Yes, ChocoHolic, sounds right to me. 5534 Avol f3 corner is 300Hz with Cc=0, f3 corner for 26dB Avcl is about 500kHz. The massive peaks are from the rising and falling edges of the rectangular signal with higher frequency and lower Avol. The input stage will clip when slew-rate limited, this can be a problem for 5534 in simple DAC I/V converters.
Reducing the transconductance of the first stage by using JFETs and/or emitter/source resistors is the classic slew-rate improvement and is commonly done in modern opamps, but noise is higher and Avol is lower (but constant to perhaps 20kHz). I have seen 5534 slew rate of about 40V/us with discrete JFET (U406) input stage at about 2mA tail current, 2nd stage prevents higher slew rate.
The 5534 on-chip input stage is NPN with no emitter resistors for low noise, its older brother LM318 has emitter resistors (and a slew rate of 55V/us), unfortunately the output stage drive current is less than 5534 and input noise is too high for phono use but OK for line level.
5534 was a masterpiece for its time (1978), perhaps the highest development of low-noise 10MHz GBW opamps relying on vertical NPN and lateral PNP IC process, but opamps on "modern" processes like AD797 surpass it in all technical parameters. Its 3-stage design, similar to LM318 described here, would not be used for new design. Despite this it is still useful due to its high performance for cost, ready availability, and for DIYers its high compatibility with custom discrete input and output stages. I have learned a lot about classic IC op amps by studying it. It is capable of fine performance if its limitations are respected, but I think there are better choices for voltage-feedback audio applications with +/-18V supply rails demanding low noise, high slew rate and high output current.
In my subjective experience, well-bypassed 5534 with non-inverting gain at line levels reduces detail and dynamic contrasts (contrast between loud and soft sounds are diminished), even more so in passive phono preamp with 2 opamps per channel, this improves a lot with external JFET input stage but OPA627 is still better. Detail is much better with inverting gain, and dynamics are better with external output stage, especially for bass. Unfortunately inverting gain forces low source impedance, limiting application.
Reducing the transconductance of the first stage by using JFETs and/or emitter/source resistors is the classic slew-rate improvement and is commonly done in modern opamps, but noise is higher and Avol is lower (but constant to perhaps 20kHz). I have seen 5534 slew rate of about 40V/us with discrete JFET (U406) input stage at about 2mA tail current, 2nd stage prevents higher slew rate.
The 5534 on-chip input stage is NPN with no emitter resistors for low noise, its older brother LM318 has emitter resistors (and a slew rate of 55V/us), unfortunately the output stage drive current is less than 5534 and input noise is too high for phono use but OK for line level.
5534 was a masterpiece for its time (1978), perhaps the highest development of low-noise 10MHz GBW opamps relying on vertical NPN and lateral PNP IC process, but opamps on "modern" processes like AD797 surpass it in all technical parameters. Its 3-stage design, similar to LM318 described here, would not be used for new design. Despite this it is still useful due to its high performance for cost, ready availability, and for DIYers its high compatibility with custom discrete input and output stages. I have learned a lot about classic IC op amps by studying it. It is capable of fine performance if its limitations are respected, but I think there are better choices for voltage-feedback audio applications with +/-18V supply rails demanding low noise, high slew rate and high output current.
In my subjective experience, well-bypassed 5534 with non-inverting gain at line levels reduces detail and dynamic contrasts (contrast between loud and soft sounds are diminished), even more so in passive phono preamp with 2 opamps per channel, this improves a lot with external JFET input stage but OPA627 is still better. Detail is much better with inverting gain, and dynamics are better with external output stage, especially for bass. Unfortunately inverting gain forces low source impedance, limiting application.
The 5534 JFET input replacement trick was documented way back in the 80's and was included in the application notes appended to the Siliconix JFET manual at the time. National documented a similar trick with the LM318 using their LM394 supermatch pair. Walt Jung started the whlole thing in the 70's with an design idea in one of the trade magazines replacing the input stage of an LM301 with a diff pair. This got around the slow lateral PNP level shifting transistors and resulted in a faster opamp.
For mixer applications the AN version excels in low noise and it can driver moderate levels on a 600R application.
I agree with the subjective statements regards dynamics but it depends what you want to do with it.
This chip can benefit from mild biasing into class A with a simple current source at the outout stage, and as mentioned earlier if more current is required place a FET at the output and bias accordingly.
The NJM versions do sound different, the NE type sounds a bit brittle.
