Understanding this MM/MC discrete phono stage

Hello fellow DIYers,

I am challenging myself at understanding my integrated amp's MM/MC phono stage so I can service it if needed and also make a DIY clone. It is a somewhat complex design and as a hobbyist I'm out of my league. The circuit looks a bit like an op-amp built out of discrete parts, so I have taken the typical op-amp block diagram as an inspiration and color-coded the sub-circuits I think I recognize but I'm not certain of anything except the constant-current sources and sinks as I recognize their LED voltage references.

The amp is a Rega Elicit mk. I; it is a very nice sounding amp overall and the phono stage is superb. Sorry for the poor image quality, the amp is almost 35 years old and all I have are those scans of the schematics (They're not even of the right board revision but are close enough). Image taken out of Hi-Fi Engine Library : https://www.hifiengine.com/hfe_downloads/index.php?rega/rega_elicit_schematics.pdf

elicit phono stage annotation.jpg


Below is the block outline color legend:

  • Blue: Darlington pair emitter follower current buffers
  • Red: Constant current sources and sinks
  • Brown: Cascode differential input stage ( Emitter follower feeding common base, current sources and sinks for bias)
  • Green: Level shifter
  • Cyan: Class AB output stages
  • Orange: Long-tail pair diffential amplifiers with constant current sources and current mirror active loads performing single-ended to push-pull conversion
  • Gray: Unknown configuration...
I would appreciate any help in correctly identifying the sections and type of the building blocks so I can study each sub-circuit in depth. I am in the process of porting the schematic into LTSpice but I see no point in trying to analyse a simulation if I don't understand the basics first.

I am also not 100% sure how is the RIAA equalization performed. R220 (2k4) and C77 (33nF) do look like a low-pass filter and the time constant is very close to the 75uS pole. R155 (97k6) and C64 (33nF) looks like the 3180uS pole, while R154 (10k) and C64 (33nF) looks like the 318uS pole, combined and fed as NFB to the second differential amplifier as an active filter?

Every piece of information will help me resolve this puzzle. Many thanks in advance.
Joris
 
Your analysis is accurate.

The pre-amp is comprised of two cascaded discrete transistor op-amps. It was designed when fully complementary transistor amplifiers were in vogue.

The op-amps use cascode stages because the supply voltages are high.

The brown box is an op-amp that provides voltage gain.

The 75uS pole uses passive equalization on the op-amp's output. There is one DC blocking capacitor.

The 3180uS pole uses active equalization in the second op-amp's feedback loop. There is a DC feedback capacitor.

The gray stage is a level shifter followed by a voltage amplifier.

The cyan stages are biased high enough to remain in class A.

This is a no-holds-barred design. The only faults I can see are the series resistors on the input, and the cyan stages could have been class A only. The high voltage is a brute-force approach to getting headroom (reduced by the series resistor in the 75uS EQ).

BTW, this is the first phono pre-amp I have seen that uses more transistors than mine. 😉
Ed
 
  • Like
Reactions: SomeJoe and 6L6
The dark blue things are meant to filter off ripple on the supplies: RC low-pass filters followed by voltage followers.

A weak point of the design is its equivalent input noise current when you use it with moving-magnet cartridges. It should work fine with moving coil, but for a 500 mH moving-magnet cartridge, the total collector bias current of the input stage is about 100 times the optimal value for minimum noise.
 
  • Like
Reactions: SomeJoe
What interesting answers! This is really stimulating as it provides new avenues to my DIY design. It also prompts new questions ! 😎

The op-amps use cascode stages because the supply voltages are high.
[...]
The high voltage is a brute-force approach to getting headroom (reduced by the series resistor in the 75uS EQ).
Thanks Ed for a very complete reply! Yes the voltage is quite high and I find there is some power wastage in there. The stage is built to use the same regulated supply as the rest of the amp, but I intend to build my DIY clone to work with a lower voltage supply - something like +/- 24V or even down to +/-15V if I can while maintaining enough headroom for the passive eq loss. That is why I need to fully understand the design if I'm to tweak it. I believe the cascode implementation is a keeper though as it has advantages in terms of high input impedance and isolation?

[...] The only faults I can see are the series resistors on the input, and the cyan stages could have been class A only.
I'll investigate the design of a class A-only stage, if I'm not mistaken one needs to have a higher bias and this can be achieved with the so-called Vbe multiplier or "rubber diode" circuit instead of the three diodes in series?
What value of input resistor would you suggest? I'm guessing it is the 1K you would lower, as the 10R is already quite low?

BTW, this is the first phono pre-amp I have seen that uses more transistors than mine. 😉
Actually I am a bit discouraged at the number of TO-92's I'll need to solder in my DIY implementation 😱 I'll probably replace the discrete darlingtons with integrated ones to try and reduce the parts count a bit, However if voltage headroom is tight I might have to revert to a discrete "Sziklai pair" to save a Vbe drop on each rail... The 10R feed resistor could be reduced as well I think.

