Hi everyone!
It's been a rather slow weekend so far, so I thought I'd see if I could do something a little different.
For a while, I've had some TL074 op-amps lying around that I've yet to use in anything. To many, these are considered to be parts not worth using due to their limitations compared to op-amps like the NE5532 which have better load driving ability, distortion performance and input noise. However, they do have some very useful characteristics that lend themselves towards a simpler and easier project build such as their lower supply current and lack of input bias current being JFET devices. This simplifies audio design with them quite considerably as various DC decoupling networks can be avoided throughout the stages.
Looking at their noise performance in particular, you wouldn't think that it could be possible to build a line preamplifier with any decent level of performance at all. This makes it a rather interesting challenge to see how it can be possible to use circuit topology to bring out the best in them. My goals here are the following...
I've attached my first revision of the schematic that I've worked out on paper and drawn up in KiCAD today. Over the next few weeks I'm going to be simulating, refining and ultimately building this preamp into a completed unit and measuring it's objective performance to see if it can meet or exceed the goals I've set out.
Take a look and see what you think
... I'll be explaining the relevant sections of the circuit in order of appearance in the signal path and simulating them where applicable to show them in action before starting on the PCB layout and physical build .
It's been a rather slow weekend so far, so I thought I'd see if I could do something a little different.
For a while, I've had some TL074 op-amps lying around that I've yet to use in anything. To many, these are considered to be parts not worth using due to their limitations compared to op-amps like the NE5532 which have better load driving ability, distortion performance and input noise. However, they do have some very useful characteristics that lend themselves towards a simpler and easier project build such as their lower supply current and lack of input bias current being JFET devices. This simplifies audio design with them quite considerably as various DC decoupling networks can be avoided throughout the stages.
Looking at their noise performance in particular, you wouldn't think that it could be possible to build a line preamplifier with any decent level of performance at all. This makes it a rather interesting challenge to see how it can be possible to use circuit topology to bring out the best in them. My goals here are the following...
- Moderately low noise less than -100dBV at line level with gain set to 0dB.
- Distortion below 0.01% from 20Hz to 20kHz driven with 7VRMS (20V peak to peak), real world audio signals will typically be an order of magnitude below this level.
- Frequency response flat to within 0.5dB, 20Hz to 20kHz.
- True logarithmic volume control with a gain of 10x (20dB).500k
- High input impedance of 1M.
- Able to drive a 10k load at full output level.
- Balance control with a range of 10dB.
- Tone control with a range of 12dB boost or cut.
- Active volume control with some sort of balanced output to aid connection to active speakers.
- Phonostage with 3rd order rumble filter, low frequency cross-feed and the all important mono switch for reducing noise and distortion on mono pressings.
I've attached my first revision of the schematic that I've worked out on paper and drawn up in KiCAD today. Over the next few weeks I'm going to be simulating, refining and ultimately building this preamp into a completed unit and measuring it's objective performance to see if it can meet or exceed the goals I've set out.
Take a look and see what you think
... I'll be explaining the relevant sections of the circuit in order of appearance in the signal path and simulating them where applicable to show them in action before starting on the PCB layout and physical build .
Last edited by a moderator:
That's certainly true, but the distortion performance does start to significantly degrade under about 2000 ohms or so.
To me this isn't a problem because their rather poor noise performance dictates that there's no improvement to be had by lowering circuit impedances to that point. All about using them correctly really .
To me this isn't a problem because their rather poor noise performance dictates that there's no improvement to be had by lowering circuit impedances to that point. All about using them correctly really
.
I can see the following potential issues:
Line stage:
1. The volume control setup is similar to Musical Fidelity A1. I would assume similar long-term issues with scratchy pots. The problem is, instead of dropouts you'd get volume spikes, and those are way more audible.
2. 100% DC-coupled? You think that's a good idea?
3. Your balance control is rather odd, giving anywhere from unity gain to about +10 dB. Are you sure you want that? Besides, the poor thing would have to output 6-ish Vrms when confronted with a 2 Vrms CD input, not necessarily the best idea for a non-inverting TL07x.
Phono stage:
1. It's likely to be somewhat noisier than average. If you can't bring yourself to add a discrete input stage (either bipolar @~100-150 µA each or JFET), consider a two-stage topology with the first opamp being made up from two paralleled TL074 channels (so the amplifier would have to be renamed 374). This would bring down noise to at least average (if still non-special) terrain and distortion performance should improve substantially, too. You might even consider going 474 (i.e. a full 074 per channel for ~6.5 nV/sqrt(Hz)), or 374 with a discrete input.
