MM-Phono Pre-amp Circuit
Update 14-Mar-2021:
The BF862, BF245C and BC560C are discontinued, but there are possible substitutes.
For the BC560C the BC559CTA is still 'active'.
The BF862 jfet can be replaced by the 2SK2394-7, and BF245C by J111.
I haven't tried any of these myself, but they are fairly close substitutes and I wouldn't expect any problems, but check the pin order, the J111 differs from the BF245C (GSD instead of DSG). Also the BF862 data sheet says drain and source are interchangeable, but this appears to be unspecified for the 2SK2394, which may or may not be a problem, my layout has these reversed on one channel. It should be easy enough to fix one device upside-down with fine wire links to the board, or try it first the 'wrong' way and check the source and drain voltages to see if they are about right, in which case it is probably working ok. Some say the symbol used for a jfet should be symmetrical, with the gate arrow mid-way between drain and source if they are interchangeable, or if drain and source differ in some way the gate arrow should line up with the source. I have rarely seen this mentioned so it is probably not widely accepted or applied, so no guarantee.Update 09-Feb-2018:
I left one problem unsolved regarding my phono preamp. I originally included 100k offset adjustment preset pots for the output stage op-amps, but found these difficult to adjust. For my own use that was of no importance, my main preamp has capacitors feeding the volume control, but for anyone needing very low DC at the phono preamp output I should suggest a solution. The cause of the problem is the interaction between channels via the 5k6 resistor (R10) forming part of the 'rumble filter'. The procedure I followed of just trying to set one control and then adjust the other is not guaranteed to work. A better idea is to short one op-amp inverting input to earth and then adjust the other channel preset for zero DC output. Then remove the short and set that channel also. That way both op-amps have their inverting inputs set close to zero, and the outputs are also both close to zero. I tried this, and could easily get both outputs down to under 0.2mV, (that's the lower limit of my voltmeter) compared to typically 2mV without the adjustments. The unadjusted 2mV offset is unlikely to ever be a serious problem, so my advice is to just leave out the 100k trimmers.Update 17-Nov-2016:
I have added a circuit diagram showing the voltages in the circuit, which may be helpful for fault-finding. The diagram is further down this page just before the 'references'.Update 30-Apr-2016:
I have a report of what could be stability problems for the input section, and I have added a comment: If there is any sign of instability I suggest increasing the 68p compensation capacitor at T4 collector up to 100p, and if there is still a problem try increasing the 3R3 resistors to 4R7. I have tried the circuit with BC560C used for T3 rather than the now unobtainable 2SA1085E, and found no problems, but the changes suggested improve stability margin in a simulation.
Input Amplifier Stage
For low current noise and low interaction with the cartridge source impedance a low noise jfet is a good choice for the input device, and the low cost and good availability of the BF862 makes that an obvious choice. Including a jfet cascode stage reduces the Vds variations and so improves common-mode rejection. There are inevitably also a few drawbacks, one of which is the small surface mount package which needs good eyesight and a steady hand for soldering. Apart from this jfets have a wide range of characteristics, and the diagram below shows that with a 100R source resistor the drain current can vary from 3 to 6mA.
One solution is to obtain matched pairs and use a differential input stage, but the solution used here is to add a trimmer pot, Vr1. The pot acts as an 'offset adjustment', and the input amplifier can be adjusted for 0V output, or for symmetrical clipping if preferred. Having measured the pot resistance needed it could then be replaced with a fixed resistor. Setting the pot for 0V output is difficult to do with great accuracy, and the voltage will drift due to thermal effects etc, but fortunately the 'rumble filter' at the input of the next stage adds capacitor coupling, so moderate levels of input stage drift are not a problem.
Another problem is the relatively low gm compared to bjt devices, so to achieve the loop gain needed for accurate equalisation a high gain second stage is needed, which uses a low noise PNP stage plus cascode followed by an output buffer stage.That 100R source resistor adds a small noise voltage, which is insignificant for use with a moving-magnet cartridge, but is enough to make this design unsuitable for low impedance moving-coil cartridges. For that application higher gain will be needed anyway, so either a step-up transformer or an additional gain stage must then be included.
