RJM Audio Sapphire Desktop Headphone Amplifier

rjm

Member
Joined 2004
Paid Member
Okay, so that went really well...

Bandwidth 125 kHz, 1.5 V rms output into 47 ohms. 10 kHz 2 V p-p output square wave passes cleanly, no sign of any kind of unpleasantness. Gain 26.6 dB, precisely as simulated.

(with C2 = 100 pF, R2 = 475 ohms, R3 = 22.1k. C2 is 100 pF C0G but I'll replace with 47 pF mica later, the smaller compensation cap as per the BOM should increase bandwidth to above 200 kHz)

aside: I checked and yes, the trim pot R6 is absolutely required
 

Attachments

  • DSC_1077s.jpg
    DSC_1077s.jpg
    229.6 KB · Views: 503

rjm

Member
Joined 2004
Paid Member
So obviously the thing that jumps out here is the THD, so lets deal with that:

It's what it is expected to be, based on the simulation. 2nd harmonic is -85~-90 dB, even unloaded. That's pretty decent, but not as good as a good op amp. My sapphire 3 sounds cleaner, the sapphire 4 sounds a little colored / warmish / fatter in comparison. Other than that there isn't much in it. The version 4 admirably keeps up in all other respects. As expected the HF rolloff is a tad steeper.

*note: ignore the comb-like patterns in the FFT, those are because the soundcard doesn't like the loopback connection with the Sapphire in/out. they are not present with the input unplugged from the PC.
 

Attachments

  • Comparison.zip
    34.4 KB · Views: 39
  • thd.png
    thd.png
    6.6 KB · Views: 439
  • comparison.png
    comparison.png
    17.3 KB · Views: 199
  • 40b sim.png
    40b sim.png
    70.1 KB · Views: 201
Last edited:
BOM 4.0b4

R6 trim resistor changed to 2k, and re-wrote part of the notes to explain the preamp configuration a little better.

*****

So, so brief comments about how it sounds:

Think a big 1970's Sansui amplifier/preamp - one of the really nice ones. Sapphire 4 sounds like what you'd imagine that to sound like: warm and heavy, smooth and groovy. It brings it's own character to the music, definitely, but not to the extent that I'd consider it a bad thing at all.

Any trace of "op amp IC sound" is banished. I'm talking about that sharpness, the edginess, that grit, that you probably know well even if you may not have had the opportunity to identify in an A/B comparison.

It also has an uncanny way of removing "digital sheen" too. AC/CD "back in black" is an album I know very well, I have very good memories of listening to the LP version at the college radio station where I hosted a show back in the day. I don't care much for the CD. Listening to the Youtube/Vevo upload last night though, my audio memories of the LP were triggered. It was actually almost freakish how close it resembled an LP played though a proaudio console. That's not a knock - I really dig that big, lazy, ever-so-slightly-fuzzy sound.

Still, no - it's not the last word in detail retrieval or transparency. There is a curtain drawn between the music and ground-level ambiance of the recording itself and the playback chain. But where that information does not add materially to the enjoyment of listening to the music, here the Sapphire 4 excels because what you do hear sounds so good.

In terms of "score" I'd put it easily ahead of all previous versions of the Sapphire amp with IC op amps, but very slightly behind the 3.1n with the Sonic Imagery Labs 995FET-Ticha 2520 module installed. I feel the discrete FET op amp - as is borne out by the distortion measurement above - gets the nod as the last word in clarity, speed, and control, all the good traits of op amp with none of the bad. That said if you asked me which I liked better, that's not an easy answer. Put it this way I'm in no rush to put the 3.1n boards back in the chassis, though it's in the back of my mind that I probably will at some point.
 

Attachments

  • pcb-sapphire-40b4-bom.zip
    158.1 KB · Views: 84
Last edited:
... brief comments about how it sounds:

Think a big 1970's Sansui amplifier/preamp - one of the really nice ones.

