New to DIYAudio Help With New Build?

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hi guys, I've been reading around here for a while but just got around to trying it myself. I would like to pose a question if anyone could help that would be awesome.

Basically I'm starting with a Douglass Self Blameless Amp design. I've got it almost exactly as he does in the online pages on "Distortion in power amplifiers". I've read through and tried to understand everything and I think I've understood most of it however I've having some trouble.

I've built a Blameless amp with CFP output stage and an Emitter Follower VAS. I have changed the NFB HF cap to 33pf (cuz its what i had) and left the LF NFB stuff alone. The transistors aren't the same but I purchased them thinking they were better than what he has suggested.
KSA916Y and KSC2316Y for the higher power stuff (Vas + Output Drivers) and KSA992F and KSC1845F for the input stage. MJL4302 and MJL4281 for the final output.
Basically I was getting some nasty oscillation (or it sounded like oscillation - just constant noise, hum and hiss) that would quiet quite a bit when I probed with my multimeter from ground to the npn driver of the output stage or when I touched certain areas with my finger.

First thing I tried was replacing crap everywhere thinking I might have melted something soldering, then tried increasing the LocalFeedback Cap on the VAS which did nothing from 100pf to 1500pf except make it sound a bit worse.
Finally I found out that putting a 220pf cap from output driver base to collector on the main output (on both output transistors) made the noise stop.
Can anyone explain this to me? All I know is that the oscillation is not audible anymore and the drivers aren't overheating like they used to (like when I got the noise) which is good, but I don't know why what I did worked? I haven't tried anymore values. Does anyone have any other possible solutions that don't involve these caps? Did I buy crap transistors?

Thanks for any help...
 
Start with something simple - is your heatsink grounded? If not it can act as an antenna and make the circuit oscillate.

CFP's - nasty little buggers in my opinion. PCB layout is highly important here. Did you use a pre-designed one, or did you make it yourself? Can we see?

The problem with the "better" transistors is that they are faster, which in itself can make things unstable unless properly compensated for.
 
is there anything that says how I can properly compensate for the cfp stage? I'm looking for higher frew output transistors because I want more high freq NFB capability. Right now I have the main output transistors on separate heatsinks (npn and pnp) because I've seen transistors with continuity from the heatplate to the (emitter?) basically the goal was to have each heatsiink have it's own charge. I did't check that though, maybe grounding it is a good idea. I'll try that tonight.

to be honest I have looked at the layout almost none. I just went from a birdnest breadboard to a radioshack special solder board to get rid of wire loops.
 
What I've done is attach the power and output directly to the output transistors via dedicated heavy gauge wire and from them to the power and them to the feedback resistors to the speaker out. Thus the entire CFP setup attached directly to the leads of the output transistor. Now I only need 2 signal wires to attach between the CFP drivers and the signal (solder) board.

Separate +- power and ground are all pulled from a tri-star like setup at the power caps, thus attempting to keep high power currents out of the signal path. I don't think that the radioshack special should need to carry any 'real' current this way, just the signal currents... Correct me if this is wrong.

I plan to get (or make) a smaller version of the board, since I'm only using 15x15 pin area on the big one and everything is still a bit spread out. Basically I didn't want to prototype a board till I get something that is working, is that some sort of paradox? I can't get a good working one until I prototype, but I will obviously prototype it wrong based on what I learned before on the simpler one since it has these oscillating features built in?

Do you think it is ok to use a generic radioshack IC connection board for the signal path? What things would you add into prototype board that would make it better? Is that written anywhere?

Thanks so much for your input!
 
Administrator
Joined 2007
Paid Member
Hmmm...

There's no quick fix for this... all your issues will be down to layout and wiring. Dougs design is a fully worked example and performs well... and it is fully stable.

Dare we ask to see a picture of your project :)

Amps like this must be wired correctly... you can breadboard it, but the wiring has to be right with separate returns for critical parts of the circuit.
 
I can post a picture if that helps, but I don't see how. I'm more interested in why than how... Thanks for the link to layout I'll look at that.

I've been searching the internet and I ant find anyone describing the oscillation characteristics of cfps... Does anyone know about something like that?
 
Circuit.jpg
Circuit picture attached...

Two wires on top right are power and ground.
Rca input is on top.
Ground and NFB are on the bottom of the circuit.

