RJM Audio Sapphire Desktop Headphone Amplifier

rjm

Member
Joined 2004
Paid Member
What's going on here?? You are putting the amp in a black-painted cardboard box...

Also did you get the discrete op amps to work?

In principle you can plug in any compatible op amp into the Sapphire 8-pin sockets. The only restriction is that the op amp shouldn't draw more than 50 mA of current from the supplies, as more will load the Z-reg.

The normal circuit consumes about 50 mA x 36 V = ~ 2W per channel. Allow 5 W for the whole amp, so far as the regenerator is concerned.
 

rjm

Member
Joined 2004
Paid Member
Only if you live in America. ;)

The bigger it is, the more you hear it. So, for best transparency the coupling cap should be no larger than you need it to be for proper bass response. Technically anything 0.22 uF is sufficient. The rest (value, brand, construction) is just tuning the sound to your personal preference.
 
Hi again!
I took photos of the new make-up process. I'm still solving problems with the diy discrete opamps; after this, I will change the condensators, maybe a pair or Jupiter HT could be a good option (I have a pair in my Icon HP8 with very good results)







I took it to a friends meeting in Madrid and the amp enjoyed a lot ;)

_DSF1431_zpsfa02e68f.jpg






¡Take a look at the photos! :-D
Fotos e impresiones de la Segunda Reunión de Auriculares-HIFI en Madrid
Greetings!
 

rjm

Member
Joined 2004
Paid Member
Thanks for sharing that. (diyaudio should really allow "+1s/likes")

1. That's... a black Sony Playstation.

2. What sort of comments did you get about the sound of the Sapphire? I'm not "fishing for compliments" ... good/bad isn't so interesting as knowing what words they use to describe their impressions. That can help me understand how it is perceived, and possibly identify areas which need more work.
 
Thanks both of you :)

Rjm, point 2: Of course; the problem is that my lvl of english language is very poor, and the technical terms that I perfectly handle in Spanish I can't tell you in a right english translation, but I will try :D

The people comments are:
Sapphire (with paper in oil russian nos condensators and default opa134), it's a powerfull amp, with enough "balls" for move VERY well the HE-500, LCD2 / 3 and HD800 for example.
It can move the (really hard to move) Hifiman HE-6, but not with the same bass and tonal control as Beta22.
Anyway, it's a very versatile amp, clear and transparent with a slight warm touch.
Another adjetive is fast; very good dinamic represantation (fast drums, rythm changes...)

For me, Sapphire was a great discovery. I enjoyed too much mainly with Senn HD800
Ok, it's not a Beta22 (probably the best dymanic headphone amp) but, Saphhire cost MUCH less...
For the price, there's no opponent :), it can beat in sound quality vs. Burson amps and company...

If the amp grows up with better condensators and better ap-amps... it will be the awesome :):)
 
repente, I use the Sapphire as the main driver for my HE-6. In my case, I replaced R4 with a 33k ohm resistor. The gain is just about perfect, with my volume pot just passed 9 o'clock. (I tried a 66k ohm resistor, too, but the gain was just too high.)

I've never hear the Beta22, so I can't compare the two, though!
 

rjm

Member
Joined 2004
Paid Member
You can read about Amb's beta22 amp here. I haven't heard it either, but from the schematic it looks pretty cool.

As for improving the Sapphire, the hard part, and why I was asking repente about the impressions, is figuring out what's wrong with it. I'm very happy with the way it is. If anything I suppose I could wish for a little more "mellowness", as I agree that the dominant impression is of power/control/clarity rather than warmth. How one might obtain a more relaxed sound that without making the amp worse is, however, not immediately obvious to me.

Thoughts:

There's the op amp, an obvious weak point but making something better than an OPA134 from discrete parts is not self-evident, since we must answer "why is the OPA134 deficient?"

There's the output stage. Again no glaringly obvious route to improvement. And there's the voltage regulation, which, here too, what is "better" and what is "worse" from the endpoint of tuning the sound is far from clear.
 

rjm

Member
Joined 2004
Paid Member
[deep breath]

So, would you believe I had no idea Lehmann even made a headphone amplifier? I did not pick it up in my many, many Google searches for "diamond buffer" or "headphone amplifier circuit schematic".

For the record my design can be dated Feb 2, 2010 (proof), while the Lehmann is at least a couple of years before that. So its a lack of due diligence on my part rather than the other way around.

Anyhow, the fact is though that the concept (op amp voltage stage + diamond buffer) is self-evident and I don't think the circuit elements can't be appropriated by anyone.

I don't know the Lehmann circuit, so to answer your question I will assume it closely resembles this "clone" published on head fi: here.

