Class A Output Stage
« on: August 06, 2019, 07:38:41 PM »
EDIT: To keep data splatter down, I've removed the initial schematics to avoid confusion.  Post # 18 contains the most up to date version of this.  Cheers.


I've been playing around with the Neve BA512 output amplifier which was a BBC specified upgrade to the Neve BA340/BA440 amps.   After some LTspice analysis of the circuit I decided to see if I could better it in terms of stability, slew rate/transient response, linearity, and noise.  Plus eliminate the crossover distortion by operating in good old inefficient class A.
What I ended up with after playing with spice analysis and a breadboard proto no longer resembles the original 512 circuit (bar maybe the two-pole miller compensation) but the slew rate, noise, linearity etc. are much improved.
Nothing ground breaking about this, pretty much a mini power amp topology.  Way more small signal transistors than the usual, but it does acquit itself quite well.
Next up is to do a PCB (maybe go to SMD for a lot of it to keep size down?) and also look at some alternate faster output devices.
Gain as shown is about 14dB and max output is +18dBu with the usual 24V Neve supply so, with a 150:600 output transformer, gain = 20dB and max output post transformer = +24dBu in Class 'A' mode.

Anyway, pdf of the circuit as it currently stands is attached. 
Hinson
« Last Edit: September 07, 2019, 09:21:51 PM by Winston O'Boogie »


Re: Class 'A' Output Stage
« Reply #1 on: August 07, 2019, 03:19:13 PM »
The circuit posted above was being powered by 3 X series 9V batteries tweaked down to 24V as I have nothing else on hand.
Today I powered it with bi-polar supplies - 18V tweaked down to 16V each -  removing the now inappropriate bootstrap biasing stuff from the input. 
Spice had given an output offset value of .55mV with perfect devices.   With random real devices, but with supplies adjusted for accurate 16V each, I measured 1 - 2 mV of output offset after an hour of being on so removed the output cap.   There is a small positive bias on the input of course so that cap stays. 

Latest schem attached:
Cheers.

Edit:  Forgot -  I added RC filtering for the front end.  The current draw is equal for each rail and the voltage drop across each 10R feeding the front end is nothing to care about really.   Total current was 66mA per rail with the bias I used btw. 
« Last Edit: September 07, 2019, 04:37:20 AM by Winston O'Boogie »

Re: Class 'A' Output Stage
« Reply #2 on: August 07, 2019, 04:11:18 PM »
Do you have some THD measurements to compare both stages?

Re: Class 'A' Output Stage
« Reply #3 on: August 07, 2019, 04:35:08 PM »
Do you have some THD measurements to compare both stages?


Heya,
Do you mean comparing the two I posted?  They're identical below clipping, the +/-16v scheme having a bit more headroom of course (2dB).
I don't have anything that can resolve low enough to measure any THD at all on this tbh! And I don't rely on Spice for THD measurements, does anyone? 
What I can say:  The input stage is pretty much perfectly balanced so H2 disappeared from there.   And the H3 was lost when the loop was closed, the two-pole miller compensation helping keep feedback sufficient as frequency increases.  The output stage will be the primary source of any THD but that also has some local feedback with the driver and output transistors cross coupled.   And with it being class A no crossover distortion.   So, with the loop closed I saw nothing until clip except  some hum pickup since the proto circuit was exposed in a rat's nest config.
Best thing would be if I built it up on a PCB and sent it to someone who has better measuring capability.
Anyway...  ?  😃
« Last Edit: August 07, 2019, 05:25:17 PM by Winston O'Boogie »

Re: Class 'A' Output Stage
« Reply #4 on: August 07, 2019, 04:39:11 PM »
P.S.  I didn't measure with the transformer included btw. 

P.P.S.  I saw no difference with or without the output cap re distortion since, as per D. Self, it was sufficiently large to begin with.  As I said in the first post, this is basically a mini power amp and some of the stuff here is directly following his "Blameless" approach.  The extra stuff on top was balancing the input stage further and the two-pole miller stuff.  And degenerating the current mirror more to further balance things and lower noise. 

P.P.S.  And I enclosed the 680uF on the inverting input with 50% of the loop just in case there was any dist. from that which I doubt but, didn't hurt 😃
« Last Edit: August 07, 2019, 04:56:05 PM by Winston O'Boogie »

ruffrecords

Re: Class 'A' Output Stage
« Reply #5 on: August 08, 2019, 04:10:28 AM »
What load did you use when simulating for distortion?

