Line Stage Transformers

I've been toying with the idea of a DHT line stage using the diyaudio 26 and/or 01a circuits that have evolved over the last several years. I've started to look at specs in order to decide on output transformers, and would find any feedback from @dsavitsk @Marvey @tomchr @ultrabike et. al. to be helpful.

So the 26 plate resistance of 7-8k and the 01a about 10k are typical high DHT values. The issue being, 3-4x load to plate resistance might be needed in order to keep low frequency distortion in check. For example, the plot below of THD and harmonics for a 45-76 amp loaded on a 7Kohms James output transformer with 40H primary inductance (which is 5k Ohms at 20Hz) shows that starting at ~100Hz the THD and harmonics are ramping up. The same thing would occur on a line stage, but perhaps the lower levels mitigate the concern?

In terms of product under consideration, Lundahl offers a LL2763 with 280H primaries, which will be around 36k Ohm at 20Hz. I was looking at Monolith as well, and probably prefer their cores in general, but the initial spec Ward provided was only 100H primaries, so substantially lower load at around 12k Ohm. He thinks this is ideal for the application, but I don't. Maybe I'm confused?!

One question would be what do we expect the subjective audible difference to be with this increase in bass distortion? Looser, flabbier bass? And how bad? I think the Monolith is likely to be the more detailed tranny, but am trying to determine the tradeoff of the 100H primary in terms of overall sound. Lundahl pricing also gets asinine pretty quickly when you start asking about options.

Thoughts?

 

For a line stage do you really need an output transformer? If you don't mind a little sand in your glass, an opamp buffer could be added. This would provide nearly zero loading on your DHT stage. Alternatively a MOSFET or JFET follower would do the same thing. Or ... keeping it all-glass ... a cathode follower. If the input capacitance of the follower is an issue, you can always bootstrap it.
If you go with the opamp buffer, you'll have to do something to prevent over-voltage of the opamp inputs during startup and power-down. That's a pretty solvable problem, though.
If you absolutely do want the iron in there, you could look at parafeed. That eliminates the DC current through the transformer, so the transformer becomes smaller, lighter, and less expensive. Now you have a coupling cap to deal with instead. Tradeoffs, tradeoffs...

As for the rise in THD at LF, I'd expect it to sound warm as you get harmonic extension at LF.

Tom

 

Thanks for the thoughts on how to redesign the circuit, but I don't want to go that route. I'll stick with the known entity that has been around the block. A little harmonic extension in the sub 100 Hz range is probably fine for the application, so long as that doesn't get sloppy for e.g. with a noticeable overhang.

 

+1 for asking the transformer manufacturers.

David/Cinemag is an excellent resource. He's very responsive on email and provides straight-up answers. Jensen is pretty helpful as well. Edcor takes longer to respond and if you know what you want they can be a good resource also.

As you probably know already, the LF rolloff of a SET transformer is determined by Lp of the transformer and rp of the tube, whereas the HF rolloff is due to rp and the transformer's parasitic capacitance. You can increase Lp by adding more turns on the transformer. That gives you better LF performance but also higher parasitic capacitance, so you get worse HF performance. Tradeoffs, tradeoffs.

Tom

 

I've sort of had the opposite experience. Unless you are very prescriptive with transformer suppliers, they will sort of wind you whatever. Cinemag made me some PP output transformers for headphones with a 32 ohm secondary that had a 20 ohm DCR, and forgot to put a center tap on the primary-- which I guess makes them crappy single-ended parafeed interstage transformers. I've had similar experiences with Edcor and Electra-Print, even if they're great when they get it right.

It's going to be very difficult to convince someone to make a gapped transformer with enough inductance to reflect 30k at 20Hz, and sufficiently low leakage inductance to manage 20kHz. Here's one design using the LL2745 custom gapped for 8mA with some measurements. It does not manage 20Hz - 20kHz, but comes pretty close. I would figure out the lowest ratio, lowest inductance, lowest DC that I could live with, and see who would wind me something for sane money. My guess would be a Lundahl interstage, if the power amp's input Z high enough for a 1:1 transformer.

 

Um. Yeah. That's pretty special.

Tom

 

@peef
Ward came back and said he can gap it towards the 300H, but for optimum performance was 100H. He didn't say much more. My amp spec is 47k input Z and 1V input sensitivity. The system has good gain, so want a low gain preamp. A 01a design with Lundahl LL2745 at 5.6:1 was an option at 3db gain. This will probably be a switched 26/01a preamp cut from the same cloth as that Bartola circuit.

@Marvey I didn't see any standard offerings or model numbers for line stages from Monolith. Ward just came back with his recommendation and said he could build it. I priced out nano cores. Would probably go Chinese. Ward says the Chinese nano cores are identical to the Hitachi nano cores, but sometimes his customers want the genuine cores for peace of mind. Hah.

 

Ward's transformers are designed for 100kHz bandwidth according to him. He means -3dB. I don't think 1:1 will work as I want to lower the output impedance.

Yeah, that's good feedback on the core choice. I am not sure on nano vs amorphous. I do want to increase resolution a bit vs what I currently have, but not to an extent that it becomes hifi analytical. Basically the goal is to inject some sweet mid tone that I can't achieve with a 6dj8/7dj8 preamp and boost resolution hopefully without overdoing it. Has anyone heard a 41PL (DHT) design by chance? In triode mode, it gets constant praise for having great tone. But it's so easy to create a switched 01a/26 circuit which probably gives more flexibility (and lower gain).

 

Thanks! Did you ever try other interstage cores in your 45 amp?

Ward did tell me this a while back:

FeSi delivers a more balanced sound to my ears.
If you are a detail freak, amorphous is the way to go.
Nano delivers both but this comes at a price.

 

I'll interpret this liberally. At a fundamental level, I just don't understand the point of the circuit. Tubes are glass and shiny and they glow and all, but you'll get the same gain and hugely better sound with less noise for considerably less money with a pair of LL1676's as a 1:2 stepup. If you want a little more color to the sound, use LL7903's instead.

That said, if you must use a tube, maybe contact Dave at Intact to see what he can do.

 

Aw come on, Doug. You've put shiny glowy things your devices too. I know you love your transformers, and they are definitely good, but I prefer the sound of active circuits. Better sound is subjective. Anyhow, thanks for the pointer to Dave!

 

100 kHz bandwidth in a SET transformer that steps from 10k (or thereabout) to common headphone impedances has me wanting for data (and test conditions!) as I'm skeptical of that claim. A 1:1 line transformer that doesn't handle any DC current, sure.

Tom

 

Where'd you get "to common headphone impedences"? Ward would build a 4:1 line stage transformer here. Edit: I guess it does get down to headphone-like impedences.... [email protected] for more data on his 100kHz bandwidth claims.

 

Sorry. Lost track of who was building which project and assumed. My bad.

The 26 tube has an amplification factor (µ) of 8.3, so I'd think you could go higher on the turns ratio and still have a line stage that approaches a buffer (assuming that's your goal).

 

You mean 4P1L? The Russian DHT pentode. I haven't heard any circuit using it but it is a super linear tube in triode mode. Some use a quad of them as a budget 300B.