Ultra Linear Driver Transformer

[KE6DF]

I have a broadcast driver transformer out of a 5KW (I think) broadcast transmitter.

It was used with two pair of PP 6550's to drive a pair 3x2500 modulator tubes.

It has screen grid taps on the primary, and a split secondary.

3x2500's are way overkill for anything I would ever need.

So I'm thinking it might be worth using this transformer to drive either a pair of 810s or perhaps 805s in a class b modulator circuit.

6550's are expensive and overkill also.

So perhaps I could use 4 807s or 4 6bg6s in ultralinear mode. The 6bg6 seems like a better choice as it can handle 350 v on the screen vs 300 for the 807.

An alternative would be to used either 2 (for the 805s) or 4 (for 810s) triodes like 6b4 and just not use the ultralinear screen taps.

Any thoughts on this?

Or would it be better to use this transformer as a door stop.

(Broadcast driver transformers are somewhat rate, though, so using it seems good. It also might make a good small modulation transformer as it has about a 1.25:1 full primary to full secondary turns ratio).

[W2XR]

I believe this driver xfmr is out of a Gates BC-5P AM BC rig. They used that ultra-linear arrangement with a quad of EL-34s/6CA7s as audio drivers to the push-pull pair of 3CX2500F3 modulators. Perhaps 6550s were used interchangably in production, and I see no reason why they should not be essentially a drop-in replacement.

It's a nice transformer, and should not be used as a door stop!

73,

Bruce

[WD5JKO]

Here is but one idea on how to drive that transformer designed by Mr. Oz, Patrick Turner of Turner audio:


http://www.turneraudio.com.au/Integrated5050.htm

[KE6DF]

Thanks for that.

One issue I was thinking about was the comment that 6L6, 807, 6bg6's etc. were not good driver tubes for Class B stages because the plate resistance is too high to deal with the highly variable load impedance encountered when driving class B grids.

Is this problem mitigated by using ultralinear connection?

[WD5JKO]

Is this problem mitigated by using ultralinear connection?

To a certain point yes. The position of the screen tap can vary. Put it at the CT, and we have full beam power tube mode with high gain, and high Rp. Put it at the plate, and we're in full triode mode with low gain, and low Rp. The ultra linear connection as done commercially was a compromise where we got decent gain and reduced Rp from full beam power tube mode.

I would still use at lest 6 db of global NFB.

Jim
WD5JKO

[W2XR]

Is this problem mitigated by using ultralinear connection?

To a certain point yes. The position of the screen tap can vary. Put it at the CT, and we have full beam power tube mode with high gain, and high Rp. Put it at the plate, and we're in full triode mode with low gain, and low Rp. The ultra linear connection as done commercially was a compromise where we got decent gain and reduced Rp from full beam power tube mode.

I would still use at lest 6 db of global NFB.

Jim
WD5JKO

In the early 1950s, it was determined by Acrosound and David Hafler of Dynaco fame, that the location on the primary winding of the output transformer for the ultralinear screen taps was at the 43% point. That point was found to provide the best compromise for output power (pentode operation) and linearity (triode operation). That was the whole point of ultralinear operation; higher output power than triodes, and less distortion than pentodes.

The ultralinear circuit is still used to this day in many hi-end vacuum tube audio amplifiers.

I think you will find that you need to use more than 6 dB of NFB in the ultralinear scheme. 6 dB is fine for true triodes (i.e. pentodes or tetrodes not strapped for triode operation) such as 2A3s, 6B4s, 6A3s, 6A5s, 300Bs, 845s, etc. For psuedo-triode operation, such as triode-strapped 6550s, EL-34s, 6BQ5s, etc., usually 10 to 16 dB is required for acceptable linearity. For ultralinear service, you typically need 16 to 23 dB of NFB for low distortion operation. If the tubes are in self-bias, you can get by typically with less NFB.

Be careful how you implement the NFB loop; I think Gates wrapped the global feedback from the 3CX2500F3 modulator plates back to the cathodes of the first voltage amplifier stage in that rig through a DC blocking cap and feedback ladder arrangement, so that the driver xfmr was within the loop. A quick check of the schematic of the driver (which I think I still have here somewhere) will confirm this.

73,

Bruce

[KE6DF]

"I think Gates wrapped the global feedback from the 3CX2500F3 modulator plates back to the cathodes of the first voltage amplifier stage in that rig through a DC blocking cap and feedback ladder arrangement"

It looks like you are right. See the "marketing" schematic below.

If you can find a detailed schematic with compent values it would be interesting to me.

[W2XR]

"I think Gates wrapped the global feedback from the 3CX2500F3 modulator plates back to the cathodes of the first voltage amplifier stage in that rig through a DC blocking cap and feedback ladder arrangement"

It looks like you are right. See the "marketing" schematic below.

If you can find a detailed schematic with compent values it would be interesting to me.

But this is good! In an audio driver circuit using tubes and a driver xfmr, this is where the NFB is usually sampled (at the plates of the modulator tubes) and the FB is typically introduced at the first voltage amplifier stage, as in the diagram you had forwarded. The important thing to bear in mind is to not have two feedback loops; a local one within the the driver, and a global one connected from the modulator plates to the driver. Two NFB loops are a good recipe for instability in this case.

