'AK Mods' SC Class A 15W Amplifier

[sean]

Still no joy - I have swapped the regulators for 3A types (1033/1085), bypassed the JLH boards, set the voltage rails to +/- 20.2V and bias current at just under 1A (240mV across 1R resistor), and the output still oscillates with the 4 Ohm Dynaudios - again, it's totally fine with an 8 Ohm load though. Is it possible that something is wrong with the amp board? I see no reason why the PSU would not be working properly, especially with these new regulators fitted.

[Deleted]

Hi Sean

Sorry to hear that you still have instability @ 4ohms which I find a little surprising.

What is strange is that the problem seems to be affecting both units.

I have home etched boards for my build of Alex's 15W class A so can't speak for Wills class A PA boards from personal experience.

However Jeff from Sunny Queensland has built with Will's boards and reported no such issues when used with his Panel speakers so I’m starting to get a little stumped as to what could be the cause.

In my experience ALL of the boards that I’ve had from Will have been top quality units which have been well thought out in terms of layout and production so I suspect that may not be the answer.

Out of interest what size reservoir caps are you using?

What are you scoping with the dummy load in place?

Can we have some measurements from around the PSU/AMP boards?

Some close up photos may be useful also if you can.

Let’s see what Alex thinks

Take care

[Deleted]

Apr 15, 2013 14:22:08 GMT sean said:

Still no joy - I have swapped the regulators for 3A types (1033/1085), bypassed the JLH boards, set the voltage rails to +/- 20.2V and bias current at just under 1A (240mV across 1R resistor), and the output still oscillates with the 4 Ohm Dynaudios - again, it's totally fine with an 8 Ohm load though. Is it possible that something is wrong with the amp board? I see no reason why the PSU would not be working properly, especially with these new regulators fitted.

Sean

This doesn't make much sense. I presume that it is higher level program content that causes this ? Remember too, that the original S.C. amplifier was never specified for 4 ohm use where it can only do 7.5W in Class A before changing to Class AB with it's much higher PSU current excursions. Some 4 ohm speakers will have impedance dips that may go as low as 2 ohms at some frequencies.
Before I changed to fully D.C. coupled, mine had no problems working into a friends big 4 ohm Ohm Walshe speakers. He no longer has those speakers for me to try with this version.
What value filter capacitors are you using ? With a 4 ohm load it would be an advantage to use closer to 10,000UF , especially as you have higher current capabilities with the 120VA toroidal transformers.
I think at this stage, with the JLH PCBs still bypassed, that you should unplug the offset corrector I.C. and fit an electrolytic cap of >100uF in the feedback network where the link is presently.
I would just cut the strap in the centre, lift the ends up and tack an electro across with the polarity as in the original design.
That should give you pretty well the original design which we are 100% certain did not have stability problems into 4 ohms.
BTW, I would have preferred a version of the original PSU design, but we had to go to 18-0-18 transformers due to non availability of the original 21-0-21 units that can still be obtained in Australia from Harbuch, or to special order. The PSUs now vary between individual constructors, with some using Will's PCBs, and others using data sheet type PSUs.
In order to assist the guys with possible problems here, I will be changing mine to Will's PSU PCBs, and when I can talk myself into going without the amp for a day or so, I will install the same main PCBs as everyone else.
Obviously, no one else who has these working into quite different speakers has problems.
It would be interesting to know what is the nominal impedance of the speakers the others are using.
What impedance are Dave's own speakers ? It is possible that the amp may be perfectly stable into his, although I understand your concerns..
Regards
Alex

[sean]

I know, I fully expected it to work after swapping the regs, bypassing the JLHs and dropping the voltage rails to 20V. The reservoir caps I am using are Nichicon KG 10,000uF 35V - easily up to the job, I use them all the time in other things that draw more current than this. The caps on the outputs of the regulators are 470uF 35V btw, the BOM specified 100-1000uF here, and I fitted electrolytic 10uF caps to the adjust pins as you suggested.

The cones in the Dynaudios start oscillating as soon as the speaker protection relay clicks in, even with no music playing through (tried no source connected and tried MP3 player on pause, no difference - originally I tried playing music via a preamp for the test, but only turned it up from zero a tiny bit before it started oscillating; I don't think source material is causing the oscillations, it seems to create them all on its own). The amplitude of oscillation is worryingly big, looks like quite a few volts worth when the cones move back and forth (15-20mm I reckon!). The oscillation seems to be to either side of 0V looking at the cone movement, it looks as if the amp itself is oscillating.

