|Elliott Sound Products||Project 187|
Moving coil (MC) phono preamps (aka head amps) are always a challenge. This is due to the very low output level, typically measured in microvolts. Typical moving magnet cartridges have an output of around 3-5mV at 1kHz, but a moving coil cartridge will typically only output 200-500µV. The output level of (most) pickup cartridges is specified at 1kHz, with a stylus velocity of 5cm/second.
Many people prefer moving coil to moving magnet (or moving iron) cartridges because of their very low output impedance, which largely negates the effects of cable capacitance. Another benefit is very low inductance, which can cause response anomalies when combined with the cable capacitance. See the article on Cartridge Loading to see why this is important.
Many (but by no means all) MC cartridges also have a lower moving mass, so they can track the vinyl grooves better, especially at high frequencies. With a DC resistance of between 2-10 ohms, most are designed to operate with a load impedance of 20 ohms or more, and 100 ohms is a 'typical' MC preamp input impedance.
There are also 'high output' MC cartridges, which may offer an output level of perhaps 2-3mV (1kHz, 5cm/s). These are designed to be used with a normal MM type preamp, as the output level is high enough to provide a satisfactory signal/noise ratio. Most MC cartridges are somewhere between very expensive and "WTF!", and the majority don't allow the facility to change the stylus, so the pickup cartridge has to be returned to the manufacturer for a replacement when the original stylus wears out. This can increase the 'cost of ownership' to the point where it requires great commitment and deep pockets.
|All noise measurements and/ or calculations assume flat response (with or without RIAA equalisation), and 'A-Weighting' has not been used anywhere. The use of
A-Weighting is very common in specifications, and can easily add 10dB or more to the claimed S/N ratio, and is (IMO) misleading. I have an article that explains why A-Weighting (usually
shown as dBA) is flawed, and it's not something I ever use. Using flat response can make measurements appear worse than they really are, but not using A-Weighting gives a far more
consistent result. See Sound Level Measurements & Reality for my reasoning on this topic.|
The use or otherwise of A-Weighting is negated by the pre-emphasis and de-emphasis circuits shown in Figure 3 and used for some of my testing. These circuits reduce noise by roughly the same amount as the RIAA phono equalisation curve, and it makes little or no difference whether the measurement is A-Weighted or not.
While there is a sub-section of the audio population that considers opamps to be 'inferior' to discrete designs, I make no apologies for using them in the designs shown here. It's actually very unlikely that anyone can hear the difference in a double blind test, but this doesn't seem to affect the rampant opinions about the supposed shortcomings of opamps. There are readily available opamps that will beat discrete designs in nearly all respects, and I have no hesitations at all in presenting opamp based designs. Trying to beat low noise opamps with discrete designs is difficult, expensive and likely to have limited success without very careful circuit design.
Of all the methods used to convert the very low output level to something that 'normal' moving magnet (MM) preamps can handle, the simplest is to use a transformer. Unfortunately, these are seriously expensive components, but they offer the advantage of 'noiseless' amplification. Because the transformer is a passive device, the only noise is due to winding resistance, which is generally very low. However, a transformer is susceptible to external magnetic fields (particularly 50/60Hz) and requires extensive magnetic shielding to prevent hum injection.
There have been countless electronic designs implemented over the years, and they are often known as 'head amps'. Many use paralleled transistors or JFETs, while others use transistors in a common base configuration, where the base is grounded (for AC), and the signal is applied to the emitter. Unfortunately, many of the very low noise parts that used to be common have now disappeared, and are classified as obsolete. While you can still get them, the chances of them being genuine are not so good. Many are available on well known auction sites, but their provenance is dubious. This isn't an approach I could recommend.
Some of the very simple designs you can find on the Net use a single common-base or perhaps a pair of common-emitter transistors effectively in parallel, but these are not really recommended. The idea looks good in theory, but testing will reveal that the gain varies with supply voltage, and because they have very little feedback, obtaining the exact gain needed may require extensive testing and adjustment. The chances of getting two channels to track perfectly is probably not very good. These simple circuits have been used commercially, but generally not in 'high-end' products.