Maintaining a high open-loop gain over the audio frequency range would be more difficult with a single dominant pole compensation, so a two pole solution is used with compensation capacitors C2 and C3. The resistor R3 in series with C2 reduces its phase shift at higher frequencies to maintain an adequate phase margin at the unity gain frequency. The output stage can be run at 10mA to ensure enough feedback network current which may be needed with high input voltages.
A stabilised supply should be used, but the supply rejection should already be reasonably good because of the cascode stages, so standard 7815 and 7915 type supply regulators may be adequate, though it may be a good idea to add simple RC supply filtering to the input section. The op-amp used in the output stage may need a capacitor, about 100n, between its supply pins as close as possible to the device to help ensure stability. I am not certain if this applies to the OPA134, but have seen it suggested for a few other op-amps. I may eventually add more detailed suggestions regarding supply regulation, layout and earthing, but as I found with the MJR7 power amplifier most of the problems can be caused by pickup of external fields, rather than supply breakthrough, so supply location is often the important factor, and keeping the pre-amp in a separate box remote from transformers or motors is a good starting point.
Adding the Rumble Filter and Output Stage.
Although good in many ways the OPA134 in common with most op-amps has measurable common-mode input distortion, which is why the inverting 'virtual earth' input configuration is chosen, there then being no common-mode signal.
The op-amp distortion also rises with output loading, and may include some crossover effects because of the class-B output stage. For these reasons a class-A output buffer transistor is added, with a 220R base-emitter resistor included to take a few mA output current from the op-amp in only one direction and prevent internal class-B switching.
R4 from the first diagram turned out not to be needed, the maximum loop gain could be used without instability, and this reduces distortion and increases equalisation accuracy.
I originally included 100k op-amp offset adjustment presets, which are still in the layout diagrams and photo, but my own preamp is capacitor coupled so minimising DC is not necessary. For anyone needing this adjustment a problem is the interaction between channels via the 5k6 resistor (R10) forming part of the 'rumble filter'. The procedure of just trying to set one control and then the other is not guaranteed to work. A better idea is to short one op-amp inverting input to earth and then adjust the other channel preset for zero DC output. Then remove the short and set that channel also. That way both op-amps have their inverting inputs set close to zero, and the outputs are also both close to zero. I tried this, and could easily get both outputs under 0.2mV, compared to typically 2mV without the adjustments.Transistors T3 to T8 can be BC560C for the PNP and BC550C for the NPN. If possible T5 should be chosen for high current gain, 500 or more is good. With low current gain the loading effect of the feedback network would have a greater effect on frequency response. T3 needs to be a low noise type, and although a BC560C should be good enough there are lower noise types such as the 2SA1085E, (which I used myself, but now practically unobtainable). The stability margin will be better if T3 has a low rbb', but my simulation suggests the ON-Semi BC560C types are good enough, and I have now tried these without any problems. If there is any sign of instability I suggest increasing the 68p compensation capacitor at T4 collector up to 100p. If there is still a problem try increasing the 3R3 resistors to 4R7.
The low gain of the jfet input stage would perhaps lead us to think that the second stage noise needs to be kept very low. The input stage voltage gain may even be slightly less than unity, but this is not a serious problem regarding noise because the output impedance of the stage is about 150R, so far lower than the typically 10k at 3kHz source impedance of the cartridge and parallel 47k, so there is low voltage gain but plenty of power gain, and the second stage noise is less important.
The 22n compensation capacitor may seem a high value, but the low 150R impedance at that point makes its effect insignificant in the audio band, and it only rolls off the open-loop gain above 50kHz. The much smaller 68p starts its effect at a lower frequency, but the exact -3dB open-loop frequency is not easily determined, it depends on transistor characteristics including current gain. The open-loop -3dB frequency could however be fairly high, maybe 20kHz, which some may believe is an advantage.The 68p input filter capacitor should probably not be omitted, the BF862 has high specified fT = 715MHz, so stray capacitance and inductance around the input are a potential problem. Some cartridges may need very low capacitance for best results, but I suggest shortening the input cable is a better approach, and at most reducing the 68p a little, maybe to 47p or 33p. If the output of the input section can be checked for signs of instability with an oscilloscope smaller capacitances could be tried. Two extra holes are included on the board for adding resistors or capacitors across the input for anyone who wants to experiment with different cartridge loads, or for use with cartridges where higher capacitance loads are recommended by the manufacturer. Remember to add cable and arm wiring capacitance, and the 68p filter capacitor, in the calculation of the total capacitance.
