The Pioneer M22 as a head amp? (and recapping guide)

Discussion in 'Headphone Amplifiers and Combo (DAC/Amp) Units' started by skem, Aug 5, 2018.

Tags:
  1. skem

    skem Friend

    Friend
    Joined:
    Nov 1, 2017
    Likes Received:
    431
    Dislikes Received:
    2
    Trophy Points:
    63
    Location:
    Cambridge, MA
    Turning a Pioneer M22 into a Head Amp
    [​IMG]

    Six months ago, I acquired this beautiful M22, albeit from an unsavory seller. Upon inspection, I discovered that although the amp was extremely nice on the outside, it was in bad repair inside: it had blown fuses, a few swelling capacitors, and terrible DC offset coming from the output. I had bought a corpse.

    This began a months-long journey to rebuild the amp with the goal of turning it into a head amp. What follows is an extensive itemization of what I learned with the hope it will benefit others, including resources for learning esoteric details about capacitors. If you don't care and just want to get to work, here is a handy DigiKey ordering guide so future M22 owners can simply skip the research and re-cap their amps based on my choices—for whatever they're worth. Which leads to an important caveat: I'm an academic physicist—not an electronics engineer. What I have done probably exceeds what was necessary and may even be sonically irrelevant at times—but since I don't have a baseline M22, I'll never know.

    Replacing Age-Sensitive Components, Keeping the Original Magic
    The journey began with simple testing of all the transistors. They, thankfully, were still fine. I presume they are the soul of the M22 and from the outset I wanted to keep the original bipolars in place even though there are notionally better-performing types today. I did not go so far as to check their curves, though one could add that to the to-do list.

    Selecting Capacitors
    Instead of just swapping out the electrolytics, as is standard practice, I began reading about capacitor aging and soon learned that it is not just the electrolytics that can go bad. Tantalums and ceramics can change with age too. There are no tantalums, but plenty of ceramics. After testing all of them, I even found one that had shorted out (this was probably the culprit of the blown fuse and DC offset). The shorted ceramic cap appeared largely normal but had a bit of dried white glue on it.
    [​IMG]
    I had not initially considered this to be indicative of a problem because Japanese electronic makers are known to have used used white glue to stabilize components against vibration—especially buzzing capacitors—but it turns out that the "dried glue" was its internal guts that had melted, squirted out, and hardened.

    The more I studied, the more I realized how far capacitor technology has come from the days of M22 (here and here and here are some starting resources.) I wondered: could I improve the sonic character of the amp beyond its original performance and still hew close to its original sonic signature? @thegunner100 once discouraged me from considering the M22 as a headamp because, he felt, that all stereo amps lacked the nuanced character of a good dedicated headamp. I have also struggled with using 2ch amps because I find many power amps don't have the noise floor needed to drive sensitive headphones without annoying hiss. I kind of took that as a challenge to see just how much I could extract from the M22. Could I quiet it down and bequeath to it the detail and nuance of a good headamp? In case you're wondering, the answer to be found at the end of this long epic will be yes—although it has specific qualities and limitations that one should be aware of before running out and buying one.

    What can be done?
    For the noise floor, the filtering caps can be improved by using special electrolytics with ten-times lower ESR than traditional electrolytics. In parallel to this choice, there are longevity considerations. The original electrolytics where Nippon-Chemi-Con "SL." Although now discontinued, those spec'ed out as standard 85°C caps. I aimed to get 105°C-rated caps with at least several thousand hours of rating time. It's possible to find caps with ratings beyond that, but the ESR properties begin to suffer—and those are paramount.

    Some of the small and mid-sized electrolytics can also be replaced with new aluminum-polymer capacitors that have extremely low ESR, further improving their filtering performance. Similarly, some smaller electrolytics can be replaced with film caps. For example, I replaced the non-polarized electrolytic that sits across the AC mains with a giant film cap. This should significantly improve the line filtering of EMI and other high-frequency crap spewed out by switching power supplies and SCR dimmers.

