Balanced, unity gain line buffer for use with Studio / Yggdrasil | Gungnir Multibit

So thinking about this over the past few weeks - realized that I have the parts I need for a solid state solution. I have a bunch of 'Wire' boards designed by Owen of diyaudio. I have a balanced to balanced boards that can be built / configured to do the trick. Essentially, I need an active component that increases the current without providing gain. Further, I need some control over both input / output impedance.

This wire board is essentially an OPA1632 differential op-amp per channel, driving a pair of LME49610 buffers in a bridge configuration. So I build a wire, first setting input / output impedance to 10K. Works like a champ. As the Rag has 10k input impedance, it's really something to hear Yggdrasil into a 10K load. There's a 'solidity' to the sound that's exceptional.

But here's the tradeoff; 10K impedance into the 600Ohm Cinemag input transformers looses a bit (insertion loss, ~.5 db). So - decided to configure my impedance to 1K input / output. Would a 1K load starve Yggdrasil's output buffers? Nope - more testing is needed, but the ~400Ohm difference seems to keep Yggdrasil output within available current. On the output side, this is more closely matched (1K -> 600Ohm), so our loss is reduced.

This configuration is the closest to a 'transparent' active component; freeing up enough current 'headroom' for Yggdrasil, while allowing minimal loss. With this device, do input / output impedance need to be matched? No. I can introduce additional gain, but output impedance cannot be below input.

So here's where I could really use some input from those with more experience with transformers. The higher the input impedance, the more 'headroom' Yggdrasil's output buffers have, affecting it's sound for the better. The higher the output impedance, the more loss. With a top notch transformer like the Cinemag 600:600, how much am I really losing with an output impedance of 10K? What about 5K?

I've attached a quick photo of my test components. You cannot but affect the sound when you introduce an active component. The wire is essentially instrumentation 'kit' (very wide bandwidth, low noise). To round it out, I'm using an a22, which has the wonderful audio property of not being current limited. While the 10k vs. 1K buffer configuration sound different, they both sound very very good. I'm not hearing any real 'penalty' for injecting this line-buffer in.

 

I am not following. The point of a buffer is that it has a high input impedance and a low output impedance.

 

Thanks for chiming in Doug; and thanks for all the patience you'll undoubtedly need in reserve as I figure this out. Yes - that's what I understand the point of the buffer to be. The balanced-balanced 'wire' headphone amp board I'm using here seems adaptable for my use, but you can probably help me make sense of the schematic. I've attached. R22, R23, R25, R27 set the input impedance; we'll call these 'base' (assuming no attenuation upstream). I have these set to 10K. R21, R24, R25, R28 give you the ability to set gain to unity (matches R22, R23, R25, R27) or say, 2K gain (6db) by using 20k (2x) resistors. These are 'feedback' resistors. To achieve the lower output impedance, I should be able to set these to say, 100Ohms. I have not attempted this yet - I've only used resistors matching 'base' or higher. I'm using a head amp board as a buffer; from what I understand of the schematic, it doesn't seem I can pull output impedance lower than input. But - I'm very much still trying to figure this stuff out Hopefully that makes some sense. Does this circuit support lower output impedance?

Doh. Ok; the feedback resistors in no way set the output impedance; they merely allow for gain if desired. Output Impedance is rated at <7mOhm; so set my input resistors to 10k, feedback to 10k unity gain, and I'm done. Save me from myself if I'm off in the weeds Doug.

 

Glad for any input forthcoming, but everything's measuring working as expected. I'm leaving input impedance at 10k - works very well with Yggdrasil. Gonna take more testing / listening, but I prefer balanced with line buffer to Yggdrasil SE.

 

Totally lost here. You're using a BAL-BAL and then into the Cinemag? Why?

 

Goal of the project was a balanced buffer that a) added sufficient current for Yggdrasil balanced out to drive the 600 Cinemags on Studio without distortion, and b) load Yggys balanced outs with a more optimal impedance (10k). I'm a bit surprised that the gains of loading Yggdrasil with a 10k impedance adds more than adding an active component takes away.

 

Ok, I see what you're trying to do. It just seems silly to me that we're adding all these components that may or may not be coloring the sound after we've already achieved our '"best digital ever" -messiah' sound....

I think Jason may (or may not) want to comment on the output abilities of the Yggdrasil's analog. I would expect a TOTL product with a solid-state output stage to be capable of driving a 600 ohm load. Have you encountered distortion with lower Z loads on the Yggdrasil?

