First attempts at some DOA designing

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

etheory

Well-known member
Joined
Mar 21, 2011
Messages
604
Location
Sydney, Australia
Hi there!

After reading the inspired stories of many forum members after that elusive DOA-design bliss, I did the clearly stupid thing and started looking into it myself.
I wouldn't call anything here original, even slightly, but I guess copying and pasting concepts is the way you get your start, and then building upon it over time with truly original ideas.

So don't expect anything good or even average, but I have a whole bunch of 550C's, 560C's and 2SK170BL's destined for around 100 Tamas BigFetBlokes, and though, "why not have a crack at it myself!".  Not to mention that as good as the BigFetBloke is, I don't necessarily need 100 of them ;-).  So with inspiration from pretty much all the main contenders - 2510, 2520, Forssell JFET opamp, and Tamas' BigFetBloke, I present, the horrifically designed eCheekyDOA's.

I expect, and hope that I will get slaughtered for these designed, all in order to learning something.  So fire away!  And tell me what is good/bad about these!  As long as they are not so bad that PRR never posts in one of my topics again, I'll feel relatively happy.

Design requirements:
1.) 24V rails - cause that's what I have
2.) Differential inputs for use in differential input stages and EQ's
3.) I need to get rid of a large number of BD140/BD139/BC550C/BC560C and 2SK170's, so it has to use those
4.) I want something warm sounding and thick - i.e. even with no transformers, I want lots of harmonic "thickness" - hopefully not distortion per-se, but some girth in the lower order harmonics - if I want clean I'll use a VST - I want something that thickens and makes it's presence known

Design 1 is far too simple, and the second JFET current source I think would explode if it were built in real-life, however, it performs OKishly on the simulator in a mic preamp, and is an interesting initial point of discussion:
 

Attachments

  • eCheekyDOA_v01_01.jpg
    eCheekyDOA_v01_01.jpg
    109.4 KB · Views: 103
Second one seems to perform a little more "realistically" - i.e. it would appear that the currents through all components are within acceptable tolerances, and whilst the lower order harmonics appear rather horrid in amplitude, they fall off very fast out past the 5th and 6th order and end up looking not too bad overall.  I almost feel this one is worth building with a hand-matched 2SK170 input pair:
 

Attachments

  • eCheekyDOA_v02_01.jpg
    eCheekyDOA_v02_01.jpg
    128 KB · Views: 61
I don't quite understand the desire to reinvent the wheel so many times but this is educational so enjoy.

My concern from a quick glance is nominal operating current...

While trimpots are not generally used inside opamps, I'd be inclined to add another one to tweak your current source feeding the input LTP so you can manage the overall operating current at a level that doesn't release smoke while delivering decent performance.

JR

 
Well if it only waas for performance then yes we could all shut down the soldering iron and go buy a hung-low brand dvd player and listen to your next best 'american idol, star academy, DSDS' pic your country's next generation star DVD....

so experiment with it, and get wiser on the way. Personally I would recommend samuels paper on opamps as a good starting lecture - I first wanted to build one of his (free use within reason) designs, but opted for a different route (first?)....

Only the tinkerer will make progress and gain knowledge......

enjoy the trip!

here's samuels page:
http://www.sg-acoustics.ch/analogue_audio/discrete_opamps/index.html

- Michael
 
I miss the good old days when IC makers published their opamp's internal schematics.

+1 to Sam's work, and IIRC Bruno Putzey has also discussed the subject in recent years.

Also the older Deane Jensen paper is worthwhile wrt practical DOA.

But I don't expect this to be a pursuit of higher linearity... more like DOAs as a special effect...

So why waste time reading the advice  for how to make it clean other than to know what to avoid?

Sorry... i'm just being me.

JR
 
I am not a DOA person, or even a semiconductor person but how about doing something a bit more radical? - how about a 100%  class A DOA?

