As for the rest, we appear to have reached a stage where I have presented a fully working amplifier, designed by proper engineering practice and in service for 6+ years, which you claim to be inferior to some mythical design. An design for which you have presented no evidence whatsoever, whether for distortion, noise or anything else - unless one includes a screen grab of a 60 year old data sheet (which demonstrated nothing at all).
Further, this so-called design changes its spots according to the nature of the debate: includes an emitter follower when the question is the stability/ageing of the EQ, loses it again when talk is of device count.
In other words, we are comparing a real fixed, working design - to an elusive and evasive one that does not exist.
If you are serious about the debate, then design your MC phono amplifier, using your best principles, and present the schematic and design reasoning here. Build it and characterise it, and I will respond in kind by dragging my amp into the lab and doing the FFT again with my latest measurement set-up. Only then will there be something worthwhile to proceed with.
reminder of the design basics:
MC in, 300uV for 1kHz, 50mm/s
gain of 82-85dB minimum, before EQ.
Use any devices you like, but keep in mind that op-amps will not support any of the arguments you have been making.
Further, this so-called design changes its spots according to the nature of the debate: includes an emitter follower when the question is the stability/ageing of the EQ, loses it again when talk is of device count.
In other words, we are comparing a real fixed, working design - to an elusive and evasive one that does not exist.
If you are serious about the debate, then design your MC phono amplifier, using your best principles, and present the schematic and design reasoning here. Build it and characterise it, and I will respond in kind by dragging my amp into the lab and doing the FFT again with my latest measurement set-up. Only then will there be something worthwhile to proceed with.
reminder of the design basics:
MC in, 300uV for 1kHz, 50mm/s
gain of 82-85dB minimum, before EQ.
Use any devices you like, but keep in mind that op-amps will not support any of the arguments you have been making.
The part you quoted was the cartridge output.
So do I gather that you don't have noise data to support your contention?
edit: Quick noise calculation for FETs. A good n-FET might be 1nV/rt Hz noise density. That calculates out to about 0.15uV over a 20kHz bandwidth. If we add in the 1/f, Merlin's number looks plausible.
I noticed a long time ago that in a cascode, the ability to drive the top tube properly was pretty important. To that end I began experimenting with paralleled tube sections on the bottom portion. This led immediately to greater gain, greater linearity and lower noise (if you also increase the current through the circuit). Not long after that I went fully differential with the whole thing.
I've experimented with high linearity transistors in the bottom section too- with dramatic improvements in gain and noise. Although distortion did not seem any higher, in practice I could not get them to sound right (dang that subjectivity!- they sounded bright and harsh, as if odd-ordered harmonics were showing up). I have tried FETs in the same location- had to hand-pick them to get one that was quieter than the tube in this location! 90% of the FETs were thus reject.
At any rate, I've had good luck with low output moving coil; 0.15mV being about the lower limit and no semiconductors. I am keenly aware of Rod's comment about painting one's self into a corner though!
I've **not** tried putting the semiconductor on top- mostly due to concerns that the tubes on the bottom might be unhappy with the idea. Seems like its worth a shot though...
I've experimented with high linearity transistors in the bottom section too- with dramatic improvements in gain and noise. Although distortion did not seem any higher, in practice I could not get them to sound right (dang that subjectivity!- they sounded bright and harsh, as if odd-ordered harmonics were showing up). I have tried FETs in the same location- had to hand-pick them to get one that was quieter than the tube in this location! 90% of the FETs were thus reject.
At any rate, I've had good luck with low output moving coil; 0.15mV being about the lower limit and no semiconductors. I am keenly aware of Rod's comment about painting one's self into a corner though!
I've **not** tried putting the semiconductor on top- mostly due to concerns that the tubes on the bottom might be unhappy with the idea. Seems like its worth a shot though...
I'm not the end all expert on this, but... what I'm saying here applies to MM style cartridges. I'm not experienced with MC cartridges.
