Phono preamp headroom - why?

[Michael Fidler]

Did you mean to state something else? The RIAA pre-emphasis of a mastering system boosts high frequencies, the de-emphasis curve rolls them off. But usually the input circuit is not able to take advantage of this fact since the EQ usually occurs slightly later in the signal path.
IME the problem is that the input circuit is not equalized and often is not in a feedback loop if one exists. Opamps are the exception and my experience mirrors the advise you stated in a later post.

In almost all series-feedback designs, the RIAA equalisation is implemented in the first stage, so the de-emphasis is applied immediately. HF overload may also be caused by excessive loading of the feedback loop at this point though. Equaliser/input stages are used in my own S5 and S15 devices. I'm not sure what you're referring to specifically, though?

 

[atmasphere]

In terms of overload behaviour: it's not so much resonances or anything that trails the clipping, more the fact that the RIAA curve features almost 40dB of gain difference from the top to the bottom of the audio band. Therefore, any clicks and pops that push the preamplifier into overload above 10kHz where cartridge resonances and high-velocity surface anomalies exist create intermodulation/distortion artefacts that will be equalised and boosted 20-30dB or so.

It was this bit. Artifacts or anything at 10KHz is going to be rolled off by the EQ. Perhaps I'm misinterpreting this?

 

[Michael Fidler]

It was this bit. Artifacts or anything at 10KHz is going to be rolled off by the EQ. Perhaps I'm misinterpreting this?

The artefacts themselves may be rolled off, but the distortion products they generate will be boosted by the EQ. The distortion can be considered to occur 'pre-EQ'.

 

[atmasphere]

The artefacts themselves may be rolled off, but the distortion products they generate will be boosted by the EQ. The distortion can be considered to occur 'pre-EQ'.

I get that- I've harped on it quite a bit.
So you're saying that despite high frequency information the distortion will be very low frequencies? I've not encountered that but I imagine its possible if the phono section has issues with IMD. Most ticks and pops I encounter seem to be HF in nature- and is why 'scratch' filters were always low pass.

 

[Michael Fidler]

I get that- I've harped on it quite a bit.
So you're saying that despite high frequency information the distortion will be very low frequencies? I've not encountered that but I imagine its possible if the phono section has issues with IMD. Most ticks and pops I encounter seem to be HF in nature- and is why 'scratch' filters were always low pass.

If you look at the waveform of surface anomalies, you can see that they're highly asymmetrical. If it were a symetrical waveform, then the intermodulation products generated by overload would not be too great. This is not the case, though, and the peaks will cause significant low-frequency artefacts relative to the signal once it's equalised with the distortion product.

 

[atmasphere]

All you have to do now is have an actual tick or pop not caused by HF or RF overload due to electrical resonance (like an actual scratch) to actually be so powerful it overloads the phono section. I imagine that can happen but I've never seen it.

 

[Michael Fidler]

All you have to do now is have an actual tick or pop not caused by HF or RF overload due to electrical resonance (like an actual scratch) to actually be so powerful it overloads the phono section. I imagine that can happen but I've never seen it.

Based on spectrograms going up to 96kHz from transfers I've been sent, it seems unlikely that cartridges can produce RF transients. In an MM system the electrical resonance of the cartridge inductor and cable/load capacitance creates a 2nd order low-pass filter, rolling off around 20-30kHz. I've seen ultrasonic output from MC cartridges go up to 100kHz or so, but acceleration limits dictate that the voltage falls off quite rapidly, at about 10dB or so per octave at a rough glance from 25kHz upwards.

 

[atmasphere]

Based on spectrograms going up to 96kHz from transfers I've been sent, it seems unlikely that cartridges can produce RF transients.

That's right. But they don't need to; all that has to happen is energy of the right type to send the RF peak (which might be 30dB) into oscillation.

In an MM system the electrical resonance of the cartridge inductor and cable/load capacitance creates a 2nd order low-pass filter, rolling off around 20-30kHz.

There's a pretty significant peak before it rolls off. Again there is an electrical resonance but owing to the lower Q of the cartridge coil, its not as profound and is a flatter peak. But it can still be up to 20dB.
More

 

[Michael Fidler]

That's right. But they don't need to; all that has to happen is energy of the right type to send the RF peak (which might be 30dB) into oscillation.