Cheers

Ian
 
What closed-loop gain are you targeting?  This will be an important consideration in stability, which perhaps you haven't considered yet (?) since there doesn't seem to be any compensation capacitor (except the one that the manufacturer supplies free with Q1).  Higher closed-loop gain = more stable (because the frequency where the feedback loop gain = 0dB is lower); if you want unity-gain stability, more compensation may be required if the Q1 collector-base capacitance isn't sufficient on its own.

As for low-order distortion, you'll get a bit of that from uneven current in the two halves of the differential pair.  Your current will surely be at least slightly unequal even if you trim the tail current to a desired value; increase the imbalance for greater distortion from this source.

The basic VAS (single transistor Q1) will also give higher distortion than a cascoded VAS, or a buffered VAS inside miller loop, but running at 10mA I think it again would be low-order.

"I need to get rid of a large number of BD140/BD139/BC550C/BC560C and 2SK170's"
I'm happy to pick up from your place...  ;)
 
JohnRoberts said:
I don't quite understand the desire to reinvent the wheel so many times but this is educational so enjoy.

It's not so much about re-inventing the wheel, as understanding what the best features of a wheel are and how to exploit them, to say, make a better car.

I feel that these days since a lot of the groundwork in audio electronics is done (if not all of it - not sure whether much new is even possible), young whippersnappers like myself find it easy to become complacent about the design the stuff around the wheel.  I am constantly after, like most people on this forum, a certain level of perfection.  Be that the perfect control over distortion components, or the literal pursuit of the perfect recording - at the end of the day, in the eye of the beholder, they kind of are the same pursuit, after all.  This is why I build rather than buy gear.  My person goal is to be, within the next 5-10 years, composing and producing music entirely using hardware and software I've built/written myself.  In a day where auto-tune and computer synths kind of rule the roost, it's something you might call a romantic pursuit.  But I digress.

One of my favorite analogies of why I tend to do things the hard way (such as designing my own amps as a part of building a pre-amp, EQ or compressor, for instance), is that I often ponder the point of view of someone who was sent back 50 years in time.  You often see films about how that person can make a fortune, but I beg to differ.  How many of you could reverse engineer penicillin, or a television, or a telephone?  Seriously?  A transistor?  Well, maybe a mute point considering the above average intelligence level of people on this forum, but I would like to understand the TV from the inside before I get sent back in time to make my fortune off of introducing it to a world earlier than it would have otherwise gotten it.

Considering my lack of formal training in electronics a lot of this stuff my never get entirely within reach, but since this is still a hobby/passion/obsession, at the end of the day it's all a bit of fun and a sense of achievement that only DIY can provide!

Thanks for the comments already, when I get home from work I'll take another look at it and see if I can improve what's there.  I think that knowing where the sources of distortion are, as some of you already pointed out, especially sources of 2nd and 3rd order harmonics and their extended series, will give me the control I want over the final sound, whether that's with the aim of clean or distorted.
 
ruffrecords said:
I am not a DOA person, or even a semiconductor person but how about doing something a bit more radical? - how about a 100%  class A DOA?

Cheers

Ian

Excellent idea - the first revision of this was, so it might be worth reconsidering the super-duper simple original class-A single transistor output stage to see how well it can be made to perform.

Unless I've read it wrong, I've always thought class-A was about ensuring the transistors are conducting through every part of the waveform cycle, which I achieved for the output stage by using 4 diodes to bias the transistors away from each other enough to put them into their conducting range all the time.  The circuit I originally "stole" that output stage from was running class A/B since I could see the current alternating between the output transistors for the positive and negative parts of the cycle and the way the output transistors switched on and off in an exclusive fashion.  I didn't like this A/B output more on a conceptual level so I messed around until I was able to get both output transistors conducting throughout the range of output voltages and loads I wanted to support.

So aside from Q1 I believe that this IS already Class-A no?  At least everything I can see aside from Q1 is conducting "all the time".  Or does Class-A have a more subtle definition?
 
etheory said:
JohnRoberts said:
I don't quite understand the desire to reinvent the wheel so many times but this is educational so enjoy.

