Greening a Classic?

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
There has been a bit of discussion on improving the Digi125 but no one seems to have picked up the gauntlet. Well, I am no expert, but I think I have followed what others have suggested in this Forum. If I can get some feedback I will do up the board and publish it.

I need to know if I screwed up the biasing arrangement with the changes I have made.

Also did I bias the caps to ground from the feedback loop properly?

The extra diodes are there for experimentation.

Is there a good place to put a pot to adjust offset? Value?

Any suggestions are welcome.

Well, here it is.......... and thanks.
 

Attachments

  • digigreen150jpg.jpg
    digigreen150jpg.jpg
    56.2 KB · Views: 2,294
I can see no fault in this

It should work resonably well
especially with regulated supply.
The currents in the input pair
are calculated to acheive equally sharing of current.

You can split the 2k21 resistor, above the 4 diodes,
into two resistors,
and put a bootstrap capacitor
from between the resistors to the output.

Or replace that resistor with a current source,
but then you lose the simplicity
of the design.

I maybe should increase the currents in the input pair.
0.3mA to drive the 15mA in the BD139
is not much.
1/10 would make the current 1.5mA
Something like 1-2 mA would be my choice
 
The one and only
Joined 2001
Paid Member
In addition to the bootstrapping mentioned above, this
amp seriously needs a few parts. (Did I just say that???)

1) Some sort of input resistance to prevent parasitic oscillation
with the source impedance. Happens all the time.

2) An output RC network to prevent oscillation with the load.
Also happens all the time

3) Flyback diodes across outputs - You never Know.

4) Decouple the V+ on the diff pair's current source to get
the ripple out unless your supply is regulated.

5) You will really need some adjustability on the bias.
4 diodes is never going to give you the exact value, and you'll
need better than that.

Do all these things and you won't be sorry.

Do all these things and you'll have an HK Citation 12.

:)

Oh, I forgot. You also need some resistance between the
emitter of the drivers to bias them up. A hundred ohms
shoud do it.

:cool:
 
djk said:
Ditch the bias on the feedback caps, all it will do is inject hum. If you really want to try this use a battery and a 1M resistor.

Don't you think it will work to just insert a 1M resistor between
the voltage divider and the caps? The caps themselves should
provide sufficient decoupling to get rid of the hum I'd guess.
If not, then a zener across the resistor to ground in the voltage
divider should help.
 
Ok,

I am taking y'alls advice. And I really appreciate it. Thanks.

Is this a board project? With the help of TOAO (Nelso Pass, The One And Only) and the rest of this forum's super contributors we may end up with a great sounding simple amp that was designed by the combined efforts of this board's posters. There is a lot of learning and experience here that can be leveraged into a great project. Let's keep it up. End of pep talk!

Here is the latest schematic.

Nelson, is this what you meant for biasing the drivers?

Because I am not one of those with a lot of knowledge and experience in the field, I am at al loss for figuring the proper DA resistors to get the DA current up and also keep proper bias on the VAS.

Also the bias circuit needs resistor values.

Help!

BTW with the RC filters for the front end dropping the front end V rails, will the amp be able to fully drive the Output to the output's rails less the output VBE drops?

Anybody here willing to put this amp into a simulator and help tweak it up?

In any event here is the latest revision
 

Attachments

  • digigreen2_150.jpg
    digigreen2_150.jpg
    72.1 KB · Views: 2,000
The one and only
Joined 2001
Paid Member
Well, I would reduce the 3K01 filter resistors down to
something like 33 ohms, and the negative one needs to
be at the right of the 59 ohm resistor.

Your 470 uF bootstrap cap (-) terminal needs to attach to
the output, not the negative input.

Personally, I don't think you need the 150 pF input or the
extra 1K input resistor.

Also personally, I would ditch the two 470 uF back to back
polar caps and the DC bias network and replace them with
just one such cap with a film cap in parallel and a 22k
resistor also in parallel to bleed off DC.

The bias transistor ??? values could be 6.8K on the CB, and
a 5K pot BE.