Transistors with a larger die can perform well (medium-power transistors for example), due to a low base spreading resistance (rbb') [ 4 ]. This is a very difficult parameter to measure, and it's (almost) never provided in a transistor's datasheet. By keeping this as low as possible, noise can be minimised to allow an acceptable signal to noise (S/N) ratio. It's essential to maintain the lowest practical collector to base voltage to keep the base spreading resistance low. It's common for very low noise circuits to use multiple transistors in parallel, and to use a collector supply voltage of less than 5V. An example of this approach can be seen in Project 25, in Figure 1.
A simple but distressing fact is that all resistive and active components create random noise. Thermal noise (aka Johnson noise) exists whenever a resistive component is at any temperature above 0K (Kelvin), roughly -273°C. Active components such as transistors, FETs, or anything else also contribute shot noise. Both of these noise sources are 'white' noise, with the amplitude rising at 3dB per octave. Semiconductors also suffer from 'flicker' or 1/f noise, which is most predominant at low frequencies. There are other sources, but they don't apply in this application.
To understand the noise figure and how it's used, the article Noise in Audio Amplifiers is worth reading. The article explains how noise is calculated for a given resistance and temperature, and how that affects a low noise amplifier circuit. In brief, noise is generally specified in nV/√Hz (nanovolts per square root of bandwidth). Given that the audio bandwidth is 20-20kHz, the square root of frequency is ...
√20,000 = 141 (it's not worth the effort of subtracting the 20Hz, so 141 is close enough)
If a very quiet amplifier has an input noise figure of 1nV / √Hz, the equivalent input noise (ein) is therefore ...
1nV × 141 = 141nV
Regardless of the circuitry used to boost the output level, noise is enemy No 1. When we look at circuits that are capable of noise levels around 1nV/√Hz, it's apparent that this is extraordinarily hard to achieve in practice. This is roughly the noise level generated by an 80 ohm resistor, just sitting on the bench with nothing connected to it. To obtain this level of input noise with any active circuitry is extraordinarily difficult. Even the source cartridge will have a limited S/N ratio. A cartridge with an output of 200µV and a coil resistance of 5 ohms limits the S/N ratio (wide band) to 74dB without any amplifier at all. S/N ratio is reduced to 71dB if the cartridge's DC resistance is 10 ohms! Adding amplification can only ever make this worse! To give you some idea of the noise problem, consider that an ideal 1k resistor will generate 574nV of noise (20Hz - 20kHz, 25°C) all by itself.
Feedback paths have to be low impedance or they add noise to the circuit, and the input resistance of the pickup cartridge will contribute noise of its own. Because most MC cartridges are very low impedance, it follows that their resistance is also low - usually less than 10 ohms. This is useful, because it minimises the resistance at the amplifier input. All resistances need to be kept to the minimum, because the noise generated is proportional to resistance. If a low resistance is in parallel with a higher resistance, the noise contribution of the higher value is (partly) short-circuited by the lower resistance, so a 10k resistor in parallel with 10 ohms contributes no 'excess' noise.
To give you an idea of what can be achieved, the following shows 20kHz bandwidth input noise voltage and RIAA equalised signal to noise ratio for various potential candidate devices, referred to the noise generated by a 5 ohm resistor as a reference. Current noise has not been considered because it's not relevant for low impedances. Not all devices are available or suitable, especially 'leadless' SMD devices, and the 2SK170 is no longer made or sold by anyone reputable. All final S/N ratios shown assume a gain of 10 (20dB) and an input level of 200µV, followed by RIAA equalisation (note that the S/N ratio of the RIAA stage has not been included). Parallel operation has been assumed to give a 2dB S/N increase (3dB is theoretical and rarely achieved in practice).
