What we are looking at here is the Fast Fourrier Transform (FFT) of the line output from my b-board buffer recorded at 24 bit, 96 kHz by an Onkyo SE-200PCI sound card. Upstream from the b-board is the Phonoclone 3 MC phono stage, connected to a Denon DL-103. The tonearm is Denon DA-307, and the deck is a Denon DP-2000.

Four recordings, taken 1) with music playing, 2) with the tonearm raised 3) with the phonoclone powered off and 4) with the b-board and all upstream components powered off.

True 24/96 data was obtained, measurements out to 48 kHz are possible, with -130 dB noise floor. (I was using Digionsound 6 to do the recording as Audacity truncates 24 bit recordings to 16 bit in Windows due to licensing issues. The FFT was generated in Audacity however.)

The soundcard's line input may have an impressive-looking low noise floor, but it's still useless for measuring line level audio devices like the b-board because the noise of the preamp/ADC...

Case report:

A set of Sapphire boards gave the proper V+, V- voltages out of the Z-reg, providing about 10.5 and -10.5 to op amp power pins. The output offsets were unusually high however, apparently at about 2 V in one board, and somewhat less in the other. Typically the offsets are in the order of +/-15 mV.

Changing out transistors and op amps did not help, and to all inspection the passive components were installed correctly and working properly. The offset voltages were extremely temperature sensitive. Measurements for the various circuit voltages were just screwy enough to be inconclusive.

I could ask for no more tests, so requested the boards be sent back to me. I found the circuit basically worked as expected, but the offsets were indeed high on both boards, though I measured 0.6 V max rather than 2 V.

***** stop here and make a guess *****

Blowing on the board through a soda straw, the offset shot up when I blew on...

Straightforward transplant. Out with the old (anyone want them?) in with the new. Re-used the OPA134 op amp and my dog-eared pair of 0.47uF Multicaps.

On powering up I discovered that with the specified 10 ohms in R9,10 the output bias current was upwards of 200 mA and things were getting a bit toasty. I paralleled a second 10 ohm resistor, dropping the resistance to 5 ohms and dialing back the output bias current to about 70 mA. Latest schematic revision has R9,10 values edited to match.

Currents stable. Heatsink temperatures around 50 C. Output offsets around 15 mV. No noise or hum.

Presently giving it some burn in time.

Update: I've ordered parts for small number of Sapphire 2.0 kits. The normal price will be $125, but as an introductory offer the first batch will be available for $100. Kit includes a set of boards and all the parts for the board. You need to supply the transformers and diodes, as well as a volume control, and the chassis hardware.

Update: boards are in stock, see photo.

Original here, diyaudio thread here.

November. That time of year for finally getting around to advancing some of my audio projects a little.

The Sapphire has remained in "rev 1+" for some time now, partly because of time constraints, partly because of the lack of popularity, and partly because it was already a re-spin of the beta version and worked just fine.

There were a few housekeeping things I wanted to add though, which have been included in the 2.0 revision.

- added a dedicated ground (GND) pad to connect to chassis...

Two headphone amplifiers sharing the same basic MOSFET source follower output stage.

When the source current and source resistance are optimized for the given headphone load and similar maximum output power (~50 mW at 1% THD), the distortion pattern vs. output power is remarkably similar.

One plot below is simulation, the other measurements. The J-Mo 2 simulation closely matched the actual measurements, it wasn't worth my while to generate a full simulated data set when I already had the measurements on hand. No reason to suspect that the Szekeres sim is inaccurate, either.

The take home message is the distortion characteristic of a MOSFET follower is what it is, and unavoidable. Take it or leave it, as it were. However - and this is key - if you don't optimize the stage for the headphone impedance, the distortion for a given output power will increase significantly.

As an aside: Greg did his homework with the original circuit....

I did up the X-reg circuit in LTSpice.

Results shown below, together with the LTSpice .asc file you can use to play around with this yourself.

First attached image shows FFT for the rectified DC (green), reference voltage (red) and X-reg output (blue) for the designed-for 10 mA output (top) and a more punishing 100 mA (bottom).

Second image shows an LTSpice screengrab for the LT1086 with bypassed adj pin under comparable loading. Input voltage in blue, output in green. This is a reasonable approximation of a "good" IC regulator.

Last image shows a plot of the exported LTSpice FFT data for the X-reg and the LT1086-12V (Cin 1000uF, Cout 100uF) both at nominal currents of 10 mA. The LT1086-12V is a reasonable substitute for a generic LM7812, i.e. a "bad" IC regulator.

A typical op amp will have sufficient PSRR to mop of the residual noise from the bypassed LT1086. The fixed LT1086-12V, on the...

This is in response to several recent emails I've received, where people were having problems with, typically, one board having a bad V+ or V- regulated voltage output.

The number of cases relative to the number of boards shipped caused me to worry that a manufacturing error might have occurred, so at my request I had a customer return the phonoclone boards he had built to me for inspection.

I'm happy to report that the problem was traced to poor soldering technique, the boards themselves are fine. What had happened was solder had cooled before the component had fully settled, and pushing the component down to the board surface then tore the trace away from the bottom of the board, breaking the circuit.

Subsequently, thinking the transistors blown, he replaced them, doing a fair bit of damage to the pads of Q1, Q2.

Fortunately, I was able to fairly quickly set everything to rights, and the boards are now on their way back to him....

I always seem to end up optimizing my headphone amplifier circuits for higher impedance headphones, this mostly happens because I own a pair of 300 ohm HD-600s and it is tedious to design for both the voltage requirements of high impedance headphones and the current requirements of low impedance headphones.

Not impossible, just, for the class-A designs I seem to be building recently, increasingly large, heavy, and impractical.

Complimentary transistor circuits, however, offer the opportunity to swap voltage for current at something close to the same design cost. They are therefore a practical topology for efficient class-A power delivery into low impedance headphones. As a design experiment, my aim is to discover how far I can leverage an ultra-low-voltage, unity gain circuit for compactness without sacrificing sound quality.

Ok. Back-of-the-envelope calculations:

A typical 16 ohm in-ear-headphone has a sensitivity of 100-105...

This isn't my first attempt. It's been on my mind for a while: how to coax a diamond buffer into giving voltage gain, without resorting to fronting it with a op amp.

After reading a particularly gregarious thread over in the headphone forum, I'm more and more stoked on giving this a real shot.

Despite the (catchy) name I'm thinking pre-amplifier rather than amplifier applications.

update: I have have a quick and dirty sim up and running in ltspice. Curiously, the output distortion is 15 dB lower when the buffer runs open loop than when it is included inside the feedback loop. Intrigued. Currently under investigation.

update: refined the sim slightly, achieved -85 dB distortion levels at 0 dB / 1 kHz / 600 ohms running the output buffer open loop. Bandwidth is just under 1 MHz, adjusted by changing the feedback resistance. As before, performance sims out notably worse with the buffer
inside the feedback loop.
...

Douglas Self writes, He's someone who should know. Anyway, it doesn't take much digging on the internet to confirm beyond reasonable doubt that bypass caps should be as close to the op amp power pins as possible. So thinking about my previous experiments with bypassing the Sapphire, by adding bypass caps around the transistors I also effectively also added a bypass for the op amp, but a rather poor one as the power-pin-to-power-pin round trip loop distance is probably 10...