Of course it does'nt hold a candle to some contemporary discrete designs like the Borbely.
Ian
I agree with the subjective statements regards dynamics but it depends what you want to do with it.
This chip can benefit from mild biasing into class A with a simple current source at the outout stage, and as mentioned earlier if more current is required place a FET at the output and bias accordingly.
The NJM versions do sound different, the NE type sounds a bit brittle.
Of course it does'nt hold a candle to some contemporary discrete designs like the Borbely.
Ian
wrenchone said:
Unfortunately there's no getting around the 2nd stage PNPs, the input replacement gets around the 1st stage NPNs, allowing customization of input stage transconductance, tail current,, input stage gain and input current. Pins 1 and 8 should be at Vcc - 2.5, this reduces input stage gain with increasing Itail. Thermal feedback from output stage to input is also eliminated, though of course an external output stage does this too. Cascoding improves common mode rejection, useful for noninverting gain. The slew rate increase comes from reduced gm and increased Itail, and (I think) is ultimately limited by the 3rd stage bias current.
Folded cascode like in AD797 is a much better topology than the 5534/118 but unfortunately is unusable with lateral PNPs and needs vertical NPN and PNP.
macka said:
My recent experience has been with a headphone amp for Sennheiser 580 using 5534 with discrete input and output stages, inverting gain with Baxandall active feedback level control, Zin about 1k, fed directly from a Cal Alpha DAC with tubes replaced by NJFET/NPN. Prior experience was with an op-amp preamp with noninverting gain. So in these cases there are more variables than just noninverting/inverting. I built the headphone amp on a whim and was surprised at how good it turned out, eventially I will build an Altoids version.
Senor Nuvistor - The lateral PNPs I mentioned were only in the case of the 301. Jung mentioned this specifically as the rationale for the original 70's circuit hack, and provided a waveform for the improved amp. I probably still have the old, smeary xerox of the original article in my rat's nest somewhere.... If you look at the schematic for the 301, taking the inputs out also takes the PNPs out of action. The rest of the transistors shown in the device schematic are NPNs. I mentioned all these cases really to point out that hacking the input stages of opamps with external devices has a long and venerable history. I even tried it back around 1979 with a 318 and a pair of FETs, but was not successful, because I didn't really understand how to use JFETs back then. I'm still learning nowdays....
wrenchone said:
My misunderstanding, thanks for the clear explanation. Never heard of the Jung hack before. I agree with your assessment, the hack should work well with external PNPs or PJFETs and an output current sink for class-A output biasing. I guess there's plenty of NOS 301s around, most designers moved to 741 and its internal Ccomp.
As for myself, I hadn't even given thought to the 301 for ages until this thread made me remember the old design idea. Jung was a clever chap (still is, of course). Around the same time he posted a design idea for using a PNP common base amp as a non-inverting summing amp, a nice little circuit.
"but I think there are better choices for voltage-feedback audio applications with +/-18V supply rails demanding low noise, high slew rate and high output current."
I disagree.
NE5534 has low enough noise for audio, especially at low freq. Some new opamps have lower noise at 1KHz or higher, but go much noisier under 1KHz.
NE5534 has high enough slew rate for audio. Higher SL is over kill and not always a good thing to audio. Based on my tests, NE5534's IMD is as good as modern opamps, if not better. A good example is the AD8397 (53V/ns) and TPA6120 (XXXXV/ns). The former is much easier to work with and sounds better.
NE5534 has enough output current for audio. The RA-1 headphone amp designed for demanding 32 ohm Grado headphones is a good example. For applications that require even higher output capacity, modern opamps are not enough neither. Usually, an output buffer is then needed and in that case NE5534 shines again because you can bypass its output and bias it into class-A. Any other opamp can do this? Only the AD744, AD829 and AD8021, as far as I know.
Note: we are talking about opamps for audio, not for cell phones. Enough is enough. I have not found an audio amp that's more close to perfect than NE5534 is. The OPA627 is crazily priced ($18/ea) and its hi-freq response (above 100KHz squarewave) is weird. AD8610 is the latest audio opamp, but can it provide more output current than 5534 and can its output be bypassed?
I read on somewhere: after decades, there is still no perfect opamp. Very true, especially for audio. There is no huge market for hi-end audio opamps and in most cases NE5534 does the jobs just fine.
I disagree.
NE5534 has low enough noise for audio, especially at low freq. Some new opamps have lower noise at 1KHz or higher, but go much noisier under 1KHz.