|Device||ein (nV√Hz)||OP Noise 1||S/N Ratio, dB||OP Noise 2||S/N Post RIAA 3|
|5Ω resistor||n/a 4||-145.7 dBu||73.8||- 4||87.8|
|LMH6629 3||0.68||-115.5 dBu||64.2||-||78.2|
|LT1028 4||0.85||-114.2 dBu||62.8||-||76.8|
|2N4403||0.90||-133.8 dBu||62.4||-135.9 dBu||78.4 (Parallel)|
|2SK170 5||0.95||-113.4 dBu||62.1||-115.4 dBu||78.1 (Parallel)|
|AD4899 6||1.00||-113.1 dBu||61.4||-||75.4|
|OPA1612 6||1.00||-113.1 dBu||63.4||-115.1 dBu||77.4 (Parallel)|
|LM4562||2.70||-105.2 dBu||53.8||-107.2 dBu||69.8 (Parallel)|
|NE5534A||3.50||-103.0 dBu||51.6||-105.0 dBu||67.6 (Parallel)|
|NJM2068||3.50||-103.0 dBu||51.6||-105.0 dBu||67.6 (Parallel)|
|NE5532||5.00||-99.9 dBu||48.6||-101.9 dBu||64.9 (Parallel)|
|OPA2134||8.00||-96.7 dBu||44.5||-98.7 dBu||58.5 (Parallel)|
|TL072||18.00||-88.8 dBu||37.5||-90.8 dBu||53.5 (Parallel)|
- Single amplifier, wide band (20Hz - 20kHz)
- Two Amplifiers in parallel (20Hz - 20kHz). Does not apply for high cost opamps.
- Signal to noise ratio, referred to 5mV output, assuming a gain of 20dB. Note that the noise contribution is not as great as it may seem at first, because the effective bandwidth of an RIAA equalised signal is only around 800Hz. That drops the effective noise by 14dB, so where you might expect (say) 100µV of noise, you'll actually get closer to 20µV. That makes a significant difference, and this is likely one reason that some of the simpler moving coil preamps can be used at all. However, in the descriptions that follow, I've assumed wide band noise (20Hz - 20kHz) so that all comparisons are equal.
- n/a - Not applicable/ not available
- While the LMH6629 is the lowest noise opamp currently available, it only comes in a 'leadless' SMD package making it very hard to mount on a PCB unless you have all the right tools. It's also extremely fast (maximum bandwidth extends to well over 500MHz !), making it more likely to oscillate if you get the smallest thing wrong in a layout. The recommended supply voltage is 5V (±2.5V).
- The LT1028 is a very quiet (albeit very expensive) opamp, but to obtain optimum noise the impedance at the two inputs must be perfectly balanced, because the IC has input bias current cancellation circuitry that will cause excess noise if the impedances are mismatched. This is difficult to achieve in real life, because the source impedance is usually unknown.
- While the 2SK170 was ideal in this role, they are obsolete and those still being sold will almost certainly be counterfeit. Unfortunately, this is no longer a viable option.
- The AD4899 and OPA1612 are available in SMD versions only.
Each example assumes that the feedback resistors are 100 ohms and 10 ohms, and that the cartridge has a 5 ohm winding resistance (that's why a 5 ohm resistor was included in the table). The signal to noise ratio is reduced by less than 1dB with a gain of 31 (30dB) as obtained with all three 100 ohm feedback resistors shown below in circuit. Note that the S/N figures shown are deliberately pessimistic. The input level is 200µV, but a gain of 11 (100 & 10 ohm feedback resistors) actually suits a cartridge with 500µV output. S/N ratio will therefore be improved by almost 8dB, although it will be degraded somewhat due to the higher winding resistance of the higher output coil.
In each case, the noise level is shown as equivalent input noise (en, in nV√Hz), and referred to 0dBu (775mV). Parallel operation of low-cost devices is feasible within a reasonable budget, and the theoretical level (assuming a 2dB improvement with two devices in parallel) is shown where appropriate. Unless you have very deep pockets, using $20 (or more) opamps in parallel to get a couple of dB lower noise isn't sensible. The TL072 is shown as a reference - it's not recommended at all in this role. Based on this, a pair (or three) 2N4403 transistors will produce (close to) the lowest noise, even beating the very expensive LT1028 opamp by 2dB. The down side is that it requires a more complex circuit to work properly. While the LMH6629 looks good, it uses a very awkward package for DIY assembly and can't be recommended.
As noted above, if the input level is increased by 6dB (to 400µV), S/N ratio is improved by 6dB, but only if the MC cartridge maintains a 5 ohm winding resistance. If the resistance rises to 20 ohms (possible but unlikely) the net S/N ratio is unchanged. In general, if the overall S/N ratio is better than 60dB, then the system will be quieter than the best vinyl pressing. Adding the RIAA preamp after the 'head amp' does degrade the overall noise level slightly, and if a head amp and RIAA preamp both have a 60dB S/N ratio, the combined pair will reduce the total to around 57dB. However, many other factors come into play, in particular the RIAA preamp is fed from a much lower impedance (perhaps 50 ohms or less) than when it's used with a moving magnet cartridge, so its noise contribution is reduced.