    But just a warning: always choosing the lowest ESR capacitors would be a bad idea. Yes, in general lower ESR is better—except in a few instances where the designer has relied on the internal resistance of the capacitor to stabilize an oscillating circuit. This meant having to tediously trace through the circuit diagram to ensure that I wasn't going to create trouble. Thankfully, it seems there was no such risk with this particular amp: electrolytics simply aren't used in those applications in the M22. Low-ESR all the way, and that should really help with noise. (If you're really obsessed, you can also replace resistors with metal-film or non-inductive wire-wound types, which have the lowest noise profiles.)

    Even bigger improvements have been made to ceramic caps. Older ceramics (Y5V, Z5U, X7R) exhibited terrible excursions in rated capacitance with temperature, and those excursions get worse with age.

    [​IMG]
    Furthermore, all ceramics, especially those with Class-2 dielectrics, exhibit a voltage-dependent capacitance. This is a real problem if they are used as Miller feedback capacitors across a transistor, because as the voltage in the signal increases, the capacitance fluctuates, introducing distortion directly into the audio path. Thankfully, the C0G/NP0-type dielectric ceramics introduced in the 1990s don't have either the temperature or voltage excursions of their older siblings. I therefore decided to replace all the ceramics with C0G/NP0-type.

    There were also a range of new film capacitors to explore—low-inductance, high impulse, high-frequency-rated components that should give better linearity in frequency response than the film caps of yesteryear. Some of these are directly in the audio path, and when I was done, the measured performance was great—but the sound seemed... off. Eventually, this was reduced to manually voicing the amp by ear—I'll write more about this in a future post.

    All told, I spent probably 30-40 hours digging through the DigiKey catalog and reading data sheets on capacitors. I selected my favorites, then ordered 50% more than I needed of each type. The idea was to select from the excess number the subset of capacitors with the best characteristics as measured with my LCR meter—at 120Hz for the mains filtering, and at 100Hz, 1kHz, and 10kHz for signal applications. Since the amplifier is a push-pull topology, I also wanted to be sure both sides were as symmetric as possible. I grouped the caps into matched pairs based on the capacitance, Q-factor, and ESR as a function of frequency, and placed them on symmetric sides of the circuit. Here is where I start exceeding my knowledge: it's not clear to me exactly how far I needed to go with all this, but I did it anyway.

    As for the 33,000µF filter caps—the big ones up on top—I found exact-size replacements. They are much heavier than the originals, suggesting to me that the originals may have started to dry out. One problem, however, is the original caps had mounting lugs separate from the terminals, the new ones do not. I devised a set of nested insulated washers to clamp the capacitors the the die-cast aluminum frame (which is grounded). I then filled the gap between the screw terminal and the frame with neutral-cure silicone. Without this, it's still possible for the caps to come loose, rotate an 1/8 inch, and short out against the frame. They were also annoyingly printed with the specs, brand, and a big white stripe to indicate the negative terminal. Since these caps are part of the M22's legendary visual aesthetic, this ugliness would not do. I carefully removed the white printing with acetone and polished up the caps—which now look great.
    [​IMG] [​IMG]


    Breaking in the electrolytic caps?
    Measuring every single capacitor may have been a bit insane, but it turned out to be good because, (1) I discovered one bad one among the new batch, and (2) I learned something new about electrolytics—they actually undergo a breaking-in process.

    Initially, I purchased Vishay 105°C electrolytics (1000µF, and 470µF) for the on-board filtering caps, but they did not measure as well on my LRC meter as the original Nippon Chemi-cons that came with the M22. So I ordered a different brand hoping for better performance, but those didn't measure well either. So I ordered a third brand. Same thing. This meant research: I started reading about the aging of electrolytics to see if the age explained their notionally superior performance. I found an academic thesis written on capacitor ageing effects and their role in nuclear-reactor reliability (where these things really matter). I learned that the shelf life of newly manufactured electrolytics is really only about 2 years. The electrolytic degrades, but it will heal itself and "reform" with steady use. I have no doubt that my newly purchased caps may have been sitting in DigiKey's warehouse for years.