There's a pretty significant peak before it rolls off. Again there is an electrical resonance but owing to the lower Q of the cartridge coil, its not as profound and is a flatter peak. But it can still be up to 20dB.
More

If you're getting a peak of 20dB, then something has gone badly awry with the design. You can avoid this by using a capacitance of not more than 100pF for MM cartridges, and 680pF with MC types, loaded with 120 ohms. This is not a matter of overload for the amplifier, but an inappropriate design choice in terms of cartridge Q compensation/loading.

If an RF transient can send the amplifier into oscillation, then it's not fit for purpose. I saw something similar on my first MC Pro prototype where the servo amplifier was an NE5532 and could start oscillating with some cartridge load combinations when the input was kicked hard enough - the output was connected to another 5532 input, which has a pair of diodes internally connected to the other input side and therefore went non-linear and destabilised the circuit once the threshold had been surpassed by a significant transient. Using an NJM2068 for the servo solved the problem entirely.

 

[markstein]

A little bit out of the context of the previous posts, but anyway...

I just recorded again the spectra of (typical?) music without RIAA applied (straight into my mic preamp) and with RIAA and rumble filter applied. This may help to understand how music is encoded on a record, and what RIAA equalization does to the spectrum.

Example: The man's too strong, Dire Straits - Album Brothers in Arms, 45 RPM, "half speed remastered"

Brothers in Arms (Half Speed Remastered 2LP) [Vinyl LP] https://amzn.eu/d/1CGN6VH

What you can see is that the spectrum as cut into the record does not rise with frequency. It is rather flat, with a maximum in the 500Hz-1000Hz area. I have seen this with other music as well, so this seems somewhat representative.

The capture does not show that single sample peak values at the mic preamp go up to -2,7dBFS (I guess this is the loudest record I own). Notice the dB scale on the left does not show the correct values as I "pulled apart" the curves is REW.

This specific example seems to be digitally remastered, as there is a brickwall like drop above approx. 20kHz. I do not see this with other records (especially on 33RPM records the highs might be lower and the transition to higher frequencies will be smoother).

I can not see any HF issues until Nyquist (96kHz sampling rate), however, my RIAA equalization does not stop at 20kHz but continues attenuating higher frequencies until Nyquist. I have as well loaded my MC cartridge with a 150R resistor inside the XLR plugs to the mic preamp, between pins 2 and 3 (signal +/-).

 

[atmasphere]

If you're getting a peak of 20dB, then something has gone badly awry with the design. You can avoid this by using a capacitance of not more than 100pF for MM cartridges, and 680pF with MC types, loaded with 120 ohms. This is not a matter of overload for the amplifier, but an inappropriate design choice in terms of cartridge Q compensation/loading.

If an RF transient can send the amplifier into oscillation, then it's not fit for purpose. I saw something similar on my first MC Pro prototype where the servo amplifier was an NE5532 and could start oscillating with some cartridge load combinations when the input was kicked hard enough - the output was connected to another 5532 input, which has a pair of diodes internally connected to the other input side and therefore went non-linear and destabilised the circuit once the threshold had been surpassed by a significant transient. Using an NJM2068 for the servo solved the problem entirely.

I think you misunderstand what I said.
The peak has nothing to do with the phono section (other than how the phono section reacts to it). In the case of a MM cartridge, the peak is the result of what happens when you put a inductance (the cartridge, in this case low or moderate Q value in the coils) in parallel with a capacitance (that of the tonearm cable). The peak can be up to 20dB depending on the Q of the coils. In the case of LOMC, because the inductance is lower the frequency is higher; because the Q of the coils is higher the peak is more sharply tuned and so is also higher, up to 30dB and usually in the MHz range. This is shown in the link I provided.

By simply adding capacitance the peak is moved down in frequency. This is why tonearm cables are low capacitance. You need a resistor/capacitor network to load a MM cartridge properly (see the link).

LOMC cartridges have inductance so low as to be unaffected at audio frequencies by loading until the load is much lower than most people might consider using. IOW you can pass a 10KHz squarewave through the coil and the output looks like the input (no rounding or ringing). So loading isn't affecting the cartridge at audio frequencies other than making the cantilever harder to move due to forcing it to do more work.