There are already a couple of discrete designs shown in the projects section, in Project 25 - Phono Preamps For All, with Figure 1 showing a hybrid circuit (transistors and opamps), and Figure 2 being fully discrete. I've not built either circuit (I don't use a moving coil pickup cartridge), but they are from respected designers and should work well. Since the output level (at 1kHz) is only a few millivolts, distortion will be minimal. The low impedance of the feedback network will not place any unreasonable demand on the opamp, as the feedback current remains well below 1mA with any typical signal.
The first preamp shown in Project 25 (Figure 1) is based on an original design by Douglas Self, but has been modified to provide three gain options. Since the gain is not something that needs changing unless you replace your MC pickup, the selection would normally be done using jumpers. A gain of 50 will raise the level of a 100µV cartridge to 5mV, but for most MC pickups a gain of 10 or 30 will be the most suitable. The circuit operates at full gain regardless of input level, but even the highest output pickup cartridge will be unable to drive the output to more than 50mV or so.
This first preamp should use ±15V supplies. Current drain (stereo) is around 17mA from the positive supply, and 22mA from the negative supply. The currents are not balanced because of the input transistors, which operate from the negative supply alone, at about 2.8mA. The remaining current is due to the opamps. The actual figure will vary, because opamp current drain is not a fixed value from one component to the next.
There is a limitation in the circuit, and that's the fact that it's liable to have a significant turn-on transient, because the transistors will show a collector voltage of -2.7V (assuming ±15V supplies). It takes time for the bias servo (U2) to work because the capacitor (C8) has to charge before the output will be at zero volts, and it may be necessary to include a mute circuit to ensure that a high level transient isn't fed straight into the RIAA equalisation circuit. In a simulation, it takes well over 10 seconds for the circuit to settle, and there's no reason to think that a physical circuit will be any different. The servo has another side effect as well, in that it can cause a low frequency peak if the input capacitor's value is changed. The value of the input cap and DC servo cap should not be changed.
The Figure 2 circuit is based on a design published by John Linsley-Hood, but it was first described in a Swedish magazine called 'Radio & Television' in 1975. The original JLH design claimed to run from a pair of 1.5V cells, meaning that a separate power supply isn't needed. With a current drain of about 2.5mA for each preamp, even a pair of AA cells should last for well over 200 hours. I have no way to verify that the noise level is within acceptable limits, but JLH wasn't known for publishing designs that didn't work as claimed, so I'd expect it to have more than acceptable noise limits.
While many MC preamps are discrete, some of the once popular transistors and/or FETs are no longer available as noted above. Very simple circuits (some run from a 9V battery) may look appealing, but most will not perform well unless carefully adjusted, and may also vary their gain as the battery discharges. This is quite obviously less than ideal, especially if the two channels don't track each other perfectly. This will upset the channel balance and ruin the imaging of your system. There's also a common misconception that symmetrical circuits are actually electrically symmetrical. Mostly, the symmetry is mainly visual - we see a circuit that looks symmetrical, and make an assumption that the symmetry is real. Unless you can get perfectly matched NPN and PNP devices, symmetry is an illusion.
It used to be that opamps were unable to provide low enough noise for use in a 'head' amp, but today there's little option. Yes, there are still discrete or hybrid circuits that can be used such as those described above, but the overall noise level ends up being only marginally better than an AD797 or similar. Meanwhile, complexity is increased, and power supply rejection becomes an issue that can ruin an otherwise quiet circuit. If designed properly (meaning that feedback resistors are the lowest practical values) there are several opamps that are suitable, but they will stretch your budget !
There are some very low-noise opamps available now, with the AD797 or LT1115 being readily available (but at considerable cost). These have typical noise levels of 0.9nV/√Hz (the LT1028 is 0.85nV/√Hz, but only if input and feedback impedances are equal), but these are all seriously expensive opamps. On a more 'pedestrian' level is the NE5534A, rated for 3.5nV/√Hz. This can be improved by operating two in parallel, which reduces the noise level by 3dB.