Cheers

Ian
www.customtubeconsoles.com
https://mark3vtm.blogspot.co.uk/
www.eztubemixer.blogspot.co.uk


'The only people not making mistakes are the people doing nothing'

Re: Class 'A' Output Stage
« Reply #6 on: August 08, 2019, 07:07:03 AM »
What load did you use when simulating for distortion?

Cheers

Ian

Hey Ian,
I used a load of 150 ohms for the complete amp sans transformer. 
I spent some time looking at the the input stage in isolation when juggling the balance of each side and used a hi-z load of an ideal buffer for that stuff, although most of the balancing was done by probing voltages and currents.
I'm by no means a simulating wizard and there's much I'd like to learn there. 
Cheers

moamps

Re: Class 'A' Output Stage
« Reply #7 on: August 12, 2019, 02:53:13 PM »
May I ask why you chose diamond buffer for output stage? Is it better than for example CFP or EF stage in that position?
And shouldn't some transistor pairs be thermally coupled on the PCB as some driver transistors with output pair in diamond buffer? Thanks.

Re: Class 'A' Output Stage
« Reply #8 on: August 12, 2019, 04:40:20 PM »
May I ask why you chose diamond buffer for output stage? Is it better than for example CFP or EF stage in that position?
And shouldn't some transistor pairs be thermally coupled on the PCB as some driver transistors with output pair in diamond buffer? Thanks.

Hi Moamps.
Please don't put too much weight on my PCB layout, it was very much a first draft and I've already changed it somewhat. 😊

I found that the diamond is a topology that is less in need of thermally coupling the drivers to output pair than an EF double or triple arrangement for instance.   Only one base emitter drop to worry about.  The weak link is probably my current sources and a more thermally stable voltage reference would help there.  Maybe an active CCS to feed the reference instead of resistors too?

I do think a CFP would probably be a more linear output stage for sure.   But there is some feedback in the slightly modified diamond  I used here which linearizes it somewhat and also, when I looked at a CFP arrangement in this it was a bit less stable and required some local treatment to be able to close the loop.
I may look at transistional miller compensation rather than the two-pole to see if there is any worthwhile benefit to including the output stage within the miller loop but, to be honest, it seems to be more than capabable of doing what I want with it.

This will directly drive a transformer which will add non linearities of its own.    And I intend on including more schmutz "to taste" via a control which will raise the DC offset from a couple of mV to some bias value which produces the required shift in the magnetic curve of the transformer.   Same as the Neve 'Silk' mode.
But I did want a low THD,  better spec than original Neve output amp to begin with. 

There will be an input stage and transformer to form a mic pre.    This will be for my son who's just starting out in recording and a pre with some character would suit what he does.

Not yet decided on the front end.  I have a nice folded cascode circuit that uses unobtainium Toshiba JFets of which I have a few on hand but, then again, maybe something Neve-esque with a current feedback single ended input.  Not sure yet...?

Suggestions and critiques  are  more than welcome so, fire away 😊

Cheers Moamps 👍

CJ

Re: Class 'A' Output Stage
« Reply #9 on: August 13, 2019, 02:22:18 AM »
what kind of output transformer?  ratio? like 1166  i bet
If I can't fix it, I can fix it so nobody else can!
Frank's Tube Page: www.mif.pg.gda.pl/homepages/frank/vs.html
Guitar Amps: http://bmamps.com/Tech_sch.html


Re: Class 'A' Output Stage
« Reply #10 on: August 13, 2019, 05:12:26 AM »
what kind of output transformer?  ratio? like 1166  i bet

Heya CJ,
Yep a 200:600 or 150:600.   The standard option  would be the ungapped Carnhill one and it's pretty cheap over here from AML but not sure, something that gets a bit squirley at the low end with some DC applied?
For the input I fancy checking out the 1:3 Sowter EMI TG copy with secondary taps for -5, -10, &-15 dB.

moamps

Re: Class 'A' Output Stage
« Reply #11 on: August 13, 2019, 02:51:30 PM »
........This will be for my son who's just starting out in recording and a pre with some character would suit what he does....
Then it should be something special.  And you should implement all your Toshiba unobtainium  stuff (2SK146 etc.) in it . :)
Thanks for detailed explanation.

Re: Class 'A' Output Stage
« Reply #12 on: August 14, 2019, 10:04:58 AM »
Then it should be something special.  And you should implement all your Toshiba unobtainium  stuff (2SK146 etc.) in it . :)
Thanks for detailed explanation.