The resistors can be 1% 250mW metal film. The 68p and 75p (or two 150p in series) can be ceramic COG/NPO, I used 50V 5% Vishay types. The 33n and 6n8 need to be 1% tolerance, either polypropylene or polystyrene if possible. The 22n and 100n can be any small polyester type. I used a Panasonic polypropylene 2u2, but polyester types are probably ok for this. The electrolytics can be rated 16V if you have a well regulated 15V supply, but maybe 25V is better for a good safety margin. The LEDs can be any red low power type, I used HLMP-1301.
The 5k6 resistor (R10) with an arrow labelled 'R' goes to the 2u2 capacitor in the other channel to produce a 'rumble filter' effect which reduces out of phase bass signals.This is described in Part 2 of the design theory page, and needs a moderately high input impedance stage, which is a problem if we want to use an inverting op-amp output stage. Checking to find how high the impedance needs to be, the impedance can be lower if we increase the original 1u to 2u2, then a 10k load will reduce gain by 1dB at about 15Hz, so 10k seems a reasonable minimum for resistor R11. The resistor noise will slightly worsen the overall signal to noise ratio, but more than 10k is only needed for medium or high output cartridges, so noise is then less important. I suggest 20k for a typical 1mV/cm/sec cartridge, and 47k for a high output type. Beyond 47k there will be little benefit in overload margin and the added noise will become significant.
With a low output cartridge such as my own Technics EPC205C-Mk3 with 0.5 mV/cm/sec the output stage resistor R11 could be 10k for both lower noise and higher gain. The cartridge inductance is only 240mH, so input resistor noise is reduced by the shunting effect of the cartridge impedance, so minimising noise in the rest of the circuit is also a good idea.The first simulation of the 'rumble filter' effect is the output level for equal inputs and for the opposite polarity inputs produced by vertical modulation rumble. Mono or in phase bass signals are reduced by only 1dB at 15Hz. (R11 is 10k for this simulation, with 20k the attenuation is less than 0.5dB at 10Hz as shown later). The opposite signals are reduced by 12dB at 10Hz and 20dB at 4Hz.
With an input on only the left channel there is now bass on both channels, and the sum of the two channels falls by only 1dB at 15Hz.
Next is a frequency response plot for the above circuit, which includes the effects of finite and frequency dependent amplifier gains. The component values used are 'exact' apart from what should be a 76p capacitor which gave a high frequency response fall, and this is reduced to 75p ( two 150p in series ) which gives a slight increase. This depends partly on the op-amp input resistor, which sets the gain. With 20k as shown the response rises slightly, but with 10k it falls a little. The 75p is probably non-optional, anyone not wanting the falling RIAA equalisation response past 50kHz should perhaps reduce the value to 22p rather than remove it completely, the op-amp input capacitance may otherwise add enough phase shift to cause stability problems.
Note that the vertical divisions are 0.1dB. Putting in the nearest resistor values using parallel pairs of standard values gave no significant difference. Errors resulting from 1% component tolerances are more difficult to estimate, the effects could add or subtract to some extent. Looking at each equalisation component individually a 1% change in all cases produces a response still flat within +/- 0.05dB from 30Hz to 20kHz.
For this plot R11 is set to 20k, which is why there is less bass reduction than in the rumble filter plots where it was 10k. With 20k there is now less than 0.5dB fall at 10Hz.
The next two plots are noise at the output in nV/sqrt.Hz, first for the 47k input resistor noise alone (with a 600mH 1k cartridge), then for a 47k value R11 noise alone. The input resistor is the dominant effect at all frequencies, but only by a few dB, so 47k is a reasonable upper limit for the gain setting resistor R11, so its acceptable range is 10k to 47k. The 47k value should only be used for high output cartridges, then the noise contribution is less of a concern.
The active input impedance circuit has not been included here. To find out whether this will make any worthwhile improvement it is only necessary to check the output noise level with a typical cartridge connected first with a 47k input resistor, then with 470k or the more complex network shown in the final part on the design theory page. If there is a useful improvement then an inverting amplifier stage can be added to reduce the impedance back to 47k, and hopefully most of the noise reduction will then remain. I find the noise is already low enough using my low output Technics cartridge so for now I have no plans to use the active input impedance anyway.