 

Those of us with Studios who discovered that our Yggys cannot drive the 600Ohm load (much research, input and confirmation from Jason) find it less silly. Much of the discussion has taken place outside the forums out of respect for these manufacturers, but these products are not compatible in balanced mode. Specifically, Yggys output cannot supply the needed current into a 600Ohm load. This is also true with Gungnir Multibit.

The desire to use our Yggdrasil balanced out with Studio has led to this research, but yea - adding an active component if Yggdrasil could drive Studio balanced would be silly.

 

That's fine. Though would you mind qualifying the effect of Yggdrasil driving a 600Ohm load? Are we talking a hard clipping or distortion?

 

It's a distortion. For hotter recordings, it's pretty constant and evident. For less hot recordings, it may only distort at peaks. Jason's taken the time to huddle with Mike B. and do the math; they didn't envision Yggdrasil/Gungnir Multibit needing to drive a 600Ohm load. A transformer swap on Studio to a 15K transformer would address the issue, but so far Craig's been unwilling to foot the bill on such a swap.

 

That sounds more reasonable I really try to stay out of such things, but I will say this much: A) A good buffer is as difficult to design as any other audio stage. The use of opamps and negative feedback is as suspect there as anywhere else -- maybe more so in a buffer as it requires 100% feedback. If you like the sound from this stage, why bother adding another one after it? It has suffecient gain and power to drive any headphone. B) 600 Ohm transformers from just about anyone will be basically blameless. They are relatively easy to make. As the impedance increases, they get more difficult. My opinion is that while Cinemag does many things right, their 15K IPTs are not my favorite. They lose a lot of detail to my ear (same with the Jensen and Sowter). I find the Lundahl LL1690 to be a better part.

 

It would make more sense to me to forgo the transformer and convert to SE with another opamp stage. You've already had to acquiesce to "evils" of monolithic solid-state electronics, so why not double down and use something like the new(ish) alpha24 to generate a single-ended signal? Same 1632, with the addition of the 1611/1612 (which I can attest is excellent sounding). If you really wanted you could bypass the input buffers for a similar configuration that you posted, but seeing as the Yggdrasil deteriorates under lowish Z loads, I would include them and load the input with 500k (1Meg total Zin) to ensure things are totally undistorted as you say.

 

Kudos to you for sticking with the Studio, @MisterRogers. I probably would have taken the refund.

 

Getting past the frustration, it's one hell of an amp Studio + Yggdrasil SE is stellar. It works with most other balanced DAC's as it should. I'm happy.

 

Thanks Doug; I greatly appreciate your input - both staying out of it, and the difficulty of designing a good buffer stage. This started as an exercise to be able to use Yggdrasil's / Studio's balanced out / in. While not an 'elegant' solution, this bal-bal 'wire' board proved to be pretty well suited to the task. The one intriguing aspect that came out, is the importance of impedance matching between components. Not a new 'lesson', but it underscores how important it is. I've done some A/B between this buffer with 10k input impedance vs. Yggdrasil's SE output (JFET summing). The key standout was the audible improvement of Yggdrasil's output when loaded with a higher impedance (10k). Beyond that, I've added components, complexity, etc. Apart from being able to show Yggdrasil a higher impedance, I haven't brought much else to the table.

But, as a DIYer that enjoys tinkering - I'm pleased to coax a bit more out of Yggdrasil, and to my ears, to not audibly degrade the signal going into Studio's balanced input transformers. It may not have any legs, but It does give Studio / Yggdrasil owners a potentially cost effective way of using Yggdrasil / Studio 'balanced'.

Just need to case it up, and it'll be ready for more 'ears' at a small meet this coming Sat.

 

Yep, that'd work. I have a balance to SE converter I helped put together that uses 6 LME49990's; extremely fast and clean conversion. But I actually wanted to use the 600Ohm Cinemags. I think Schiit did a pretty good job of their SE circuit. Good implementation, fewer parts - I'd leave it alone if I didn't want to load the 600Ohm Cinemags

 

Maybe something like that http://glass-ware.stores.yahoo.net/acf9pin.html.

 

Well, it's certainly not an elegant solution from the perspective of using the fewest possible components to get a job done.

On the plus side, it presents Yggdrasil with a 10k load, which I believe is more optimal for its output design (Jason's comments / listening). Sonically, it sounds very good, though with extended listening it does seem to mask a bit of the nuance in the signal.

 

You know there is only one solution to this.