I'm using smallish active speakers (Klein + Hummel O110) as my home stereo set, which saves the space and hassle of a separate power amplifier. The speakers have got a balanced input (impedance 14 kOhm). Together with my firewire computer sound card, they work nicely for classical/baroque music.
Occasionally, I'm taking one or both of the speakers on the road to historic dance events. (I find that the boom-boxes usually employed in the absence of live music commonly result in a dreadful rendition of the music.) Using the laptop computer+firewire sound card approach at these venues is usually not an option, because the user interface for playing a track on a CD is unexpected for non-computer people, to say the least. For example, having a real knob for tuning the volume would certainly be appealing.
Unfortunately, the combination "transportable", "balanced output", "nice user interface", "reasonably priced", and "CD player" seems to be unavailable in the market. In particular, the desire for a balanced output seems to propel the price into the 2000 EUR+ range fairly easily, plus comes in a heavy 19-inch enclosure, which does not satisfy my definition of transportable.
I've got a simple small CD player (walkman-style) with a (non-balanced) line-out and an MP3 player. Unfortunately, the manufacturers of these devices are less than forthcoming with specifications such as frequency range or THD+N. And the MP3 player's headphone output seems to delivery only about 150mV at full volume, far away from 0dBu.
So I thought I'd build a small box with stereo volume control, balanced output and the usual inputs (3.5mm TRS stereo, Cinch, and probably 1/4 inch TRS balanced input just in case).
Option 1: unbalanced internals
For balanced input, use INA134 to reduce to unbalanced. Use SSM2402 switches to choose between the inputs. Use a PGA2310 volume control circuit, then a pair of DRV134 to provide the balanced output.
Option 2: balanced internals
Either use OPA1632 or DRV134 to convert from unbalanced input to balanced. Use SSM2402 switches to choose between the inputs. Use two PGA2310 volume control circuits. Either rely on the opamp inside the PGA2310 or add a OPA1632 (gain=1) output opamp for each channel. Having at lest one OPA1632 in the fully-balanced signal path would help for common-mode rejection right there, including compensation of different resistor values in the two PGA2310 channels.
In either case, for controlling the PGA2310, I'd need a microcontroller such as Atmel ATmega8 receiving input from a rotary encoder. Once the microcontroller is there, it would be easy to add an LC display to show the current attenuation and input choice. (If I get carried away, I could use the A/D converter in the ATmega8 to provide a VU-meter, too.)
Power supply would be via a 5V plug-type switching power adapter plus a small integrated 5V -> +-15V switching converter that fits on the PCB. (Yes, that's a bit of switching noise, but the opamps are specified with decent PSRR.)
I'd like to hear your comments about the plan, and I've got these specific questions and concerns:
- Prefer option 1 or option 2? Option 2 needs a lot more power, because the OPA1632s eat 17mA quiescent current each.
- The data sheets have ESD warnings all over. Would it be advisable to apply suitable Z-diodes between all inputs and ground?
- I'd like to DC-couple all inputs (no capacitor in the signal path). How bad an idea is this?
- I'd like to avoid relays for the switches, because they eat a lot of power and need more PCB space.
- It seems prudent to have an opamp in all signal paths before reaching the PGA2310, because the latter requires a low source impedance per the data sheet.
- The data sheet for the THS4131 (very similar to the OPA1632, some claim it's a re-branding) suggests using only 390 Ohms in the feedback paths of the opamps to achieve gain=1; this seems to result in a fairly low input impedance (spice says 780 Ohms). My understanding is that 10kOhms input impedance would be standard for audio transmissions. Increasing the resistor values will increase the Johnson noise, though. It seems that building an instrumentation amplifier (i.e. adding buffer stages in front of each input of the OPA1632) as shown on page 3 of TI's application note SLOA064 would be a way out, at significant expense of board space and effort. Any recommendation which opamps to use as buffers? AD797?
- Would it be advisable to shorten the opamp inputs that receive a signal from the outside with a 100pF capacitor to kill RF noise?
- Are there any chip-based precision rectifier circuits available that I could use for the VU-meter (i.e. the A/D converter of the microcontroller)? Or do I just sample a lot of values and do the AC->magnitude conversion in software?
