at the cathode. the level shift adds a DC bias shift when the plate is off. more dead zone. that's not right. it needs to be symmetrical and that will ensure it isn't. the bias has to be at the grids.
my skin is thin, my skull is thick.
here is the money shot: for ONE tube. current in the plate and cathode windings with the screen current pulled away from the plate from the opposing screen. you can see the reduced plate current in the plate winding... blue, with the resultant opposing pulse... for the opposite phase in the other tube. a third order error term. this would not be there with an isolated screen winding. crazy thing is that the feedback still does a good job of dealing with it, but i bet the third harmonic is lower in the tertiary winding scheme.
Last edited by nanana on Wed Jan 20, 2021 12:54 am, edited 2 times in total.
my skin is thin, my skull is thick.
just to be clear... these wave forms are only associated with one 6L6 and one pair of windings... blue is plate, green is cathode, and red is screen. remember that current only goes in one direction, more or less. so the current in the cathode and plate windings should be the same... if everything was symmetrical. which it isn't in class B. it's a mostly "half duty cycle". it can't be that purely unless things are highly coordinated, which is difficult. but the transformer and the feedback circuits (2 splitloads) really do work together in this circuit.
my skin is thin, my skull is thick.
it should be clear that the plate current is pulled down from the opposing tube's screen grid, clearly an impedance change... at the plate winding that doesn't happen at the cathode. the amplitude of the drop in current in the plate winding is identical to the peak screen current! and it introduces a 3rd order error term. a cost of the cross coupled screen supplied from the plates. the Mac tertiary winding for the screen is obviously superior to this solution. but it did solve the problem of dead zone operation that cannot be controlled by the feedback mechanism inherent in the splitload inverter. the core stores energy for 180° which together with screen current pulling at the cathodes while the plates are off, keeps enough in the system to error correct. you need energy to make error correction... if both plates are off for a short time, it falls apart. but you can get really close to falling apart without doing so here. there is NO sign of the perturbations of either tube in itself, showing up in the load current... except for the highish 3rd order harmonic distortion near full power. which is typical of ALL push pull operation.
my skin is thin, my skull is thick.
-
- Posts: 2092
- Joined: Sat Jan 22, 2005 3:54 am
- Location: NYC
- Contact:
I too really love these kind of conversations... spice is an incredible tool to let you know what is really going on from the conceptual side of things.
(I resized your images... dropped them to 100dpi most image viewers will allow for size changes... i find dropping to 100dpi from 144 is enough)
The cross coupled cathodes are the heart of this. In the Chokertina where the windings are in the plate and cathode of the same tube in order for the bifilar to be phased properly the DC must also add. By taking one of the bifilar pairs to the other tubes cathode the DC currents null and the AC sums. As a bonus it really helps the class B aspect.
I applied bias to the cathode to quickly adjust class of idle operation. swapped it to the grids and same thing. the minute current hits 0 in a coil the impedance goes squirrelly. Triode wiring didn't change things then I switched to a 6L6GC model and the squirrel is gone or at least has taken some valium.
Now I get somewhat more sane values for Z-out and it appears as one tube shuts off the Z-out goes way up. How to move from static bias points to judging dynamic behavior is my next question....
I also have been assuming the ideal for coupling. try dropping the K factor to 0.9995 and another squirrel shows up. have you looked at output Z?
(I resized your images... dropped them to 100dpi most image viewers will allow for size changes... i find dropping to 100dpi from 144 is enough)
The cross coupled cathodes are the heart of this. In the Chokertina where the windings are in the plate and cathode of the same tube in order for the bifilar to be phased properly the DC must also add. By taking one of the bifilar pairs to the other tubes cathode the DC currents null and the AC sums. As a bonus it really helps the class B aspect.
I applied bias to the cathode to quickly adjust class of idle operation. swapped it to the grids and same thing. the minute current hits 0 in a coil the impedance goes squirrelly. Triode wiring didn't change things then I switched to a 6L6GC model and the squirrel is gone or at least has taken some valium.
Now I get somewhat more sane values for Z-out and it appears as one tube shuts off the Z-out goes way up. How to move from static bias points to judging dynamic behavior is my next question....
I also have been assuming the ideal for coupling. try dropping the K factor to 0.9995 and another squirrel shows up. have you looked at output Z?
- Attachments
-
- Output Z in Ω for various static bias points
- Screen Shot 2021-01-20 at 9.51.25 AM.png (48.71 KiB) Viewed 7741 times
Get Your Fix
www.hifiheroin.com
www.hifiheroin.com
-
- Posts: 2092
- Joined: Sat Jan 22, 2005 3:54 am
- Location: NYC
- Contact:
Here is the output Z for various bias points and less than ideal coupling between the windings.... could this really be an amp that segues from voltage dive to current drive as frequency goes up :-0
I wish I knew a way in spice to get different coupling K's for windings on the same core. The bifilar airs could indeed be 1 but the others have to be less than ideal.... Time to point stephie to this
dave
I wish I knew a way in spice to get different coupling K's for windings on the same core. The bifilar airs could indeed be 1 but the others have to be less than ideal.... Time to point stephie to this
dave
- Attachments
-
- Screen Shot 2021-01-20 at 9.59.52 AM.png (52.56 KiB) Viewed 7739 times
Get Your Fix
www.hifiheroin.com
www.hifiheroin.com