Dave does have 8 Ohm speakers I understand, but i'd still like to make sure that it works on as many speakers as possible. I'm not really happy to give them to him as they are, they might be OK initially but i'd be afraid that it migth start to oscillate again if he drives them a bit louder than normal. I'm not sure what the Dynaudios are like in terms of their impedance curve, they might well drop down to something quite low - I have used them in the past with quite a few amps, including class D stuff and a gainclone, so I didn't think that they were too hard to drive. I was under the impression that the amp much prefers 8 Ohm loads, but it can drive some 4 Ohm speakers so long as they're not too loud. I did wind the inductor with a value suited to 4 Ohm speakers, as per notes given somewhere in the depths of this thread.

If it helps, I can copy in the complete BOM that I used for the main amp board and the PSU. I don;t think the JLHs are at fault seiing as they are bypassed at present; to recap, the base resistor is 1k8 and the emitter resistor is 1R, as per your recommendations near where I first posted on this thread.

I will remove the DC offset IC and put the feedback cap back in tomorrow and see how it goes - i'm using anOPA134 with a cap of 3n3 fitted across the 1M resistor BTW. I'll get the scope on it as well, sorry not had time to do that today. I'll scope the power rails, the output, anything else? I can post up a pic of the internals too, now that it is all wired up. The signal input and mains cables are specifically routed as far from the DC rails as I can practically manage.

Thanks again for the help so far, I do feel like we're getting closer to the reason why it's oscillating. It might well be that my Dynaudio 50s are just too hard for it to drive and Dave's speakers will be fine forever on it, but i'd at least like to have an idea why it's happening.

[Deleted]

‘’It would be interesting to know what is the nominal impedance of the speakers the others are using.’’


PMC FB1+

‘’The stated frequency range of this two-way design is from 25khz down to flat response at 25hz and on down further to 22hz with a 3dB roll-off by that point. Let me say that again—the stated frequency range of this two-way design is from 25khz down to flat response at 25hz, and on down further to 22hz with a 3dB roll-off by that point. This is low for a two-way design! Efficiency is a specified 90dB, making them fairly tube friendly. Peak SPL@1 meter is stated as greater than 110dB, (and that's loud!). The FB1+'s are designated as an 8 ohm load, and the crossover is in the 3kHz neighborhood, employing a moderate 12dB/octave slope.’’

Taken from Positive feedback issue 25 written by Sasha Matson

So in short a fairly easy 90DB 1W@1M flat 8ohm transmission line speaker

Here is the full review should anyone be interested

www.positive-feedback.com/Issue25/pmc_fb1.htm

Take care

[Deleted]

Hi Sean

RE the Dynaudio 52’s

I had a look up on the net and they are quoted as nominally 4ohm but most times as 3-4ohm speakers. from what I have read they are a beast to drive.

Just about everything I’ve read suggests that they are a difficult load.

TBH it should not make much difference to the Class A and as Alex has pointed out it just reduces the amount of class A available.

But

Once things slip out of class A then the amp needs to work harder especially on transients

I’m not totally convinced that the above has much bearing but it could be worth thinking on.

I used 100uf caps on the reg OP. regs can be a bit picky about cap size on their OP.

I’ve just taken a very quick skim through the DATA sheet for the LM1085 and it Seems to suggest that 150uf is the max OP cap wise (same for the 317 if i remember rightly)and I have 100uf in that spot.

Could be worth checking the DATA sheet of the regs you have just to make sure on that.

what value did you wind the inducter to?

BTW I’ve used Alex’s class A with my Sonus Faber Concertino Home (The low 6ohm carbon fibre versions) without any issues. I used those for quite some time before switching to the PMC.

Just out of curiosity

What OP DC offset are you measuring?

Yup it’s just a question of trying to cover all of the bases and that’s really going to help others later on so well done for sticking with it.

We are almost there

I can feel it in my bones.

good work

Take care

[sean]

Hi Shaun, the speakers are actually Dynaudio 50, not 52 - they're extremely similar though AFAIK, same drive units and cabinet design. And more than likely, they're just as hard to drive as the newer 52's.

Re. regulator OP cap size, it is a bit bigger than the ones in the datasheet examples, but i've never had problems myself with a 470uF cap - when it gets above 2200uF or so then the 108x and 317 regs can get a bit picky about whether they will work or not, but in those cases they usually just go into shutdown mode and not supply any power at all, so I get the feeling that is not the problem here. But I'm running out of things to try, so I will look at those tomorrow as well and swap them for some 100uFs as a test. I won't be using low ESR types here BTW.

