Output transformer question

Hey all. Could use some help in mapping out a project.

Let's say I want a transformer coupled 300B output stage, and the ability to drive 24 Ohm and 300 Ohm HPs effectively, full bandwidth.

Do you feel that something like a 10W, 100ma, 5k:8 single tap OPT with parallel 12 Ohm, 10W resistor is going to be sufficient for both loads?

In the case of the 24 Ohm headphone, the 12 Ohm resistor provides exactly 8 Ohm at the secondary (nominally). But now with the 300 Ohm headphone, I worry about it being underpowered with the parallel resistor dissipating nearly all the power.

Thanks!

 

This strikes me as the wrong way to about things. In order to swing 100mW into 300R, you need 5.5V on the output. With a 5K:8 OPT, that's a 25x voltage ratio which means swinging 140V on the primary. That's a lot - close to 4W of actual power, which means lots of gain and distortion and expense. Additionally, a transformer wound to handle that kind of power and current will hugely under perform a transformer made to do what you want. You'd do much better with a transformer wound to meet the load you actually want to drive.

 

@dsavitsk
Yeah, thanks Doug. I arrived at the same conclusion offline. I'm going to ditch 24 Ohm loads and have the OPT custom wound for 300B --> 5K @ 100ma to 300 ohms, 10 watts. Then I would stick with HDx00 headphones, which I want to do anyhow!!!

The more I look in to design details, the more I realize application-specific is the way to go.

 

PS, he's saying no upcharge whatsoever for this particular winding, and it's very close to the IMD offering as well because they would wind it with a lower IWC.

 

Hmm, looking at this a little more, maybe indulge me one more time as I'm thinking aloud... The HD800S is 102dB/1Vrms, so 102dB at 3.3mW given its nominal 300R impedance. The 100mW in your calculation is then going to be around 117 dB, and much too loud for me. So let's talk about more normal listening levels:

Although 102 dB is football stadium crazy loud level, and still too high for me for very long, let's talk about swinging 3.3mW into 300R with the 5k:8 because 1V is a nice round number. 1V on the output now means swinging 25V on the primary, or about 125mW of actual power. Not so bad, right?

This particular amp I'm building has a gain of around 10 dB. If we decreased the turns ratio from 25:1 to 4:1 for a custom 5k:300 then the gain increases 6x to roughly 28 dB ... doesn't it? That seems very likely to be a noisy amplifier, no?

 

I'll respond more fully later, but LL1671 for 4:1. You won't find a better part.

 

OK thanks, will check back in later ... because I thought I convinced myself that 5k:8 or 25:1 turns ratio was going to be better here - i.e. lower noise, plenty loud at negligible power with HD800S. And with the 12 Ohm parallel resistor on the output, you can switch between speaker and headphone outputs easily, such as this where you would wire the top terminal to 8R speakers and the bottom terminal to headphones (even the 300R HD800S) with the resistor dissipating most of the power:

 

BTW, I think the gain increase is actually about 16 dB if using the 4:1.

20*log(sqrt(300/5000)) = -12.2 for 5k:300
20*log(sqrt(8/5000)) = -28 for 5k:8

 

To make the best possible transformer, you want the smallest core, the thinnest wire, the fewest turns, and the lowest turns ratio. In order to make a transformer capable of powering several watts into speakers, you need a huge core, lots of turns, a big ratio, and thick wire. On the other hand, a transformer for phones, or line level can be small and appropriately made.

I'll add that a single feed transformer is useful for speakers because it increases circuit efficiency at the cost of fidelity. You don't need efficiency here since you only need a fraction of a watt. Parafeed will sound better.

Yes, the larger turns ratio will reduce noise more that a small turns ratio, but it will also require more gain which will add more noise and distortion. I can't speak to the rest since you have not posted a schematic, or said what driving stage you are planing to use.

 

Thanks, @dsavitsk . It'll be the DG300B (Neurochrome) adapted for headphone use. When I ran it by Tom, his thought was something along the lines of the gain from the 4:1 turns ratio being akin to using a 100W amp to drive headphones. "You will hear every electron flowing in the input tubes, every cat that walks by outside, and your ears will bleed every time your fridge turns on. That’s my fear."

So even though it's not a lovely method, he seems to prefer burning most of the power in a parallel resistor to allow headphone use with the same OPT winding intended for speakers. When I inquired about the custom winding with Jack, he seemed to have no preference - either using the 5k:8 with L-pad or 5k:300 without.