2. I would suggest a more standard feedback network topology as I'm not at all sure about the RIAA accuracy of this simpler one - definitely a case for simulation.
3. With the input bootstrapping provided by the active highpass, I wouldn't be sure that input impedance actually is what you intended it to be. A two-stage circuit (or a passive subsonic on the output side) would eliminate this issue.
4. I would give the supplies a bit of RC filtering (220R - 100µ each?). When going two-stage, the first more than the second. Could be some fun with your quads though. Making the regulator section more fancy would be worth considering (maybe a 317/337 combo with Elvee's denoiser?).
Speaking of supplies, if you need filtering or bypassing to ground use a dedicated power GND return to central GNDREF. I see you have only rail-to-rail bypassing included now, presumably to avoid injecting rail noise into your audio ground? V+, V- and PGND should be kept close together but I'm sure you know that.
Line stage:
1. The volume control setup is similar to Musical Fidelity A1. I would assume similar long-term issues with scratchy pots. The problem is, instead of dropouts you'd get volume spikes, and those are way more audible.
2. 100% DC-coupled? You think that's a good idea?
3. Your balance control is rather odd, giving anywhere from unity gain to about +10 dB. Are you sure you want that? Besides, the poor thing would have to output 6-ish Vrms when confronted with a 2 Vrms CD input, not necessarily the best idea for a non-inverting TL07x.
Phono stage:
1. It's likely to be somewhat noisier than average. If you can't bring yourself to add a discrete input stage (either bipolar @~100-150 µA each or JFET), consider a two-stage topology with the first opamp being made up from two paralleled TL074 channels (so the amplifier would have to be renamed 374). This would bring down noise to at least average (if still non-special) terrain and distortion performance should improve substantially, too. You might even consider going 474 (i.e. a full 074 per channel for ~6.5 nV/sqrt(Hz)), or 374 with a discrete input.
2. I would suggest a more standard feedback network topology as I'm not at all sure about the RIAA accuracy of this simpler one - definitely a case for simulation.
3. With the input bootstrapping provided by the active highpass, I wouldn't be sure that input impedance actually is what you intended it to be. A two-stage circuit (or a passive subsonic on the output side) would eliminate this issue.
4. I would give the supplies a bit of RC filtering (220R - 100µ each?). When going two-stage, the first more than the second. Could be some fun with your quads though. Making the regulator section more fancy would be worth considering (maybe a 317/337 combo with Elvee's denoiser?).
Speaking of supplies, if you need filtering or bypassing to ground use a dedicated power GND return to central GNDREF. I see you have only rail-to-rail bypassing included now, presumably to avoid injecting rail noise into your audio ground? V+, V- and PGND should be kept close together but I'm sure you know that.
Last edited:
OK, this is interesting... Let me lay out my reasoning for the choices made in the design...
.
The whole point of the circuit is to see how much performance and real world practicality can be squeezed out of a single chip per channel with a nice low component count.
As far as the supply goes, the TL074 has good enough supply rejection to bring a 78xx/79xx supply noise injection well below the noise floor of the audio circuitry itself, especially at low frequencies where there'll be lots and lots of feedback in the phonostage. The grounding arrangement will be relying on a ground plane type PCB so is acceptable for the schematic diagram. There is no power ground as there are no power supply currents flowing to ground (except for the regulator decoupling which occurs ahead of the audio ground) so a single ground is quite appropriate.
This won't be the case as the problems you describe are a result of (usually dirty) potentiometer contacts leaving the track in this case and going open circuit. In this case the amplifier retains feedback through R12 so it doesn't cause any open loop misbehaviour that can occur through this means, although there will still be dropouts...