If the cartridge has an output voltage of 400µV, the wide band (20kHz) signal to noise ratio of an NE5534A is ...
20 × log ( 400µ / 493n ) = 58dB (~72dB after RIAA equalisation)
If we use a quieter opamp (e.g. 0.9µV / √Hz), the signal to noise ratio can be improved to 64dB (200µV cartridge) or 70dB for a cartridge with an output of 400µV. Using a pair of NE5534 opamps in parallel will improve the worst case example shown above to 61dB, which will generally exceed the S/N ratio of most vinyl discs. Note that the references here are for input noise, and both the opamp's input noise and cartridge output signal are amplified by the same amount. The ultimate noise level is much lower than these figures would indicate though, because the RIAA equalisation removes the high frequencies above 2,100Hz at 6dB/ octave. It also boosts the low frequency noise, but this is less obtrusive and has little influence on the audible noise performance.
For a phono preamp with a total mid-band gain of 40dB (a gain of 100 at 1kHz), you'd expect a wide band noise input signal of 1µV to be 100µV at the output, but it's not. The exact figure depends on a number of factors, but you're more likely to measure less than 20µV (14dB less) at the output. The signal is 100 times (40dB) greater, but noise is only increased by 26dB, so the noise from a MC preamp stage is not simply amplified by the mid-band gain of the RIAA equaliser stage. The estimated 53.6dB signal to noise ratio using paralleled NE5534 opamps will translate to about 67.6dB after the equalisation stage has done its work.
Note that the above calculations may be pessimistic, because all noise sources have been factored into the final circuit and the table above is for a gain of 10 with a 200µV output cartridge. The feedback resistors add noise, and it usually won't be possible to get an overall S/N figure that's much better than 60dB before the RIAA equalisation is taken into account (74dB after EQ). This might seem rather poor, but it's actually unrealistic to expect much better [ 5 ].
Figure 1 - Opamp Design Using AD797
Figure 1 above is one suggested circuit, ideally using AD797 opamps. While this is not an inexpensive design, it should perform very well in practice. I don't have a MC pickup cartridge, and nor have I built the circuit to test its performance. There may be a small noise penalty compared to a design using multiple parallel transistors or JFETs, but with a calculated S/N ratio of better than 62dB before the RIAA equalisation (around 76dB S/N after EQ) it's more than likely to be as good as you'll ever need. Note that the opamp runs with a gain of 31 (29.8dB) regardless of the output jumper, and for lower gains the signal is simply tapped off the feedback series resistors. Contrary to expectations, this neither improves nor reduces S/N ratio - it remains relatively consistent regardless of the gain selected.
Another alternative is the AD4899, a low noise SMD opamp. These are cheaper than the other three suggested, and have an input noise level of 1nV/√Hz. With a corresponding EIN of 141nV across the audio band, this will give a S/N ratio of 69dB for a 400µV cartridge, or 63dB for a 200µV cartridge. After RIAA equalisation, that will work out to around 83dB or 77dB S/N ratio respectively. This IC is rated for a maximum supply voltage of ±6V, but it would be prudent to use ±2.5V supplies. These are easily obtained from the more 'traditional' ±15V supplies using precision voltage reference shunt regulators. The AD4899 is a very high speed opamp, so it's essential to minimise the impedance of the feedback network, both to minimise noise and the effects of stray capacitance. Being SMD (surface mount device), it's harder to use than a more 'traditional' DIL (dual in-line) package because it's so small (which makes it hard to solder to a PCB).
The bit marked 'FB' is a ferrite bead, which is intended to minimise RF interference. Because we are struggling to keep all resistances as low as possible, my usual trick of including a resistor in series with the input will degrade the noise performance rather badly. Even as little as 10 ohms will reduce S/N ratio by up to 3dB, which is unacceptable. The ferrite isn't essential, but it's a small (and cheap) option that should help.
