Digikey supplies most of the parts needed to build Nelson's latest Bride of Son of Zen preamplifier, featuring no negative feedback, balanced operation and single MOSFET gain stages. The most important lesson of Pass' designs is that single-device, single-ended (per polarity) topologies are elegantly simple and tend to sound very good.

As mentioned below the parts to build a balanced pre-amp are hard to get. One alternative would be to buy a commercial product like the Adcom GFP-750. This is essentially a revision of Bride of Son of Zen done by Nelson Pass for Adcom. Like Bride of Son of Zen, gain in the 750 is provided by a single differential pair with active current sources, and no follower stages. The 750 differs from Pass' published DIY designs in having current sources to the positive rail, where the DIY have current sources only to the negative potential and a resistor load to positive (or resistors in both positions). The second current source should further improve power supply and common mode rejection, etc. The 750 is capacitor coupled with film bypassed electrolytics on input and output, a no-no in my book. Electrolytic capacitors should never be used in the signal path. An easy mod would be to change these to film caps as in BoSoZ. Pass himself describes the 750 thusly in a forum at AVS:

"the circuit is very much based on the single differential pair topology of the Aleph P and the Balance Line Stage which appears as a DIY on our web site."
All devices are IRF HexFETS. According to Wes Phillips' review in the in March 1999 Stereophile, the GFP-750 uses relays for input selection, and it appears to have a 4 gang potentiometer as a passive attenuator, followed by the MOSFET pair, with a repeated MOSFET pair presumably buffering the processor/tape loop. It also features a switchable passive mode where the gain block is probably bypassed or shorted across. Wes called the $1250 Adcom one of the best preamps he's heard, comparing favorably to his $6.5k Mark Levinson. He almost seemed afraid to recognize such a good product at a relatively low price. I've seen a used GFP-750 advertised on the net for about $800 and it's tempting to buy one rather than build my own. A recent issue of Stereophile placed the Adcom GFP-750 as the least expensive preamp in Class A.

My Next Preamp

If I do choose to build a preamp, my current thinking is to use an audiophile stepped attenuator (rotary switch with resistors), 4 pole rotary switch for input selection and MOSFET differential pair a la Bride of Son of Zen, but with current sources like the Adcom. The likelihood of getting and maintaining balance and channel tracking over the rotation of the attenuator is much greater with a stepped attenuator than with ganged pots. It is somewhat doable using the pots, but really difficult and expensive. Stepped attenuators are also expensive but should be more consistent. Relays for input selection affords such features as remote/computer control, but it also adds more power supplies, electromagnetic noise, back-EMF, and often digital complexity. Presumably for a relay-selected balanced preamp you use pairs of 2-pole relays per input, with the coils wired together. (I haven't been able to find high-quality, miniature 4-pole relays, but maybe I'm not looking hard enough.) However if the goal is input selection and you don't mind turning a selector knob, a rotary switch should work fine and have a lot less complexity. Good 4 pole rotary switches are surprisingly expensive, however, at about $50. A rotary switch is also very compatible with hand-wired, point-to-point connections, whereas relays are more naturally used with production-based printed circuit boards.

I've started to get some of the key parts for building my own balanced preamp, after the Pass designs. I have a 10k 4 section stepped attenuator and 5 position balanced input selector rotary switch from DACT (Danish Audio ConnecT). Saw their ad in Audio Electronics. Their U.S. distributor wasn't yet set up when I ordered directly from Thailand! They use the Swiss-made ELMA rotary switches with their own surface-mounted metal film resistor arrays on proprietary PCBs. They claim 0.05 dB matching and tracking precision, which as I mention next is critical for balanced circuits. I have a hard time believing this kind of precision could be attained by the hobbyist except by hand-selected precision resistors (which few have access to), and it could almost never be obtained using ganged potentiometers.

I'm going to try to use 3 inch tall, u-channel aluminum extrusions and 1/8 inch sheet aluminum to build the chassis. I plan to build the input selector and attenuator into their own box with passive outputs within the larger preamp chassis. Also want to incorporate the Headroom headphone amplifier module. It will probably be switchably tapped off the attenuator. The headphone module will have it's own power supply, but don't know if the main gain stage should have an internal or external DC supply. The latter offers the possibility of noise pickup through the power cable, but isolates line noise, transformer hum, etc. from the active electronics. Having the power supplies in the same box makes for mechanical simplicity and shorter cable runs.

Update regarding chassis: Front Panel Express offers aluminum chassis with panels custom designed using free CAD/CAM software and made using CNC equipment. Looks promising.

Nelson Pass and others argue against passive attenuation at the input of a control amplifier for several reasons. The impedance seen by a signal is much more favorable if it sees an active input stage first. Attenuation after the gain stage, they argue, has the desirable affect of attenuating the input stage's noise along with the signal. In contrast, attenuation at the input means impedance changes more as the attenuation is adjusted, and the gain stage that follows amplifies all the noise. On the other hand, placing the gain stage after the attenuator should have the advantage of presenting a stronger, better-controlled signal to to the amplifier that follows.

Balanced Preamp Practical Challenges

A practical balanced pre-amp requires either 4 gang pots or attenuators and input selectors, or input relays or transistor switches and digitally controlled attenuators. In order to maintain good signal-to-noise the positive and negative halves of a balanced circuit need close matching. The same is true across channels to maintain channel balance throughout the attenuation range. Getting 4 mechanically ganged pots to track well together seems inherently troublesome. Crystal and others make analog-domain attenuators (sometimes called Multiplying DACs or MDACs) that are preferred over digital-domain attenuation which commonly throws away bits. Even when implemented with adequate bits to preserve the original resolution, digital domain attenuation literally recomputes the signal, altering the information, which is inherently un-hi-fi. Another inherent flaw in digital attenuation is that the unless the DAC can scale its output (which I've never seen) you still lose bits. A DAC with a fixed output level must use fewer bits to create a lower amplitude (attenuated) signal. On the other hand, electronic analog attenuators are usually transistor switched, which of course adds semiconductor junctions for the signal to go through. I'm a minimalist; the less a signal needs to pass through, the better.

Some digitally controlled balanced preamps use attenuators on inputs and outputs. This adds electrically stable programmable attenuation per signal source/output but also adds more layers of circuitry. Unfortunately it looks like there's no easy answer or perfect solution. Input relays at the jacks and digitally switched analog attenuators on the outputs seem the most practical. Off-the-shelf MDACs and switched analog attenuators reportedly sound excellent in applications like Mark Levinsons (Madrigal) and Sonic Frontiers. Levinson/Madrigal may be using the Analog Devices AD7541A MDAC or a relative. I seem to recall from a review that SF is using Crystal Semiconductor. Presumably it's the CS3310 Stereo Digital Volume Control. The former is 12-bit and the latter 16, but the AD has no built in output stage, which may permit a better off-chip buffer implementation. The 7541 family appears mature and has second sources, such as Linear Technology's LTC7541A.

Acquiring the parts as a hobbyist may be difficult. Minimally using them requires some trivial digital work (up/down counters). Using a rotary knob to control the counters requires a photo-interrupter and associated quadrature detector, also fairly minor. Adding multiple functions to the knob (balance, variable source attenuation for level matching) requires some function selection logic and memory, perhaps in the form of latches. Microprocessors/controllers should be avoided for their radiated noise, but at some point the control functions start to get easier with one.

Here's a page with schematics for the Jensen 990 discrete op amp, and others. Looks like Nelson Pass has just published an article in Audio Electronics about building simple discrete Op Amps from MOSFETs or tubes. That sounds interesting....