Thanks for your thoughts.
Occasionally, I'm taking one or both of the speakers on the road to historic dance events. (I find that the boom-boxes usually employed in the absence of live music commonly result in a dreadful rendition of the music.) Using the laptop computer+firewire sound card approach at these venues is usually not an option, because the user interface for playing a track on a CD is unexpected for non-computer people, to say the least. For example, having a real knob for tuning the volume would certainly be appealing.
Unfortunately, the combination "transportable", "balanced output", "nice user interface", "reasonably priced", and "CD player" seems to be unavailable in the market. In particular, the desire for a balanced output seems to propel the price into the 2000 EUR+ range fairly easily, plus comes in a heavy 19-inch enclosure, which does not satisfy my definition of transportable.
I've got a simple small CD player (walkman-style) with a (non-balanced) line-out and an MP3 player. Unfortunately, the manufacturers of these devices are less than forthcoming with specifications such as frequency range or THD+N. And the MP3 player's headphone output seems to delivery only about 150mV at full volume, far away from 0dBu.
So I thought I'd build a small box with stereo volume control, balanced output and the usual inputs (3.5mm TRS stereo, Cinch, and probably 1/4 inch TRS balanced input just in case).
Option 1: unbalanced internals
For balanced input, use INA134 to reduce to unbalanced. Use SSM2402 switches to choose between the inputs. Use a PGA2310 volume control circuit, then a pair of DRV134 to provide the balanced output.
Option 2: balanced internals
Either use OPA1632 or DRV134 to convert from unbalanced input to balanced. Use SSM2402 switches to choose between the inputs. Use two PGA2310 volume control circuits. Either rely on the opamp inside the PGA2310 or add a OPA1632 (gain=1) output opamp for each channel. Having at lest one OPA1632 in the fully-balanced signal path would help for common-mode rejection right there, including compensation of different resistor values in the two PGA2310 channels.
In either case, for controlling the PGA2310, I'd need a microcontroller such as Atmel ATmega8 receiving input from a rotary encoder. Once the microcontroller is there, it would be easy to add an LC display to show the current attenuation and input choice. (If I get carried away, I could use the A/D converter in the ATmega8 to provide a VU-meter, too.)
Power supply would be via a 5V plug-type switching power adapter plus a small integrated 5V -> +-15V switching converter that fits on the PCB. (Yes, that's a bit of switching noise, but the opamps are specified with decent PSRR.)
I'd like to hear your comments about the plan, and I've got these specific questions and concerns:
- Prefer option 1 or option 2? Option 2 needs a lot more power, because the OPA1632s eat 17mA quiescent current each.
- The data sheets have ESD warnings all over. Would it be advisable to apply suitable Z-diodes between all inputs and ground?
- I'd like to DC-couple all inputs (no capacitor in the signal path). How bad an idea is this?
- I'd like to avoid relays for the switches, because they eat a lot of power and need more PCB space.
- It seems prudent to have an opamp in all signal paths before reaching the PGA2310, because the latter requires a low source impedance per the data sheet.
- The data sheet for the THS4131 (very similar to the OPA1632, some claim it's a re-branding) suggests using only 390 Ohms in the feedback paths of the opamps to achieve gain=1; this seems to result in a fairly low input impedance (spice says 780 Ohms). My understanding is that 10kOhms input impedance would be standard for audio transmissions. Increasing the resistor values will increase the Johnson noise, though. It seems that building an instrumentation amplifier (i.e. adding buffer stages in front of each input of the OPA1632) as shown on page 3 of TI's application note SLOA064 would be a way out, at significant expense of board space and effort. Any recommendation which opamps to use as buffers? AD797?
- Would it be advisable to shorten the opamp inputs that receive a signal from the outside with a 100pF capacitor to kill RF noise?
- Are there any chip-based precision rectifier circuits available that I could use for the VU-meter (i.e. the A/D converter of the microcontroller)? Or do I just sample a lot of values and do the AC->magnitude conversion in software?
Thanks for your thoughts.