DIY build of the Pioneer C-21 preamplifier

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.

rjm

Member
Joined 2004
Paid Member
As you probably know already, I download and study audio schematics for fun as well as for reference. The C-21 circuit has long struck me as being unusually elegant, and I've already borrowed the power supply design for my headphone amplifier. Although I can't follow it all, after a while you can sense when the designer is someone who really knows what they are doing.

Lately though I've been interested in discrete transistor pre-amplifier circuits, and of the ones I've seen the C-21 seems like the best place to start.

The circuit, edited to remove clutter, is shown below. Would welcome any opinions and comments.
 

Attachments

  • Pioneer C-21.png
    Pioneer C-21.png
    143.6 KB · Views: 2,044
Last edited:
Hi,

Lately though I've been interested in discrete transistor pre-amplifier circuits, and of the ones I've seen the C-21 seems like the best place to start.

I would not say so, but arguably there is worse.

I would suggest to remove or reduce the compensation capacitors (470pF) after the second differential amplifier, add a 10pF Miller cap to the uncompensated transistor in the second differential and to give each differential amplifier emitter degeneration resistors. In fact, I'd probably pull all but the main miller compensation and dial up the degeneration until the circuit is stable with a a 100pF load and a 10KHz squarewave.

Another area of concern, if you are using a conventional volume pot, these really do not like to have their wiper loaded plus, neither pot or stepped attenuator like to have non-linear base currents from BJT inputs injected into their wiper.

Maybe the best solution would be to flip the polarities and use a cascoded JN-Channel J-Fet input. Of course then we would have something that looks like a Goldmund Preamp.

Ciao T
 
... 2 CENTS OR 2 POINTS ...

one for the power supply ... i am not really thrilled ...obviously there are much better ways to stabilize clean power an yes small details there are very important .

two as about the main circuit i have absolutely no issues except that as said is not something that special ...we have seen things like that before .... be sure that gain is too high for now days equipment and be sure that Japanese designers of that time had an absolute and perfect knowledge of the parts used .... that will apply to performance specs durability and better possible use .... that will mean that this preamp will not work that perfect if you use a pair of BC 546 -556 or similar and even if the specs are very very close ...

kind regards sakis
 

rjm

Member
Joined 2004
Paid Member
"obviously there are much better ways to stabilize clean power"

Aha, that's twice in one thread now. Please don't make statements like that unless you are willing to back them up.

The Zener/pass transistor is indeed a very basic, simple voltage regulator, but it sounds very good. In my opinion better than when feedback + error amplifier is used, at least for line level where the last word in noise filtering is not required.

Returning to the C-21 line amplifier circuit, I realize that it is "nothing special", at least in concept. Execution looks exceptionally solid though. I should have stated at the outset that I was looking for something relatively simple to play with, rather than the latest DIY fashion. Set myself the goal of 7-8 transistors: (long tail pair + voltage amp + push-pull output)/feedback loop. I still suck at reading transistor amplifier schematics, the idea is to make a very simple, yet working, "op amp" from discrete components to see how far I can get. Plan to do a fair bit of benchtop testing.

No, I will not build the C-21 circuit unmodified. Obviously the loudness control is going to go, and the gain reduced. Maybe reduce the voltage rails, hopefully get the offsets down so I can get rid of some of the coupling capacitors...
 
Hi,

No, I will not build the C-21 circuit unmodified. Obviously the loudness control is going to go, and the gain reduced.

If you change the gain you will need to change the compensation. Plus as the circuit lacks any degeneration it is going to have a miniscule linear range. It look at an estimate that DC Gain open loop is around 115dB as drawn and strongly load dependent.

Putting 470 Ohm degeneration into the input stage and some into the VAS will lower this open loop DC gain but make the circuit much more linear.

Then again, I would probably consider the Goldmund Pre design or the Spectral DMC-10 Design (which can be found with some effort) better choices if high quality is desired.

Maybe reduce the voltage rails, hopefully get the offsets down so I can get rid of some of the coupling capacitors...

Reducing the rail voltages would remove anything that makes this circuit even a little special and you may as well use a NE5534.

Ciao T
 

PRR

Member
Joined 2003
Paid Member
> Please don't make statements like that unless you are willing to back them up.

Then you should back-up this strong statemet:

...after a while you can sense when the designer is someone who really knows what they are doing.

----------------------------

> Would welcome any opinions and comments


Every designer has different goals and intents and tricks. What I see is a damm-the-cost brute-force design.

Two 37V rails? What power amp needs 30V, even 15V, input swings?

It is of course from the Very high NFB era. It probably has SO much NFB as to be benignly neutral. Fashion changes.