You can see that all the main elements are basically the same, right down to the op amp and output transistors. I use BD135,6 transistors throughout the diamond buffer, the 'clone uses smaller BC550,560 on the input.

Here are the differences.

The Sapphire is dual mono, single op amps are used, and nothing is shared between channels. That's a big difference.

Also the clone uses LM317/337 voltage regs. Blech. Boo. Hiss. Sapphire has Zener regulated BD135,6 pass transistors. I think that makes a big improvement.

The clone runs at 15 V rails, the sapphire about 10.6 V. Sapphire is biased to 30 mA. I don't know what the clone runs at, you can't tell from just looking at the circuit, but given the 10 ohm emitter resistors it's probably the same or a bit less.

So, looking at it, the big thing is that the Sapphire has completely independent power supplies and signal paths for each channel due to its dual mono (one board per channel) construction.
 

rjm

Member
Joined 2004
Paid Member
Lately I've been wondering if some soft of damping was needed in the Sapphire circuit between the driver and output transistors of the diamond buffer.

I added an RC filter stage (475R 1nF low pass) before the output transistors limiting the buffer bandwidth to 335 kHz.

The voltage drop across the resistor lowers the bias current in the output stage. I hadn't planned for that, and the side effect of considerably less class A output power is unwanted. In future this will have to be compensated for by adjusting R9, R10.

The effect of the RC filters on the sound is surprisingly large, and not a result of the change in bias current (I already confirmed independently what low bias current sounds like.) It really does seem to take the "edge" off things. I think I may have overshot it a little, and a smaller filter or perhaps just damper resistors alone might give the best balance - but the interesting thing is that the position does really seem to be a "pain point" for the circuit, so its worth looking into it in some depth.
 

Attachments

  • DSC_7111_02.jpg
    DSC_7111_02.jpg
    337.3 KB · Views: 383
  • DSC_7120_05.jpg
    DSC_7120_05.jpg
    324.4 KB · Views: 352
  • DSC_7122_06.jpg
    DSC_7122_06.jpg
    326.2 KB · Views: 344
  • DSC_7126_09.jpg
    DSC_7126_09.jpg
    460.7 KB · Views: 366
  • Sapphire 20f5 mod 01.png
    Sapphire 20f5 mod 01.png
    29.6 KB · Views: 395

rjm

Member
Joined 2004
Paid Member
Sapphire 3.0

I'm starting to ramp up things towards the next spin.

Upcoming features being considered:

1. Gain switch, optional, configurable, hi/lo.
2. Current sources on the driver emitters.
3. Better optimization of the bias current ratios.
4. A well-defined, resistor-programmable output bias current.
5. Some sort of damping between driver and output stages - the exact circuit is currently under evaluation.
 

rjm

Member
Joined 2004
Paid Member
A word or two about bias currents.

The diamond buffer is a sequence of two emitter followers, the driver and the output. Both have minimum h_FE current gains of 100. (BD135/6)

The output drives headphones, the "design maximum" output current into the headphone load is about 50 mA. So the output base needs 0.5 mA, provided by the driver. The driver runs in class A, so the standing bias has to be slightly in excess of this value, say 1 mA minimum, to avoid clipping. 2-3 mA is a reasonable headroom. In my rev. 2 version, it was 9-10 mA.

In terms of impedances, 30 ohm headphones translates into an output driver input impedance of (30xh_FE) 3 kohms. The driver stage emitter resistor (R7,R8) was 1 k in rev.2, a substantive load, for a total of 750 ohms, translating to an input impedance of the buffer of (750xh_FE /2) 37.5 kohms.

It's ok, but the circuit is working unnecessarily hard to put a large current through the driver transistors. This lowers PSRR and increases distortion, as well as lowering the input impedance. The solution is to replace the 1k emitter resistors with a) a larger value or b) a current source.

A current source would allow a large current AND a large impedance, at the expense of noise and complex distortions. A larger resistor means lower bias currents, unless the driver stage power rails are increased of course. I haven't fully committed to one or the other solution yet.

As the input impedance of the output transistors is not especially high (particularly with low impedance headphones) its probably not a good idea to put large resistors in from of the base. A maximum of about 220 ohms seems reasonable. The small "boost" resistors on the driver emitters (R9,R10) need to be included in this total. They are 4.7 ohms in rev. 2, they could be made significantly larger with no real detriment to the performance.

In the schematic below I've made some tentative changes, adding the switchable gain feature, reducing the driver bias currents, and adding small resistors to the base of the output transistors. I haven't replaced the driver stage emitter resistors with current sources yet.
 

Attachments

  • sapphire 30a 00.png
    sapphire 30a 00.png
    19.9 KB · Views: 244
Last edited: