Projects‎ > ‎

High Gain, Low Noise, Low Distortion Microphone Preamplifier

Introduction

One of the most crucial aspects of capturing a quality recording of an audio signal is using a microphone preamplifier that can perform to desired standards. The desired specifications for a preamp may vary according to the sound source being recorded and the microphone used to capture the signal, and several crucial specifications to be considered involve gain, noise, and distortion. Specifically, gain must be adequate to amplify a given signal to a level suitable for recording, noise should be minimized, and distortion at the desired gain levels should either be minimized or pleasing to the user (more on this in a moment).

Commercially available preamps range widely in price, and a casual perusal of spec sheets reveals dramatic price increases for incremental advances in performance. The goal of this project is to explore, through implementation and measurement, the design differences between reasonably good preamps and excellent ones, and how well theoretical improvements in circuit topology perform in practical application. Because "pleasing" distortion cannot be objectively defined and limits the versatility of a preamp, the ideal for the purposes of this project will be the minimization of distortion (in other words, the most accurate reproduction possible of the microphone signal entering the preamp).


Specifications

Tentatively, specifications are chosen according to the performance of two commercially available preamps generally regarded for clean, accurate performance: the Grace Design M1011 and the Millennia HV-3C2. Also, the preamp will be specifically designed to perform well with quiet sound sources (such as quiet vocals) captured with the Audio-Technica AT4050 condenser microphone3. 
  • Gain: up to 80 dB
  • Total Harmonic Distortion (THD) + Noise @ 60 dB gain, +20 dBu output level: < 0.005%
  • Input: Balanced XLR
  • Output: Balanced XLR
  • Input Impedance: 1k Ohms
  • Output Impedance: 300 Ohms
  • +48 V Phantom Power


Members

Todd Rohrer <todder1@vt.edu>


Mentor

Bob Lineberry


Design

Overview

Microphone preamp design approaches vary, but might reasonably be broadly grouped into two categories: transformer-coupled and transformerless. Although transformers are commonly used in preamp inputs and simplify certain aspects of the design process, they are heavy, expensive, and, most importantly, tend to "color" the audio signal (add distortion). As one of this project's primary design objectives is to minimize distortion, a transformerless input was chosen. 

Two different topologies were considered for the preamp circuit. The first, based on the classic differential amplifier circuit, utilizes several op amps (often on a single chip) to accept a differential input and output a single-ended signal, which can be cascaded with additional single-ended gain stages as needed to attain a desired signal level, which can then be re-balanced in a final line-driver stage. 

The second topology considered involves keeping both the positive and negative sides of the differential signal separate throughout, such that the signal is never unbalanced. This requires more components, but should improve common-mode rejection of the circuit. The design for this project will use the "through balanced" approach with the goal of optimizing noise performance.

Microphone Specs and Signal Levels

The preamp's target noise, gain, and input impedance specifications are chosen in accordance with the specifications of the AT4050 microphone. A common rule-of-thumb states that a preamp's input impedance should ideally be at least 10x the output impedance of the microphone being used. The AT4050 has an input impedance of 100 Ohms, so the target input impedance for the preamp design will be 1k Ohms3

Target values for peak-to-peak input noise levels can also be calculated from microphone specifications. One rule-of-thumb states that a noise floor 100 dB below "normal" listening volumes is effectively inaudible. The following calculation uses this figure as a reference to calculate target equivalent input noise:

Estimated quiet vocal volume level: 60 dB SPL = 0.02 Pa
AT4050 Sensitivity = 15.8 mV/Pa
RMS signal level of quiet vocal using AT4050 = (0.02 Pa)(15.8 mV/Pa) = 316 uV
Peak-to-peak signal level = (316 uV)(1.414)(2) = 894 uV
Target noise floor = -100 dB = 20log(Npp/894uV) => 10^(-5) = Npp/894 uV
Npp = 8.94 nV

Therefore, the target equivalent input noise level for the desired performance is 8.94 nV. This level of performance may prove difficult to achieve. 

Gain Stages

The OPA1612 op amp4, selected for its excellent distortion and noise characteristics, will be utilized for the input stage, as well as all subsequent gain stages prior to the final, line-driver stage. According to the datasheet curves, the op amp has at least 60 dB of open-loop gain up to almost 100 kHz; allowing for 30 dB negative feedback (a rule-of-thumb minimum based on Bob's personal experience), a single stage should therefore be able to comfortably supply 30 dB of gain. What is not immediately apparent is whether better noise performance can be achieved through, for example, using a single 30-dB gain stage as opposed to two 15-dB gain stages.

A SPICE simulation, using sufficiently accurate models, should provide a good clues as to which approach to pursue. The OPA1612 TINA-TI (Texas Instruments' SPICE software) model ostensibly models noise performance. Using Art Kay's op amp model testing method5, I compared the simulated results with the datasheet curves (see Fig. 1 for simulation results). While not perfect, the results are adequately similar to the datasheet curves to lend some credence to circuit simulation results. Figs. 2 and 3 show the two simulated test circuits, each producing approximately 30 dB of gain.

Fig. 1: Simulated voltage and current noise for the OPA1612 TINA-TI model.
 

Fig. 2: Single 30 dB gain stage.


Fig. 3: Dual 15 dB gain stages.

The simulation results, shown in Figs. 4 and 5, favor the single gain stage, though the difference is not particularly dramatic. Achieving 80 dB of gain in a through-balanced configuration will therefore require 3 gain stages, followed by a unity gain line driver, as depicted in the block diagram shown in Fig. 6. Although fine-tuneability is not a design goal for this project, the ability to at least select between one, two, or three stages of gain would enable the preamp to be used and tested with a range of sound sources of varying volume. The gain stages will be switchable via 3-position pin headers and jumpers.  

The Burr-Brown BUF634 will be used for the line driver stage, due to its excellent performance driving capacitive loads. Typical microphone cable has a capacitance of about 100 pF/m; the BUF634 has excellent gain and phase performance at frequencies beyond 1 MHz when driving 1 nF loads. Therefore, the BUF634 should allow the preamp to drive microphone cables in excess of 30 feet without any problems6.


Fig. 4: Output noise for single 30 dB gain stage.


Figure 5: Output noise for dual 15 dB gain stages.

Figure 6: Basic preamp block diagram for through-balanced topology.

Through-Balanced or Single-Ended?

Although a fully-differential signal path seems, in theory, to be the best approach for optimizing common-mode rejection, there are several factors that could potentially impact the performance of such a design in unexpected ways. Perhaps the most concerning is the fact that maintaining two effectively separate signal paths throughout the amplifier circuit could result, through inadequate thermal tracking, in gain asymmetries. However small these effects may be, they may be sufficient to at least negate any benefits in common-mode rejection attained through the employment of the through-balanced topology, and could in fact result in worse performance than attained with single-ended gain stages.

In the event that a through-balanced architecture is employed, there is some question as to whether better noise performance can be achieved through the use of single op amps (which lend themselves more readily to the use of guard rings) or dual op amps (such that both the positive and negative sides of the signal can be amplified on the same chip at each gain stage). A third option for implementation of a through-balanced topology is the use of a single fully-differential op amp, such as TI's OPA1632, for each gain stage8. This third approach could potentially provide better common-mode rejection than the other two options, as each gain stage would be implemented on a single substrate, maximizing interchannel thermal tracking.

Comparing Topologies: Simulation

In order to determine the best design approach for the desired level of performance, I will design a test board consisting of a single gain stage, utilizing jumper-selectable circuit topologies: one utilizing single op amps, one utilizing a dual op amp, one utilizing a fully-differential op amp, and one utilizing a differential amplifier chip followed by a unity-gain balanced output driver. The purpose of this design will be to measure which approach yields the best noise performance. This will determine which topology is used in the final preamp design.

Simulating the test circuits in Tina-TI will provide some insight as to what results can be expected from this test.

Passive Component Selection

The proper selection of passive components can make a significant difference in audio performance. Different types and values of resistors and capacitors can vary widely in linearity and noise performance for a given application. Douglas Self discusses some of these considerations in his book Small Signal Audio Design7. This design will utilize MELF resistors and acrylic film chip capacitors to minimize noise9 10.

Power Supply

+/- 15 V Supply Rails

A well-designed power supply is essential to achieving excellent noise performance in an audio circuit. For this design, two 24 V "wall wart" supplies will provide unregulated DC to a Texas Instruments TPS7A30-49EVM-567 evaluation board, employing TPSA4901 (positive) and TPSA3001 (negative) voltage regulators11 12 13. This evaluation board is inexpensive and features an optimized layout; construction of a quiet power supply should only require housing the evaluation board in a steel enclosure with the appropriate connectors.

+48 V Phantom Power

The AT4050, like many condenser microphones, requires a +48 V "phantom power" supply, which is applied to the signal lines. This design will utilize TI's TPS43060 DC-DC boost converter to step the +15V supply rail up to the +48V required for this purpose14. As in the case of the selected voltage regulators, an evaluation board, which can simply be purchased and housed in a steel enclosure, is available for this boost converter15.

A Note About Grounding

A topic worthy of significant consideration in this design process is grounding. Noise minimization in an audio circuit such as this one requires painstaking attention to proper circuit layout, with consideration given to ground return currents. In particular, return currents from should never be allowed to modulate ground impedances seen by the critical first gain stage, as this could result in significant noise or distortion. Circuit layout will therefore be of critical importance to the performance of this design. As the author has no prior experience in PCB layout, a great degree of mentor guidance will be necessary in this endeavor.    


Build

Power Supply Prototype

In order to minimize noise, the power supply and audio circuitry will be isolated from each other in separate steel boxes. Connections are made between the boxes by way of isolated BNC jacks that can be connected using jack-to-jack connectors when the boxes are placed next to each other. Figs. 7 and 8 depict  the first prototype build for the power supply. Note that the chassis ground has not yet been connected; this connection will be made via a small machine screw and star washer located roughly between the ground terminals at the input side of the board and connected to the terminals via the shortest length of wire possible. 

Several issues were encountered in building this first prototype. The initial decision to locate the BNC jacks close to the bottom of the enclosure wall beneath the indicator LEDs resulted in a mechanically stable and visually appealing design, but may likely lead to crowding issues in the enclosure containing the audio circuitry. This is due to the fact that the jacks of the power supply and audio circuitry enclosures must align with each other, and the audio circuitry will require more space inside its enclosure. A better solution would have been to locate the jacks near the top of the enclosure, in line with the indicator LEDs, allowing the audio circuit to occupy the space beneath the jacks in its enclosure.

Figure 7: Power supply prototype.

Figure 8: Power supply wiring.

Power Supply Noise Measurement

Because the noise specifications for the voltage regulators used in this supply are in the microvolt range, acquiring accurate measurements of noise performance with available test equipment poses a challenge. The smallest amounts of electromagnetic interference can cause errors that become significant due to the extremely low levels of the signals being measured. Jim Williams and Todd Owens of Linear Technology have designed a bench setup that minimizes error in such measurements16. This design includes a 60-dB amplification stage (to amplify noise to a sufficiently high level that it can be accurately measured) and a filter circuit (to bandlimit noise to the spectrum of interest). The amplification and filter circuitry is enclosed in a separate enclosure from that containing the circuit under test (a paint can or similar tightly-sealed steel enclosure). Fig. 9 shows Williams and Owens' noise measurement setup.

Figure 9: Noise measurement bench setup16.

Williams and Owens' documentation of this setup includes a schematic for the filter and amplification stages, but no layout. This circuit appears to be a good low-level noise measurement solution, but a good layout will be critical to its effectiveness. One possible alternative is to modify a circuit presented by TI's John Caldwell as an active crossover for two-way loudspeakers17. As is the case with the Williams/Owens filter design, Caldwell's crossover features Butterworth highpass and lowpass filters, as well as a suggested layout. However, the circuit would need to be modified in several ways. Obviously, the filters would need to be cascaded for the measurement circuit, as opposed to connected in parallel, as in the crossover circuit. Also, the component values of the filters would need to be changed in accordance with the desired high and low corner frequencies. The tweeter time delay and attenuation stages of the crossover circuit serve no purpose to the measurement circuit, and 60 dB of gain would need to be incorporated into the circuit (perhaps through modification of the attenuation stage). A perusal of the crossover schematic and layout suggests that this might be achieved through cutting several traces on the TI board, adding several jumpers to reroute the signal accordingly, and changing or omitting a handful of components. The input buffer and baffle step compensation stages of the crossover circuit could, with minimal difficulty, be converted into two cascaded 30 dB gain stages in order to obtain the desired 60 dB of signal gain. However, Bob feels that in order to ensure lowest possible noise performance through maintaining control of component selection and layout, the best option is to lay out the 60-dB measurement amplifier/filter from scratch. As Bob is building a low-noise power supply for one of his own projects, he will design and layout the measurement amplifier/filter. Clearly, the noise measurement setup has become a design project in its own right.

In order to reduce the number of possible noise sources, the amplification/filtering stage will be powered by batteries, using voltage regulation for improved performance. The test circuit will also be battery-powered, but will not require additional regulation. Connections between the enclosures housing the test and amplification/filter circuits will be made via BNC-BNC adaptor connectors to reduce the possibility of noise introduced on cables.


Sources

2. Millennia HV-3C spec sheet: http://www.mil-media.com/hv-3c_ps.html
3. Audio-Technica AT4050 spec sheet: http://eu.audio-technica.com/en/resources/AT4050.pdf
7. Self, Douglas. Small Signal Audio Design. Amsterdam Oxford: Focal Press, 2010. Web. 
9. MELF resistor datasheet: http://www.vishay.com/docs/28713/melfprof.pdf.
10. Acrylic film capacitor datasheet: http://www.cde.com/resources/catalogs/FCA.pdf.
11. TPS7A30-49EVM-567 user's guide: http://www.ti.com/lit/ug/slvu405/slvu405.pdf
15. TPS43060 Boost Evaluation Module datasheet: http://www.ti.com/lit/ug/slvu828a/slvu828a.pdf.
16. Williams, Jim and Owens, Todd. Linear Technology Application Note 83: Performance Verification of Low Noise, Low Dropout
17. Caldwell, John. TI Precision Designs: Analog, Active Crossover Circuit for Two-Way

Project Notes

Feb. 12, 2015

During the meeting with Bob this week, we compared the noise floor of my AT4050 with that of his reference mic; my mic was significantly noisier. The datasheet indicates that its self-noise is respectably low; however, its sensitivity is lower, which raises the effective noise floor when sufficient gain is applied to bring a signal up to the desired level. Bob suggested that the practical solution to the problem that inspired this project (insufficient clean gain for recording quiet sound sources) might be better addressed with a more sensitive microphone. Be that as it may, I plan to continue with the preamp build.

Bob has also brought it to my attention that the problem with my current preamp may not actually be a lack of sufficient gain, but a lack of adequate headroom. It seems hard to believe that device supposedly designed to output a professional line-level signal would have insufficient headroom to handle a vocal amplified to this level without clipping. Another possibility is that the preamp's design attempts to achieve too much gain in a single stage, resulting in the harmonic distortion that I've detected. I've borrowed the AMP Lab's Yamaha MG06 mixer (featuring 2 on-board 60-dB mic preamps with phantom power) in order to conduct a bit of subjective testing: if I can amplify a signal to line level with the Yamaha preamps and connect this to one of my interface's line inputs, a clear output signal (evaluated by listening) would indicate that headroom is not the problem. If I dial in the Yamaha preamp for a small to moderate amount of gain and connect the output to my interface's mic preamp input, I can obtain a line-level signal using varying amounts of gain from each device. How well I am able to produce an undistorted line-level signal in this manner may shed some light on the benefits of utilizing multiple stages of lower gain. I also plan to visually inspect the preamp circuit in my interface to get an idea how the gain staging is achieved.

Feb. 14, 2015

I experimented a bit with different combinations of gain using the Yamaha preamp/mixer cascaded into the mic pre input on my Echo AudioFire 8 interface (owner's manual: http://files.echoaudio.com/manuals/audiofire_mac_manual_v2.2.pdf). At the signal levels I've been recording with, I'm still hearing some soft clipping (nothing particularly hideous--just a little upper-midrange "hair," to put it in subjective terms), regardless of how the gain is distributed between the devices. This suggests to me that the issue may indeed be with the headroom of the interface, and that maybe I simply haven't been noticing it on other sources I've been recording (the only analog instruments I commonly record are electric guitar and bass, and a certain amount of soft clipping on these signals may be less noticeable and even complimentary). I will experiment with recording acoustic guitar at similar signal levels and see if I am able to detect any clipping.

The fact that the interface appears to have insufficient headroom for the levels I am trying to record (averaging as high as possible without lighting the red overload meters in Cockos's Reaper digital audio workstation) makes me suspect that I have simply been setting my levels too high. Although this is tangential to my project, I'm now interested in determining how a true professional line level signal (+4dBu or 1.228 VRMS) reads on Reaper's input meters. I will try to devise a way to connect a function generator outputting a sine wave at this level into one of the AudioFire's line inputs. 

Feb. 17, 2015

I've simulated several different circuit topologies in Tina-TI, each with two 30-dB gain stages (for a total of 60 dB) and with a 30-dB gain stage followed by a 6-dB gain stage (for a total of 36 dB). The goal of these simulations is to get an idea what topology will likely yield the best noise performance in the first stage; the reason for the simulation of the 36-dB circuit is to ensure that the contribution of the second stage to total noise can be safely ignored.

The four topologies tested are as follows: a fully balanced input stage consisting of two OPA1611 op amps, a fully balanced input stage consisting of one OPA1612 op amp (these two topologies yielded identical simulation results), a fully balanced input stage consisting of an OPA1632 op amp, and a balanced-to-single-ended input stage consisting of an INA163 instrumentation amplifier (datasheet: http://www.ti.com/lit/ds/symlink/ina163.pdf). Each of the balanced input stages is followed by an INA163 for the second stage, and the INA163 input stage is followed by an OPA1611 for its second stage. See screenshots of circuits and simulation results below.

60-dB circuit with OPA1611 input stage.

60-dB circuit with OPA1632 input stage.

60-dB circuit with INA163 input stage.

Output noise of 60-dB circuit with OPA1611/OPA1612 input stage.

Total noise of 60-dB circuit with OPA1611/OPA1612 input stage.

Gain plot of 60-dB circuit with OPA1611/OPA1612 input stage.

Phase plot of 60-dB circuit with OPA1611/OPA1612 input stage.

Output noise of 36-dB circuit with OPA1611/OPA1612 input stage.

Total noise of 36-dB circuit with OPA1611/OPA1612 input stage.

Output noise of 60-dB circuit with OPA1632 input stage.

Total noise of 60-dB circuit with OPA1632 input stage.

Gain plot of 60-dB circuit with OPA1632 input stage.

Phase plot of 60-dB circuit with OPA1632 input stage.

Output noise of 36-dB circuit with OPA1632 input stage.


Total noise of 36-dB circuit with OPA1632 input stage.

Output noise of 60-dB circuit with INA163 input stage.

Total noise of 60-dB circuit with INA163 input stage.

Gain plot of 60-dB circuit with INA163 input stage.

Phase plot of 60-dB circuit with INA163 input stage.

Output noise of 36-dB circuit with INA163 input stage.

Total noise of 36-dB circuit with INA163 input stage.

The RC lowpass filter between stages, designed for a cutoff frequency of about 159 kHz, is intended to reduce gain above the frequency of interest in order to suppress noise at those frequencies, which, although not directly audible, can cause intermodulation distortion in the audible range.

Surprisingly, the INA163 is the clear winner for noise levels in this simulation. Bob suspects that this may have to do with noise at the input stages of the INA163 not being statistically correlated. I may build a test circuit as planned, or I may simply change my design approach and build a preamp utilizing single-ended gain stages with the INA163 as the first stage. One option to reduce noise levels in the first stage is to utilize parallel gain stages; this causes input noise, being uncorrelated, to cancel to a degree, and yields a slight reduction in equivalent input noise. However, this technique is costly in component count and cost relative to the marginal improvements it yields.

Feb. 22, 2015

I'm currently working on the schematic for my single-stage test board. I now have a well-defined purpose for the test board. Bob has been working on a silent A/B switch that can accept single-ended audio signals from a number of different sources and switch between which one is routed to an output without pops or clicks. My test board will include the 4 separate topologies described previously, with a total of about 36 dB of gain each. They will all be on the same PCB and driven by a voltage follower buffer (I haven't yet selected an op amp for the buffer). There will be 4 separate BNC outputs. 

In order to test the board, we will first measure the noise of the four circuits. We will do this with the aid of the 60-dB measurement amplifier Bob designed and built. It would be valuable to measure total harmonic distortion; I still need to research how this can be accomplished. The second part of the test will be to connect various audio signals to the board input, connect the outputs to Bob's A/B switch, and then perform blind listening tests to determine if any difference can be perceived in the sound of the different circuits, and if so, which circuit yields the most "pleasing" sound. We will then compare the results of the subjective and objective tests to see what measurable performance characteristics correspond with our preferences (if any).

Feb. 24, 2015

I performed a series of measurements of the line output signal from the AudioFire 8 corresponding to different combinations of signal levels and preamp gain. Pictured below are several scope captures from this experiment:

Line in, 1kHz sine wave @ +4dBu (1.736 Vp)

Preamp input, 1kHz sine wave @ 50mV peak amplitude 

It appears that there is not a headroom problem with the device; it does not clip noticeably until the output signal amplitude reaches about 8.5 V. 

March 24, 2015

I'm finally getting back to this project after a somewhat-longer-than-intended hiatus. Below is a block diagram for my topology test circuit:

I've got a schematic almost completed for this design. The input buffer stage will actually consist of 8(!) OPA1611 voltage followers in parallel. There are several reasons for choosing this configuration. One is that the four parallel preamp circuits will not present a suitable input impedance and will draw more current than a single buffer (per channel or signal "side") can source. 

Another is that, as previously mentioned, connecting input buffers in parallel produces a degree of noise cancellation.

April 6, 2015

The schematic is finally complete, and I'm currently working on the layout. One previously unforeseen issue that has become apparent is that there is absolutely no way I'm going to be able to fit a circuit with this many parts on the relatively small board size allowed by Eagle Light, which is the layout software I am using. I will either need to find some way to divide the circuit into multiple boards, or perhaps consider purchasing the full version of the software.

Pasted below are schematics of the six basic functional blocks of the test circuit, along with brief descriptions. Nodes connecting the functional blocks are labeled. Each of the 4 topologies features roughly 60 dB of gain in order to facilitate measurement of the (hopefully) low noise of the circuits. 

Input Switching/Buffer Block: Includes a pin header to switch phantom power in and out of the circuit, as well as Schottky diodes to protect the input buffer stage from any transient voltages in excess of the supply voltages. Pin header blocks allow selection of the balanced microphone input or a 51-Ohm resistor (to be used for a noise measurement reference). Each polarity of the signal is buffered with four parallel OPA1611 voltage followers (not all are visible in the picture).

One necessary trade-off of this circuit is the use of electrolytic capacitors to isolate the +48 V phantom power supply voltage from the input buffers. Electrolytic capacitors are known to cause significant harmonic distortion. Cyril Bateman discusses this at length in a series of Electronics World articles (http://www.scribd.com/doc/2610442/Capacitor-Sound), and finds that bipolar electrolytics exhibit much better distortion characteristics than unipolar ones.. Douglas Self also discusses the use of electrolytics as blocking capacitors in Small Signal Audio Design. His experiments show that even when blocking capacitor values are chosen for cutoff frequencies well below the audio spectrum, the capacitor can have sufficient impedance at the bottom of the spectrum to introduce harmonic distortion. His prescription for this problem is to use very large capacitances; his measurements indicate that blocking capacitors of 1000 uF should introduce practically no distortion in the audio range. I have selected 470 uF, as the distortion performance as measured by Self is still quite good for this value, and these capacitors (while rather large) will not take up quite as much board space as the larger ones.


OPA1611 Gain Stage: Pin 1 of each op amp (having no internal electrical connection) is connected to ground in order to allow the use of guard rings around the inverting inputs in order to keep stray currents from introducing noise. The three RF vias shown connected to the noninverting input of each op amp will be placed around its guard ring to provide a low impedance path to the ground plane. The .1 uF bypass capacitors used at the supply rails of the INA163 are tantalum per the recommendation of that part's datasheet. A pin header after the INA163 allows the signal to be sent to the Output A bus or the Output B bus. If no jumper is applied, the output of the stage will float. In this case, the pin headers at the input to the OPA1611 should be set to terminate the inputs to ground to ensure that the stage is sufficiently "off."


OPA1612 Gain Stage: This is essentially the same as the OPA1611 stage, only using a dual op amp package instead of two singles. Guard rings may prove difficult to implement for this configuration.


OPA1632 Gain Stage: Per the OPA1632 datasheet's suggestion, 10 uF tantalum capacitors are used in parallel with the .1 uF ceramic bypass capacitors. 

INA163 Gain Stage: An OPA1611 is used to provide the additional 30(ish) dB of gain in the second stage.


Output A Additional Gain/Buffer Stage: An additional 20 dB of gain is available from the OPA1611 amplifier. The pin header block at the input to this amplifier uses two jumpers to bypass the gain stage: one to terminate the input to ground and the other to connect the Out A bus to the noninverting input of the OPA177. The pin header connection on the OPA1611's output is left open in this case. In order to insert the gain stage into the circuit, the jumper terminating the input to ground is moved to the pin header at the op amp's output, and the other jumper is moved down to connect pins 3 and 4 of the input pin header. 

The OPA177 was chosen to accompany the BUF634 in the buffer circuit as the BUF634's datasheet mentions it as a stable option. The extremely wide bandwidth of the BUF634 has the potential to cause stability problems; C54 should help ensure stability by reducing gain at very high frequencies. A pin header is included to allow the option of operating the BUF634 in wide bandwidth mode, which may be beneficial for driving headphones, but could cause instability. The output pin header allows the output to be sent to either the BNC-connected Output A or the left channel of the headphones. The audio signal will, of course, be in mono, so only the signal on Bus A will be available for headphone output. 


Output B Additional Gain/Buffer Stage: This section has the same selectable +20 dB stage as the Output B section. The additional pin header allows selection of either the Output A or Output B (post- +20 dB stage) bus for input to the output buffer. This enables the Output A signal to be buffered and sent to the right channel of the headphones.

April 18, 2015

I'm working on the layout for the test circuit. At Bob's suggestion, I have been experimenting with layout ideas for a smaller subcircuit (just the OPA1611 input stage block) to help get ideas for the larger layout, which will be quite complex. Part of this process is learning tricks to work efficiently in the Eagle layout editor. 

One minor change I've made to the circuit is the replacement of electrolytic phantom power blocking capacitors with metallized polypropylene film capacitors, which, as documented by Cyril Bateman in the aforementioned articles, feature much better distortion performance than electrolytics, and have a much longer lifespan as well. The main drawbacks to these capacitors are their high cost and considerable size. They are also only available with axial leads, which will require much more PCB space than the radially-leaded electrolytic capacitors previously considered. However, for the purposes of this design, these trade-offs are acceptable. 
Comments