PC Preamp Project

I looked for the ideal PC to audio interface and was unable to locate an off-the-shelf solution. It should have tone controls including volume, balance and loudness compensation. It needs a high quality headphone output (high power, low noise and distortion, good impedance match) for 1/4 “ and 1/8” headphones. It needs a balanced line level output to drive a remote stereo as well as a balanced line-level input from a remote stereo. Balanced is needed to eliminate ground loops.

I also looked at WiFi and networked PC interfaces. These solve the PC to stereo problem and the ground loop problem. But they do not provide the return signal (stereo to PC), and also I suspect that they don't lend themselves to listening to the same material local to the PC and froma stereo. If they do, then they would require a short delay to the stereo. Ever listen to two channels of audio with a delay? This doesn't enhance the listening experience. I have a house-wide stereo and can play the same or different sources at several points in and out of the house.

It should have a high quality phono preamp to allow ripping vinyl to the PC line input. A selector switch to pick the sound source.

In addition, a 25-35 watt per channel amplifier to allow quality speakers to be used. THis can be bypassed so active speakers can be used.

Phono Preamp: Holman Design

For a phono preamp, I have a lovely board I built in the late 70’s. I owned an Advent 300 receiver at the time. This minimalist receiver uses a Holman designed phono preamp. The bipolar transistor design uses 6 transistors per stage instead of the usual 2. It sports a differential input configured as a current source loaded high gain stage, followed by a darlington X10 voltage amplifier, followed by a 2 pole, high pass filter with emitter follower output for rumble rejection. In 1977 I 'borrowed' the circuit and laid out a single-sided board for it. A friend and I built 4 of these boards for use in home stereos and in a disco mixer. With better components the noise performance was as good or better better than the Advent 300. Fortunately I still have a couple of these boards and one is now located in the upper right of the PC Preamp. It still sounds great. I worked at Hewlett Packard Medical division in Waltham at the time. Many of the parts are from HP lab stock. I thank HP for their generous lab stock policy.

Here is the schematic in .PDF for my version of the Holman preamp. Advent put the original schematic in the owners manual and it is now available on-line from various Advent 300 repair and upgrade web sites so I am not giving away any secrets. I changed all the original 5% resistors to 1% metal film, and use tantalum capacitors in place of the original cheap electrolytics. The Advent 300 upgrade web sites report degradation or failure of the original circuit due to the electrolytic caps drying out over time. Also they used 16V caps on 12V power supplies. This is a bit marginal and accelerates failure. For best reliability I use tantalum caps which don't exhibit this problem and perform better in most power supply circuits. Also I use 2:1 voltage derating: 25V caps in a 12V circuit. The original design used noise selected 2SC1345E for the NPN differential pair stage I used T0-92 2N5088 transistors for these. These provided a nice low noise level on the front end. When I A/B tested my design vs. the Advent 300 by listening with headphones and the volume cranked up, my noise was a tad lower. The original circuit used 2N5087s for the PNP constant current stage. I used metal can 2N2907s for these. Nowadays I'd use plastic 2N3906s for the PNPs. Holman used a darlington PNP MPSA65 for the voltage amp stage. I used two PNPs for that stage. For simplicity I used the same 2N5088 NPNs on the emitter follower output stage. The original design used non-noise-selected 2SC1345Es here. They had to use the rejects somewhere.

The PC artwork is an ancient hand-taped (pre-CAD), single sided desgn which I cannot readily publish since the artwork is long gone. The layout is not particularly critical as long as you keep signals short, flow from the inputs at one end to the outputs at the other, keep the power traces away from the inputs, and run a nice fat ground trace down the middle, between the channels. An ExpressPCB mini board would do this circuit very nicely. I laid one of the two channels but no the other. The other channel is a mirror image except for the transistors and power inputs, so be careful. This layout has not been checked for cap sizes, etc, so use at your own risk. Here are the ExpressPCB Schematic and PC layout. The software to edit them is at ExpressPCB.

Eliminating Ground Loops With Cheap Transformers

When coupling a PC to a stereo, one should be careful to avoid ground loops. A ground loop will generally cause annoying 60Hz hum in your speakers. They are caused by connecting two pieces of equipment with single ended cables when there is also a power ground or 3 wire power cable on each equipment. The house ground wiring causes small amounts of AC voltage to between the grounds of two pieces of equipment. Then when an audio cable is connected, that voltage induces a current in the audio ground lead ( the shield ) which induces a voltage at the equipment input. A decent stereo wants to see less than -80db of hum or about 100uV on a 1V signal. This is nearly impossible to achieve in the presence of a ground loop.