So the RF likely isn't inducing an oscillation- it instead is the result of the peak going into excitation (which is another word for oscillation) independently of the phono section. How the phono section reacts to RFI injected into its input is another matter altogether. Some simply make more distortion but do so without overloading. For this reason if a phono section is capable of LOMC operation, I recommend judicious use of RF beads or coils to filer RFI as well as stopping resistors for a base, grid or gate of the input device to limit bandwidth via Miller Effect.

When you see loading provisions in the LOMC phono section its a good indication the designer ignored the implications of putting an inductance in parallel with a capacitance.

 

[Multicore]

@Michael Fidler Thank you for your generous contributions here. I didn't imagine I'd learn so much when I asked the question and certainly didn't think we'd get so much circuit design advice. I was never an analog designer so some of it is over my head.

That phantom-powered phono to mic-preamp driver looks ideal for my purposes. I have a friend who plays his vinyl records for fun and frequently just for background while he's in another room. He doesn't clean them properly, or his stylus, or replace the stylus often enough. These habits are wearing out the grooves. Not me. I want to play each of my LPs only once again and then enjoy them digitally after that.

 

[Michael Fidler]

I think you misunderstand what I said.
The peak has nothing to do with the phono section (other than how the phono section reacts to it). In the case of a MM cartridge, the peak is the result of what happens when you put a inductance (the cartridge, in this case low or moderate Q value in the coils) in parallel with a capacitance (that of the tonearm cable). The peak can be up to 20dB depending on the Q of the coils. In the case of LOMC, because the inductance is lower the frequency is higher; because the Q of the coils is higher the peak is more sharply tuned and so is also higher, up to 30dB and usually in the MHz range. This is shown in the link I provided.

By simply adding capacitance the peak is moved down in frequency. This is why tonearm cables are low capacitance. You need a resistor/capacitor network to load a MM cartridge properly (see the link).

LOMC cartridges have inductance so low as to be unaffected at audio frequencies by loading until the load is much lower than most people might consider using. IOW you can pass a 10KHz squarewave through the coil and the output looks like the input (no rounding or ringing). So loading isn't affecting the cartridge at audio frequencies other than making the cantilever harder to move due to forcing it to do more work.

So the RF likely isn't inducing an oscillation- it instead is the result of the peak going into excitation (which is another word for oscillation) independently of the phono section. How the phono section reacts to RFI injected into its input is another matter altogether. Some simply make more distortion but do so without overloading. For this reason if a phono section is capable of LOMC operation, I recommend judicious use of RF beads or coils to filer RFI as well as stopping resistors for a base, grid or gate of the input device to limit bandwidth via Miller Effect.

When you see loading provisions in the LOMC phono section its a good indication the designer ignored the implications of putting an inductance in parallel with a capacitance.

OK - I was confused because you're referring to resonance as oscillation. If you have a peak at RF, that's resonance, but if it's self-sustaining (necessarily powered unless we're breaking the laws of thermodynamics), that would be oscillation.

MM cartridges should not have an electrical peak exceeding a few dB at most. Manufacturers will often use the electrical peak to add an extra bit of boost to compensate for mechanical roll-off at the top of the audio band.

LOMCs are loaded with low resistance to absorb energy in the system and prevent peaks. With a 120 ohm loading resistor and 800pF of capacitance, it's quite easy to make sure the response never peaks more than 3dB or so with the vast majority of cartridges. Adding more capacitance creates a lower frequency of resonance and also increases the amplitude of the peak. The peak is proportional to the relative impedance of the LC circuit, against the loading resistor which absorbs the energy. Adding RF beads don't do much to prevent RF peaks, as the resonant frequency is much too low for them to be effective, and if the layout is done carefully the loading capacitor will make short work of VHF/UHF interference through simple shunt action. The problem with LOMC RFI is typically within the region of 1-5MHz.

As far as HOMC cartridges go, the inductance is around 1-2mH, with 100-200 ohms of dc series resistance. This can produce a peak of around +10dB to +20dB at 300-600kHz when loaded with 50k/100pF, which is luckily low enough for most high GBW-product ICs to handle without non-linearity or detection of ambient RFI pickup, so long as you make sure you have a continuous RIAA response to attenuate it for the next piece of gear in the signal path. ICs have much better immunity to RFI than discrete transistor designs, and using an IC such as the 5534 without compensation makes an input that's highly robust. I can speak from experience here in many such home applications.