Input stage current is quite high. Perhaps on the high side for lowest noise in a 100K volume pot. Yes, the added output pot can put the part-volume hiss super-low, but I'm not sure it is supposed to be used that way.

Mismatch between 100K bias at input and 1K resistance at backside of the pair will cause some DC offset, which does incidentally bias the big electrolytic appropriately.

C13 suggests GBW near 80MHz, however C15 C17 may be rolling-off by 0.2MHz. C11 against 13K says something happens near 3MHz. I'm not seeing elegance here. Sometimes compensation is not elegant, but it is nice when it is.

The two 470pFd loading caps are red-flags to my eyes. The thing was unstable, and was hammered into submission with BFCs. On the napkin, 7mA forcing ~~1000pFd is 7V/uS slew-rate. Which is fine for any line-level audio, but cheap chips do this well with crummy lateral PNPs; selected discrete devices "can do better".

There is a hidden 2nd-harmonic nonlinearity. Because it is worked as non-inverting there is common-mode at the input. Input pair is long-tailed with a resistor. Current in first and second stages varies with input. Second-stage current will change more than first-stage current. At the extreme (+/-2.5V inputs) there could be 20% change of current and maybe 10% change of gain peak-to-peak, hence 2nd-harmonic. I don't think this is a sonic sin, and it sure will be reduced by NFB.... but the whole pair-pair-mirror topology is often thought to null 2nd-harmonic, not be dominated by it.

I wonder how much "better" (that word again) this is compared to NE5532. (Unless you really need 20+V swings.)

The power supply seems unremarkable to me. Follower on Zener with RC filter is an old-old workhorse. 0.01u across the rectifiers may-or-may-not improve the switching noise; stout layout may help switching hash more than adding parts.
 

rjm

Member
Joined 2004
Paid Member
Thorsten, PRR;

Thanks for your comments. And feed free to name drop: any alternatives (transistor, with gain, relatively simple) will be given full consideration.

The 37 V rails are unusual and "costly" so there must have been a reason. Improved linearity? The output is specifically shown on the schematic as "(1V)" so it doesnt seem to have been for the high voltage swings. Also mysterious to me why then chop that to +15V for the top of the input long tailed pair.

"The two 470pFd loading caps are red-flags to my eyes."

It seems to me the circuit was designed for low distortion rather than high bandwidth. The bandwidth/slew rate was curtailed intentionally down to "sufficient" levels and stability ensured by the multiple capacitors distributed over the whole circuit...
 
Hi,

The 37 V rails are unusual and "costly" so there must have been a reason. Improved linearity?

Cost increase is not really large, most of the cost is there with +/-15V and with +/- 37V. Transistors have many undesirable properties, which incidentally are poorly if at all modelled in common simulation software, most of those reduce with more voltage across the transistor. It also allows more voltage headroom for things like current mirrors, which get better the greater the degeneration.

Also mysterious to me why then chop that to +15V for the top of the input long tailed pair.

Economy and sound. While resistive tails do create common mode distortions, many like the sound quality it gives. The voltage is dropped mainly to clean it up and make sure noise is low.

Personally I'd probably stick J500 CCS into the tail.

"The two 470pFd loading caps are red-flags to my eyes."

It seems to me the circuit was designed for low distortion rather than high bandwidth. The bandwidth/slew rate was curtailed intentionally down to "sufficient" levels and stability ensured by the multiple capacitors distributed over the whole circuit...

Hard to say.

Basically, they did not degenerate any of the gain stages. This creates low noise (but given the 100K Volume control this is the greatest noise source anyway), the result is very high gain. As remarked, using bipolars connected to a potentiometer will create a substantial but hidden and variable distortion (and noise) source. Using J-Fets would be better on both counts.

The output stage bias is reasonable for a linestage (max 20mA Peak into 1K = 14V) but inappropriate for headphones and if the circuit drops into class AB sound quality will suffer.

As I have remarked before, the topologically similar Spectral DMC-10 (but with Fet's on input and output and much greater output bias) would be a better choice IMNSHO, or if extra complexity can be allowed the Goldmund circuit (needs a hefty output stage though due to +/- 60V Rails).

If you can live with Op-Amp's I'd probably try an OPA627 or similar followed by a SE Output stage with a nice Mosfet and a big heatsink that runs at 0.3A or so bias and a suitable CCS (even a 317 is not as bad as it's reputation).

Ciao T
 
Hi,

DMC-10 simplified schematic, for reference.

As you can see, it is essentially the same kind of circuit as the Pioneer, but with J-Fet inputs (NPD5566 like so many US Pre's of the time, matched 2SK246 can be substituted with minor changes, according to Kaneta and some of his followers 2SK30 is best for input pair) and lateral Fet outputs (VN.. and VP.. parts long obsolete - nowadays I'd use 2SK214/2SJ77 or related).