I wouldn't want to drive any RIAA EQ parts with a drastically changing impedance, so no drive off a plate, and I'd want as much gain in the first stage as possible for lowest noise, and of coarse use all metal film R's and polyprop or polystyrene caps. I'd use a current source (yes solidstate) in place of the plate resistors not so much for the increased gain and linearity as much as for the power supply hum rejection.
One of the most important things to do though is to verify and tweek the final circuit using a test record, because the input capacitance needs to be right, and is highly unlikely to be right without this final verification/tweek. Using a variable capacitor is recommended. You can then replace it with a fixed cap if you want.
Having said all that, I would consider using an EF86 pentode for the front end because it has all the gain you'll need for the whole circuit. I'd follow that with a triode follower which would drive the RIAA EQ (not split), and then one more triode follower output buffer. Personally I'd use the 6SN7 for the triode buffers. If you use SS current sources on the followers too in place of the cathode resistors (I would), power supply hum should be very minimal. Grounding will have to be done just right. Put a shield over the input tube, but try to get one that doesn't ring (the spring inside those can ring, and modulate the sound). Putting a vented faraday cage over the tubes might be worth it.
I'd use a mutistage RCRCRC filter in the power supply, and low noise diodes with .01uF caps across them for rectification. I don' t think you'd need to have hi-voltage regulators if you use current sources in the circuit as mentioned above, but I could be wrong about that. IXYS makes good current sources.
Using fancier topologies doesn't appeal to me. More to go wrong. It's already complicated enough. I would not use any negative feedback anywhere.
I wouldn't want to drive any RIAA EQ parts with a drastically changing impedance, so no drive off a plate, and I'd want as much gain in the first stage as possible for lowest noise, and of coarse use all metal film R's and polyprop or polystyrene caps. I'd use a current source (yes solidstate) in place of the plate resistors not so much for the increased gain and linearity as much as for the power supply hum rejection.
One of the most important things to do though is to verify and tweek the final circuit using a test record, because the input capacitance needs to be right, and is highly unlikely to be right without this final verification/tweek. Using a variable capacitor is recommended. You can then replace it with a fixed cap if you want.
Having said all that, I would consider using an EF86 pentode for the front end because it has all the gain you'll need for the whole circuit. I'd follow that with a triode follower which would drive the RIAA EQ (not split), and then one more triode follower output buffer. Personally I'd use the 6SN7 for the triode buffers. If you use SS current sources on the followers too in place of the cathode resistors (I would), power supply hum should be very minimal. Grounding will have to be done just right. Put a shield over the input tube, but try to get one that doesn't ring (the spring inside those can ring, and modulate the sound). Putting a vented faraday cage over the tubes might be worth it.
I'd use a mutistage RCRCRC filter in the power supply, and low noise diodes with .01uF caps across them for rectification. I don' t think you'd need to have hi-voltage regulators if you use current sources in the circuit as mentioned above, but I could be wrong about that. IXYS makes good current sources.
Using fancier topologies doesn't appeal to me. More to go wrong. It's already complicated enough. I would not use any negative feedback anywhere.
Last edited:
JFET en values vary by as much as 30:1, depending on the device and its geometry, so "normal JFETs" means little in terms of low-noise design.
My design adopts the super low noise K369 or K372 parts - which offer the highest degree of low-noise performance available in a practical component.
From the characterisation data, this works out at about 100nV per device.
This family of devices was expressly designed for the first stage of of Moving Coil amplifiers, so it is unsurprising to find that it works very well.
Oh, so when you berated me for being "off by 20dB, if not more" you actually meant "less than 10dB off", yes?
I'm guessing you mean the 2SK369 as I can't find any info on a K369.
http://www.mouser.com/ds/2/408/6937-30733.pdf
Annoyingly, and despite being advertised for audio, the 2SK369 datsheet doesn't include a noise spectral density graph. However, it does include some noise figures:
NF = 5dB 100Hz, and 1dB 1kHz relative to 100 ohms.