Thanks Moamps, appreciate the interest.

I hope I can live up to my own expectations and ideas for this particular mic amp.
Throwing a few more cheap transistors at it (if it benefits anything) is no big deal in a one off device.
Same for matching complementary pairs + spares of obselete Toshibas since I have them.

Cheers.


Re: Class 'A' Output Stage
« Reply #13 on: August 19, 2019, 02:25:55 PM »
Hi Winston,

Can you explain to me how this is Class A? (Or am I missing something? - I noticed the "A" in "Class A" is in quotes in the thread subject)

To my understanding the Diamond Buffer is Class AB but like most typical Class AB output stages can be biased into Class A.
Is that what is going on here?

One other detail I noticed is that Q4 and Q6 are tied to Q14 and Q15 emitters instead of the voltage rails - is that the trick?

Lucas

Re: Class 'A' Output Stage
« Reply #14 on: August 19, 2019, 03:08:32 PM »
Hi Winston,

Can you explain to me how this is Class A? (Or am I missing something? - I noticed the "A" in "Class A" is in quotes in the thread subject)

To my understanding the Diamond Buffer is Class AB but like most typical Class AB output stages can be biased into Class A.
Is that what is going on here?

One other detail I noticed is that Q4 and Q6 are tied to Q14 and Q15 emitters instead of the voltage rails - is that the trick?

Lucas

A Diamond buffer can indeed be class AB if biased for a leaner standing current.  In that case, and in reference to Oliver, there is an optimum voltage drop across the emitter resistors to minimize gm doubling when the stage moves from class A to B and switches each pair on and off as more current is demanded of the stage.     This switching leads to distortions, sometimes inadvertently more so as our intrepid DIY'er has added a bit more class A bias but not full class A into the mix,  moving away from the optimum and causing little bumps in the gm. 

With a pure class A, this distortion mechanism simply isn't there.   

I biased this for class A up to clipping into the lowest load it will see - which here is about 200 ohms reflected to the transformer primary in parallel with about 2K from the -tve feedback.   This condition seems to keep the output devices equidissipational and keeps any thermal distortions down.

The tying of the buffer's input transistor collectors to output transistor emitters has a number of advantages over the regular connection to the other side's V rail: Dissipation is kept lower in the input pair.  The input impedance is raised.  It reduces the voltage variable CB capacitance and Early effects on beta.  It costs nothing to do it 😊

Hope that covers everything. 

Edited for clarity.

« Last Edit: August 19, 2019, 03:28:39 PM by Winston O'Boogie »

Re: Class 'A' Output Stage
« Reply #15 on: August 19, 2019, 03:11:59 PM »
That covers it.

Thank you.

Lucas

Re: Class 'A' Output Stage
« Reply #16 on: August 19, 2019, 03:14:21 PM »
OK.  Welcome to the forum by the way Lucas  :)

Re: Class A Output Stage
« Reply #17 on: August 19, 2019, 03:29:40 PM »
P.S.  Are you by any chance a Lucas Burden of Rupert Neve Designs?

Re: Class A Output Stage
« Reply #18 on: August 22, 2019, 01:21:48 PM »
For the sake of completeness, final schem with slight changes.
Cascode biasing voltage from led returned to VAS emitter rather than ground. 
VAS current limiting BJT removed due to a slight increase in 2nd harmonic when in place.  Current limiting with VAS emitter R @ approx.  1/3rd of VAS current source R seems to be about right here so 33R:100R in this instance.

Over and out on this for now, onwards and upwards...
« Last Edit: September 07, 2019, 09:22:58 PM by Winston O'Boogie »

CJ

Re: Class A Output Stage
« Reply #19 on: August 22, 2019, 01:53:32 PM »
KSC1845 same as 2SC1845, 120 V 50 ma 200-1200?

let me know if you need the real outputs with the rev wind!  :D
If I can't fix it, I can fix it so nobody else can!
Frank's Tube Page: www.mif.pg.gda.pl/homepages/frank/vs.html
Guitar Amps: http://bmamps.com/Tech_sch.html


 

Related Topics

  Subject / Started by Replies Last post
6 Replies
3361 Views
Last post June 12, 2004, 03:02:39 PM
by sismofyt
2 Replies
1186 Views
Last post February 11, 2005, 01:58:55 AM
by pstamler
3 Replies
1799 Views
Last post October 12, 2007, 04:20:28 PM
by Gus
1 Replies
1282 Views
Last post November 26, 2007, 11:37:17 AM
by pstamler