I wound the inductor with (I'm pretty sure) 18 1/2 turns of 1.0mm wire - they measured to be somewhere around 3.3uH, as that's what is noted on my BOM. I think Alex referred to this being a good choice to drive 4 Ohm as well as 8 Ohm speakers. The 2 inductors were fairly well mathced to each other too.

DC offset with no speakers connected is 18mV - not perfect but it's not bad either, it's certainly within acceptable limits.

Alex's suggestion of taking out the DC offset op-amp and adding in the 100uF feedback cap again sounds like it might be a good idea at this stage, maybe the DC offest correction is over-shooting and causing the massive low-frequency oscillation. The oscillations seem to build up very quickly from the second the speaker protection relay switches in, which is what you'd expect if some element of feedback is overcompensating somewhere.

Busy days tomorrow and Wednesday for me, but i'll try to get some more work done on the amps and report back when I get chance.

[Deleted]

DC offset with no speakers connected is 18mV - not perfect but it's not bad either, it's certainly within acceptable limits.

Sean
If that is with the Offset Corrector working, then you have some kind of common problem.
DC Offset in that case is typically within + - 1mV. This is what other constructors are achieving too.
Did you use 78L15 and 79L15 with the Offset Correctors? Some makes of 78L15 have documented failures due to reverse biasing at switch on.
Regards
Alex

[sean]

Ooh, sounds like we might be getting close now! I used L7815CV and L7915CV - quick test and they are giving +14.87 and -15.03 respectively. With the DC offset corrector in circuit, it's giving 18.5mV of offset. With the op-amp removed and a 100uF cap fitted in CFB, I get 2.4mV of offset. Hmmmm! Is it safe to test the amp with no DC offset correction on speakers?

[Deleted]

Apr 16, 2013 10:36:39 GMT sean said:

Ooh, sounds like we might be getting close now! I used L7815CV and L7915CV - quick test and they are giving +14.87 and -15.03 respectively. With the DC offset corrector in circuit, it's giving 18.5mV of offset. With the op-amp removed and a 100uF cap fitted in CFB, I get 2.4mV of offset. Hmmmm! Is it safe to test the amp with no DC offset correction on speakers?

Hi Sean
You appear to have a similar problem with both offset correctors.
Seeing you have a speaker protector fitted, and assuming it was tested O.K. , then you should be able to see how it goes. If all O.K. then you could let the JLHs back in too.
After you are sure all is O.K. then we can try to see what is wrong with the offset correctors.
I will repost a diagram of an offset corrector.
Regards
Alex

URL=http://imageshack.us/photo/my-images/40/akoffsetcorrectorrevise.jpg/][/URL]

[Deleted]

It would be interesting to know what is the nominal impedance of the speakers the others are using

Mine are Naim SBL;

Sensitivity 88dB/1W/1m
Impedance 6 ohm (min)
Power handling (music programme) 75W
Frequency response 30Hz - 20Hz plus or minus 3dB (in room)

Only I'm nowhere near built amps yet

I would have preferred a version of the original PSU design, but we had to go to 18-0-18 transformers due to non availability of the original 21-0-21 units that can still be obtained in Australia from Harbuch, or to special order.

Due to memory failure, again, I ended buying 22-0-22 also at an enormous 160va! Shouldn't run out steam there then

@ Sean,

Closing in now

[sean]

Thanks for the schematic Alex, i'll check it against my build. Just to clarify, the 22n cap in parallel with the 1M resistor is the right value? I ask because I fitted a 3n3 cap in parallel with the 1M resistor on the underside of the board as per instructions further back in the thread (page 37).

[Deleted]

Hi Chris
My DCM QED 1A TL speakers are 87dB/1W/1M
They are 8 ohms minimum.
Are you using just a single transformer ?
Regards
Alex

[Deleted]

Apr 16, 2013 11:54:30 GMT sean said:

Thanks for the schematic Alex, i'll check it against my build. Just to clarify, the 22n cap in parallel with the 1M resistor is the right value? I ask because I fitted a 3n3 cap in parallel with the 1M resistor on the underside of the board as per instructions further back in the thread (page 37).

OOPS !
I need to correct that diagram, Anything from 2.2nF to 4.7nF will be fine.
I also just noticed it said mA DC at the output instead of mV.