NE5534 has high enough slew rate for audio. Higher SL is over kill and not always a good thing to audio. Based on my tests, NE5534's IMD is as good as modern opamps, if not better. A good example is the AD8397 (53V/ns) and TPA6120 (XXXXV/ns). The former is much easier to work with and sounds better.
NE5534 has enough output current for audio. The RA-1 headphone amp designed for demanding 32 ohm Grado headphones is a good example. For applications that require even higher output capacity, modern opamps are not enough neither. Usually, an output buffer is then needed and in that case NE5534 shines again because you can bypass its output and bias it into class-A. Any other opamp can do this? Only the AD744, AD829 and AD8021, as far as I know.
Note: we are talking about opamps for audio, not for cell phones. Enough is enough. I have not found an audio amp that's more close to perfect than NE5534 is. The OPA627 is crazily priced ($18/ea) and its hi-freq response (above 100KHz squarewave) is weird. AD8610 is the latest audio opamp, but can it provide more output current than 5534 and can its output be bypassed?
I read on somewhere: after decades, there is still no perfect opamp. Very true, especially for audio. There is no huge market for hi-end audio opamps and in most cases NE5534 does the jobs just fine.
If anyone ever made the "perfect opamp", I am pretty shure something would be quite imperfect about it: the price.
There is no need for an opamp to simultaneously excel in all parameters. Applications are specialized, so are op-amps. I can't think of an application that would require a part to at the same time be a good audio amp, and a MOSFET driver, and an instrumentation amp, and a good comparator, and an rf amp, and an active filter, etc etc. In the opamp world, we have specialized parts, for spacialized tasks, which are very, very good at what they do.
I agree with fixup about the OPA627. We are diy'ers, not spaceship designers. My diy budget needs to be kept to a negligible level, and the 5534 helps there.
There is only one case, when I felt the need to surpass that chip, and that's when I built an array of mike amplifiers, for an electret microphone array, intended for miking a videotaped round-table conference.
I chose the INA103 for this, and was very happy with the results. The price is lower than the OPA627, and the audio specs are better. Noise is 1nV, compared to 4.5nV for other opamps. The only pain with the INA103 is the insane pinout, which makes it impossible to drop-in different parts, for comparison.
Andy
There is no need for an opamp to simultaneously excel in all parameters. Applications are specialized, so are op-amps. I can't think of an application that would require a part to at the same time be a good audio amp, and a MOSFET driver, and an instrumentation amp, and a good comparator, and an rf amp, and an active filter, etc etc. In the opamp world, we have specialized parts, for spacialized tasks, which are very, very good at what they do.
I agree with fixup about the OPA627. We are diy'ers, not spaceship designers. My diy budget needs to be kept to a negligible level, and the 5534 helps there.
There is only one case, when I felt the need to surpass that chip, and that's when I built an array of mike amplifiers, for an electret microphone array, intended for miking a videotaped round-table conference.
I chose the INA103 for this, and was very happy with the results. The price is lower than the OPA627, and the audio specs are better. Noise is 1nV, compared to 4.5nV for other opamps. The only pain with the INA103 is the insane pinout, which makes it impossible to drop-in different parts, for comparison.
Andy
Fixup said:
And I thought this thread had run its course...
Actually I suspect we agree more than we disagree.
If cost were no object, I would always choose OPA627/637 or AD8610 for noninverting high-gain high-impedance applications, I have compared these to 5534 with class A biased output in the 1st stage of my opamp passive RIAA phono preamp for MM cartridge, Acl = 30, and the subjective difference is not subtle. If I used MC cartridge I would consider AD797. For low gain noise is not an issue
I have tried driving headphones with low-gain inverting 5534 and find that a discrete output stage "opens up the soundstage" and improves bass definition, I prefer the classic class AB NPN-PNP EF with Vbe multiplier and FET CCS driven from pin 5.
I agree that 10V/us is more than adequate for high quality audio, and that wide bandwidth opamps can be difficult to apply, but an LM6171 for line level noninverting gain is more rewarding for me to listen to, and it does happen to have a lot more slew rate and bandwidth than audio op amps.
So I have two points here, based on my experience -
Modern op amps with "excessive" specs for audio can amplify with more impressive subjective performance (clarity and dynamics) than older ones, if properly applied, with respect for their characteristics, and with the test equipment (at least 30MHz function generator and 60MHz oscilloscope) and knowledge to verify stable and linear gain and frequency response. I acknowledge that the difference may not justify the cost, and that measurements may not favor one device, DIY allows more freedom of choice.
