2A3 Linestage IT with filament cancel current.

Using tertiary windings to offset DC.
Post Reply
dave slagle
Posts: 2085
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

2A3 Linestage IT with filament cancel current.

Post by dave slagle »

Just got finished running the basic tests on these and everything worked as expected. I'm suggesting a gap of .004 inches to give a fair window of carefree operation. This gives just below 30hy of inductance and of course shrinking the gap will up the inductance and narrow the acceptable operating range. With a simple butt gap and 100% offset i measured -1dB points of ~4 and 400Khz. I'll measure and post the FR plot for the gap below, the 400Khz number shouldn't change, but the larger gap will drop the L and put the -1dB point at 8-10hz.

the cancel winding can be wired to offset 7 different levels of plate current with the following possibilities of offset are available assuming 2.5Adc of filament draw.

0.008
0.017
0.025
0.033
0.042
0.050
0.058

so running the 2A3 at 60ma and the filaments at the 42ma cancel winding will net 18ma total DC on the core and the graph for +18ma looks pretty good.

i think it will be quite interesting to see if there is any sonic difference between No offset, partial offset, and full offset. If you drop the Plate current you can even give a -biased core a listen.

i'll run the sweeps for a few different gap sizes tonight and they should go out by the end of the week.

dave
Attachments
dowdy.jpg
dowdy.jpg (16.61 KiB) Viewed 34192 times
Raj Gupta
Posts: 37
Joined: Fri Feb 11, 2005 2:19 am

Post by Raj Gupta »

Dave: do you wind these on a smaller core than you would a typical two-winding 2A3 OPT? Or do you just get more LF headroom, on the same sized core?

Also:
running the 2A3 at 60ma and the filaments at the 42ma cancel winding will net 18ma total DC on the core and the graph for +18ma looks pretty good.
Why wouldn't you set it up for 58mA if you were running at 60 mA? Is there an advantage to a small offset? It seems to me you'd want it to offset as nearly as possible your actual operating current, or am I missing something (a frontal lobe, perhaps)?

Just curious.

Also:

What do you think of this bad whamma jamma?

-j
Attachments
CCFL.jpg
CCFL.jpg (44.75 KiB) Viewed 34184 times
dave slagle
Posts: 2085
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Post by dave slagle »

Raj Gupta wrote:Dave: do you wind these on a smaller core than you would a typical two-winding 2A3 OPT? Or do you just get more LF headroom, on the same sized core?
i use a nickel core which would really be a stretch for this situation, plus it buys you some much needed headroom so you aren't constantly flirting with the knee of the loop. (and some extra inductance) I think the best use is to consider it a "helper" rather than rely on it to do all of your work.
Why wouldn't you set it up for 58mA if you were running at 60 mA? Is there an advantage to a small offset? It seems to me you'd want it to offset as nearly as possible your actual operating current, or am I missing something (a frontal lobe, perhaps)?
this is why we need some ears on testing, leaving some DC on the core will assure single quadrant operation. If you notice the slight dip in inductance near 100% offset comes from the AC flux exceeding the DC flux (zero crossing) and the core is not held together the entire cycle. when testing you can hear the null from the buzz the transformer makes. I would actually like it if the full offset sounds the same as some DC bias, then a slightly smaller gap could be used for the same window (DC wise) and another myth challenged!
What do you think of this bad whamma jamma?
nice... for the life of me i cannot figure what the feedback from the 2A3 cathode to the 6C45 plate would be.

you know, if one were a cheapskate, he could use a power supply toroid for the iron in this amp those large ratios shouldn't be a problem and a tapped cancel winding should let you null it out nicely. You could also make the resistor in series with the 6C45 plate choke variable to tune the null of the output tube.

your biggest problem is going to be a good 2.5A CCS :-)

dave
Raj Gupta
Posts: 37
Joined: Fri Feb 11, 2005 2:19 am

Post by Raj Gupta »

if one were a cheapskate
A pretty big "if," wouldn't you say?

Not that I would ever consider such a thing, but a 6V toroid would give you a 3200:8 impedance ratio if one were so inclined, pretty much the balls for loading a 2A3. I think I even have a few of those lying around here. . . who am I kidding? I *know* I have a few of those lying around here!
your biggest problem is going to be a good 2.5A CCS
The way my daughter says it is: "Duh!"

PS: I think the cancel winding on the plate choke may be overkill, but ya never know. I figured as long as I was drawing it anyway . . . .

-j
Gary P
Posts: 5
Joined: Sun Feb 06, 2005 5:20 pm

Post by Gary P »

Hi Guys,

What parameters do you need for the 2.5A CCS?

I've got some bigass mosfets I picked up for doing filament CCS's that are still in their shiny silver bag. Might be fun to give this a shot.