The J-Fet inputs improve input level handling and linearity, as well as RFI resistance. The Fet outputs largely remove the loading on the VAS (some nonlinear capacitance remains, but very little).

With the Hitatchi TO220 laterals I recommend above you can easily run a few 100mA quiescent current to get pure class for even low impedance headphones.

Ciao T
 

rjm

Member
Joined 2004
Paid Member
A couple of questions.

The DMC-10 schematic is missing a few details. What's a good (best?) way to generate the bias for the output pair?

If I'm getting this right, the Pioneer and the Spectral share the basic topology: differential input pair, second differential pair as the voltage amp, bias, and output pair. The difference is one of substituting the bipolars for fets.

Is this basically true of all discrete solid state preamplifiers or is there any completely different way to approach the matter?

[that's a little bit disingenuous of me: I just got the 47 labs 0247 headphone amp kit and was quite thunderstruck by the circuit used. Think diamond buffer with voltage gain and feedback, direct coupled to a Szilkai pair running open loop as a buffer. So I'm naturally curious now to know what else is lurking out there.]
 
Hi,

The DMC-10 schematic is missing a few details. What's a good (best?) way to generate the bias for the output pair?

If using laterals a resistor, if using vertical fets a VBE multiplier, if using J-Fets at Idss, nothing will do fine.

If I'm getting this right, the Pioneer and the Spectral share the basic topology: differential input pair, second differential pair as the voltage amp, bias, and output pair. The difference is one of substituting the bipolars for fets.

Yes, that is so. You can add many more Preamp's including Sony TA-88ES, Goldmund and many others.

Is this basically true of all discrete solid state preamplifiers or is there any completely different way to approach the matter?

There are probably more different approaches than you can shake a stick at.

[that's a little bit disingenuous of me: I just got the 47 labs 0247 headphone amp kit and was quite thunderstruck by the circuit used. Think diamond buffer with voltage gain and feedback, direct coupled to a Szilkai pair running open loop as a buffer. So I'm naturally curious now to know what else is lurking out there.]

The "diamond buffer with voltage gain and feedback" is known as "current feedback" It is quite common, as is a Sziklai Output.

Ciao T
 
Hi,

"The "diamond buffer with voltage gain and feedback" is known as "current feedback" It is quite common, as is a Sziklai Output."

Are you sure about this? A diamond buffer has nothing to do with current feedback.

Yes, I am quite sure about that.

The description in parenthesis was taken from the OP's post and his description of the circuit. The actual circuit is the one I commented on, a current feedback system with diamond transistor front-end (e.g. AD844 in IC's).

Of course, to understand this would actually require reading posts and knowing a few things about basic electronics, prior to commenting.

This is something that long gave up from certain people, these days I only expect from those are educated, well mannered, reasonable and knowledgeable...

Ciao T
 
www.hifisonix.com
Joined 2003
Paid Member
My question was asked in a perfectly civilized manner and without any other intentions.

However, since you chose to interpret my question as a hostile challenge, let me restate it it in terms you may better understand:-

You don't need a diamond output stage to create a CFA.

Do you understand how a CFA works, and why?
 

rjm

Member
Joined 2004
Paid Member
Thorsten is correct, it's a current feedback circuit element. (see here) There are symmetrical current mirrors on the collectors of the diamond buffer output transistors, with the output load of the mirror tapped to return a proportion of the signal back to the diamond buffer output.

When I was parsing the circuit earlier I saw the feedback resistors but my brain did the I-V conversion automatically :) and I saw it as voltage feedback rather than current feedback. Sorry if I was unclear.
 

Attachments

  • xxxx.png
    xxxx.png
    16.5 KB · Views: 571
Last edited:
rjm,

Not sure why you want to build a C-21 when you can find a used Pioneer at reasonable price. I got one and it's a very decent preamp. I can put it up against Threshold all FET (FET1 MKII) preamp which I also have from that era. Sometimes, simpleton circuit is all it needs to sound GREAT! I checked out mine to see if anything needs to be done since it's over 30 yrs old. I found the Parts Quality is really good and Step Attenuator adds to the simple BJT circuit. Even the old lytics are a OK on my LCR meter.
 
Last edited:

rjm

Member
Joined 2004
Paid Member
James, your post reads like a classified add, but I agree it would be simpler and cheaper (in the long run) to buy one rather than build a copy.

The electrolytics in a 40 year old component are of questionable status, despite your checks. Let's face it they must be on their way out by now. That's not what's holding me back though; it's just a matter of purpose - I'm more interested in the circuit than the preamplifier per se.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.