In relation to surface transients exciting RF resonances: considering that the mechanical rolloff of the cartridge begins between 20-40kHz in most cases, and that the worst-case resonances that will have serious peaks (the HOMC case) of a high Q from 300kHz upwards, I'm quite happy to categorically say this will never be an issue. The mechanical response of the cartridge is simply unable transmit anywhere near a high enough frequency to remotely tickle a 300Hz RF resonance. RF peaks with high and low output MC devices are problematic because they are excited by ambient/environmental interference picked up by cabling, not because they can be set off by a miraculous cantilever and stylus that are capable of transmitting 300kHz signals.

 

[Michael Fidler]

@Michael Fidler Thank you for your generous contributions here. I didn't imagine I'd learn so much when I asked the question and certainly didn't think we'd get so much circuit design advice. I was never an analog designer so some of it is over my head.

That phantom-powered phono to mic-preamp driver looks ideal for my purposes. I have a friend who plays his vinyl records for fun and frequently just for background while he's in another room. He doesn't clean them properly, or his stylus, or replace the stylus often enough. These habits are wearing out the grooves. Not me. I want to play each of my LPs only once again and then enjoy them digitally after that.

Let me know if you need a PCB! I might do a proper layout for this one if enough people are interested. I was doing quite a few P48 sketches after doing some electret-condenser designs earlier this year that worked far too well to not be explored further.

 

[atmasphere]

OK - I was confused because you're referring to resonance as oscillation. If you have a peak at RF, that's resonance, but if it's self-sustaining (necessarily powered unless we're breaking the laws of thermodynamics), that would be oscillation.

MM cartridges should not have an electrical peak exceeding a few dB at most. Manufacturers will often use the electrical peak to add an extra bit of boost to compensate for mechanical roll-off at the top of the audio band.

LOMCs are loaded with low resistance to absorb energy in the system and prevent peaks. With a 120 ohm loading resistor and 800pF of capacitance, it's quite easy to make sure the response never peaks more than 3dB or so with the vast majority of cartridges. Adding more capacitance creates a lower frequency of resonance and also increases the amplitude of the peak. The peak is proportional to the relative impedance of the LC circuit, against the loading resistor which absorbs the energy. Adding RF beads don't do much to prevent RF peaks, as the resonant frequency is much too low for them to be effective, and if the layout is done carefully the loading capacitor will make short work of VHF/UHF interference through simple shunt action. The problem with LOMC RFI is typically within the region of 1-5MHz.

As far as HOMC cartridges go, the inductance is around 1-2mH, with 100-200 ohms of dc series resistance. This can produce a peak of around +10dB to +20dB at 300-600kHz when loaded with 50k/100pF, which is luckily low enough for most high GBW-product ICs to handle without non-linearity or detection of ambient RFI pickup, so long as you make sure you have a continuous RIAA response to attenuate it for the next piece of gear in the signal path. ICs have much better immunity to RFI than discrete transistor designs, and using an IC such as the 5534 without compensation makes an input that's highly robust. I can speak from experience here in many such home applications.

In relation to surface transients exciting RF resonances: considering that the mechanical rolloff of the cartridge begins between 20-40kHz in most cases, and that the worst-case resonances that will have serious peaks (the HOMC case) of a high Q from 300kHz upwards, I'm quite happy to categorically say this will never be an issue. The mechanical response of the cartridge is simply unable transmit anywhere near a high enough frequency to remotely tickle a 300Hz RF resonance. RF peaks with high and low output MC devices are problematic because they are excited by ambient/environmental interference picked up by cabling, not because they can be set off by a miraculous cantilever and stylus that are capable of transmitting 300kHz signals.

You might want to rethink some of this. If the phono section is immune to RFI you'll find the LOMC cartridge does not need a load to work properly. This can be important since the compliance of the cartridge forms a mechanical resonance in the tonearm that typically best between 7 and 12 Hz. If you load the cartridge you make it do more work (100 Ohms will be over 2 orders magnitude more as opposed to 47K) and can cause it to be outside this window if you chose the cartridge to work in your arm based on its compliance (there are calculators online that assist with this).

IOW they don't have a need for for loading- the loading is for the benefit of a phono section that can't handle the RFI. The only peak you're really dealing with is the electrical resonance which generates RFI as it goes in an out of excitation. All you need is something that can generate a bit of flyback which normal cartridge tracking can provide.