The bias to the diff pair could be around 33K, and the
resistance off the input drain in the region of 220 ohms -
adjust either of these for DC offset.

Output swing? With the bootstrap it will be much higher
than it was.

At this point the open loop gain is low and stability is probably
high enough that you can consider reducing the 150 pF
lag cap to between 0 and its current value, looking for
oscillation as you do so.
 
My thanks to Nelson, Halojoy, dik and Christer. With their help the project has moved along.

Below you will find the third installment of the saga. Indeed, if you squint at the schematic real hard, you should see the beginnings of a PC Board!

The schematic has been shuffeled a bit for thinking about the PC Board layout. Please check me out for errors. I have made a bunch so far.

I moved the +V 33R resistor to the right side of the bootstrap resistors so that the VAS gets quiet DC on both ends. Is that ok?

The 22R resistor/Jumper on the collector of the DA is there for PC Board layout ease. I understantd the resistor will not hurt. If it does a jumper can be used. That resistor is a spare anyway.

The 500R pot on the other leg of the DA is for offset adjustment. A cermet multi-turn should be just fine, don't you think?

Point A on the schematic connects to point A. I did it this way for clarity and to show that in the practical amp, point A will be the positive speaker lug on the amp. Any thoughts on this technique?

You will note that I have used caps that are overkill. But this allows the builder to use the same caps throughout the amp. If bought in bunches (like at DigiKey) they will be cheaper and there is less chance of error for a beginning builder.

I am sure there is more that can be done. But, since this is the first time I have attempted such a project (design rather than straight diy) I have no clue what should be done next.

When I get feedback on this version I will post the schematic again. Then tweak, tweak, tweak. Someone might try to model it on their simulator for shaping it up before I do the PC Board.

So, here is version three:
 

Attachments

  • digigreen3_150.gif
    digigreen3_150.gif
    14 KB · Views: 1,843
I also would be interested to see a Simulation

The result of a Simulation could be quite nice!
If anyone have some time, so please
gives us some results :)

You have not so much to gain by putting a film
parallelled to ANY big elektrolyte.
The total resulting qualities will to depend on
the big one.
A good rule is that for parallelling should begin
to do any real good, it should be at least 1/3 of the electrolytic.
If you bypass 470uF with 10uF FILM (and that is a big one)
you get 1/47.
So the qualities will be 98% electrolyte.
And if the effect of the 470uF is small in the circuit
the 2% difference will not even be noticed.

It is different when you put them
in parallell with long supply line between them.
Then you have some little wire resistance/inductance
to give some effect. RC LC filters.

I never do bypassing of an electrolyte. I put rather my money on some other thing where it will have greater effect.
Or some new CDs to listen to. ;)

This project is beginning to look quite interesting
for DIY builders, aborza
 
The most recent version #4

Thanks for the support halojoy. With your help and the help of others (and THANK YOU SO MUCH NELSON) this could work out well.

As to the cap issue, although I am not an EE, or even close, what I understand is that all caps have frequency dependent capacitance.

By that I mean that a big electrolytic in a power supply should be able to sink all frequencies from giga hertz to just above DC. But they cannot. They have an innability to deal with the higher frequencies. Therefore, a film across such a big cap can bypass higer frequencies the larger cap cannot see. So it is not, for the instance I describe, a proportional issue as you state.

If the electrolytic cap acts as an open circuit at some higher freq and the film cap can pass it, then the relative sizes are not material. Indeed, that is one of the reasons there are so many different cap materials. They have applications at different frequencies.

My Question was wether it made sense in that instance (the bootstrap cap) to bypass it with a film type. If there are no frequencies that need to be passed (at that point) that the electrolytic cannot handle, then the answer is no. Discussion any one?

Nelson pointed out that the open loop gain of the amp is quite low and that the compensation cap might be diminished or eliminated. I understand that if local degeneration is judiciously used each stage will be more linear and the open loop gain is reduced. Therefore there is less distortion for global feedback to clean up and the amp's stability is enhanced. Does this not result in a better sounding amp? Discussion anyone?