A 100 ohm resistor generates 180nV Johnson noise, so the above figures roughly imply 92nV white noise (hey you were right!) But you forgot the pink noise which is somewhere around 260nV. The EIN over 20-20kHz is going to be around this figure. Not far off my estimate.
Last edited:
Yes, The K369 = 2SK369 (all Japanese transitors prefix 2S-).
Your noise voltage is still off - but to be fair, if you are not familiar with these low noise JFETs (how no?) you will have been led astray by the error in that noise figure data.
The 5dB at 100Hz value should read 5dB at 10Hz. As the various graphs on pages 3 and 4 show, the NF for 100Hz is more like 1.3dB.
Your noise voltage is still off - but to be fair, if you are not familiar with these low noise JFETs (how no?) you will have been led astray by the error in that noise figure data.
The 5dB at 100Hz value should read 5dB at 10Hz. As the various graphs on pages 3 and 4 show, the NF for 100Hz is more like 1.3dB.
There's a graph of NF versus frequency which indicates about a 1kHz 1/f corner. It's a pretty quiet device for an FET, but crippled for MM use by a very high input capacitance. Cgs is about 75pF, and with a typical first stage gain of 40dB, add in another 1500 pF for Miller. That's OK for MC use, but the noise is a bit high for that. A cascode could help the input C problem.
It's a very low noise device for a FET. And BJTs are hardly suitable. We need not consider valves, at this signal level - without the aid of expensive transformers.
Attempting to design a first stage to be compatible with MC and MM interchangeably will simply result in compromise in the performance.
Shunt Cascode effectively solves the capacitance problem, at least for MC - for which use it is intended.
Attempting to design a first stage to be compatible with MC and MM interchangeably will simply result in compromise in the performance.
Shunt Cascode effectively solves the capacitance problem, at least for MC - for which use it is intended.
That part I agree with, though a transformer at the input of a low Cin MM preamp will work very well indeed. Input transformers aren't cheap, but many excellent ones aren't that expensive, either, especially compared to the cost of some tubes. For example, the Sowter 9570 is under 70 pounds, and the 8055 that I use are under 80 pounds. Compare that to the going rate for quiet tubes like Telefunken ECC83 or Siemens CCa!
Why aren't BJTs suitable? They are commonly used for MC inputs owing to their lowest possible voltage noise.
Cartridge outputs are always given at 1kHz 5cm/s unless explicitely stated. Noise voltages are understood to be 20-20kHz integrated (sometimes 22-22kHz for some reason) unless explicitely stated.
They may be commonly used in that position, but show us a sample schematic, and we can compare.
An externally hosted image should be here but it was not working when we last tested it.
Last edited:
I've measured 200pF in several common circuits, probably a reasonable average. Cgp seems to vary a lot between brands and always seems to measure higher than the datasheet value. Add in the cable capacitance and you're pushing the issue for many MMs.
We design-in three parallel low noise transistors, and then couple them to the cartridge with an electrolytic capacitor?
If that meets your design criteria for quality, you are welcome to it.
Even if you believe that electrolytics - directly in series with a ~3uA input circuit - are transparent (I will leave that to individuals to judge), you would have to accept that their performance, especially leakage, will degrade starting from the day you build it. Having only 600mV of polarisation certainly will not help.
The squidgy electrolytic internal structure will introduce electromechanical noise at that level, too - just try measuring the output of this MC amp while tapping the PCB the capacitor is mounted on.
The specified leakage into the cartridge will be up to 40uA steady-state, increasing as lifetime expires, and yet more during warm-up. Transient switch-ON pulses may cause unwanted magnetisation of the cartridge.
In all, a circuit designed to meet a headline specification in a cheap consumer product of the 1970s... No need for DIY if this level of performance is good enough - a few pounds of ebay money on a 70s consumer box will do.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.