As a result if the phono section can handle the RFI its plug and play with no need to pay attention to 'cartridge loading'; the 47K value is sufficient.

WRT to MM cartridges- are you saying the calculator at the link I provided is incorrect? I'm dubious as I became aware of this resonance about 35 years ago trying to solve an employee's playback problem of ticks and pops. He was using the same Grado MM cartridge I was which was high output. The same LP played on his system yielded ticks and pops while it got none on my system. Turned out his preamp was the culprit; a Japanease opamp based design from the late 1970s. The resonance causing it to overload was ultrasonic. I'm assuming the designer simply thought enough gain and proper EQ would do the trick. It didn't.

We use RF chokes rather than beads; I mis-spoke that bit.

I agree 100% with your 4th paragraph.

 

[Michael Fidler]

You might want to rethink some of this. If the phono section is immune to RFI you'll find the LOMC cartridge does not need a load to work properly. This can be important since the compliance of the cartridge forms a mechanical resonance in the tonearm that typically best between 7 and 12 Hz. If you load the cartridge you make it do more work (100 Ohms will be over 2 orders magnitude more as opposed to 47K) and can cause it to be outside this window if you chose the cartridge to work in your arm based on its compliance (there are calculators online that assist with this).

IOW they don't have a need for for loading- the loading is for the benefit of a phono section that can't handle the RFI. The only peak you're really dealing with is the electrical resonance which generates RFI as it goes in an out of excitation. All you need is something that can generate a bit of flyback which normal cartridge tracking can provide.

As a result if the phono section can handle the RFI its plug and play with no need to pay attention to 'cartridge loading'; the 47K value is sufficient.

WRT to MM cartridges- are you saying the calculator at the link I provided is incorrect? I'm dubious as I became aware of this resonance about 35 years ago trying to solve an employee's playback problem of ticks and pops. He was using the same Grado MM cartridge I was which was high output. The same LP played on his system yielded ticks and pops while it got none on my system. Turned out his preamp was the culprit; a Japanease opamp based design from the late 1970s. The resonance causing it to overload was ultrasonic. I'm assuming the designer simply thought enough gain and proper EQ would do the trick. It didn't.

We use RF chokes rather than beads; I mis-spoke that bit.

I agree 100% with your 4th paragraph.

Good luck building an input with less than 0.6nV/sqrtHz with RFI immunity. I'm working on a class-A 'high-end' type to deal with this issue for the future. You only need the tiniest amount to generate enough non-linearity to be obtrusive as the voltage is so low.

I have no comment to make on other sources on websites, other than that I do my own simulation of these thing either in LT Spice, or a mathematical plot. Regarding Japanese op-amps of the late 1970s, it's highly likely that impedance overload from the RIAA network and slew rate are problem for this period. They like a very low impedance RIAA network, something like 3 times lower than I prefer in contemporary equipment using the 2068 IC. In your case I would posit that it's not resonance, rather just transients causing overload. The higher the output of an MM cartridge, the more curtailed its frequency response is generally speaking, as this usually means more inductance and therefore a lower electrical cut-off point.

I am also experimenting with RF chokes for better immunity. Something like 100uH, although care will need to be taken to make sure that there's enough immunity to external magnetic fields given the astronomical gain at 50Hz. Luckily, you can buy SMT devices with full magnetic shielding, although further testing is needed.

In my experience, a low-impedance load is absolutely necessary for LOMC with the kind of input amplifiers that give a reasonable noise figure. You can of course just use a high-GBWP device like a 4562 and wing it, but the noise figure is very poor indeed.

 

[atmasphere]

Good luck building an input with less than 0.6nV/sqrtHz with RFI immunity. I'm working on a class-A 'high-end' type to deal with this issue for the future. You only need the tiniest amount to generate enough non-linearity to be obtrusive as the voltage is so low.

I have no comment to make on other sources on websites, other than that I do my own simulation of these thing either in LT Spice, or a mathematical plot. Regarding Japanese op-amps of the late 1970s, it's highly likely that impedance overload from the RIAA network and slew rate are problem for this period. They like a very low impedance RIAA network, something like 3 times lower than I prefer in contemporary equipment using the 2068 IC. In your case I would posit that it's not resonance, rather just transients causing overload. The higher the output of an MM cartridge, the more curtailed its frequency response is generally speaking, as this usually means more inductance and therefore a lower electrical cut-off point.