Nelson, I do have access to a 20MHz scope. How should I stress the amp to determine its stability re the comp cap?
 
The AKSA fans like a sleazy electrolytic with no bypass for the bootstrap cap, claim it sounds better than a good cap in this position.

I have not played with this myself.

As regards the 100 ohm base resistors, I would remove the pair and replace with only one running between the pair of output transistor bases, with no connection to the amplifier output. This resistor should have a 0.1µF~1.0µF film cap in parallel (as seen in Leach, Levinson, SAE, AKSA, etc).
 
At the risk of being the one-too-many cook that spoils the broth...

1) Unlike AKSA and Pass I don't favour bootstraps being used as a constant current source substitude. IMHO it is an old-fashioned method that was invented decades ago primarily to save on transistors. They are very frequency dependent and involve an extra electrolytic in the signal path and present an impedance to the driver that is affected directly by the linearity of the output stage. My advice: replace it with a CCS - you can buy these as FET diodes that only need a single bias resistor if you want to keep it simple to build.

2) Also use a CCS on the LTP emitter supply. This will improve both the CMRR and PSRR of the critical input stage. AKSA likes a resistor because it adds 2nd harmonic distortion and improves the perceived sound of his design. That's logical. But I'm an extreme oppoent of anything that adds distortion.

3) Get rid of the 22k that is in parallel with the 470uF on the inverting input. Not needed.

4) Like Halojoy says you need to be cautious about paralleling electrolytics with high frequency caps. This is due to resonances. In general I would avoid this unless you have a network analyzer to check out the circuit. This won't show up in simulations.

5) I'd put a 470pF cap between non-inverting input and ground to supress RF and parasitic feedback.

6) The output zobel ought to have a lower corner frequency. I suggest a 220nF cap in series with the 10-ohm resistor. Don't use wire-wound resistors because they are inductive.

7) Get rid of the emitter degeration resistor on the VAS transistor to improve linearity. Yes, you heard me (see my comment on feedback below).

8) "As regards the 100 ohm base resistors, I would remove the pair and replace with only one running between the pair of output transistor bases, with no connection to the amplifier output."
I would not do this if you are intending to operate it in class AB. Keep it as Nelson suggests.

A few nuggets to ponder:

You can only hear sounds between about 20Hz and 20kHz. You may well ask why some people say it is important to have a wide bandwidth design, like 200kHz or more. Keep asking yourself why this should be. What is the underlying reason? You can only hear voltage signals below 20kHz. Let me say this again: you can only hear voltage signals below 20kHz. And again: you can only hear voltage signals below 20kHz.

Local vs global feedback. Every amp design is a mix of these anyway. A darlington has local feedback. The Miller cap across the VAS stage is local feedback. The important thing is to minimize sources of distortion everywhere regardless of feedback. Some localised feedback arrangements introduce more distortion than they remove.

Simulators are pants at predicting sound quality. They are good at checking basic circuit functionality and approximate frequency response. They are "too ideal" to predict audible distortion with precision and to do so would require you to teach them more detail about real circuit performance.
 
Hi traderban,

Thanks for your input and welcome to the fray. I hope others join you.:up:

There cannot be too many cooks for this broth. The whole point is to obtain as much good info as possible so as to develop a nice sounding, clean, simple and stable design (with possibilities for experimentation) for the diy community and that the amp be simple to build from readily available parts.

Hopefully this thread will be a learning experience for me and the diy community at large. After all, where has it been done that an amp has been designeed, on line for all to see, with the likes of Nelson, yourself and others contributing to the final product?:xfingers:

It has been said that a camel was the result of committee design. Well, that beast, while ugly, surely works well in the desert! Hopefully the result of our mutual efforts will be an audio amp that is as successful as a camel.;)

Indeed, I hope to design the PC Board with options that allow conversion from one mode of operation to another if the basic parameters of the system are not compromised (i.e. conversion from resistor current sources to CCS). Making this kind of experimental change may introduce some second harmonic distortion, but the user may like it. This sort of option can be made available with a little help from learned posters like yourself. And many will learn by it.