For consumer audio gear, single ended line-level audio on RCA jacks works OK as long as the system is simple enough and most of the system uses 2-wire line cords and decent power transformers. This is the case with most home audio gear. It’s when you hook up a PC or video equipment that ground loops can occur. It is very likely that a PC chassis and its I/O connectors are grounded, even a laptop that uses a 3 wire power adapter. It is also very likely that the stereo has a chassis ground or 3 wire line cord somewhere. Some stereo systems are entirely 2 wire, but sneak ground paths through a cable box RF shield are common. In any case, balanced inputs and outputs eliminate the ground loops. If you have ever tried to trace down a ground loop and the stereo are grounded somewhere you will know the pain. Planning for them in advance is good.

There are a couple of ways to reduce ground loops when passing audio from equipment to equipment. Differential inputs with good common-mode rejection will do it. The problem with this approach is that an unbalanced line output often will have two different impedances in its ground path (near 0 ohms) and its signal path: typically 100 to 1,000 ohms. So to get 80db of CMRR with this approach, the input impedance must be 80 dB or 10,000 x the difference in impedance or 10,000 x 1,000 = 10M ohms. It is hard to build such an amplifier. It requires clever bootstrapping, precision matched resistors, trims, or worse. The other solution is to use audio transformers. Unfortunately these generally need to be shielded, and high quality audio transformers are in the $50 and up price range. You need 4 to receive and send stereo. They are large and heavy.

For coupling transformers, I have previously used 600:600 ohm modem transformers with good success. Most modem transformers are only specified for about 400 Hz to about 5KHz, but have actual response far beyond. The challenge is in the low frequencies: a transformer that can handle 20Hz full amplitude sine waves is hard to find. It would seem impossible to find a transformer that is specified for 400Hz but can pass 20Hz well, but there is a loophole. Many modem transformers are designed to pass about 50mA of DC current on the telephone line. This requires a larger core and heavier wire that can avoid saturation while still passing the 400Hz. So most transformers designed for DC can pass 50Hz at 1.0V p-p and some can pass 20Hz without the DC.

At the high frequency end, all the modem transformers I have tested passed up to 25 KHz before cutoff. As far as shielding, I have used them unshielded within 6" of a CRT monitor without hearing adverse effects. They do pick up hum if they are not driven by a low impedance though. But when there is sound there is drive. These particular transformers are Atech ATS-166, purchased from surplus dealer All Electronics for $1 each. All Electronics used to have other Atech models available but they distort on 20Hz test signals at 1V p-p. The ATS166 is by far the best $1 audio transformer I have found. Seriously, they work very well.

As far as driving and receiving impedance, they work well with a low impedance drive (0 to 200 Ohms) and a load impedance of 1K Ohms or greater. Get a couple, test their frequency response. Stick them between any line-out / line in and have a listen. I think you'll be impressed.

Tone Controls

Here is the schematic for the other boards, the switching and the connectors in .pdf. Here is the original file in .SCH format for ExpressPCB. Download their excellent free software to edit it.

The tone circuit is a classic treble and bass Baxandall circuit built with the excellent NE5532 audio Op-amp. Like most active filters, this is succeptible to source impedance, so a unity gain buffer stage precedes it to provide a low source impedance. Since the tone circuit uses an inverting configuration and I want all output signals to be non-inverting, the balance circuit re-inverts the signal. I generally prefer to have the volume control last in the signal chain. That way when the volume is reduced, any noise from the previous stages is also reduced. But this application requires a low impedance output driver after the volume to drive up to three amplifiers: the headphone amp, the external powered speakers, and the main amplifier.

For loudness compensation, there are simpler circuits but they typically require a tap on the volume control. This one is pretty reasonable. I DC couple its input and output and then AC couple only after the balance stage. The DC bias errors accumulate for the tone, balance and volume circuits. Worse they are a function of the Bass, Volume, Loundness and Balance controls. Then all magically fixed by the AC coupling cap at the output. The problem is that this 100mV or so of DC has an unknown polarity. So the 4.7uF coupling cap should be non-polarized. So much for using a tantalum. Could use two 10 uf tantalums in series, back to back. Could use a film cap (big and expensive). Could use a NPO electrolytic. I use a 4.7uf tantalum and figure it can handle 100mV of reverse polarity.

Headphone Amplifier

For convenience, the headphone amp circuit is mounted on the tone board. It consists of my favorite NE5532 wired as a X4 non-inverting amplifier with a discrete, complimentary emitter follower stage to boost the drive current. The follower uses 3 diode drops (2.1V) to bias 2 transistor Vbe junctions plus two 33 ohm resistors. So the 66 ohms of resistors have 0.7V across them causing a bias current of ~10mA to flow. This is enough to reduce crossover distortion to a very low level. This circuit will drive a nice clean 0.5 watt into 50 ohm headphones so be careful with the volume adjust else you may damage your hearing.

The current limt is a simple 47 ohm resistor in series with the output. Crude but effective. The 47 ohms also isolates the amplifier output from capacitive loading caused by long headphone cables. It's a good idea to add a resistor to any audio line output for this reason.