I think you can build up a really nice audio driver circuit with a global NFB loop with this Gates xfmr. The primary plate-to-plate impedance is probably quite low, as the quad of 6550s are connected in push-pull paralell, so you will most likely have to use the same or similar tube complement. This is critical in a driver to a class B modulator; you want to have the highest possible step-down ratio possible between the driver tube plates and the modulator tube grids, to improve the driver regulation and keep the distortion to a minimum. The Gates engineers designed this transformer and tube line-up with all of this in mind. Whatever tubes you select in lieu of the 6550s, just make sure that the plate-to-plate Z presented by the driver tubes is equal or similar to that presented by the quad of 6550s. I'd probably go with the current production  Russian EL-34s, as they are pretty inexpensive and are of excellent quality; I think they are actually based upon the original Mullard design.

I'm using the audio driver xfmr in my rig from a Gates BC-1F 1KW rig. All of these driver xfmrs are becoming scarce and hard to come by, since they were largely designed out of the lower power (1 KW and lower) BC rigs in the '50s and in the higher power rigs (>1KW) in the '60s.

73,

Bruce

[steve_qix]

Or would it be better to use this transformer as a door stop.

Naaaaaaahh.  I wouldn't use it as a door stop.  Someone might get a broken toe.  I find they work out very nicely for holding up shelves in my make-shift work bench 

[steve_qix]

Oh, and by the way, that particular transformer is bad (shorted).  Just so you know 

[Opcom]

In here:
http://www.turneraudio.com.au/Integrated5050.htm

Under what conditions would they expect those diodes on the transformer ends to conduct? I have never seen that done before.

[KB3DKS]

If you can find a detailed schematic with compent values it would be interesting to me.

Try this. I can send a higher res version if needed. It is a large file. I have this deck for the 2X2500's.

[WD5JKO]

In here:
http://www.turneraudio.com.au/Integrated5050.htm

Under what conditions would they expect those diodes on the transformer ends to conduct? I have never seen that done before.

Patrick,

   For those diodes to conduct the plate of either output tube must go negative with respect to ground. Under normal conditions, this cannot happen. Under some form of extreme overloaded pulsed input, maybe the diodes on one side or the other will conduct. Another scenario is if the output transformer is unloaded, the diodes might offer some protection, especially if the NFB becomes unstable and the AMP turns into a 50 watt ultrasonic oscillator.

You will find Patrick Turners circuitry quite unconventional, but at the same time after following this mans work for many years, I think he has literally advanced the state of the art with his designs. Just one mans opinion here. Mr. turner cannot be accused of being a 'copy-cat'. Spend some time on his web-site looking around. He presents a lot of theory to back his methods.

Look for his posts in rec.audio.tubes

Jim
WD5JKO

[k4kyv]

I would use the transformer with a quad of 2A3's, 6A3's or 6B4's, ignore the ultra-linear tap, and not bother with negative feedback. The step down ratio, 1.25:1 full primary to secondary is perfect for a quad of the above tubes to a pair of 810's or similar triodes in class B, and because of their low plate resistance, feedback is unnecessary.  That's exactly the turns ratio I use with my UTC LS-49 to match 2A3's to class B 810's.

I wouldn't recommend using that transformer with 845's; they need a lot more step-down ratio than that.

If you decide to go with tetrodes, I would suggest a quad of plain old 6L6's, with plenty of feedback round the driver stage only. Less likelihood of ultrasonic oscillation because you are working with less overall gain than with global feedback. No need to go to the extra expense of 6550's. 6L6s would be more than adequate.

I have never heard of anyone using the ultra-linear configuration for class-B driver tubes before, but it might be worth experimenting to see if it gives a better compromise between output power, gain and distortion.  Generally, connecting tubes like 6L6s as low mu triodes still does not result in as  low plate resistance as that of the 2A3/6A3/6B4 series, and the power output capability is less than that of the full-fledged tetrode mode.

Gates used to make a 250 watt BC transmitter that used a pair of 6L6's (with negative feedback) to drive a pair of 810 modulators.

The 1950's ARRL handbooks claim that a pair of 6L6s with feedback gives about the same performance as a class B driver as a pair of 2A3s.

If you decide to use that transformer as a doorstop, let me know.  I have a nice concrete block lying in the mud next to my driveway that I'd be glad to trade for it.  

Out of curiosity, what is the physical size and weight of the transformer?  Also, I think it would be a waste to use it as a modulation transformer.  Good class-B driver transformers are extremely hard to find, but the transformer is the weakest  link in a class-B driver/modulator system.

[KE6DF]

It's about 3 x 3.5 x 5"H

Weighs about 9 lbs.

It has a thin steel case and is potted.

The case and potting material can't weigh more than a few ounces.

I did find a picture of the modulator deck from a BC-5P and sure enough, it's the right transformer -- it looks identical.

It is a bit lighter than an LS-49, but the LS series uses heavy cast cases.

I'd probably use four of them with a plate (and screen) voltage of 350V in ultra linear if I went that way.

That would be less power than the driver transformer was intended for, but I'm only driving 810's (or 805's) rather than 3x2500s.

I suggested 6BG6's only because I have a box full of them and it seemed to me they were pretty much the same as 6L6's -- the RCA spec sheet, in fact, says they are a "mechanical redesign" of a 6L6 for higher voltage use.

I also have a bunch of 810s.

Perhaps I'll spring for a quad of 2A3's and go that way.