The input stage and tone control are DC coupled, but with a properly decoupled source, there will be no issues and I've yet to encounter a 'serious' source that has these sorts of issues. DC decoupling is mainly done with BJT input types to avoid bias currents from generating nasty clicks while switching between inputs. In this case, there are no bias currents (at least not significant ones) to worry about in this case. The volume control may seem at first DC coupled but is actually decoupled through R12 and C6. The only offset appearing on the output will be the rather small offset (typically 3mV or less) voltage of the volume control op-amp. Using an output decoupling cap is less feasible here due to the impedance balanced output... That's something to go over later
.In my experience balance controls in practice are put to use boosting up a weak channel rather than attenuating a strong one. Using TL072s we really can't afford to attenuate and amplify as the amplification we have at hand has noise limitations that will make themselves painfully apparent otherwise... A TL072 can happily handle 6V RMS as long as it's not driving a significantly heavy load (less than 2k as a rule of thumb, see datasheet). I take it that the reference to non-inverting (series feedback) topology refers to common mode effects? Have a closer look and you'll see there's some subtle cancellation of these woes going on with the line input (very similar impedances on each input). The boost of the balance control (about 2.5dB centred) will help to reduce the SNR in following stages without compromising headroom too badly. It also means that clipping and some clamping will occur in the line input before some more sinister things will...
The topology is a series feedback type here, but with some Sallen-Key trickery on the input. Input loading actually only matters at frequencies where cartridge inductance makes itself more apparent, so you can actually get away with this as I'll show later... With the values shown, RIAA accuracy will be within 0.5dB with 5% capacitors and 1-2% resistors, it's very fortuitous ratio that even includes a 10pF parasitic margin across the network. The noise performance will not be great here, but it may be interesting to see how usable it is in practice with real source material (vinyl records). I'm aiming for a noise figure of 3dB when the needle goes down here.
The whole point of the circuit is to see how much performance and real world practicality can be squeezed out of a single chip per channel with a nice low component count.
As far as the supply goes, the TL074 has good enough supply rejection to bring a 78xx/79xx supply noise injection well below the noise floor of the audio circuitry itself, especially at low frequencies where there'll be lots and lots of feedback in the phonostage. The grounding arrangement will be relying on a ground plane type PCB so is acceptable for the schematic diagram. There is no power ground as there are no power supply currents flowing to ground (except for the regulator decoupling which occurs ahead of the audio ground) so a single ground is quite appropriate.
Nice circuit. The balance control is quite similar to what D. Self has been suggesting. 10dB of range seems like a lot though.
Considering the opamps are an absolutely marginal cost in such a project (compared to the switches, pots and so on), why bother with the tl074 and not go for the opa1644 (or even the opa1654 at half the price) ? I can understand it for a manufacturer which has to scrape the last penny but is it worth it for diy ?
For the rest, I would have considered a proper Baxhandall volume control (which needs two opamps though). But then you can use a better opamp for phono. Having a separate dual opamp for the phono section will probably make your life easier for pcb layout too. Unless you're really stuck with the idea of using as few of those tl074 as possible.
I don't see why you can't have an output cap ? Plenty of pseudo balanced outputs out there with AC coupling. You just need two caps instead of one. It would protect you against a power supply catastrophic failure (which would worry me more than DC offset at the input of the preamp).
Considering the opamps are an absolutely marginal cost in such a project (compared to the switches, pots and so on), why bother with the tl074 and not go for the opa1644 (or even the opa1654 at half the price) ? I can understand it for a manufacturer which has to scrape the last penny but is it worth it for diy ?
For the rest, I would have considered a proper Baxhandall volume control (which needs two opamps though). But then you can use a better opamp for phono. Having a separate dual opamp for the phono section will probably make your life easier for pcb layout too. Unless you're really stuck with the idea of using as few of those tl074 as possible.
I don't see why you can't have an output cap ? Plenty of pseudo balanced outputs out there with AC coupling. You just need two caps instead of one. It would protect you against a power supply catastrophic failure (which would worry me more than DC offset at the input of the preamp).
The idea is seeing what sort of result is possible with just one of these chips per channel with the lowest component count possible... You could use the same topology with the OPA1644 with better results but then your power supply requirements would increase along with the cost of the devices in question. The power supply itself is only capable of about 40mA or so which keeps the cost of the transformer down. We could start a 'performance preamplifier' thread somewhere else though...
The Baxandall volume control is less logarithmic and 3dB noisier at minimum than the one described in the schematic (summing the input noise of two op-amp stages plus the resistor noise for the second). Its only benefit is immunity from tolerances in the absolute value of the potentiometer in question (which tend to be quite miniscule between channels in my experiences with dual gang potentiometers). The phono section will have both channels on one side of a 14 pin package in the PCB layout (the other side being the line input) so should not lead to any problems in reality.