Amp

The space on the left side is set aside for the 25W per channel amplifier. This will be a board I designed, based on two LM3886 amplifier ICs mounted to the left side. An external heat sink will also be mounted on the left side. The toroidal AC transformer will need to be increased to about 50W RMS which is considerably larger than the present transformer. It will stil fit in the 1U chassis size though. I have a design and layout for an ExpressPCB MiniBoard board to do a stereo LM3886 amp. Just haven't gotten around to building it yet. The board design sports:

Power 20 to 50+ W into 8 ohms: maximum power is supply voltage dependent
2 Stereo amps fit on one MiniBoard: they need to be cut in half the long way. $10 per board ($59 / 6)
Uses National LM3886s
Power supply is separate: Toroidal transformer, diode bridge and 2 electrolytic caps. When I finalize these parts I'll post them here.

Here is the .PDF for the schematic, plus the files for the schematic and PCB artwork. If you build it up, please let me know how it goes. You'll have to cut-and-paste the amplifier onto the other half of the board to get 2 amps per mini-board. Make sure you leave room for a saw cut down the middle. G10 is tough to cut. I use a bandsaw or hacksaw and then file the edges smooth.

For a heat sink I am thinking of a larger PC type heat sink sliced to 1.5" wide. The LM3886s are available in an isolated case so they can simply be screwed to the side of the chassis with a 4-40 screw and some heat sink compound.

I currently don't need the amp section since I am using old Cambridge Soundworks powered speakers which sound quite decent for their amazing low cost of $49. Also I don't have a set of small bookshelf speakers at this time. Another project.

Fortunately the +/- 12V supplies can operate off of the Amplifier unregulated power as well so the smaller transformer can be removed. I've built a couple of amps based on the LM2886 and am very happy with the performance as well as the cost, and size.

Construction and The Chassis

Since this project will live in my computer room, I wanted a decent looking package to hang from the shelf above my PC monitor. I like the idea of 1U rack size (1.75"), and this provided space for all the circuitry, controls and I/O. The enclosure is 10" deep x 15" wide, and has room for an external heat sink along the left side. It will mount via the rack ears on simple angle brackets with 2 screws. The rear will be supported by a bracket. This same size package could be used under a monitor as well.

For construction I used sheet and angle aluminum. Home Depot has 2" x 2" lightweight (0.060") angle bracket. I cut this down to 1.7" x 0.25" for the sides and back. The front is a 1 3/4" (1U), 1/8" aluminum panel. These can be purchased in a nice brushed aluminum, black anodized finish. I chose basic sheet metal and have not decided how to finish it yet. The bottom and top are 0.060 aluminum. To mount the front to the sides and base, I used 1/4" x 1/4" bar stock. Unfortunately I could not locate aluminum so I used steel. I drilled and tapped holes in it to accept 4-40 hardware. Drilling and tapping steel is no fun. I recommend buying 1/4" square stock from a local metal vendor or from McMaster Carr. And to prevent breaking your 4-40 taps, use a tapping lubricant or at least 3in1 oil.

This approach was a lot of work. In retrospect, next time I think I will use 1/4" or 5/16" X 1.5" aluminum bar stock for the sides, and drill and tap it for the front, top and bottom screws.

For board construction, I used prototype boards where a PC board was not available. When I need to build more than one of something, ExpressPCB is the way to go. But for one-offs, hand wiring is fine. I like the Radio Shack breadboards for building small DIP prototypes. Their smaller board can take 1-2 DIPs, the larger one was used for the preamp / headphone amp board. These can support up to 3 rows of DIPs. I needed many connections to the front panel controls, so used the center row for these connectors. I use the Molex / Waldom single row 0.1" connectors for most board-to-wire connections. These work well with hand built breadboards and are available in straight and right angle versions. To build the cables, the pins are crimped on #24 insulated wire and poked into the housings. The Radio Shack crimp tool works well with these pins.

The transformer board was made from a scrap of G-10 proto board and hand-wired. The input and output conections are both 6 pin Molex.

The +/- 12V power supply board is courtesy of my employer. We built this board because an internet search for a small and simple bipolar linear supply board came up dry. It is that funny trapezoid shape because it was built on the scraps of an ExpressPCB mini board from another design. With ExpressPCB mini boards, all boards are the same size so if you want a smaller board, you cut off the excess. But the excess can be used for other board functions and I hate to see it go to waste.

Input cables are shielded. Output cables are unshielded. As I don't consider R-L channel separation a real important spec in a stereo. I use three wire cable everywhere: 2 shielded condustors. In fact I use this stuff for my 30' PC to stereo runs. Sure, there will be some hight frequency crosstalk between R and L, but probably not much compared to my head and ears.

All boards are mounted on 3/8" 4-40 spacers. The balanced input and output cables still use RCA jacks, but these are insulated from the case by plastic shoulder washers.