In support of the LM4562 opamp, National Semiconductor published Application Note AN-1651, Keeping Up with the Expanding Demands of High-Performance Audio. The author, Joe Curcio, showed designs for two phonostages and an electrostatic headphone driver. I built the second phonostage design, a two stage moving coil preamp with split active/passive RIAA equalization and a DC-nulling servo.
Here’s the schematic from the application note, complete with original errors. Mine is slightly different, as described later.
I built two versions of the circuit.
The first one was done point-to-point, aka “deadbug” style, aka “Manhattan” style. In this technique, a copper-clad board is used as both a platform for the parts and the ground plane. Some components are glued to the board. Leads are soldered directly to each other. Some components are held airborne by their leads and the leads of the other components they’re soldered to.
This version had problems with a bit of RF pickup. I helped a friend build a copy of mine and his produced mysterious intermittent thumping noises. We could never find the cause, though it was probably an elusive cold joint.
For the second version, I designed a PCB and had it made by expresspcb.com. My PCB still had RF pickup problems. My friend built the PCB version too and his has fewer RF issues than mine. At least the thumps are gone.
In an attempt to reduce susceptibility to RF pickup, the circuit was modified by
- adding RF gain compensation caps around U1 and U2
- adding ferrite beads on the inputs
While reading Jung’s 1995 voltage regulator articles, I found an interview with Allen Burdick (Benchmark Media Systems) in TAA 1995/4 that discussed RF gain in op amps and distortion. He has a criteria to leave at least 20db of unused gain at all frequencies in a gain circuit so there is enough open loop gain available for negative feedback to correct for any RF-induced errors in the circuit. This is accomplished by a gain compensation cap across the feedback resistor that creates a low pass filter. The pole sets the upper frequency below which the gain is flat and above which the gain curve decreases, roughly paralleling the gain bandwidth curve 20db above it.
In my preamp, the HF rolloff for the two gain stages combined is about 0.06db at 20kHz, 0.23db at 30kHz, and 3db at 100kHz. The compensation cap values were determined graphically using a plot of the OPA1611’s gain bandwidth, and then verified by modeling the filters in SPICE and combining them in a spreadsheet.
Even with these changes, RF was still a problem, so I have since tried
- enclosing the board in a box made of Permalloy
- adding RF filtering on the power supply input
There is still a little RF pickup that can be heard as impulse noise patterns at random. I live very close to a cell phone tower. The noise can be conjured at will by bring a cordless phone or a cell phone close to the phonostage. My friend’s build doesn’t have this problem, at least with ambient RF. I plan to eliminate two more possible causes: replacing the power rail circuitry U4/U5 with one of my Hubble regulators, and adding load isolator filters on the outputs.
Component Selection and Noise
Since I was so satisfied with the sound of the deadbug version, I decided to go for the lowest noise and distortion that I was willing to pay for with the PCB version. The RIAA caps are boutique polystyrene caps, but not too expensive. The cost of available Teflon caps was too high for me to stomach. The cap tolerance was not that great, so a friend measured them on a lab meter and I recalculated the RIAA resistors for the same network response as the original design.
The noise of the deadbug build was too high for my taste: With focus, I could barely make out the noise signature of the preamp below the surface noise level of the vinyl when playing back low level passages. I purchased and built Jim Hagerman’s Piccolo MC headamp board and put this in front of the AN1651/Curcio after dropping the gain of the first stage. This was a failed experiment as there was no noticeable change in the noise level.
I improved the noise performance by about 8db (on paper) by scaling down the resistors in the input stage by about 4.7x and switching to the OPA1611 opamp. The OPA1611 opamp has better noise specs than the LM4562, so I used it for U1, U2, and U3. This change alone should improve the noise in U1 by 3db. Rescaling the resistors and caps in the first stage reduced the current noise through the resistors for another 5db noise improvement. This required changing C4 from 100nF to 470nF, which is a pretty large cap in polystyrene. Because the lower value resistors could have loaded the first stage too heavily, I did some current calculations. The original network requires 45µA from U1, while the scaled network requires about 200µA, well below the maximum output current for the OPA1611. Listening confirms that there is a very worthwhile improvement in the noise, but the precise reduction is not known.
A friend suggested using IRC tantalum nitride thin film resistors because of their non-magnetic materials and construction and excellent performance. They’re rather pricey at $0.70 each from Mouser in the 1206 SMD package, but I am glad I used them.
You many find this spreadsheet I made useful for calculating the noise of an inverting opamp stage and for comparing different opamps.
My previous phono preamp/reference was the Yamaha C-2a. The AN1651/Curcio preamp has no apparent difference in tonal balance. The highs are much cleaner, but neither higher nor lower in level. The Yamaha had a significant amount of power supply noise and hash, AC (power line) buzz and hum, as well as component noise. This phono preamp is vastly cleaner sounding in these areas.
Ticks and pops are much less apparent. No or less ringing, faster recovery? Several huge pops that I know and love (!) on some records aren’t nearly as obnoxious as they used to be. I always thought if ticks and pops are less prominent it meant that the cartridge was not extracting enough low level signal. But the cartridge didn’t change, so the problem was related to the Yamaha’s electronics.
Certain sibilant and hashy torture tracks I use for testing are very much cleaner if not completely undistorted. These have female voice, hand claps, loud continuous snares, etc. They sounded like sandpaper when not reproduced right.
Whether because of the lower inherent noise or less distortion due to layout and parts, I hear more and clearer low level information. Sounds sound more real.