    So, I went through the process of manually reforming all the big electrolytics, both the new and original ones. It took two full days to go through the process: A voltage is applied across the capacitor and slowly raised in steps until the rated voltage is reached. At each voltage step, the capacitor is allowed to sit until its leakage current levels off to its design value.
    [​IMG] [​IMG]

    Thankfully, my life was much eased by borrowing a programmable laboratory power supply that would automatically adjust the voltage throughout the day and keep the current at the target level. I used some backup resistors to dampen the circuit and ensure there were no current spikes on startup and discharge that might damage the caps. Reforming the big 33,000µF was especially interesting—it took about six hours for the leakage current on those to drop to a working level. That's when I realized: this is a real, bona-fide "burn-in" effect. One often hears about burning in audio gear, but here finally was a physical explanation for why it might actually happen.

    Not only did this process pre-burn in my caps, it gave me an opportunity to measure something that one cannot measure with an LCR meter: the true leakage current at operating voltage. This measurement is the most informative with respect to capacitor health and, indeed, it was here that I finally discovered the superiority of the newly purchased caps over the old: the original caps all had much higher leakage currents at operating voltage, and a few were really far out of spec—something not detected with the LCR-meter measurements. Before going through the reforming process, I had been tempted to leave the original electrolytics in place since they initially had better ESR and Q-factor measurements. However, after reforming all the caps, I retested them with the LCR meter now it was the new caps that had better ESR and Q-factor (and leakage current) measurements. The choice was now clear—all new caps throughout. Here, by the way, is an excellent guide to checking out old caps and on recapping in general.

    Below is how the re-capped boards look. The one on the right has the original orange coupling cap in the top-right corner for testing purposes, but I quickly determined that my new choice sounded better and replaced it with one of the gray boxes seen on the top left. Nevertheless, these two caps remain a place for exploration.
    [​IMG]


    Other Maintenance Tasks
    In addition too all this nutty capacitor stuff, I also replaced the fuses (as one had blown) with Littel fuses, which I like for their very low resistance and high quality build.

    I spent literally hours polishing contact posts in all the connectors and terminals. Some of the oxidation was pretty bad. In this task, I found Deoxit to be a total waste of money. I resorted to fine-grade scotch pads and a lot of elbow grease.
    [​IMG] [​IMG]

    The only contacts I was not able to clean were the relay contacts. I would have replaced the relays, but I could not find drop-in replacements. In testing the amp, I took a stethoscope and went around listening for oscillating capacitors. All the caps seemed fine, but the relays were clearly singing. This suggests that at high volumes they may be introducing distortion from sympathetic mechanical vibration. Replacing the relays is on the to-do list.

    Initial Tuning
    If you read around on DIYAudio, you'll find people talking about setting the DC offset and bias voltage. The manual specifies a DC offset across the output terminals of 80mV, and to set the bias voltage by setting the voltage across test terminals to 850mV. Many people on DIYAudio think this is a typo, because the logical place for the DC offset to be set at is zero volts. Initially, I followed the DIYA community's suggestions, but in listening tests, it seemed like something was wrong.

    The music sounded slightly crispy to me—suggestive of odd-order distortion. Since this is a push-pull design, odd-order distortion tends to emerge preferentially as even-order cancels. I traced the problem to the DC-offset adjust. What this offset actually does is to change the bias on the input-stage pair. If you study the circuit diagram, there's an annotation that the input-stage should be at -20mV DC with no signal—which is why this amp needs a coupling capacitor. I checked the Japanese-language data sheet for this transistor and it seems to confirm this. Instead of setting the DC offset via the output terminal, I measured the bias voltage on the transistors directly and set it that way. The crispy quality went away.
    [​IMG]
    I also adjusted the bias on the other transistors to ear, rather than just adjusting to the recommended level. I hope soon to re-do this using a distortion analyzer.