Top left section is the current source
Middle 4 are the differential and current mirror

Top middle is vas current source
Right side is bias and below it is EF VAS

Two wires in from the left are ground and NFB
Wire leaving down and right are CFP driver inputs
 

Attachments

  • Circuit.jpg
    Circuit.jpg
    152.3 KB · Views: 297
Administrator
Joined 2007
Paid Member
Hi buzz... you've not been triggering the thread notifier... still under moderation I guess for first 5 posts.

I'll have a look later :)

Very briefly, part of the problem is that any amp has to be taken as a "whole", not just the circuit design but the way that design is built... the problem isn't the cfp stage, it's all the inter related coupling and "contamination" of critical points in the circuit by currents causing a volt drop across conductors and that volt drop modifying the actual currents in the amp... that's not explaining it very well is it :)

The wiring or PCB has to be designed so that any signal related current (speaker, zobel network etc) doesn't cause a volt drop along a conductor that other critical points are attached to. If it did then that volt drop may be seen as "a signal" by that part of the circuit and depending on the phase relationship could cause positive feedback (that's how oscillators work) and instability.

It's a hugely complex subject, all I can say is Dougs design is stable if constructed with the correct parts, and that instability will be down to layout and wiring.

Edit... and designs like Dougs (the topology... not meaning Dougs examples) are in some ways the worst for breadboarding as they are the type most prone to highlight any such problems by breaking into instability.
 
Last edited:
Administrator
Joined 2007
Paid Member
One other thing I forgot, the wiring around the drivers and outputs has to be kept short for stability, using wires to connect the drivers to the outputs is a no no really.

Don't be discouraged though... perhaps have a go at something less ambitious.

Dougs "blameless class b" in the original articles in Electronics World used a conventional EF type output stage which is much more docile than a CFP. His trimodal design used a CFP, and possibly others too. Do you have a link to the circuit you are using ?

Edit... this appeared today which might give you some encouragement as to what is possible,
http://www.diyaudio.com/forums/ever...your-vero-board-designs-here.html#post2264434
 
Last edited:
So then if I understand you correctly... would a star topology for the signal section power and ground be better then the rail topology that you get on your standard radio shack special? Then I would stil want to isolate signal power from cfp power somehow, right? Maybe with a lowpass filter of some sort? But I should attach the output transistors to the board in either case because the driver leads need to be short.

bottom line is that I have a amp that works right now, but I'm on here to figure out a bit more about the topology so that I don't need to add weird capacitors that I'm sure are degrading the quality of reproduction that could be had.

Thanks a ton mooly, the piece that didn't explain so well, I think I got it. Just need to figure out the best way to do it. I had a few misconceptions about where those drops were going to be important. I figured if the cfp stage driver and output were directly attached, they would be stable on their own
 
I am using his blameless amp but I don't have a full schematic. I just cluged together the different parts. I don't have one of his schematics with a cfp in it. I read 'distortion in power amplifiers' parts 1-7 and just tried something. Although it's almost exactly his EF version with a cfp in it's place.
 
AmpLayout.jpg
Ok, I have devised a slightly different amp layout to get rid of the vas input wire crossing over the whole circuit and to get rid of the mess of ground and nfb crap at the bottom. I drew up a high level and then tried for a circuit level. Now I'm suggesting a single point ground and a Rail + and Rail -.

I tried to design the circuit level like I would prototype it, all the crossing leads happen under components, and I have dashed them to indicate such.

Any suggestions to make it better? Does this look like a feasible setup or is there something missing? I was worried about the NFB lead being long but I can't figure a way around it... I thought about moving the HighFreq NFB cut over by the output though, maybe it would has less adverse effects running under the other components?

Obviously there will be local capacitors from signal power to ground on the signal/output board. Also a resistor between the two power sources (like a low pass filter), but that is not shown here because I plan for it to happen on a power circuit board by the main capacitors.
 

Attachments

  • CFP Amp Topology.png
    CFP Amp Topology.png
    45.3 KB · Views: 147
  • AmpLayout.jpg
    AmpLayout.jpg
    49.7 KB · Views: 140
Administrator
Joined 2007
Paid Member
I think if you are to stand any chance of this working you have to use a much larger board, put the drivers on the board too, then very short wires to the outputs.

It's the ground returns that are vitally important. You must return the input ground and the main "earthy" end of the feedback loop to a common point on the PCB then return that to a star ground at the PSU.
The speaker return must go back to the PSU star and so must the zobel network.