Maybe I can shed a little light on where that warm, heavy, smooth Sansui sound of decades past comes from. I think it comes mainly from Sansui’s use of carbon composition resistors throughout the circuitry of its big ’70’s and ’80’s amps. At least, this is what I decided after hopscotching modifications to two Sansui AUD-11 II integrated amps (1984 vintage). I’ve documented my work on these amps elsewhere, so let me just summarize what was done: they’ve both been recapped with better components, rebuilt & modified, but in leap-frog fashion: I wanted to compare changes in sound from one change to the next, one to the other. The most startling change to my now ‘best’ AU-D11II was switching out all the carbon composition resistors for metal film resistors, matched in value at each position, channel to channel. The improvement in sound definition and spatial imaging was so dramatic that I can no longer listen to the other AU-D11II that still has its original carbon comp resistors - even though it has all the other revisions but for the metal film resistors.

I used cheap Xicon MF resistors (I had to buy a lot of resistors), so I wonder what "better" resistors would yield.

I also changed out the carbon composition resistors in my recapped Marantz 2270 receiver and it sounds amazing, too. Walking into a dark room in which this old girl is playing, you wouldn’t recognize it as a Marantz. Taking out the carbon comps has excised that two-martini vintage Marantz audio signature.

In the photo of your boards, I can see a variety of resistor types. Could this be part of what you’re hearing?
 

rjm

Member
Joined 2004
Paid Member
Interesting, but alas no: it's built with a mix of Vishay and KOA Speer, all metal film resistors.

But even after your mods, your Sansui still sounds like a Sansui right? In any event, don't take what I wrote above too literally. All I meant to indicate there is I heard a smoothness, a kind of laid back warmth that I personally associate with that kind of equipment.
 

rjm

Member
Joined 2004
Paid Member
I spent a little effort in LTSpice to try and finesse the circuit values with the aim to lower the distortion.

What I found is what you'd expect: you can trade stability (phase margin) for lower distortion. This is done by keeping a high open loop gain (R2 = 475 ohms) while simultaneously dialing in a low closed loop gain (R3 = a low value). The harder you push this the more the circuit relies on the compensation capacitor C2 to maintain enough phase margin to stay stable. The upside is distortion becomes vanishingly small.

The values below show what is possible, at least in theory. -107 dB 2nd harmonic for 15 dB voltage gain, 0 dB output level.

Other than that, the other circuit values seem to be well optimized already.
 

Attachments

  • sapphire 4.png
    sapphire 4.png
    169.3 KB · Views: 204
Last edited:
... your Sansui still sounds like a Sansui right?... All I meant to indicate there is I heard a smoothness, a kind of laid back warmth that I personally associate with that kind of equipment.

I'm not so sure it sounds like a Sansui anymore - I can't listen to the carbon comp resistors any more, they sound marinated. My ears could do with more education. We do fall susceptible to the characteristics of our customary haunts, they become our "normal". I listen to a lot of modestly compressed files in itunes over a home wireless network, through hot-rodded vintage electronics (the Sansui mentioned earlier is my favourite). Comparing the itunes files to what I can play on a re-clocked, modified Marantz CD63 player, I can't tell the difference. I'm probably bumping up against the limits of the resolving power of what I normally listen to.

My Sapphire 3 sounds great, and while it dialled up the resolution over my non-headphone listening, it offered up sound quality that didn't completely leave behind what I was listening to (of course, imaging in headphones is different than stereo imaging in a basement room), when I took the time to listen critically. I'm still using AD8610 op-amps in Sapphire 3, I haven't yet tried the OPA827's I bought. But I look forward to trying 4.0b.
 

rjm

Member
Joined 2004
Paid Member
My Sapphire 4 seems to be still breaking in. I can't quite pin it down yet as my impressions vary day-to-day. They only thing I can latch onto is this recurring sense of "analog-ishness". Warmer, less crisp, less impactful bass compared to the Sapphire 3... but both the treble and bass do seem to be filling out over time, to the point where I'm no longer confident which rendition is the more accurate: is the '3 slightly overemphasizing the transiensts, or the '4 under-emphasizing them?