    Shut up already and tell me how it sounds!
    Is it good? Yes. The amp still needs to be voiced and fussed with, but it sounds pretty good. In its current state it is a serious contender for powering planars, which I think clearly benefit from the beefier treatment of a power amp.

    [​IMG]
    When listening with HEK, it fixed many of the overly-soft shortcomings I previously associated with these headphones. It easily bested the $3K Benchmark HPA4 (which may or may not be saying anything). It sounds better than the $4K Leptoni DT-Alpha by what I would say is a substantial margin. It sounds MUCH better than my former EC Zana Deux Super (although that is not a surprise given that the ZDS isn't really great at driving planars).

    [​IMG]
    I am also comparing the M22 to the $6,800 Pass Labs XA30.8—here we find a contender. Some would say no contest: the XA30.8 has less veil and more slam and sparkle. In all technical respects, the Pass Labs is more impressive. However, I think the M22 actually has a more honest timbre. I heard more subtle tremolo on piano and string sustain with M22 than on XA30.8. Breaths and aspirations of the players sounded more natural/live on M22. Listening to Pink Floyd's The Wall, M22 had better separation and less hash during loud passages, and more details in the quiet parts; XA30.8 has more... well... more of everything else. If I were only listening to acoustic/classical, I'd say the M22 was currently better; but for most people's taste, the XA30.8 is a better technical performer. It's like HEK vs. LCD2C. Depends on who you are.

    A few important notes: This amp clips promptly at 1V RMS on a sine wave. It's easy to over-drive this amp with a variety of DAC outputs. This alone will be responsible for shit sound if you don't pay attention!

    In its original configuration, it doesn't have enough voltage to drive HE-6 without going into clipping. I expect I can squeeze another 3–6 dB out of it by changing the output impedance, but that probably isn't enough. Future experimental results will be posted below.

    Despite what is already a very decent performance, there are a few things I'd like to tweak on the M22. I want a bit more presence and a bit more slam. So the project isn't done. I've ordered a whole pile of coupling capacitors for testing, and I'll report back on that—as well as the experiment with changing the output impedance to see if I can get more voltage into the high-impedence headphones.

    Stay tuned for parts 2 (voicing) and 3 (measurements) in the next couple of weeks ...
     
    Last edited: Aug 7, 2018
    Xecuter, mphuongz, Golmang and 47 others like this.
  2. skem

    skem Friend

    Friend
    Joined:
    Nov 1, 2017
    Likes Received:
    431
    Dislikes Received:
    2
    Trophy Points:
    63
    Location:
    Cambridge, MA
    Voicing the M22

    After the re-cap job, the M22 sounded fine but it didn't sound right. It had more detail and better extension than any other amplifier had given my HEK—but it seemed to be slightly cold, had a bit of veil on some of the treble, and yet was also a bit etched at the same time. I didn't exactly love the sonics.

    I figured I would experiment with different coupling capacitors. I had tested the original coupling cap (an "orange drop" type) against the gray-box replacement visible in the photos above and heard a definite improvement in detail with the replacement. But perhaps my replacement choice was not good in other respects.

    I did not want to go buy the most expensive audio caps out there since I figured they had unreasonable mark-up and might even be snake oil. Many people on DIYA advocate for using more mainstream capacitors. I decided to experiment with a range to see what I might learn.

    Here's a summary of what I had gathered from websites before starting out:


    With this as a starting point, I ordered a whole lot of film capacitors to examine these claims and see what I could find.

    [​IMG]

    Here's the bottom line. (1) Coupling caps really do sound different. I don't want to sound like a fool and suggest the differences are huge. They don't change the dynamics, for example, but the differences are there and can be big enough to change an amp's sound signature from "ick" to "mmmm!" (2) the specialized audio cap I tried sounded the best. Damn-it! So much for my attempt at discovering some money-saving secret.