[sean]

YES! Results, of sorts - put the amp onto the 4 ohm speaker with the DC corrector bypassed and fired it up for a short time - played a bit of low volume music and sounded fine, no oscillations appeared. DC offset was 57mV, not ideal at all but not enough to kill anything (I tried it on the 8 ohm stunt speakers first). So looks like the problem is with the DC corrector section - I suspect it's how i've built it, some component value is probably wrong.

edit - yep, I spotted the mA instead of mV too but I knew what you meant Alex, thanks for the clarification.

edit2 - just tried again, feeding the amp from a Naim preamp instead of a cheap MP3 player, and offset was down to 10mV - looks like the MP3 player has a bit of offset itself, which is not usually a problem of course. Anyway, things look a bit better with the amp now.

[Deleted]

Apr 16, 2013 11:54:39 GMT @sandyk said:

Hi Chris
My DCM QED 1A TL speakers are 87dB/1W/1M
They are 8 ohms minimum.
Are you using just a single transformer ?
Regards
Alex

Hi Alex,

No, when completely done I'll have two stereo power amps running actively into the 'speakers, each stereo amp has twin transformers in each of the two PSUs

[Deleted]

Apr 16, 2013 12:08:18 GMT sean said:

YES! Results, of sorts - put the amp onto the 4 ohm speaker with the DC corrector bypassed and fired it up for a short time - played a bit of low volume music and sounded fine, no oscillations appeared. DC offset was 57mV, not ideal at all but not enough to kill anything (I tried it on the 8 ohm stunt speakers first). So looks like the problem is with the DC corrector section - I suspect it's how i've built it, some component value is probably wrong.

edit - yep, I spotted the mA instead of mV too but I knew what you meant Alex, thanks for the clarification.

Sean
At least nothing is blown up. A few of us learned the hard way with those little 78L15.
If I were to hazard a guess, I would check that the 75K isn't 7.5K.
Then again, you might just be unlucky with your particular amp and corrector combination.
It would be a first, though.
Regards
Alex

P.S.
It's a shame you didn't mention earlier that rather high (for an offset corrector) D.C. Out, due to our previous experiences there.

[Deleted]

Apr 16, 2013 12:13:50 GMT @cjarchez said:

Apr 16, 2013 11:54:39 GMT @sandyk said:

Hi Chris
My DCM QED 1A TL speakers are 87dB/1W/1M
They are 8 ohms minimum.
Are you using just a single transformer ?
Regards
Alex

Hi Alex,

No, when completely done I'll have two stereo power amps running actively into the 'speakers, each stereo amp has twin transformers in each of the two PSUs

Have you recently jacked up the prices in your restaurant to pay for all that ?

[Deleted]

edit2 - just tried again, feeding the amp from a Naim preamp instead of a cheap MP3 player, and offset was down to 10mV - looks like the MP3 player has a bit of offset itself, which is not usually a problem of course. Anyway, things look a bit better with the amp now.

Hi Sean
A working offset corrector would have taken care of most of that too .
Regards
Alex

[Deleted]

Apr 16, 2013 12:20:05 GMT @sandyk said:

Have you recently jacked up the prices in your restaurant to pay for all that ?

That would be nice

It has been another reason for the slower progress, play time funds have been a bit trickle-feed on the off season. Summer's a-comin' so hopefully I'll get this beastie pumping tunes sooner rather than later.

[sean]

Apr 16, 2013 12:18:31 GMT @sandyk said:

Apr 16, 2013 12:08:18 GMT sean said:

YES! Results, of sorts - put the amp onto the 4 ohm speaker with the DC corrector bypassed and fired it up for a short time - played a bit of low volume music and sounded fine, no oscillations appeared. DC offset was 57mV, not ideal at all but not enough to kill anything (I tried it on the 8 ohm stunt speakers first). So looks like the problem is with the DC corrector section - I suspect it's how i've built it, some component value is probably wrong.

edit - yep, I spotted the mA instead of mV too but I knew what you meant Alex, thanks for the clarification.

Sean
At least nothing is blown up. A few of us learned the hard way with those little 78L15.
If I were to hazard a guess, I would check that the 75K isn't 7.5K.
Then again, you might just be unlucky with your particular amp and corrector combination.
It would be a first, though.
Regards
Alex

P.S.
It's a shame you didn't mention earlier that rather high (for an offset corrector) D.C. Out, due to our previous experiences there.