Just went back and looked at the schematic again. Don't know if you want to try and run the filament current through the 2A3 bias resistor. Would be nice from the angle of the low resistance needed. Down side is the power disapated in the resistor-- 380 watts!! More than the rest of the amp combined...

You could run a smaller current loop for the filament circuit, just doing the fliament of the 2A3 and the cancle winding. Lots less power but now you are back to the problem of high value cathode resistor and the need for a bypass cap.

Gary
Raj Gupta
Posts: 37
Joined: Fri Feb 11, 2005 2:19 am

Post by Raj Gupta »

380 watts!!
sssssss....sound of balloon deflating....ssssssss

oops.

-j
dave slagle
Posts: 2085
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Post by dave slagle »

Gary P wrote:Hi Guys,

What parameters do you need for the 2.5A CCS?
the only real parameter is that it can handle an inductive load and behave reasonably well throughout the audio band. The few people i have talked to about designing a filament CCS have all mentioned filament inductance as an issue. In this particular situation, the filament L is nothing compared to the L of the cancel winding.

a dumb thought I have had lately is to somehow "soften" the response of the CCS to make it behave more like a magnetic circuit.
I've got some bigass mosfets I picked up for doing filament CCS's that are still in their shiny silver bag. Might be fun to give this a shot.
i think we want high impedance through the audio band, but the key is the LF behavior and if the Z drops at 20K that isn't an issue. the stability with an undamped inductive circuit is.
Just went back and looked at the schematic again. Don't know if you want to try and run the filament current through the 2A3 bias resistor. Would be nice from the angle of the low resistance needed. Down side is the power disapated in the resistor-- 380 watts!! More than the rest of the amp combined...
do you have something against a few 12V halogen bulbs in series :-)
You could run a smaller current loop for the filament circuit, just doing the fliament of the 2A3 and the cancle winding. Lots less power but now you are back to the problem of high value cathode resistor and the need for a bypass cap.
yeah.. but that is where we are at this point... ya gotta crawl first and a 10W ccs is a bit easier than a 500W one... I just know that nobody is going to try the cancel winding idea unless they can do it cheaply and in a small package.

the good news is, people want CCS's for their filaments anyways. I'll stick with LCLCL filament supplies, but i have a line on cheap iron :-) If you come up with a killer stable filament CCS, the cancel winding will just fit right in and give you a huge bonus. Just design the CCS, and the iron to test it will show up.

remember i see this as a simple solution not a complex one. What is needed for the cancel winding is already in place to power the filaments of a DHT (or according ot some should be) since the LCLCL supply can become quite large, just build the best sounding CCS that you can. I tried guido's boards and they worked beautifully until the cancel winding entered the loop. Then thy nolonger behaved as they should. (this was my fault since i asked them for something they were not designed to do... the boards worked great powering just the filament of the tube.)


dave
shinebox
Posts: 88
Joined: Fri Jan 28, 2005 9:44 am

Guido's board as CCS

Post by shinebox »

Yo Dave,
My understanding of Guido's board is that it behaves as a CCS at HF and voltage source at LF - that way the tube essentially gets constant voltage operation (which is the correct way to operate them filaments) as far as DC goes but the audio signal sees a nice high impedance.

It may be that whacking it in the cancel config is messing up the feedback that the module uses to achieve the above. If you were to modify the module so it was just a CCS, it might do that job rather well. You'd need to readjust periodically for the correct filament voltage etc, still...
cheers
Gary P
Posts: 5
Joined: Sun Feb 06, 2005 5:20 pm

Post by Gary P »

Hi Dave,

Spent some time this evening and built up a CCS that might work. I used the self bias circuit and board. Just changed the resistor values around for low voltage high current operation and changed the mosfets to Fairchild FQPF50N60L's. These are 60 volt 36 amp mosfets rated for 45 watts disapation.

Results so far. Hard clipping voltage minimum is 3 volts. Performance starts to drop below 5 volts. The performance drop was an increase in ripple showing up across the load inductor (35mH Stancor filament choke). At 5 volts and above there was less than 10mv ripple across the choke. Across a 1 ohm resistor (2A3 filament simulator) the ripple was much lower, too low to measure with my scope. The maximum voltage across the CCS for power disapation is about 17 volts.

To get the drop-out voltage as low as possible the CCS needs a positive supply to connect the reference to. I just installed Q4 the same way as used in pentode mode. The "s" terminal needs to be connected to a voltage source that is 20 volts or more higher but not more than 500 volts higher than the positive end of the cancel winding. Pretty loose voltage requirements and only needs .35ma. B+ should work fine.