    Lessons

    In general, I found that the (physically) larger the capacitor was, the better it sounded. To a good approximation, for any given capacitance (set by the design), the voltage rating determines the physical size. Although the voltage of the signal is only ~1 volt, I found that caps rated at 300Vac and above sounded the best. One conjecture is that this simply assures the field strength per unit area of conductive plate is low, so the capacitor plates doesn't oscillate and produce signal-related distortions. Interestingly, most of the audio-grade caps are also high voltage, which suggests the size-quality correlation is a real thing. (EDIT: or maybe they’re high voltage so they can be used as interstage caps in tube amps. In any case, bigger still sounds better).

    Another insight: I could not identify a trend between PP and PET in my rankings. While my top non-audio grade caps were indeed PP, there was no statistically significant correlation between dielectric type and sound. I could not substantiate the claim that PP sounds the best, and indeed, there were plenty of PP caps that did not sound as good as some PET caps.

    Some motor-run caps sounded good, others did not. The one that sounded good was an oil-immersed one. I did not test more than three, however. These caps often have big metal cans and were more sensitive to RF interference. With one of them, I could literally tune in to different AM radio stations simply by lifting the ground and moving my finger across the body of the capacitor.

    The best non-audio-grade cap I tried was the Cornell Dubilier 930C and 935 types, which incidentally look very similar in construction to the audio caps (giant axial cylinders). None of the plastic-box-package caps sounded good compared to these. These are also a touch more expensive than the mainstream film caps.

    Here's a picture of the amp rigged up for capacitor testing.

    [​IMG]

    The cap I wound up picking for the final build, however, was an Audyn Reference Series audio cap. This was actually the only audio-grade cap I tried, because I didn't want to spend a lot testing a whole bunch (and frankly I was pretty convinced at the outset that one of the cheap caps was going to win, but I was wrong). I am very happy with the Audyn Reference and have no desire to change anything. If you want to go beyond this simple recommendation, however, Marv has written impressions about different audio caps here.

    The Audyn Reference was significantly different from all the non-audio caps in that it seemed to warm up the sound and smoothed out the etched quality without sounding veiled. It wasn't quite as crisp as the Cornell Dubilier, but it was still VERY detailed and lovely... and dare I say... "fluid." I really love it and think it's perfectly suited to my tastes.

    This cap does not fit on the board, so I retrofitted it into the amp by placing them underneath the RCA jacks and using very thin coaxial cable to run the leads to/from the board. Only the inner conductor of the coax is used for signal; the outer shield of the coax is connected to ground so as to prevent the long wire runs from acting like antennae.


    Why do caps sound different?

    I was very curious about why the caps were affecting the sound. The short answer is I'm still not sure. Certainly the amplifiers frequency response is not measurably affected by the capacitor brand. And yet there really is a perceived change in the frequency response from different caps.

    I could measure some differences in the distortion spectrum, but they were very small. A few dB here and there. It was not a visibly a huge effect, but they certainly do sound different. One theory is that the ratio of second and third harmonic distortion might shift at different parts of the frequency spectrum, and that is responsible for the overall tonal effect. Unfortunately, I did not feel crazy enough to begin a systematic mapping of all the capacitor sounds. It will have to remain a mystery for now.

    To round off this entry, I'll share some measurements of the final setup made with the shielding still removed. You'll see the 60Hz influence because of the lack of shielding. Noise floor is down at 1.6 uV and distortion is at 0.0005%. I'm quite happy with these numbers.

    [​IMG]

    A more complete set of measurements to come later once the amp is re-assembled.
     