The FET diode solution you suggest is a perfect example. Could you (or anyone else) elaborate on its use in this application?

Assuming the bias levels at this stage of the design are appropriate (Let us not assume, are they appropriate? Are other operating points better? Hmm?) can you suggest diodes and values as well as placement on the schematic? I would greatly appreciate it and will modify the schematic and post it for corrections and comment. The PC Board will allow for either option.

As for your "nuggets":

I agree that an audio amp need not go from DC to visible light. But flat from perhaps 5Hz to 100KHz should be more than sufficient and will keep phase from 20-20K from getting ugly. Discussion any one?

I do not know what the response of this design might be. I a not an engineer (or even a technician). Perhaps you or others might elaborate on that issue as well.

As for simulators, yes they use ideal active and passive components and those things do not exist in reality. But, they do address functionality. And I need all the help I can get. :D
 
There are other ways to get some CCS performance without the CCS "complexity", if you just have to avoid the CCS. How about having the best of both worlds (effects of a CCS and simplicty of a resistor) by using a resistor and raising the +V rail on the driver stage. I always thought one did the bootstrapping thingy to avoid having to use a higher voltage supply for the driver stage.

How you get the extra rail voltage is, of course, left as an exercise for the reader. :D (I've always wanted to say that)

One other thing, might there be some benefit to running the drivers in class A by connecting the emitters through a resistor that does NOT touch the output line? In other words, taking Nelson's advice literally. I see that Tradebaum suggests otherwise (in his #8 below). You can do both easily and see which you like better. This is DIY, after all ;)

good luck!
mlloyd1

traderbam said:
...
1) Unlike AKSA and Pass I don't favour bootstraps being used as a constant current source substitude. IMHO it is an old-fashioned method that was invented decades ago primarily to save on transistors. They are very frequency dependent and involve an extra electrolytic in the signal path and present an impedance to the driver that is affected directly by the linearity of the output stage. My advice: replace it with a CCS - you can buy these as FET diodes that only need a single bias resistor if you want to keep it simple to build.
....
8) "As regards the 100 ohm base resistors, I would remove the pair and replace with only one running between the pair of output transistor bases, with no connection to the amplifier output."
I would not do this if you are intending to operate it in class AB. Keep it as Nelson suggests.
....
 
aborza said:
I moved the +V 33R resistor to the right side of the bootstrap resistors so that the VAS gets quiet DC on both ends. Is that ok?

The 500R pot on the other leg of the DA is for offset adjustment. A cermet multi-turn should be just fine, don't you think?

You will note that I have used caps that are overkill. But this allows the builder to use the same caps throughout the amp. If bought in bunches (like at DigiKey) they will be cheaper and there is less chance of error for a beginning builder.

I am sure there is more that can be done.
I try to use 1/20 to 1/10 current for a stage that drive next stage.
VAS stage use 15mA.
I suggested 1-2 mA in each input transistor.
Nelson raised the 0.3 mA to 0.5mA
0.5mA can make lower noise than 2 mA.

The voltage at BD139 base, is something like 1.5volt
So you need 3000R instead of 500R, if you use 0.5mA
For my choice 1.5mA this resistor would be 1000R

After bootstraping the size of the 59R resistor
has not so negative effect.
But here I am not so sure what is best to do.
I like to have emittor resistors, to lower the open-loop gain
and use less amount of feedback

in the transistor setting the idle current
15 mA will also flow
I would use at least 1mA to bias
6k8 will let something like 0.3mA flow
it will surely work, but wouldn't be my choice
I would use 2.2k

about the positive line 33R
this is not a symetrical amp, so there are several choices
of RC-filtering in the positive line
the most important is to have clean current
for the input pair.
this could be done with 2-steps filter.
Or only one filter to the left of the VAS-stage
the negative and positive line filter components need not to have same values

this is what I figured out, using my present knowledge.
I tried my best
but if that will do, I do not know
/halojoy
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.