The CCS seems to be stable connected to the choke. No funnies were observed. I need to figure out a way to inject AC into the load to see if the CCS behaves. A SE tranny would work but I don't have any suitable. The 2 pair of SE trannys I have are setup for parafeed. I might be able to borrow one from someone in the OTS. Figure that connecting the ccs across part of the secondary winding and feeding some AC into the primary should check out how CCS handles AC voltage under load.

What will be the P-P voltage maximum across the cancel winding at full power?

Just for grins I started cranking up the current with the power supply at 6 volts. Ran out of adjustment range when the current hit 7.7 amps. Had to back it down as R1 (current set resistor) was 180% of it's rating and the run on the board between the mosfets started smoking! Guess the little 15mil runs just are not big enough for this kind of abuse.

I'll try to learn more about this setup this weekend. If things look good I'll send you a pair to experiment with.

Gary
dave slagle
Posts: 2085
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Post by dave slagle »

Gary P wrote:
These are 60 volt 36 amp mosfets rated for 45 watts disapation.
cool!
Results so far. Hard clipping voltage minimum is 3 volts. Performance starts to drop below 5 volts.
so a one ohm cancel winding would be a good thing :-)
The performance drop was an increase in ripple showing up across the load inductor (35mH Stancor filament choke).
in the case of the 2A3 IT, the inductance of the cancel winding seems like a good number to use and would be the worst case. In this the cancel winding is 18mhy but if you shrunk the gap for a perfect null, you could get 5X that. (hmmm)
The maximum voltage across the CCS for power disapation is about 17 volts.
so the filament bias 2A3 CCS is a project for another day :-)
Figure that connecting the ccs across part of the secondary winding and feeding some AC into the primary should check out how CCS handles AC voltage under load.
i'll send you a 5K OT with a cancel winding for a 2A3. It will be an odd doorstop from the trash heap, but will serve your test needs nicely.
What will be the P-P voltage maximum across the cancel winding at full power?
for the linestage. 2V/40 for an output, 100V or so/40 (40 is the turns ratio, so worst case for a 2A3 would be 2.5Vac on the offset)
I'll try to learn more about this setup this weekend. If things look good I'll send you a pair to experiment with.
very cool! also feel free to ask for anything you need iron wise to make your tests easier/more relevant to the situation.
TonyB
Posts: 30
Joined: Mon Feb 14, 2005 7:11 pm
Location: Toronto, Canada

Post by TonyB »

Hi guys,

So where did you get with the filament current cancellation scheme? I drew a couple of current sources capable of driving 2.5A and 18mH. My only concern is that any signal I insert with a current source will be amplified 40x. Three permalloy chokes for the LCLCL filament power supply would be nice but it is getting a bit heavy and expensive.

Dave, if you have enough space on the bobbin, how about adding a cancellation winding with something like half the turns of primary (maybe with 2 taps). I would put this winding into the cathode together with a cathode resistor. And either shunt cathode to ground with (evil) capacitor or use 2 capacitors as per WE connection. The idea is to stay with DC flux at half the saturation flux. This way the signal does not saturate the core and does not cross into another quadrant. I may be game to try one transformer like this since it is better than Tango NC-20. Right, Dave? :)

Regards,

TonyB
dave slagle
Posts: 2085
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Post by dave slagle »

TonyB wrote:Hi guys,

So where did you get with the filament current cancellation scheme?
the 1:1 is being built as we speak... the offset winding will be fed by LCLCL
I drew a couple of current sources capable of driving 2.5A and 18mH.
cool... care to post them?

My only concern is that any signal I insert with a current source will be amplified 40x.

the LCLCL seems dead quiet. the 1mv or so of ripple does not show up as 40mv on the plate.
Three permalloy chokes for the LCLCL filament power supply would be nice but it is getting a bit heavy and expensive.
agreed... but if that is what you need to do for a DHT filament supply anyways... I agree that the 20 pounds of extra stuff keeps a lot of people from trying the idea. My thinking is still the best DC supply for a DHT is a current source at which point the offset winding fits in nicely.
Dave, if you have enough space on the bobbin, how about adding a cancellation winding with something like half the turns of primary (maybe with 2 taps). I would put this winding into the cathode together with a cathode resistor.
space can always be found :-) Shienebox wanted to try something similar with his dual primary S&B tvc. i'm not sure if he used a negative supply to get the resistor big enough. Keeping with the 2A3 and assuming a cancel DCR of 100 ohms, that leaves a 650 ohm resistor in series which will reflect back only 2600 ohms (or is it 3000?) in either case... thats a bit low. a negative supply will make the resistor larger or maybe use it on a direct coupled output stage? otherwise you could add a second choke in series to set bias (this might have been what shines was going to try)
The idea is to stay with DC flux at half the saturation flux. This way the signal does not saturate the core and does not cross into another quadrant.
i like the thought, far too many people would say just use a PP transformer and assure 100% cancellation then next thing you know you are parafeeding with an external DC bias supply to get single quadrant operation javascript:emoticon(':shock:')