5534 is a fine performer in low-gain inverting applications, and with pins 2 and 3 connected to V-, a cascoded JFET pair connected to pins 1 and 8 and class-A output biasing rivals OPA627 subjectively with noninverting gain into line loads, and may actually be better if one's tastes run towards warm and fatigue-free, and low-level detail is not as important.
Also LM318, LM301, AD846, ua748, and perhaps others with "sufficient" audio bandwidth and external Ccomp. The comp pin for these and those you mention is a high-impedance node from the 2nd stage VAS and cannot drive low impedance, but when connected to an external buffer, their output stage is not in the signal path (except for residual loading).
5534 is unique to my knowledge in that its comp pin can drive low impedance, and (if one favors class A CFP output) output from pin 5 with bias current into pin 5 should have the lowest output stage distortion. In 5534 you can only bypass the upper part of the output stage (Q13, R18, R19, D5) and output current limit, the lower NPNs (Q18,17) and PNP feedback follower (Q22) remain and will sink current to destructive levels if not limited externally.
telewatt said:
Yes, particularly since there are so many types available.
Is there a peer to high-end FET opamps like OPA627/637 for low-noise high-gain high-impedance noninverting gain using a single device? The AD8610/20 is the only serious competition I know of, though OPA132 comes close. Perhaps AD745? Bipolar input such as 5534 with higher input current introduce current noise and voltage offset across high source impedances giving higher input noise.
As a DIYer I value my time as well as my cash, so I don't mind paying a premium for meaningful performance improvement. BTW, if you are patient, OPA637 on eBay can be reasonably priced (<$3 in lots of 5) especially SOIC due to its Avcl>5 requirement.
The INA103 looks like a great choice for high gain from low-impedance sources, by the numbers certainly better than OPA627, at less than half the price. I've been looking for these on eBay for awhile now without success.
I can sympathize with the notion that choosing op amps based only on specs orders of magnitude outside of bandwidth and slew rate needs for audio is misguided, and that the idea that a 3000V/us-1GHZ op amp must be better than a 10V/us-10MHz op amp is probably based on "bigger is better" and not on direct experience. But you really won't know whether a new opamp is better in your application than a familiar one if you haven't used it, assuming its specs are compatible with your application.
Although it's true that you cannot beat the OPA637 for high impedance applications, I try to avoid high impedance applications altogether.
It's a personal preference, but years of fighting tape hiss in pro audio recordings in the 70's and 80's (using Dolby "B", "C", dbx, selective pre-emphasis) have made me more aversive to noise than anything else.
FETs were in general noisier than bjt's 25 years ago, and there is not a lot to indicate that this has changed radically today. FETs have evolved, but so have BJTs. I still cringe when I see a FET in the audio signal path. I know, it's prejudice, but I can't help it.
Let's face it, in the 1950's and 60's high impedance was all around us. Ceramic and crystal microphones, ceramic and crystal phonograph cartridges, outputs from vaccuum tube devices, you name it. There are not that many high impedance sources around us any more. (unless you're amplifying industrial sensors/transducers of some sort).
Everyone has their preference, I guess, but I've been sticking with INA103, INA163, and INA217 for a while now, and they seem to ouperform the OPA637 in the measures that matter to me. Again, I try to avoid high impedance. I try to re-engineer the big picture "how can we do this and get around the high impedance".
If I can't get around the Hi-Z, in goes the 637, but I have yet to find such a situation.
But of course, this is a 5534 thread. Excellent amp, by the way, if your signal is already at 100mv. Can do a lot with it. Not everything, but a lot.
It's a personal preference, but years of fighting tape hiss in pro audio recordings in the 70's and 80's (using Dolby "B", "C", dbx, selective pre-emphasis) have made me more aversive to noise than anything else.
FETs were in general noisier than bjt's 25 years ago, and there is not a lot to indicate that this has changed radically today. FETs have evolved, but so have BJTs. I still cringe when I see a FET in the audio signal path. I know, it's prejudice, but I can't help it.
Let's face it, in the 1950's and 60's high impedance was all around us. Ceramic and crystal microphones, ceramic and crystal phonograph cartridges, outputs from vaccuum tube devices, you name it. There are not that many high impedance sources around us any more. (unless you're amplifying industrial sensors/transducers of some sort).