It's not so much about re-inventing the wheel, as understanding what the best features of a wheel are and how to exploit them, to say, make a better car.
I manage to find new things to put already round wheels on...
I feel that these days since a lot of the groundwork in audio electronics is done (if not all of it - not sure whether much new is even possible), young whippersnappers like myself find it easy to become complacent about the design the stuff around the wheel.  I am constantly after, like most people on this forum, a certain level of perfection.  Be that the perfect control over distortion components, or the literal pursuit of the perfect recording - at the end of the day, in the eye of the beholder, they kind of are the same pursuit, after all.  This is why I build rather than buy gear.  My person goal is to be, within the next 5-10 years, composing and producing music entirely using hardware and software I've built/written myself.  In a day where auto-tune and computer synths kind of rule the roost, it's something you might call a romantic pursuit.  But I digress.
whatever works for you...  while I guess it depends what is the product of your effort? making music or making gear?

I don't try to make my own screw drivers and wrenches to work on my car. 
One of my favorite analogies of why I tend to do things the hard way (such as designing my own amps as a part of building a pre-amp, EQ or compressor, for instance), is that I often ponder the point of view of someone who was sent back 50 years in time.  You often see films about how that person can make a fortune, but I beg to differ.  How many of you could reverse engineer penicillin, or a television, or a telephone?  Seriously?  A transistor?  Well, maybe a mute point considering the above average intelligence level of people on this forum, but I would like to understand the TV from the inside before I get sent back in time to make my fortune off of introducing it to a world earlier than it would have otherwise gotten it.
A rigorous education might allow you to prosper 50 years ago... while 1960 doesn't seem very far in the way back machine...to some of us.  8)
Considering my lack of formal training in electronics a lot of this stuff my never get entirely within reach, but since this is still a hobby/passion/obsession, at the end of the day it's all a bit of fun and a sense of achievement that only DIY can provide!

Thanks for the comments already, when I get home from work I'll take another look at it and see if I can improve what's there.  I think that knowing where the sources of distortion are, as some of you already pointed out, especially sources of 2nd and 3rd order harmonics and their extended series, will give me the control I want over the final sound, whether that's with the aim of clean or distorted.

There is nothing wrong with learning how stuff works. I recall studying an IEEE paper about the (heaven forbid 741 opamp) which was a significant achievement at the time.

JR
 
etheory said:
So aside from Q1 I believe that this IS already Class-A no?  At least everything I can see aside from Q1 is conducting "all the time".  Or does Class-A have a more subtle definition?

Q1 & Q4 are a single-ended stage, so this must be class A (and already is in your schematics) or it all falls in a heap.  Your output stage idles at around 100mA so maximum class A output current is 200mA; class A for loads of about 100 ohms or higher at maximum output voltage.  That's 2W output power at 100 ohms.


 
From a quick glance, you're in for a meltdown. The output stage is still a class A/B topology, but you have increased the quiescent current so that it stays "class A" into heavier loads. Your four diode bias will run the output stage at around 140mA and your output devices are asked to dissipate like 3W idle! This calls for some massive heatsinking! 15-20mA would be sufficient and more sane. The VAS is also running unduly rich at 11mA.
The input stage currents are very unbalanced at that tail current (=odd harmonics) and your trimpot won't be able to trim away the offset. With that 1k collector load, i think around 1.5mA in the tail ccs would be more in the ballpark. Then you can get rid of R19, which will help to increase your open loop gain. Myself, i'd bias that input stage at least twice your current.
Stability is another big concern, as steveh pointed out. You will need some miller compensation around your VAS at least.
LF open loop gain is likely to be very low as-is, in the 60dB range - so not much is available for feedback, unless you're running it at very low closed loop gain. But i guess that could be seen as a feature by your book? ;)
Most of my suggestions for further improvements would just clean it up, and since you're heading the other direction i'll rest my case for now... :)

EDIT: Added some stuff...
 
> use in differential input stages and EQ's
> lots of harmonic "thickness"


You do NOT want a hundred cascaded "thick" stages.