i am also a bit uneasy about floating all of those turns in the winding somewhere. essentially we want the cancel winding to only couple at DC and while logic says it shouldn't matter, my gut feel is this is easier to to with a low turn high current winding.
I may be game to try one transformer like this since it is better than Tango NC-20. Right, Dave? :)
the tango NC-20 is imo the pinnacle of design from an engineering standpoint. It was my reference until one day i heard some bifilar 1:1's on nickel cores. these just sounded musical to me and confirmed my love for nickel. Many say they do not like the sound of bifilar iron. On paper i don't like the idea of that capacitance, but i love the bandwidth! It is impossible to get bifilar like bandwidth in a layered design, but the NC-20 sure comes close! I have heard stories of people trying ot reverse engineer the nc-20 and they gave up because keeping track of the details as one was unwound showed far too complex of a design to ever replicate for less money.

my gut feel is a NC-20 type winding on a nickel core would be awesome if it could only handle the current... thats where the cancel winding comes in... to push you back down into the center of the linear range and not operate it up at the edge. (chose your own definition of linear range here... single or 4 quadrant)

dave
TonyB
Posts: 30
Joined: Mon Feb 14, 2005 7:11 pm
Location: Toronto, Canada

Post by TonyB »

Hi Dave,

Sure, I can post the current sources. I have spoken with our electronics PhD and implemented his suggestions. I will scan it and either post it (if I figure how) or email it to you. Do you want to trade a working current source for 3 permalloy inductors for LCLCL filament power supply? I am not sure who would get a good deal since it may take a while to get the current source working. :)

In my case the driver tube (45) would likely be DC coupled to the input tube. That implies a larger cathode resistor, say 4kOhms and up. This resistor and and the cancel winding (carrying cathode current) would be bypassed with a cap. The bypass cap therefore does not have to be that big, not an electrolytic. This cathode current cancel winding would have more turns than just a filament cancel winding. But it would carry the same current as in the plate winding, with nothing artificial injected. Only today I realized that there was a similar preamplifier circuit published in Sound Practices.

The idea of running the magnetic transformer flux at half the saturation value came from my telephone conversations with Herb Reichert in the last century.

I always thought than NC-20 used permalloy core. Am I wrong? Now there is NC-20F which has inproved top end response compared to NC-20.

Regards,

TonyB
EvilMoose
Posts: 2
Joined: Thu Apr 07, 2011 2:04 am

Perhaps an LED current source instead?

Post by EvilMoose »

I have never seen LED current sources specifically used in audio, but I do not see why it would not work. 1A sources seem plentiful and cheap - hence, for Raj's design, perhaps swap out the 2A3 for a 6A3/6B4G. May need to toss an LCL network in there to get rid of high frequency hash from the SMPS, but a couple of 10mh air core inductors from a T-amp would do, keeping it reasonably light, compact, and inexpensive.

I found this thread while looking for ways to use cheap mains toroidals and OPTs. With evenly split secondaries and a 3200:8 ratio should give 20:1 step down on current, so 1A gives 50ma of DC offset. Therefore, could switch to an IT based design and use an operating point of 300v/50ma to get near-perfect DC offset.

Thoughts?
dave slagle
Posts: 2085
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Post by dave slagle »

it should work fine with the toroids, just remember you will need to keep the null set properly. Maybe a servo and some appropriately scaled sense resistors?

the nice thing opposed to PP with toroids is that you get a primary that can handle 240V @ 60hz which puts you comfortable in the 8W output range.

dave
Get Your Fix
www.hifiheroin.com
EvilMoose
Posts: 2
Joined: Thu Apr 07, 2011 2:04 am

Post by EvilMoose »

Perhaps I could perma-null the set with a double CCS approach? 1A CCS for the filament and a ~50ma CCS providing cathode bias, bypassed by a cap.
dave slagle
Posts: 2085
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Post by dave slagle »

that should work nicely.

dave
Get Your Fix
www.hifiheroin.com
PET-240
Posts: 23
Joined: Sun Jan 29, 2012 1:30 pm
Location: BrisVegas, Austalia

where did this get too?

Post by PET-240 »

Hey Dave,

Where did this get too?

Would possibly like to try for a 6CL6 Spud amp.

Know it's not a lot of current, but then they aren't a big cost item to see how they go?

Thanks,

Drew.
dave slagle
Posts: 2085
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Post by dave slagle »

hey,

my guess is that at the currents you are going to use (30ma??) there reslly isn't much to be gained by the offset winding.

dave
Get Your Fix
www.hifiheroin.com
Post Reply