The next task is to determine the gain needed by the MC preamplifier. The target output level is around 5mV at 1kHz, so you need to know the output level of the cartridge. For a cartridge with an output of 200µV, you need a gain of 25 (5mV / 200µV), and with a 500µV type the gain is 10 using the same formula. It doesn't have to be exact of course, because moving magnet cartridges aren't all the same anyway. For a gain of 30 (a little over the 25 suggested), the closest using readily available 1% resistors is 300 and 10 ohm resistors, providing a gain of 31 (29.8dB).
As shown the gain is a maximum of 30, but a gain of 50 can be obtained using 510 and 10 ohms (not shown in the circuit). These are quite alright in all cases, and will provide a little over 5mV output with a 100µV cartridge. Note that the feedback resistors are much lower than you'd normally find in an opamp circuit, but since the maximum output voltage is only around 15mV RMS (less than 22mV peak), the opamp only needs to be able to supply less than 100µA through the feedback resistor network.
It's actually unlikely that any discrete circuit can better the performance of the suggested very low noise opamps. Since the source S/N ratio is already 'only' 76dB for a 200µV cartridge, every transistor and resistor that's used degrades the noise figure, and simple circuits without feedback will be unable to meet the performance standards expected. Dealing with very low level signals has always been difficult, and getting good noise performance is particularly irksome.
The cost of these devices is high, but they offer the simplest way to get high performance. Using a pair in parallel is tempting, but this adds considerable extra cost and will be hard to justify. Using a MC cartridge with a higher output voltage helps greatly, because you start with more signal and need less amplification. If at all possible, I'd suggest a cartridge with at least 400µV output, as this instantly increases the overall S/N ratio by 6dB.
If the cartridge you use has a low output (200µV) it's probably already a very expensive item, and the added expense of a transformer is warranted. There are actually two transformers - one for each channel. Most of the available transformer based interfaces cost over (sometimes well over) AU$500 or so, but of course there is a wide range. Cheap units should be avoided unless you have personal experience with them, as some are made in Asia and may overstate their performance. Buying second-hand is fairly safe, as there's not much that can go wrong with low level transformers. You could even try making your own, but that's not something I'm about to try to design so you'll need to look elsewhere for details.
Note that in the circuit shown below, the input and output caps are connected with their +ve terminal towards the preamp. The actual direction depends on the opamps you use, so build the circuit without the caps first, and measure the DC voltage at both input and output. Orient the two caps so they have the correct polarity. Different opamps have differing input configurations, so you may measure either positive or negative offset at the input and output, depending on the opamp's input stage configuration. Input offset is amplified by the gain you set, because the opamp itself has full gain down to DC (so input offset is amplified by 30 at full gain).
Figure 2 - Alternative Low-Cost Opamp Design
The above shows a version that some constructors may wish to try. The NJM2068 is a little-known part, but it's quieter than an NE5532 and roughly the same noise as the NE5534 (about 3.5nV/√Hz). By running two in parallel, all impedances are effectively halved, and the opamp noise is reduced by up to 3dB (I've assumed 2dB in the above table). Although the feedback impedance appears way too low, the maximum output will be less than 15mV at any frequency, so the current won't exceed 100µA. Any opamp can provide this easily, so distortion performance isn't compromised by the very low impedance. Signal to noise ratio should be about 53.6dB with a 200µV MC pickup (S/N of just over 67dB after RIAA equalisation). This noise figure is obviously improved if a higher output pickup is used so the gain can be reduced. A lower noise opamp will allow more gain with less noise penalty.
C1 and C3 are shown with the correct polarity for the NJM2068. If you use an NE5532, these two caps must be reversed, because the 5532 will always show a negative offset. I measured the NJM2068's DC output at 30mV, and the NE5532 was -31mV. This is significantly more than the AC signal, but it's immaterial in real terms. Note the ferrite bead ('FB'), which serves the same purpose as for the Figure 1 circuit.
If preferred, the circuit can be run from ±5V, although that should make little difference to the noise performance. Only a single channel is shown (despite the dual opamp), and of course two are required for stereo. Because this is not a 'super low noise' opamp, trying to use this circuit for very low output cartridges is not recommended, and it's designed primarily for pickups with an output of 300µV or more. The ×30 position is there if you need it, but it will be noisier than the Figure 1 circuit. If preferred, the same circuit can be used with the LM4562, and that will increase the S/N ratio by around 3dB. While that's a more expensive option than the NJM2068, it's still cheaper than an AD797 or LT1115, and far easier to solder than SMD parts.