My usual distortion measuring methods using test signal nulling are not easily applicable, but distortion testing is not really essential, the distortion will be primarily second and third harmonic, and the high feedback loop gains will ensure it is at a very low level. The RIAA accuracy, as I mentioned earlier, will depend mostly on component accuracy, so testing my own version only shows how accurate my components are, so I just did a check on the relative gains at 100Hz, 1kHz and 10kHz to be sure there are no serious errors. What is easily checked is input overload level, and the peak inputs before clipping are about 800mV at 20kHz, 200mV at 1kHz and 50mV at 100Hz, all of which are very good. Assuming a cartridge with typical output 1mV/cm/sec and referring to the Shure plot of maximum observed recorded velocities the peak input voltages likely to be encountered at the three frequencies are about 55mV, 40mV and 5mV respectively, so there is a good safety margin, particularly with my own low output cartridge. I used a 20k gain setting resistor, and in practice this should ensure sufficient overload margins even for high output cartridges.
Voltages
References
BF862 Low noise jfet data sheet.
BF245C jfet data sheet.
OPA134 Data and application information for OPA2134 and OPA134.
'Small Signal Audio Design' by Douglas Self, P134 Fig.4.38. OPA2134 distortion with different output loads, showing that high frequency distortion is far lower with no load.
BC560C Low noise PNP.
BC550C Low noise NPN.
Supply Regulator.
One reason for designing a discrete component regulator for this application is that to avoid an additional transformer it would be convenient to use the power amplifier power supply, which may be +/- 50V or more for high power types, and this will be too high for many regulators, for example the LM340 15V regulator has a maximum input of 35V.
Here is an example of a +/- 15V regulator capable of operating with a wide range of input voltages. There are no doubt better circuits, but this one was just designed to use components I already had in my spare parts collection. It has one potential problem, which is that the 600mV Vbe of a transistor is used as the voltage reference, and this has a temperature coefficient around -2mV / degC. Multiplying the 600mV by 25 to give the 15V output gives -50mV / deg.C. For this application it is not too serious, but if the temperature falls below freezing the output voltage could exceed 16V, so the output capacitors should be rated 25V rather than 16V, as should the 220u electrolytics in the pre-amp. The OPA134s are ok up to 18V. Using a 14V zener instead of the 2k4 resistor may be a better choice. Adding a capacitor in parallel with the 2k4 will reduce output noise, and using a cascode instead of the single KSC3503 will improve ripple rejection, but it should be more than good enough without these improvements.
The 2k4 resistors give 15V output if we assume the BC transistors have exactly 600mV Vbe. The actual measured voltages at 20deg.C were +15.42V and -15.75V, so parallel resistors 82k and 47k were added, which gave +/- 14.8V. The maximum output currents were +164mA and -173mA, which is far more than needed for the RIAA pre-amp, but I planned to use it to also power a complete pre-amp with optional tone controls, and may reduce the 8R2 and 5k6 resistors to increase the available current further to power a headphone amplifier.
Photos.
The first photo is the finished board. I was unable to get 20M resistors or 75p capacitors from my usual supplier, so they are series pairs of 10M resistors and 150p capacitors, which looks a little untidy, but as usual appearance is not one of my highest priorities. The second photo is the board track layout, which is not identical to the layout diagram on the 'Board Layout' page, which has a few small adjustments to make some components fit better. The final photos show how the BF862 jfets and one of the OPA134 op-amps are soldered to the copper-side of the board. I rarely use surface-mount components, and the small size of the jfets was a little disconcerting, but it was not as difficult as I had expected to solder these after using a clip to hold them in position. There is a surface-mount version of the OPA134 available, but I used a DIP type with pins bent and cut shorter. The large capacitors near the centre of the board are 2u2 polypropylene. I would normally avoid such physically large coupling capacitors because of observed problems with interference pickup, but I had these left over from a capacitor test project, and the board will anyway be in its own screening box well away from transformers and other sources of interference, so for once I made an exception. The 6n8 capacitors near the top were originally intended to be 1% polypropylene, but again availability was a problem and I had to substitute axial lead polystyrene, which have a different lead spacing.