It's too complicated a question to answer really with a stripboard layout. Yes it can be done... but it's quite an advanced project :)
 
I dumped the previous stripboard, I got a larger board that has a ground plane on one side and circular solder pads only. I thought the ground plane might help me out... I dunno. I'm trying to find a way to get the output on the board as well. I think the new board has large enough holes.

With regards to your thread on ground contamination, I read your thread, but I cant really see what you have hooked up on the breadboard. I assume you are just showing that (the wrong way) is to have your output ground in-between your signal ground and the psu ground. I'm trying to follow this advice from the distortion in power amps guide on how to hook it up properly. (Picture below)

I.E. Have a single ground for the entire system. I planned to gather all the signal grounds on the signal board to a single spot (as in the picture) and run a single wire back to "Ground" where the main capacitors are attached. I would also run the speaker and local caps back to "Ground" directly.
 

Attachments

  • self_ch21_fig6t.jpg
    self_ch21_fig6t.jpg
    14 KB · Views: 141
Administrator
Joined 2007
Paid Member
Just take your time building it up and follow the basic layout in your picture above. You shouldn't get too many problems with the small signal stages and interaction, it's the driver and outputs where problems occur, so keep the wiring short to these.

You really are trying to run before you can walk with this :)

A ground plane isn't needed for audio work, it's more applicable to very high frequency design.

If you want a copy of the blameless amp with EF outputs send me a PM with an email address.
 
Thanks for the help so far, you've given me a lot to work from. I have a copy of the standard blameless amp online, EE or something like that where I got all the other info. I don't know if what you have is any different from that or not...

I think from what I've heard that I will hopefully be able to eliminate those pesky extra driver caps once I get everything on a single board. If I get really mad at it I'll go to the EF output setup, but I just like the idea of the CFP and I like the output linearity plus I think I can keep the drivers cooler that way. I'm not usually one to choose the easy way into anything... I'll probably get a used oscope pretty soon.

Although what I've been noticing a lot lately is which of my mp3's are well recorded and which aren't. I was just listening to a song and was getting some distortion on my amp and test speakers and got pretty angry at the sound of it. But when I switched back to the regular onkyo amp and speakers I heard it there too. Never noticed it before...:no: Now I cant listen to that song anymore...
 
AmpLayout.jpg
Ok, I have devised a slightly different amp layout to get rid of the vas input wire crossing over the whole circuit and to get rid of the mess of ground and nfb crap at the bottom. I drew up a high level and then tried for a circuit level. Now I'm suggesting a single point ground and a Rail + and Rail -.

I tried to design the circuit level like I would prototype it, all the crossing leads happen under components, and I have dashed them to indicate such.

Any suggestions to make it better? Does this look like a feasible setup or is there something missing? I was worried about the NFB lead being long but I can't figure a way around it... I thought about moving the HighFreq NFB cut over by the output though, maybe it would has less adverse effects running under the other components?

Obviously there will be local capacitors from signal power to ground on the signal/output board. Also a resistor between the two power sources (like a low pass filter), but that is not shown here because I plan for it to happen on a power circuit board by the main capacitors.

You really should go EF2. CFP made my AM radio go crazy. :D:D They (the cfp's) are separate feedback loops (another amplifier) from that of the main amplifier. This not only creates much more difficulty in compensation but the cfp's themselves can amplify radio frequency "garbage" from the room independantly. Almost all home stereo is EF2 based.

as far as layout , I have colorized 3 examples of DIY amps. (RED = V+ / BLUE = V- / GREEN= ground /light green=signal ground / YELLOW=output)

As you can see by these top 2 boards (quasi and symasym) , while not bad (they do separate dirty zoble/output and signal ground) are far from perfect. The EF2 lets them get away with this. They also do not have the output close to the input circuit except where the NFB resistor comes back to the input transistors (LTP). 3rd board is mine , A "star looking star" with a second star where the voltage amp plugs in.

BTW , good transistor choices , ksa992/sc1845 all the mouser items.. cheap!
here is the "blameless" with EF's and your trannies , about the most used and abused design on "planet AB" :spin:
 

Attachments

  • quasiPCB.gif
    quasiPCB.gif
    41 KB · Views: 137
  • sympcb.gif
    sympcb.gif
    82.1 KB · Views: 107
  • PB60PCB.gif
    PB60PCB.gif
    50.1 KB · Views: 60
  • AX1.gif
    AX1.gif
    21.6 KB · Views: 68
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.