    Last edited: Sep 10, 2018
  3. skem

    skem Friend

    Friend
    Joined:
    Nov 1, 2017
    Likes Received:
    431
    Dislikes Received:
    2
    Trophy Points:
    63
    Location:
    Cambridge, MA
    Reserved for Part 3
     
  4. thegunner100

    thegunner100 Hentai Master Chief

    Staff Member Friend
    Joined:
    Sep 25, 2015
    Likes Received:
    2,190
    Dislikes Received:
    2
    Trophy Points:
    113
    Location:
    NYC
    Admittedly I skipped most of the post since the engineering talk would be incomprehensible to me. But wow! Glad you were able to tune the M-22 to your liking for headphones. I still stick to my opinion that most people would be better off with a well designed tube headphone amp, especially for dynamic headphones.

    @skem How does the M-22 compare to the ZDS with your dynamic headphones?
     
    Dr. Higgs likes this.
  5. famish99

    famish99 Friend

    Friend
    Joined:
    Dec 26, 2017
    Likes Received:
    187
    Dislikes Received:
    1
    Trophy Points:
    43
    Location:
    Austin, TX
    I'm interested how much improvement the boutique capacitors (Auryn, Jensen, Jupiter, Duelund, etc.) would have to replace the coupling cap (since in the signal path and all).
     
  6. skem

    skem Friend

    Friend
    Joined:
    Nov 1, 2017
    Likes Received:
    431
    Dislikes Received:
    2
    Trophy Points:
    63
    Location:
    Cambridge, MA
    @thegunner100: No dispute there. I don’t have dynamic headphones, so can’t really say (Never found one I liked). I don’t doubt you could be right, however.

    @famish99 I’ll tell you how Audyn sound. That’s the only super premium cap I bought. If there’s a big gap between those and the others I’ve bought, I might explore other audio caps. I’ve seen extensive reviews of these audio caps that purport significant differences, but the reviews contain information that make me think the reviewer was suffering from psychological bias. (E.g., oil filled caps sound liquid, Teflon caps sound slippery, etc).
     
    Last edited: Aug 6, 2018
    SoupRKnowva and bengo like this.
  7. Thenewerguy009

    Thenewerguy009 Friend

    Friend
    Joined:
    Oct 11, 2015
    Likes Received:
    224
    Dislikes Received:
    13
    Trophy Points:
    43
    Isn't that a HD800 & HD650 in one of your pictures?
     
  8. skem

    skem Friend

    Friend
    Joined:
    Nov 1, 2017
    Likes Received:
    431
    Dislikes Received:
    2
    Trophy Points:
    63
    Location:
    Cambridge, MA
    Those belong to a friend. We didn’t connect them.
     
  9. skem

    skem Friend

    Friend
    Joined:
    Nov 1, 2017
    Likes Received:
    431
    Dislikes Received:
    2
    Trophy Points:
    63
    Location:
    Cambridge, MA
    Last edited: Sep 10, 2018
  10. Tachikoma

    Tachikoma Rando

    Joined:
    Oct 4, 2015
    Likes Received:
    20
    Dislikes Received:
    0
    Trophy Points:
    8
    About the DC offset adjustment, how much DC did you get across the terminals in the end? Also, isn't it important to get that down to zero if you're going to use it as a headphone amp?

    Also looking forward to new updates :)
     
  11. skem

    skem Friend

    Friend
    Joined:
    Nov 1, 2017
    Likes Received:
    431
    Dislikes Received:
    2
    Trophy Points:
    63
    Location:
    Cambridge, MA
    So yes. I wound up re-doing this to get and close to zero as possible. I was initially tuning it with HEK planars and I think the original dc bias might have helped tighten the sound, but for reasons that are unorthodox.
     
    Last edited: Nov 28, 2018
  12. skem

    skem Friend

    Friend
    Joined:
    Nov 1, 2017
    Likes Received:
    431
    Dislikes Received:
    2
    Trophy Points:
    63
    Location:
    Cambridge, MA
    I’m still undecided, but I might wind up selling this amp. I’m building a new Pass designed amp from scratch for my HE6, and that would make this amp superfluous.
     
    porkfriedpork likes this.

Share This Page