I am also experimenting with RF chokes for better immunity. Something like 100uH, although care will need to be taken to make sure that there's enough immunity to external magnetic fields given the astronomical gain at 50Hz. Luckily, you can buy SMT devices with full magnetic shielding, although further testing is needed.

In my experience, a low-impedance load is absolutely necessary for LOMC with the kind of input amplifiers that give a reasonable noise figure. You can of course just use a high-GWBP device like a 4562 and wing it, but the noise figure is very poor indeed.

Yes, the gain and RFI immunity at the same time is challenging. We did it by using tubes in a differential embodiment, the first of its kind back in the 1980s. I think you are correct about that old preamp. I don't think 4558s helped matters any either, or the lower supply Voltages.

Isn't the load of the LOMC cartridge itself low enough for an opamp input? If you parallel it with a 100 Ohm resistor the impedance is nearly unchanged.

Couldn't you use Miller Effect to your advantage for RF immunity? Or is Johnson noise the concern?

 

[Michael Fidler]

Yes, the gain and RFI immunity at the same time is challenging. We did it by using tubes in a differential embodiment, the first of its kind back in the 1980s. I think you are correct about that old preamp. I don't think 4558s helped matters any either, or the lower supply Voltages.

Isn't the load of the LOMC cartridge itself low enough for an opamp input? If you parallel it with a 100 Ohm resistor the impedance is nearly unchanged.

Couldn't you use Miller Effect to your advantage for RF immunity? Or is Johnson noise the concern?

4558 - only about 1V/uS, so you're only going to be getting 5.6V RMS maximum at 20kHz.

You use the parallel resistor as an absorber to reduce the Q of the LC resonance of the cartridge for LOMC, in much the same way as an MM cartridge.

IME Miller Effect is of limited use, as we have to excite an amplifier to use it, which sort of defeats the purpose. If you wanted to filter using a series element, then the best way to implement the capacitance would be by parallel action to ground.

 

[atmasphere]

You use the parallel resistor as an absorber to reduce the Q of the LC resonance of the cartridge for LOMC, in much the same way as an MM cartridge.

IME Miller Effect is of limited use, as we have to excite an amplifier to use it, which sort of defeats the purpose. If you wanted to filter using a series element, then the best way to implement the capacitance would be by parallel action to ground.

Again, if you use a resistance to detune the LC network, you cause the cantilever to be less compliant. Jonathan Carr, a noted designer of LOMC cartridges and I had a conversation about this topic several years ago at the Munich show. He has a different perspective of course but we both arrived at the same conclusions. Were it not for the ideal tracking issue using a resistor (which can also affect HF tracking ability) would be nice as its a lot easier.

I've not used Miller Effect in a solid state embodiment for a phono section yet as I've only been considering such a preamp in the last year or so. But its been very handy to deal with this issue in a tube phono section.

 

[Michael Fidler]

Again, if you use a resistance to detune the LC network, you cause the cantilever to be less compliant. Jonathan Carr, a noted designer of LOMC cartridges and I had a conversation about this topic several years ago at the Munich show. He has a different perspective of course but we both arrived at the same conclusions. Were it not for the ideal tracking issue using a resistor (which can also affect HF tracking ability) would be nice as its a lot easier.

I've not used Miller Effect in a solid state embodiment for a phono section yet as I've only been considering such a preamp in the last year or so. But its been very handy to deal with this issue in a tube phono section.

The idea that the coils are able to absorb any power and affect compliance is a myth once you crunch the number, I'm afraid. Even if the coil is shorted, it is only able to absorb a fraction of a percent of the power stored in the moving mass of the cantilever itself alone every second, not even considering the mechanical compliance of the damping material the cantilever is mounted on. I'm constantly bombarded with questions regarding current-input/transimpedance MC stages that represent a very low impedance input and feature the claim that they are able to mechanically damp the system by absorbing its energy - simply not true. I will write this up and show my workings when I make all the points in my case against this, but have been aware of it since well before 2022.

If you're using the Miller effect in valve front end (lots of even order distortion), then you are exciting the distortion mechanism (non-linear gain) to generate the capacitance, which somewhat/entirely defeats the purpose. For valve devices, a transformer is mandatory for LOMC if you want an SNR higher that's approaching as low as 60dB (best noise I've measured on a valve input was 7nV/sqrtHz).