There's plenty of room for modification to suit the cartridge you have, and the gain can be changed easily by varying the value of R7(a, b). For example, with R7 as 15 ohms, the gains will be 21 (26dB), 14 (23dB) and 7.7 (17.7dB) rather than 31, 21 and 11 as shown. Reducing R7 isn't recommended, because the circuit will almost certainly be somewhat noisy with a gain of over 30 times. Having said that, it will probably still be better then the S/N ratio available from vinyl, so it's worth a try if you have a low-output cartridge.
To be able to test the circuit properly and get a representative set of measurements, the following circuit was put together. There is a pre-emphasis circuit at the input, that boosts high frequencies, which are then cut by the same amount by the de-emphasis circuit at the output of the Project 158 low-noise preamp. This allows a reasonably close approximation to the noise performance with full RIAA equalisation. The noise bandwidth is still higher than a true RIAA EQ system, so noise measurements are a little pessimistic. The voltages shown are all at 1kHz. The time constant used (10k, 10nF & 1k, 100nF) is 100µs rather than the 75µs for RIAA (1.59kHz and 2.12kHz respectively), so there is a (very) small improvement over reality with the values I used.
Figure 3 - Test Circuit
The process of pre- and de-emphasis is similar to that performed by high-frequency part of the RIAA curve, and while there is a small difference, it's immaterial as far as noise levels are concerned. The overall response is flat to well over 20kHz, with a -3dB frequency of over 60kHz (opamp dependent). This increases the measured noise, although most of it is well outside the audio range. The initial tests were performed using a 1kHz sinewave to measure signal to noise ratio, with an effective 'cartridge' output of 106µV at the input of the MC preamp stage. This gives a signal of 3.3mV at the output of the preamp at 1kHz. This was then amplified by my low-noise preamp so that measurements could be made easily. De-emphasis reduces the 1kHz voltage slightly, and restores flat response.
A simulation says that even with 490nV of input noise (ein of -124dBu), the total output distortion plus noise measures 0.045% (all of which is noise, not distortion). A listening test confirms this, as noise was effectively inaudible using my workshop speakers. That 'distortion' level translates to a S/N ratio of 67dB. Reality proved to be no different, and as shown in the table above, the actual measured S/N ratio is also 67dB. That's pretty good, and far better than vinyl will ever achieve (around 50dB appears to be the accepted figure, but 60dB is apparently achievable with some high quality pressings). While there are many claims that vinyl can achieve up to 75dB, this is often referring to dynamic range (we can hear signals below the noise floor, but not with any fidelity). Unfortunately, it seems to be almost impossible to get any real information on the S/N ratio of vinyl. There are countless opinions, but few facts.
During testing, I used both sinewave and music tests, the former for measurements and the latter for listening. There is no doubt that circuit noise is audible when the input is disconnected, but it was not intrusive (my workshop music signal sources are an FM radio and a CD player). I know from listening to vinyl that the record surface noise is greater then the noise I could hear with the test circuit shown. Even with only 100µV input, the background noise was not intrusive when music was playing. At times (with quiet material from CD), the input level was less than 10µV or so (10mV before the pre-emphasis/ attenuator circuit), and some noise was audible, but not intrusive.
Without the pre- and de-emphasis, I measured a S/N ratio of 56dB using the NJM2068 circuit, and 52dB with an NE5532. That was measured flat, and with the preamp sitting on the workbench with no shielding. When the pre-emphasis and de-emphasis circuits were included, that improved the NJM2068's S/N to 67dB, referenced to an input of only 200µV, and with a 10 ohm source resistance. The NE5532 measured 63dB S/N with the pre- and de-emphasis circuits in place, slightly less than shown in Table 1 (64.9dB). Overall, these figures are in fairly good agreement with the theoretical values shown in Table 1, with the NJM2068 being only 0.6dB shy of the calculated figure. Since the input resistance was increased from 5 ohms to 10 ohms, this is actually better than the calculated value! Since the measurement bandwidth is substantially greater than 20kHz, I consider this to be a success in all respects.
