[TzeYang]

Dear DIYers and great teachers of Audio DIY,

I'm about to receive those bjts and jfets for the above diamond buffers.

Thing is, which of the two are much more stable for open loop no opamp configuration? Sound quality is kinda important too. The Millette hybrid diamond buffers also uses a lot more parts as compared to the PPA v2 diamond buffers.

Thanks you.

 

[FallenAngel]

Uhm... and here I was thinking they were the same buffer.

Diamond Buffer for Millet Hybrid

PPAv2 Buffer

Only difference I see is that the Diamond buffer doubles up on the output transistors and uses a 2N5486 vs a PN4392 CCS.

 

[TzeYang]

oops, i just realised this too.

Anyway, the stacked output transistors improve stability of the open loop configuration?

 

[amb]

Quote:

Originally Posted by TzeYang /img/forum/go_quote.gif Anyway, the stacked output transistors improve stability of the open loop configuration?

No, it just doubles the maximum theoretical output current capability. Or, put it another way, for a given bias setting, the MH dDB's output transistors flow only half the current that of the PPAv2's, so they run cooler and would be a little less prone to thermal runaway. Since there is no room for heatsinks on the MH dDB, this may be of benefit.

 

[TzeYang]

i'm going to build either on a perfboard.

So heatsinking the output transistors isnt a problem.

Any suggestions on which one i should build? I don't really need opamps. I just want a unity gain stable diamond buffers. Also, the CCS JFETs are different. Which one i should use instead?

Again thanks everyone.

 

[tomb]

Personally, I'm biased (pun intended), but cetoole's DB design for the MAX is better than both, IMHO. The revMH DB's were a wonderful solution for an extreme case of a constrained and limited board area. The PPA's were in a similar situation, I believe, because of the existing v1 board design, but not nearly as severe as the revMillettHybrid. Even so, the PPA DB's have no heat sinks and are limited to 30ma bias. Interestingly, as Amb pointed out, the revMH DB's actually move more current - about 40-60ma per channel, potentially.

Colin's design on the MAX beats them both, with full-blown extruded heat sinks and up to 60+ma bias possible on the output trannies - in a single pair transistor/channel mode. There may be some that might ask why is such current needed(?), but it allows the output transistors to operate in a much more favorable regime with less distortion. Sound background is very black and transparent, with tremendous impact and detail, IMHO.

 

[amb]

Quote:

Originally Posted by TzeYang /img/forum/go_quote.gif Any suggestions on which one i should build? I don't really need opamps. I just want a unity gain stable diamond buffers. Also, the CCS JFETs are different. Which one i should use instead?

Assuming you will power the buffer with a dual-rail PSU, without an opamp through which a global negative feeback loop is applied, there is no mechanism to keep the output DC offset in check. Even if you match all devices carefully it would likely to still be unsatisfactory, so you should use an output coupling capacitor. It's evil but that's the price you pay for omitting the opamp and global NFB.

The Millett Hybrid is powered by a single-rail PSU, so an output coupling cap is mandatory anyway, because the output DC offset is about 1/2 the supply voltage (depends on tube bias).

The CCS JFET is chosen for an Idss such that the trimpot could achieve the desired range of bias adjustment. PN4392 has a higher Idss spec than 2N5486, so the PPAv2's bias adjustment range is higher that of the Millett dDB.

 

[TzeYang]

hmmm, i would be using TREADs or batteries to power it up. Any suggestion for an ideal PSU for the buffers?

EDIT: Now i seriously understand why the open loop sijosae diamond buffers cannot stand 2x 9V. Thanks alot amb. I really appreciate the help.

 

[amb]

Quote:

Originally Posted by TzeYang /img/forum/go_quote.gif hmmm, i would be using TREADs or batteries to power it up. Any suggestion for an ideal PSU for the buffers?

Dual TREADs would work fine.

 

[tomb]

OK - missed the part about open loop/no opamp part. (I'm not reading carefully!) Amb is correct - serious offset occurs and this is only alleviated in the MH and MH Max by the sizable ouput coupling caps.

However, he's being modest again, I suspect.

His JISBOS buffer design may be the best available when offset potential is present:

Through-hole JISBOS Buffer

As Amb notes in that very same thread,
"I don't know if it's obvious, but this buffer is also suitable as a standalone unity-gain headphone amp for low-Z headphones. Just add a volume pot and power supply."

 

[TzeYang]

Quote:

Originally Posted by amb /img/forum/go_quote.gif Dual TREADs would work fine.

dual TREADs as in combined dual supply? (into V+ V- and a ground reference). Or just one tread for one channel of the buffer?


Quote:

Originally Posted by tomb /img/forum/go_quote.gif OK - missed the part about open loop/no opamp part. (I'm not reading carefully!) Amb is correct - serious offset occurs and this is only alleviated in the MH and MH Max by the sizable ouput coupling caps. However, he's being modest again, I suspect. His JISBOS buffer design may be the best available when offset potential is present: Through-hole JISBOS Buffer As Amb notes in that very same thread, "I don't know if it's obvious, but this buffer is also suitable as a standalone unity-gain headphone amp for low-Z headphones. Just add a volume pot and power supply."

I'm considering RuneEight's DC Servo solution. Hopefully it'll work.

 

[amb]

Quote:

Originally Posted by TzeYang /img/forum/go_quote.gif dual TREADs as in combined dual supply? (into V+ V- and a ground reference). Or just one tread for one channel of the buffer?

The former.

 

[TzeYang]

amb, your jisbos buffer is awesome. unfortunately i only have parts for the millett hybrid db.

Also, i don't quite understand why combined dual treads will help reducing dc offset. This way, it'll only support up to two channel of buffers only right? (i'm suspecting the total amp draw is playing a significant role in affecting the dc offset.

 

[amb]

Quote:

Originally Posted by TzeYang /img/forum/go_quote.gif Also, i don't quite understand why combined dual treads will help reducing dc offset.

It doesn't. If you want to use the TREAD (and since it's a single-rail supply), then you need two of them to make a +/- dual rail supply (see tangent's TREAD docs about how to do this). If you power the buffer with only a single TREAD, then you'd either have to make a virtual ground, or the output of the buffer will sit at approximately 1/2 the power supply voltage relative to real ground. That's a very large amount of DC and must be blocked with an output coupling capacitor.

Quote:

This way, it'll only support up to two channel of buffers only right?

No, you can use the dual TREADs to power as many buffers as you like, limited only by the LM317 voltage regulator (and more importantly, the amount of heatsinking on it).

 

[TzeYang]

okay thanks amb. I was confused XD.

I'm thinking of using Jisbos' dc servo scheme. Do you think it'll work. Operating voltage would probably be from a 24V steps, splitted with TLE2426 into +12/-12. I personally feel that the 12 volts is good enough supply that does not neccessary require a dc servo circuit.