I've checked the component values in the schematic for the corrector against the amp PCBs and also against the PCB layout here...

i100.photobucket.com/albums/m29/Fwapp/PAmp.png

..and all the components seem fine (the 3n3 is definitely across the 1M and not the 4k7 too). Not checked the rest of the amp though but that seems to work OK. Do you think maybe the op amp could benefit from some 100n caps adding on its supply rails? It seems strange that it only oscillates with the lower impedance speakers. Alex, what did you mean by "unlucky with your particular amp and corrector combination" - do you mean that in my case the corrector might need to be modified somehow? Is it possible that something that i've done wrong in the amp is affecting the corrector? Although the amp works perfectly fine with the corrector out of circuit...

HOWEVER, I have spotted on the amp PCB that the feed to the offset corrector is taken from the common pad of the eight 1R resistors - according to the amp schematic, that isn't right, the feed to the corrector should be taken from the output. Is that likely to be an issue here? The amp schematic I refer to is here:

img13.imageshack.us/img13/8766/15wclassaamplifiernew20.jpg

[Deleted]

Hi Sean
I can't even hazard a guess at the moment why this is happening. This offset corrector is the same as used in all the previous HA/preamps and 15W Class As that have already been built, without any problems. Have you measured the offset with the Corrector in circuit when using say a 3.9 or 4.7 ohm 5W resistor as a dummy load ?
What is the output voltage of the Offset Correctors at the output pins of the OPA134? My original DC coupled 15W Class A has about -60mV at that point.
I would be surprised if the problem is caused by the take off point for the offset corrector coming from the other side of the Zobel network. I have yet to finish building up Will's PCBs for this version,
so I am unfortunately not much help there at present, but others such as Jeff C. have had no problems .
Regards
Alex

P.S.
When measuring the DC out with the Offset Correctors back in circuit, are the inputs terminated with something like 1K resistors, or a preamp connected as before ?
The Offset Corrector is likely to hunt without the inputs having some kind of termination.

[jeffc]

However Jeff from Sunny Queensland has built with Will's boards and reported no such issues when used with his Panel speakers so I’m starting to get a little stumped as to what could be the cause.

so I am unfortunately not much help there at present, but others such as Jeff C. have had no problems

Hi Sean, Alex, Shaun,

Sorry that this oscillation problem is proving so difficult to pinpoint. Only really frustrating problems I had when finally all was connected were with silly transistor misplacements in the speaker protector and having to change resistor values to try and get the front-end balancing close to 0 mV. My DC offset correctors had, from memory, around 18 mV offset with no speakers connected, but this dropped nicely to <0.5 mV when connected to a sacrificial speaker driver pulled from an old TV. Based on your descriptions Sean, your build is close to identical to mine with an 18.5 turn - 1 mm wire inductor, OPA134 corrector opamps, 2n2 across the 1M feedback resistor, LM7815/LM7915 corrector regs, 10,000 uf/50V PSU supply caps, LT1033/LT1085 PSU regs, 100 uf PSU output caps, 1K8 and 1R in the JLH current limiters, etc etc.....

Why have I been spared? Not sure, but the impedance presented by my DML panels is at best a guess. With 4 x 4 ohm exciters connected parallel-series-parallel, with 3 x cheap piezo tweeter cones connected in parallel, one between each pair, I’ve just assumed it to be nominally around 4 ohm. DML panels are quite unusual though in that their impedance doesn’t spike at driver Fs like normal speakers and tends to be somewhat flatter across a broader frequency range before increasing at higher frequencies, with small spikes occuring at more prominent panel resonance frequecies dictated by its structural proprieties. So they're not that ideal for comparing impedance effects to conventional speakers.

I haven’t noticed the panels/exciters going into oscillation or doing anything else spastic when connected to the Class A amp. However, for various non-technical reasons I've been driving them lately with my tweaked Class D amplifier, seeing I had some time to tinker with it while testing the Class A.

Anyway, with the Class A amp sitting idle, I’ll challenge it with whatever 4 ohm (or slightly lower) conventional speakers I can cobble together to see whether my build might also have this oscillation issue.

cheers.. jeffc

[Deleted]

Hi Jeff
Seeing you like your music a little more "feral" sounding, you could always sell your Class A to Phil at cost price !
Kind Regards
Alex

[Deleted]

Sean
Are you able to confirm that with the feedback capacitor bypassed, speakers disconnected, and a 1K resistor across the input terminals,
that DC out is <50mV ? (2mV of previous x gain of 16 ?)
If that's O.K. it could be helpful to measure the DC voltage at pin 6 of the I.C.s with the 1K input resistors and something like a 4.7 ohm load.

Regards
Alex