It should work fine with moving coil, but for a 500 mH moving-magnet cartridge, the total collector bias current of the input stage is about 100 times the optimal value for minimum noise.
Interesting! What value of collector current would be a good compromise between MM and MC performance? I'll integrate this mod in my design, as it is basically merely a change in resistor values. Can you provide links to some reference texts regarding the relation of collector current and cartridge? Thanks in advance.

No selectable input capacitance either, so wide-open to RFI.
This is another tweak I can easily integrate in a DIY build, and I already had this in mind. Have to check some implemenatations, if you have pointers that would be greatly appreciated. Off the top of my head I guess that's probably just a 3-position dual pole switch hooked to some polystyrene caps.

Thanks to all for your replies!
- Joris
 
I have to get back to you regarding good compromise collector currents. I derived a good compromise a couple of years ago, but can't quite remember how.

Regarding equivalent input noise current and moving magnet, see "Noise and moving-magnet cartridges", Electronics World October 2003, pages 38...43, https://worldradiohistory.com/UK/Wireless-World/00s/Electronics-World-2003-10-S-OCR.pdf Mind you, Electronics World drew one of the sections of the gain switch in the wrong state in figure 5 and I mixed up the terms spectral density and power spectral density.
 
  • Like
Reactions: SomeJoe
The cascode has multiple benefits including input shielding and elimination of the Miller effect.

My pre-amplifier uses only class A stages. Class AB is a vestige of IC op-amps. Biasing it into class A eliminates the crossover notch but leaves the components. You should not change this unless you are capable of designing an op-amp.

I can't make out "K" versus "Ω" on the scan. The input should not have series resistors.
Ed
 
I have to get back to you regarding good compromise collector currents. I derived a good compromise a couple of years ago, but can't quite remember how.

Regarding equivalent input noise current and moving magnet, see "Noise and moving-magnet cartridges", Electronics World October 2003, pages 38...43, https://worldradiohistory.com/UK/Wireless-World/00s/Electronics-World-2003-10-S-OCR.pdf Mind you, Electronics World drew one of the sections of the gain switch in the wrong state in figure 5 and I mixed up the terms spectral density and power spectral density.
Thanks for the link, I couldn't get better advice than the author of a published article on that particular subject! I'm not through the article yet but I feel this theory will help me move my game up. I see your example circuit implements variable capacitive loading, I will integrate this in my design (albeit perhaps with less values).

Are you still writing? Just curious 🙂
 
My pre-amplifier uses only class A stages.
Ed, I would be curious to see that design!

I see on your profile pic you have Maggies, I have a pair of MG12s here I used to enjoy a lot but the wiring has delaminated from the mylar on one, I have to open and re-glue the wires. Haven't got the confidence to operate on them yet...
 
What value of collector current would be a good compromise between MM and MC performance?

About 516 uA for all six transistors together, so 86 uA each. That's assuming a 500 mH moving-magnet cartridge and a moving-coil cartridge with negligible impedance producing 10 times less signal voltage than the moving-magnet cartridge, and again using A- and RIAA-weighting.

That is, the noise contributions that you can change by tweaking the collector current are the collector shot noise, which translates into (mostly) equivalent input voltage noise, and the base shot noise, which translates into (mostly) equivalent input current noise. The voltage noise will mostly affect the moving-coil case and the current noise the moving-magnet case.

When the moving-magnet cartridge produces a signal voltage VMM while the moving coil cartridge produces a signal voltage VMC under the same circumstances (same record, that is), you can make the loss of signal-to-noise ratio due to the base shot noise in the moving-magnet case equal to the loss of signal-to-noise ratio due to the collector shot noise in the moving-coil case. That boils down to optimizing the input stage for VMC/VMM times the impedance of the moving-magnet cartridge.

There are alternatives, of course, such as high-transconductance JFET input stages (at the expense of a bit more silicon and current), bipolar input stages of which the collector current can be switched or different input stages for MM and MC.
 
Last edited:
About 516 uA for all six transistors together, so 86 uA each.
Marcel, thank you SO MUCH for taking the time to compute this. I will apply this to my design for sure. Also thank you for the detailed explanation. I will have to study the subject in detail as this borders the limits of my knowledge, but knowledge can expand and your article and explanations are invaluable!

That's assuming a 500 mH moving-magnet cartridge
I couldn't find the inductance figures readily in the manuals for my MM carts but I'll do more digging later. I assume however that this is a ballpark value for most MMs? What I'm curious about is the figure for my Denon DL-110 "high output" moving coil that outputs voltages comparable to MM but is an actual MC...

I have finished drawing the schematic in KiCAD and tried a tentative layout to see if I could cram the design on a 10x10cm PCB so it can be had cheaply as a prototype board from JLCPCB and the like. Man that's a lot of transistors - 38 count for a single channel without power supply. There is ample space around the caps to account for size variations of actual parts when I source them, but looks like I can also fit a few more small caps and a DIP switch for capacitive loading choice too.

Elicit Mk. I revision 3 Phono Stage v.0.1.jpg


The clone is needed as a backup for my main amp as the old thing needs to have its electrolytics replaced; but with the tweaks we are talking about I believe it will be a potent preamp in its own right. It will also serve as a platform to safely test component upgrades before porting them to the Elicit.