The better logarithmic approximation of this topology is alluded to on the ESP site also ESP - A Better Volume Control
There's not much need for it and it would be adding another 4 components that would need to be quite costly (about 100uF is needed) and wouldn't necessarily provide protection from the effects that you describe as they would be conductive against their polarity in the event of a significant fault. There's not much chance of such a fault occurring with the power supply described in the schematic and the TL074s themselves are not too capable of supplying enough current to do too much damage on 15V rails. In any case the offset voltage will be 3mV in 99% of cases and up to 10mV in the remaining 1% or less, which I believe is sufficiently low to forgo such measures which are necessary when BJT input bias currents generate 10-500mV of offset voltage. There's also the fact that capacitor tolerance for the kind of capacitors that we would need to use would play a fairly large part in unbalancing the impedance balancing network (a 20% difference in tolerance could limit rejection to -26dB or so at 50Hz with 100uF parts) so they would need to be rather large to mitigate this, consuming valuable PCB space in the process.
I'm going to simulate the phonostage tonight to show that it might not be as bad as it seems.
Michael
I get some of your points and I'll follow this with interest, really. Still playing the devil's advocate, the opa1654 has about the same power draw as the tl074 and better overall specs. And the RS pro transformer is an expensive option to start with. 4VA transformers can easily be found for cheaper (while not being much bigger physically speaking).
As for the Baxhandall being less logarithmic, I'd have to sim the one you have with your values. In Self's book, both topologies have extremely similar control laws.
4* 100uF/25V bipolar caps are about 1.2€ from mouser, so yes, it's kind of costly compared to tl074.
As for the Baxhandall being less logarithmic, I'd have to sim the one you have with your values. In Self's book, both topologies have extremely similar control laws.
4* 100uF/25V bipolar caps are about 1.2€ from mouser, so yes, it's kind of costly compared to tl074.
That's fair enough. I think if we started deviating in the direction of such components, then we'd be well on our way to designing a modern performance preamplifier.
Unfortunately op-amps like the OPA165X and OPA164X with it's clever input capacitance linearisation circuitry are a little more expensive and only available in tiny SMT packages which makes them a bit of a pain to use and replace if they get zapped at any point...
Rod Elliot and myself have simulated the single op-amp version and it is a little better towards the top end than the Baxandall version. I think Self likes he Baxandall version more due it's immunity from the absolute value of the potentiometer, but it is an inferior circuit with a higher component count and noise for a given potentiometer value.
There is, however, a rather nifty trick you can do with it for a couple more resistors to generate a balanced output with lower noise that works very well indeed, but that's a discussion for another time...
In regards to the transformer, I chose it as the preamplifier will be 'always on' connected to the mains. Having a power switch means having to deal with turn on/off transients with the associated muting circuitry adding a fair amount of additional cost and complexity to the preamplifier. The transformer in question draws a very low level of quiescent current at no load so will pay for itself in a year or so of use. A toroidal transformer also offers the advantage of reduced field levels which will be especially useful in a smaller enclosure that I'm planning for, where a EI core type may be unworkable with.
Things will start to make more sense once they get built into a completed unit.
Unfortunately op-amps like the OPA165X and OPA164X with it's clever input capacitance linearisation circuitry are a little more expensive and only available in tiny SMT packages which makes them a bit of a pain to use and replace if they get zapped at any point...
Rod Elliot and myself have simulated the single op-amp version and it is a little better towards the top end than the Baxandall version. I think Self likes he Baxandall version more due it's immunity from the absolute value of the potentiometer, but it is an inferior circuit with a higher component count and noise for a given potentiometer value.
There is, however, a rather nifty trick you can do with it for a couple more resistors to generate a balanced output with lower noise that works very well indeed, but that's a discussion for another time...
In regards to the transformer, I chose it as the preamplifier will be 'always on' connected to the mains. Having a power switch means having to deal with turn on/off transients with the associated muting circuitry adding a fair amount of additional cost and complexity to the preamplifier. The transformer in question draws a very low level of quiescent current at no load so will pay for itself in a year or so of use. A toroidal transformer also offers the advantage of reduced field levels which will be especially useful in a smaller enclosure that I'm planning for, where a EI core type may be unworkable with.
Things will start to make more sense once they get built into a completed unit.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.