One thing that does stand out though: the noise floor is so low I can listen with the volume to 100% and still not be certain if the unit is powered up or not. Just a tiny rush sound at max.

Your boards should arrive later this week I expect. I'm anxious to get your second opinion.
 

rjm

Member
Joined 2004
Paid Member
BOM 4.0b5

For R2=475 Ohms, the circuit should be stable without any capacitance in C2, at least for gains more than 6 dB.

Removing the capacitor increases the open loop gain at high frequencies, so there is more feedback applied, and less distortion. For my configuration with R3 = 22k it amounts to about 8 dB more feedback applied for audio frequencies above 10k. Since there was only 15-20 dB applied in that region to start with, that's a pretty significant increase.

I tried removing the 47 pF mica cap I was using and, wow, "now we're cooking with gas!" The treble really lights up and the bottom end starts to pack some serious wallop too. The amp throws off it's previous lethargy and wakes up, in other words.

So, C2 is dropped as a required component from the BOM. It should not be needed for any of the suggested R2/R3 combinations.
 

Attachments

  • sapphire 40b rjm.png
    sapphire 40b rjm.png
    31.1 KB · Views: 384
  • pcb-sapphire-40b5-bom.zip
    158.2 KB · Views: 83
Excellent.

By the way, what gain are you looking to use? As you can see in the last BOM I posted above, there is some range of choice involved in R2 and R3 for a given gain setting.

*****

The compromise is,

R3 and to a lesser extent R2 should be large enough that they do not load down the current stage sources Q5-9. These transistors are biased to deliver a maximum of 1 mA, and so the minimum resistor values can be determined from the output voltage swing. Roughly, R3 should not be smaller than 5k, while R2 should not be less than 5k/gain.

but

R2 should be as small as possible to maximize the open loop gain.

In other words there is a window for R2 where the overall performance is maximized.
 

Attachments

  • pcb-sapphire-40b4-sch.png
    pcb-sapphire-40b4-sch.png
    20.7 KB · Views: 380

rjm

Member
Joined 2004
Paid Member
Continuing on from yesterday's post, about the values of R2 and R3 and where I said there were limits on how small they can be because the current sources were only set to 1 mA...

Well, there's a way to get around that, and that's to power the feedback loop from the output of the buffer rather than the internal current sources Q5-8.

It's interesting, normally we think that wrapping the feedback loop around the output stage reduces the distortion of the output stage, but here at least that's not what is happening at all. The buffer provides drive to R2 and R3, so more open loop gain can be dialed in, more feedback applied, and lower distortion obtained.

It's a simple modification. One end of R3 is disconnected taken to the T+ pad on the board instead.

Oh, the shiny attractiveness of global feedback! Such sweet benefit for no downside!

And everyone knows how that story ends.

First the circuit is now reliant on C2 for stability.
Second, the feedback now has to deal with the inductance and capacitance of the headphone cable and drive units.
Third, as a consequence generally of more applied feedback and higher bandwidths, the circuit will be more susceptible to the "unsimulatables" such as parasitic capacitance.

Still, it's an interesting avenue for experiment, which, evil genius that I am, is provided for with the existing board layout.

Probably should get in touch with me to discuss before going ahead with it though. I can give you some suggestions and guidelines at least.
 

Attachments

  • pcb-sapphire-40b4-R3mod.png
    pcb-sapphire-40b4-R3mod.png
    55.3 KB · Views: 357
  • Sapphire 4b2 buff.asc
    5.8 KB · Views: 37
  • sapphire 40b buffer mod.png
    sapphire 40b buffer mod.png
    74.9 KB · Views: 360
Continuing on from yesterday's post, about the values of R2 and R3 and ... to power the feedback loop from the output of the buffer rather than the internal current sources Q5-8... One end of R3 is disconnected taken to the T+ pad on the board instead.

Probably should get in touch with me to discuss before going ahead with it though. I can give you some suggestions and guidelines at least.

I'll try it. Reminds me of a resistor on the backside of Sapphire 2, but I'd have to go back and see what that was all about.

Shall I do this as a later mod, or start with it?