Here's the driver board schematic (I've seen it posted elsewhere in public threads which Tom C participates, so there should be no reason not to post it here).

 

Tom is a fine circuit designer, and his 300B amp is well regarded. But it is not a headphone amp. It is a speaker amp, and it is intended for driving a transducer that needs literally 3 orders of magnitude more power than what you are driving. When one discovers that their large square peg will not fit into a tiny round hole, the solution should not be to grab a bigger hammer.

If your goal is to have a box with a a couple of big light bulb sticking out the top, then this is a way to go. That's fine. But to be clear, if your goal is fidelity, it isn't. Headphone amps are neither big preamps, nor little power amps. They are their own thing, and designing them is its own challenge. And indeed, you can do it with 300Bs in a (sort of) reasonable way - this just isn't it.

 

@dsavitsk are you saying you don't find either approach reasonable - a custom wound OPT nor a 5k:8 with L-pad. I had seen one headphone conversion of this amp for Grados which used, I believe, a custom 60 Ohm secondary. If memory serves, it may have been at a SoCal meet some time ago, which @Marvey and others attended. Either way ... the goal is a DHT SET amp. I had entertained making it switchable for speaker vs headphone usage initially, but that's no necessarily a primary goal. Relegating it to headphone usage with a custom OPT would be fine by me if it turns out well in the end. Your points are well taken though, and if it's not the right design to begin with, I probably ought to abort now!

 

That is part of what I am saying. I think both adding gain to throw it away and having too much gain are indicators of poor design. There are theoretically reasonable-ish ways to use 300Bs for phones.* Those involve a driver stage with an appropriate amount of gain. If you look at the Western Electric datasheet, and the point you are going to have to run the 300B for enough gain for your Senns with the 25:1 stepdowns, inappropriate and compromised OPT design aside (see below), you are going to generate 5% THD just in the output stage. A 6N6 swinging that much voltage is going to add a bunch as well. I am not, as Steve might say, a number chaser. But that's too much. And distortion drops off quickly as you lower the voltage swing.

As for compromised output transformer design, this is really the biggest issue. Too much gain is not ideal, but speaker transformers simply do not have the fidelity needed by headphones. Full stop. And the problem with 300Bs is that you need so much current that the OPT has to be made to accommodate it and that will diminish your fidelity. It is the reason I suggested parafeed here - you can load the tube with a CCS and run as much current as you like, and you can still use a small nickel core OPT in a reasonable ratio that will sound better than a big single feed OPT. Seriously, the little Edcor $20 7K:300 parafeed OPTs we had made will run circles around even the best 300B single feed OPT.

I am also saying that corners that can be cut in a speaker amp cannot always be cut in one for phones. If the designer suggests that there might be any reason to think there are noise issues, take that concern seriously. I don't know the amp well enough to know.

The 300B has a mu of ~4. Actual gain from just the output stage, including a 5K:300 (4:1) step down and accounting for various copper and impedance losses is probably going to be -6dB. I think you probably want at least 5X gain overall into Senns to account for low recordings, different DACs, etc. So that means you want a driver with around 20dB of gain.

The above with a 6N6 is not actually that much higher and might be just fine - depends on your source and source material. With the D3a it is likely too high. There are a multitude of ways to accomplish this.

*Honestly, using 300Bs for efficient phones is never reasonable. But within the universe of unreasonable things, there are things that are less unreasonable.

 

Every time i read @dsavitsk posts. I feel like i learn something, even though i know nothing about amplifier design or electronics.

 

Yeah, great stuff from Doug as usual here.

I talked to Tom. We agree. Using the 8-12R parallel load resistor is the square peg here. Obviously when you're intentionally burning off a whole crapton of power in a resistor to make it all work... you're starting with the wrong design in the first place. We tossed around the idea of making something work with a 2A3 that would be interesting with a 10k:300 OPT and 200V B+. The input stage would be redesigned to work with the lower B+ and we'd no longer really be adapting a speaker amp for headphone usage.... So quite possibly we might produce the first DG2A3-HP amp here... Should be fun!

 

By the way, @dsavitsk If you know of a schematic or further reference for the 300B parafeed amp you mentioned, I’d be very interested.

 

Have you read this? http://diy.ecpaudio.com/p/parafeed-tutorial.html

Using a DHT is not that much different, you just have to deal with the filament current.