Everyone has their preference, I guess, but I've been sticking with INA103, INA163, and INA217 for a while now, and they seem to ouperform the OPA637 in the measures that matter to me. Again, I try to avoid high impedance. I try to re-engineer the big picture "how can we do this and get around the high impedance".
If I can't get around the Hi-Z, in goes the 637, but I have yet to find such a situation.
But of course, this is a 5534 thread. Excellent amp, by the way, if your signal is already at 100mv. Can do a lot with it. Not everything, but a lot.
funberry said:
I guess that for modern audio, condenser microphones are about the only hi-z amplifier application left, and a hi-z op amp isn't needed with a tube or FET follower in the mic. Unfortunately I'm addicted to MM phono cartridges so I'm stuck with hi-z at home, MC is lo-z and I wouldn't use an OPA637 (or 5534) for a head amp. Then there's the ubiquitous line stages with a 50k pot followed by a non-inverting op amp
I agree that lo-z should be used as much as possible, but not all sources can drive 1k loads well.2SK170, 2SK369? Voltage noise competitive with BJTs but high input capacitance. AD745 is 2.2nV/rtHz but again high Cin.
Allright, Question:
I'm a bit fuzzy on this. Is the audio noise performance of an opamp with a nf of 5nV/rtHz and BW of 10MHz different than that of an opamp with an nf of 5nV/rtHz and 100MHz BW (all else being equal)?
In other words, does a wider bandwidth cause an opamp's noise spec to be skewed (better or worse) than you might experimentally find at audio frequencies?
If noone knows, I'll just have to whip up a circuit and plug in all the opamps I have to determine this empirically.
Adrian
I'm a bit fuzzy on this. Is the audio noise performance of an opamp with a nf of 5nV/rtHz and BW of 10MHz different than that of an opamp with an nf of 5nV/rtHz and 100MHz BW (all else being equal)?
In other words, does a wider bandwidth cause an opamp's noise spec to be skewed (better or worse) than you might experimentally find at audio frequencies?
If noone knows, I'll just have to whip up a circuit and plug in all the opamps I have to determine this empirically.
Adrian
Hi,
Nuvistor said
"MM phono cartridges so I'm stuck with hi-z at home, MC is lo-z and I wouldn't use an OPA637 (or 5534) for a head amp. Then there's the ubiquitous line stages with a 50k pot followed by a non-inverting op amp I agree that lo-z should be used as much as possible, but not all sources can drive 1k loads well."
Can you clarify?
When talking about Lo-Z and Hi-Z do we mean source impedance? or as Nuvistor implies we mean both source and load impedances?
The load impedance could be as low as 1k but can be anywhere upto 100k and when driven from a low source impedance would still be treated as a Lo-Z system. Is this correct?
I thought a source impedance of an MM cartridge was about 500r and this falls into Lo-Z (towards the high end) not Hi-Z which I thought would have started above maybe 2k ohms.
Nuvistor said
"MM phono cartridges so I'm stuck with hi-z at home, MC is lo-z and I wouldn't use an OPA637 (or 5534) for a head amp. Then there's the ubiquitous line stages with a 50k pot followed by a non-inverting op amp I agree that lo-z should be used as much as possible, but not all sources can drive 1k loads well."
Can you clarify?
When talking about Lo-Z and Hi-Z do we mean source impedance? or as Nuvistor implies we mean both source and load impedances?
The load impedance could be as low as 1k but can be anywhere upto 100k and when driven from a low source impedance would still be treated as a Lo-Z system. Is this correct?
I thought a source impedance of an MM cartridge was about 500r and this falls into Lo-Z (towards the high end) not Hi-Z which I thought would have started above maybe 2k ohms.
When I brought up the issue of impedance, it was in reference to the input impedance of INAxxx part numbers, and their suitability for various tasks.
The INA103 input impedance is specified at 60MOhms, by the way.
I think of a Hi-Z source as something that would be severely loaded by a conventional AUX amp input of 10K. Something that constantly makes you think twice about where you connect it.
Adrian
The INA103 input impedance is specified at 60MOhms, by the way.
I think of a Hi-Z source as something that would be severely loaded by a conventional AUX amp input of 10K. Something that constantly makes you think twice about where you connect it.
Adrian
Good comments, I guess I used the terms "hi-z" and "lo-z" without defining them, and I don't think there is a clear and generally understood definition.