The DC offset at the outputs of each opamp (NJM2068, NE5532) are within expectations. Both have bipolar input stages, with the NJM2068 using PNP (positive offset) and the NE5532 using NPN (negative offset). Because I don't have any AD797 opamps in stock, I was unable to test the Figure 1 circuit for noise or offset (the latter should be less than 20mV), and there is no reason to expect it to be any different from the calculated figures provided. With an expected S/N of 76dB after RIAA equalisation, it's obviously a quieter preamplifier, but the cost may well outweigh the benefits. Be careful with this device, because it's available on ebay (from China and elsewhere) at unrealistically low prices. The AD797 normally sells from authorised distributors for AU$23.00 or more each, so it's unrealistic to expect that paying AU$3.00 each will get a genuine device. The only thing you can count on is that 'AD797' will be printed on the IC, but what's inside could be anything.
Just for a laugh (and to make an absolute comparison, I tried a TL072 and a 4458 dual opamp in place of the NJM2068. The TL072 managed 42dB S/N before EQ, slightly better than the 39dB expected, and the 4558 was a bit of a surprise, measuring 48dB. I didn't bother re-testing with the pre- and de-emphasis circuits connected, but I did listen to the TL072 with them in place, and noise was audible. It wasn't terrible, and was still probably better than you'd get from a high quality compact cassette player (these were once popular with many people before CDs became common). With the EQ in place, S/N would be around 53dB. Now you know why I suggested that the TL072 is not recommended, even if I did measure less noise than claimed in the datasheet.
I've been asked any number of times about MC preamps, but I've resisted because I don't use one so I'm unable to run proper listening tests. However, the Figure 2 circuit shown will work well and has been tested with representative signal levels. The parallel NJM2068 circuit is also used in the front-end of my low noise preamp (see Project 158 for details). However, it's intended to be used at levels somewhat greater than a millivolt, so the feedback resistors are a higher value. It was only because I have the P158 high gain preamp that I was able to conduct proper listening tests and measurements - without it this project would never have eventuated. The Figure 1 circuit has been simulated but not tested, but I have exactly zero doubts about its performance with a genuine AD797 opamp.
With the values shown and/ or changed as desired, you can get almost any gain you wish, with a noise level that should be well below that of vinyl, and with low cost opamps. While I don't have a moving coil cartridge, I do have the high gain preamplifier that allows me to make fairly accurate measurements of output noise and signal to noise ratio. Using RIAA equalisation after the preamp reduces the audible noise contribution by 14dB, so while the S/N ratio may not look too impressive by itself, once equalised it is improved to the extent that even a low output cartridge can be accommodated. Listening tests confirm this, and I doubt that anyone will be disappointed if they use a quiet opamp to start with.
Of the two circuits shown, my preference is for Figure 2. It can be used with any dual opamp, and if you elect to use the LM4562 it will give a surprisingly good account of itself. I tested the circuit using an NJM2068 and an NE5532 with the circuit shown in Figure 3, and also with no pre-emphasis (just a simple 1,000:1 voltage divider from my workshop sound source). Even without the RIAA equalisation (but with considerable extra gain added by my low-noise test preamp), circuit noise was just discernable on quiet passages with material having a large dynamic range. It's obvious that including the circuit in front of Project 06 (for example), noise is reduced even further by the equalisation network.
Both measurements and listening tests confirm that the performance of the Figure 2 circuit is actually better than calculated, especially when using the NJM2068 dual opamp with both sections in parallel. Even without the improvement of overall noise provided by the RIAA equalisation, the noise is only audible during 'silent' parts of the CD I used. Listening to the FM radio no 'excess' noise could be heard, because radio is not a perfect music source, and there's always something happening to avoid 'dead air' (no sound). When using the full test circuit shown in Figure 3, I had to reduce the input level dramatically before the circuit noise was audible through my workshop system.
|Copyright Notice. This article, including but not limited to all text and diagrams, is the intellectual property of Rod Elliott, and is Copyright © 2019. Reproduction or re-publication by any means whatsoever, whether electronic, mechanical or electro- mechanical, is strictly prohibited under International Copyright laws. The author (Rod Elliott) grants the reader the right to use this information for personal use only, and further allows that one (1) copy may be made for reference. Commercial use is prohibited without express written authorisation from Rod Elliott.|