You want many dead-clean for utility and EQ.

You want at most one, two, maybe three "thick" stages between mike and CD.

The "thick" stage itself does not need differential inputs or hefty outputs.... you can buffer in and out of a "thick" naked-JFET or starved LM308 with dead-clean amps.
 
PRR said:
> use in differential input stages and EQ's
> lots of harmonic "thickness"


You do NOT want a hundred cascaded "thick" stages.

You want many dead-clean for utility and EQ.

You want at most one, two, maybe three "thick" stages between mike and CD.

The "thick" stage itself does not need differential inputs or hefty outputs.... you can buffer in and out of a "thick" naked-JFET or starved LM308 with dead-clean amps.

I absolutely agree with everything above.
However I am a dreaded "bedroom producer" and hence I have control over those 1 or 3 stages within a space of about 30cm of not much gear sitting next to me ;-)
So for me you could kind of call recording the processing of being able to choose anything from an active wire to a distortion box - depending on which bit of analog synth gear I am recording.

Having said that, I revised the design already, and it's stupidly better now anyway with regards to distortion - in fact it's the opposite; as clean as I can get.  Now I'm playing the game of experimentation.
I feel you will be right and I'll end up liking the cleanest sound possible better, but I want to be able to tune stuff to dial in some harmonic content for recording my Oberheim SEM or Juno 106 or whatever is sitting next to me.
It's a little more limiting going straight from analog synth to computer, since it's unlikely in my case to get a heck of a lot of external processing (as I've been mostly ITB, but have never really liked it - hence getting the analog synths and building my first bit of DIY gear - the SB4000).

Modification to circuit follows, however I think potentially Q1/Q4 running at around 250mA will be in flames, BUT it's significantly better on the distortion front in the simulator, and playing with resistor values is giving me a much better concept of how the harmonics and their relative amplitudes are tied to signal topology.
Thanks again!
 

Attachments

  • eCheekyDOA_v03_01.jpg
    eCheekyDOA_v03_01.jpg
    110.4 KB · Views: 47
I would replace J3 with a bipolar transistor, just like Q4. Otherwise you will either have to measure the JFET outside the circuit and choose the appropriate source resistor, or use a trimmer. Without that the currents will be all over the place due to the big variations in the transistors.

If you are determined to use a JFET CS, a 2SK170 isn't really the optimum part for a current source in the first place.
 
volker said:
I would replace J3 with a bipolar transistor, just like Q4. Otherwise you will either have to measure the JFET outside the circuit and choose the appropriate source resistor, or use a trimmer. Without that the currents will be all over the place due to the big variations in the transistors.

If you are determined to use a JFET CS, a 2SK170 isn't really the optimum part for a current source in the first place.

Yup I already suggested a trimpot there but if he is looking for experience he will get more of it by ignoring advice.  8)

JR
 
I'm not ignoring anyone JohnRoberts ;-)

There is a much more comical reason for not replacing that current source.

I tried to last night and just was too tired (aka brain was in standby mode) to get it to stabilise (using a BJT) to the current I wanted it to, so, as I was more interested in improving the overall performance of all areas of the opamp, and just decided to just leave it temporarily (knowing full-well it would be relatively easy to replace).  Next step will be to do just that!

The other mods had such an incredibly drastic improvement that they took all of my focus.
 
JohnRoberts said:
Yup I already suggested a trimpot there

Trim pots are a bit of a pain though and J-Fets are never well-defined enough in Idss in my opinion to use as current sources/sinks anyway.
A BJT is much more predictable.  Although, a J-Fet as a cascode (self-biased) on top of a BJT is an improvement that is sometimes useful.
 
I am generally not enthusiastic about the project with or without trimpots, I suggested a trimpot in the spirit of an opamp that already has one, and looked like the nominal operating current could vary wildly depending on actual parts used.

If the poster is just working in a sim environment, who knows how the sim deals with real component variance. No actual smoke will be lost during simulations. 

YMMV

JR
 
Back
Top