Now back to LTSpice for simulations and study. I also have to figure out what's the best PCB stack-up for low noise. I have currently opted for a single top-side ground plane.

Thanks again,
Joris
 
500 mH is pretty normal for MM; some are around 200 mH, some 450 mH, some 700 mH.

According to https://www.denon.com/en-us/product/turntables/dl-110/136623.html your Denon DL-110 high-output MC cartridge has a nominal impedance of 160 ohm. If that is at 1 kHz, the inductance must be less than 25.5 mH. The optimum for this specific cartridge must then be somewhere between 1 mA and 4 mA for the sum of the collector currents of the six input transistors - 1 mA if it is mainly inductive at 1 kHz, 4 mA if it is mainly resistive at 1 kHz.
 
Last edited:
A very nice design. The input stage aims for lower noise while the second stage aims for lower distortion. They are basically the same amplifier with different goals in mind. The input stage employs masking and some how mimicks a jfet, you should probably have a switchable input stage between bjt masking and jfet masking and see which you prefer in AB testing. However just as it is its a superb phono stage. No need to match the input parrallel input transistors, the purpose is to assist them correct each other so they essentially should have different characteristics.
 
Last edited:
500 mH is pretty normal for MM; some are around 200 mH, some 450 mH, some 700 mH.

......If that is at 1 kHz, the inductance must be less than 25.5 mH. The optimum for this specific cartridge must then be somewhere between 1 mA and 4 mA for the sum of the collector currents of the six input transistors - 1 mA if it is mainly inductive at 1 kHz, 4 mA if it is mainly resistive at 1 kHz.....
When searching the net, I found that most manufacturers seem to specify impedance at 1K without stating resistive and inductive components.
The exception seem to be Hana, specifying total impedance and inductance. Assuming that the inductance value of 0,61mH is typical of HOMC pickups it seems the higher current value is the one to go for.

1699913581995.png
 
  • Like
Reactions: MarcelvdG
Marcel,
Two further questions where I would appreciate your input.
1 - Assuming that one would like to optimize the design above for a Ortofon 2M series MM pickup that sports an inductance of 700mH / DCR=1300ohm. Would you recommend using 2x3 parallel transistors running 90uA per transistor or single NPN/PNP transistors (2x1) running 270uA?

2 - The explanations you've given elsewhere of minimizing noise at a spot frequency of 5179 Hz is useful for analyzing the input stage. However, I have seen little discussion on noise properties of the second stage. Given a two stage deign of this type with passive high-cut and active low-boost, the second stage source impedance increases towards 43K at low frequencies on this Rega design. Other designs I've seen vary from 75K (Rotel 970bx) to 2K8 (NAD S100) or less (e.g Borbely). Given that the second stage will only have low frequency boost and most amplifiers have increasing 1/f noise with decreasing frequencies I assume that the 5179Hz spot frequency is not relevant to looking at the second stage in isolation.
Do you have any advice or cues on what to look for when optimizing the second stage for noise?
 
If that is at 1 kHz, the inductance must be less than 25.5 mH
So this is just the regular formula for inductive reactance. Sometimes I can't see what's right in front of me... 😎
1 mA if it is mainly inductive at 1 kHz, 4 mA if it is mainly resistive at 1 kHz.
So if I understand you correctly, 516 uA total current is the sweet spot for MM, but around 4-6 mA for MC or high output MM. I have simulated the first gain stage (before passive eq) in LTspice today, with zero input. Fiddled a bit because my models for BD139/140 seem innacurate, had to tweak resistors R241 and R245 from 470R to 1K to get the voltages specified on schematics at points around the inputs. I'm a bit puzzled as the models were freshly downloaded from ST.com...

Anyways once the voltages were right I messed around with values to get a grasp of the design. Looks like resistors A1 and A3 set the bias for the input transistors; the simulation gives about 300uA per transistor (1.8 mA total) with the specified 470R value; going up to 550R and the collector current went down to 165uA each (about 1mA total). Beyond that LTspice seems to struggle with "stepping the source" (whatever that means 😡 ).

So I guess it would be possible to insert DIP-switched resistor(s) in series with those resistors to get a super-duper variable op point. If this is at all possible this would give the design an unprecedented level of adjustment...

Speaking of adjustments, I have checked what kind of capacitive input settings are present on commercial offerings, and I've seen just about anything in range. Top-end Rega preamp goes from 1000p to 5000p.... A long way from the 100p in my Elicit.. And Marcel's example in Electronics World goes up to about 550p. What would be the best range? I was thinking DIP-switcheable 50p, 100p, 220p plus the hard-wired 100p but that goes up only to 470p maximum when all are in circuit.

Nice thread going on, thanks all for your valuable input!
Joris
 
Last edited:
Thank you OnAudio for your reply.
The input stage employs masking
What do you mean by "masking"? The high input isolation offered by the cascode gain stage or ?
No need to match the input parrallel input transistors
You saved me a lot of time there as I was going to do some binning on these! However I guess there is still a need to match the complementary pairs in Hfe?