To me, "lo-z" refers to source impedance of 5k or less impedance, "hi-z" to 5k or more. "source impedance" is the impedance at the amplifier input terminals "looking into" the source, amounting to a zero impedance voltage generator in series with the source impedance, the amplifier's feedback network and the output impedance of a previous amplifier stage is included. These ranges roughly correspond to source impedance ranges for minimum input noise from bipolar and FET input opamps respectively. Note that impedance is frequency dependent for inductive and capacitive sources, frequency assumed to be 20Hz to 20kHz for audio.
So for me "z" is source impedance, and amplifier selection for optimal noise performance depends on knowing what "z" is. My viewpoint is noise performance oriented, but matching of amplifier stage input/output impedances for controlled loss is another viewpoint.
Amplifier noise at line level (0.5 - 2.5Vrms) may or may not be so important but amplifier noise at mV or uV levels is important when using transducers like mics and phono cartridges, hopefully one has a goal for signal/noise ratio.
It's best to refer to a schematic, but I consider your system as hi-z since source impedance referred to amplifier input can be 100k. A shunt attenuator volume control could present this impedance range.
Actually an MM cartridge and its load network have a varying source impedance with frequency. The cartridge impedance is a resistance in series with an inductance, the load network impedance is a capacitance in parallel with a resistance, the source impedance is the cartridge and load impedances in parallel. For a Shure V15VMR, L=0.33H, R=815 ohm, with load impedance of 47k and 220pF the source impedance is about 4k at 2kHz, 10k at 5kHz and 24k at 10kHz. Calculating the input referred amplifier noise in this case by hand is complicated since the RIAA curve as well as the source impedance must be considered, this was done in a good National Semi app note in 1977, much easier today with a spreadsheet or trustworthy SPICE model. I regard MM cartridge as hi-z because of the rising impedance with frequency, but others may differ.
For lowest noise, the source impedance should be as low as possible, but the lowest noise factor is at source impedance equal to Vn/In, for INA103 at 1kHz this is 500 ohms since Vn=1nV/rtHz, In=2pA/rtHz. A single-ended signal from 1Meg source impedance would not be significantly affected by the INA103's 60Meg + 7pF input impedance, but input-referred noise including source would be about 16x higher than the source noise alone. Ott's "Noise Reduction Techniques in Electronics Systems" is a great reference in this area.
To me, "lo-z" refers to source impedance of 5k or less impedance, "hi-z" to 5k or more. "source impedance" is the impedance at the amplifier input terminals "looking into" the source, amounting to a zero impedance voltage generator in series with the source impedance, the amplifier's feedback network and the output impedance of a previous amplifier stage is included. These ranges roughly correspond to source impedance ranges for minimum input noise from bipolar and FET input opamps respectively. Note that impedance is frequency dependent for inductive and capacitive sources, frequency assumed to be 20Hz to 20kHz for audio.
So for me "z" is source impedance, and amplifier selection for optimal noise performance depends on knowing what "z" is. My viewpoint is noise performance oriented, but matching of amplifier stage input/output impedances for controlled loss is another viewpoint.
Amplifier noise at line level (0.5 - 2.5Vrms) may or may not be so important but amplifier noise at mV or uV levels is important when using transducers like mics and phono cartridges, hopefully one has a goal for signal/noise ratio.
It's best to refer to a schematic, but I consider your system as hi-z since source impedance referred to amplifier input can be 100k. A shunt attenuator volume control could present this impedance range.
Actually an MM cartridge and its load network have a varying source impedance with frequency. The cartridge impedance is a resistance in series with an inductance, the load network impedance is a capacitance in parallel with a resistance, the source impedance is the cartridge and load impedances in parallel. For a Shure V15VMR, L=0.33H, R=815 ohm, with load impedance of 47k and 220pF the source impedance is about 4k at 2kHz, 10k at 5kHz and 24k at 10kHz. Calculating the input referred amplifier noise in this case by hand is complicated since the RIAA curve as well as the source impedance must be considered, this was done in a good National Semi app note in 1977, much easier today with a spreadsheet or trustworthy SPICE model. I regard MM cartridge as hi-z because of the rising impedance with frequency, but others may differ.
For lowest noise, the source impedance should be as low as possible, but the lowest noise factor is at source impedance equal to Vn/In, for INA103 at 1kHz this is 500 ohms since Vn=1nV/rtHz, In=2pA/rtHz. A single-ended signal from 1Meg source impedance would not be significantly affected by the INA103's 60Meg + 7pF input impedance, but input-referred noise including source would be about 16x higher than the source noise alone. Ott's "Noise Reduction Techniques in Electronics Systems" is a great reference in this area.