Amplifier Design - A Documentation

[Bunkey]

I don't think there's much you could learn from me tbh Iain, even less I haven't already posted here.

I'm glad you like it.

I've no frame of reference by which to compare so I don't know what good is. I'm pretty sure it's not finished though.

[Bunkey]

Oh hello... me again... Seeking some advice...


I feel an update is overdue on this but I've just been playing around with different ideas really and not much has changed; the search for better tends to loop back to where I started more often than not.
Still, the amp is progressing and I'm ever closer to committing it to a 'finished' form.

There are some problems which need to be solved first-


I'm having a bit of an issue with grounding/oscillations and unwanted global feedback. I've pondered this for a few weeks now and wanted to put it out here to see if anyone can offer some input or identify a problem/solution.

This is directly related to master volume. The amp sounds killer at low levels but the higher the output the more signal seems to be finding its way back into the pre-amp, either affecting the tone & sustain of the signal as negative feedback, or at times becoming additive and causing an almighty shriek depending on the fundamental frequency of the noise floor and oscillation.

It was first apparent as NFB - I'd lose sustain, fuzz and the harmonic content the higher the output volume - but the amp was quite useable at low levels, such as my previous video. The bass has always been quite ragged - It produces the low frequencies but distorts unpleasantly with any significant gain.

Most recently I've experimented with adding a dual triode tube stage between the original pre-amp and power amp, to bring some second-order harmonic warmth to the otherwise mid-scooped and slightly thin sound from the preceding stages.
There are a few other issues with this tube stage (such as the separate SMPS heater hum), that I'll ignore for the time being - What it has done is highlight the NFB and oscillation problems at a level and frequency that's useful to demonstrate. The prominent hum has become the fundamental frequency of the oscillation, which is much more palatable than the piercing shrill of additive BJT noise!

Some key points/observations:

The zobel network on the LM380 output; I'm using a 10uF cap with a 4.2 ohm resistor. This combination sounds good and balanced but is obviously letting a lot of audible frequencies to ground. Increasing the value of R to 10 ohm increases the signal at the speaker and reduces slightly the NFB acting on the preamp, but sounds harder so isn't favourable. It doesn't do much to curb the spontaneous oscillation either.

I was measuring a pretty significant audio signal on pre-amp ground a while ago, likely coming from this zobel. I've since separated the pre-amp & power-amp grounds to star at the regulator (reference) ground - this reduced the AC component in the pre-amp ground <10x, however it's obviously still present to some degree.

The intermediate tube stage is fed and grounded pre-regulator to make use of the full 25v (18v loaded) supply.

All stages have 10uf local decoupling (LM380 additional 100nf across supply pins)

The pre-amp features a pair of Ge clipping diodes to ground, preceding an emitter follower, so this is a likely candidate for where the feedback signal is coupling.

The supply itself is a double insulated 18v DC wall wart, so there is no solid earth reference in the amplifier (no chassis ground) - just the +ve and -ve input from the DC supply.

This issue persists after removing the 10r dropping resistors between reservoir and amplifier ground.

See below for a block diagram of the layout. I'm afraid I don't want to publish the full schematic online at this point but I've tried to be as informative as necessary - it's all very simple.



Hypothesis:

Assuming there is nothing wrong with the ground scheme, I'm wondering if these problems could be due to a lack of reservoir capacitance. Given there is no true earth anchoring the amplifier ground at a fixed point and less capacitance soaks up less AC component; the signal being dumped to ground from the power amp output zobel remains in part on the -ve rail (acting ground in this case) as AC ripple and gets coupled back into the pre-amp via the conducting clipping diodes causing the NFB..? A lack of reservoir capacitance might also account for the bass distortion and hard clipping of large input transients too, if my experience with leaky HiFi electrolytics is anything to go by.

This is speculation. I require some asstance.

The thing sounds quite good otherwise, honest!



PS: I should point out, the amp sounds far more 'passable' without the dual triode - as in the video clip at the bottom of the previous page - It still has the same issues though, they just appear later in the volume travel as there isn't the additional gain on the pre-amp signal. NFB killing the tone/sustain is the main offender here. I'll maybe get another clip of this comparing low & high volumes without the tube stage, if need be.

[iainpunk]

sounds to me like power line wiggle or ground resistance/ground differentials.
three quick fixes that might be easy to try is splitting the power supply lines like you did with the grounds,
adding bigger reservoir caps,
adding a decoupling cap at every gain stage, just to keep it stable.
another possible but controversial thing to try is coupling the power amp and pre-amp grounds like one continuous ground path. it might case ground loops tho, so that's a possible negative side effect.

[OT]

i really like your cabinet by the way, it looks really nice, my small diy guitar amp looks a lot more janky, being a cigar box with a big black heat sink sporting a good loking TO-3 transistor, a huge DC choke and a modern industrial looking switching power supply mounted on top. it really looks quite dangerous, and i'd advice everyone to not touch the choke's exposed contacts when sound is comming out of the speakers, it kind of zaps you like one of those zapping toys on steroids. it doesn't really hurt, but it does zap.
mine is almost infinitely less complex than yours, having only 3 discrete transistors, of which one is a Bazz Fuss circuit, acting like the pre-amp. fuzz all the time baby!
the other two are that DTS-410 and a luxury BD139 ** in a darlington configuration.
> that;s not really true, i have no idea what electronics are in the power supply, i bought a affordable 60w 12v DC supply block. its way less whiny than i anticipated, i originally put in 10.000uF 25v caps, and there is no power supply whine at all, when running the amp's 4.5A of quiescent current, it doesn't get hot at all!

[/OT]

cheers

* i chose it for its looks, its a DELCO DTS-410, and the red lettering and the font and the shiny case, i really like its looks.

** its gangster looking. gold plated legs and heat-sink plate in a grey plastic body.

[Rob Strand]

Try connecting the ground side of the cap on the - input of the LM380 to the quite ground (light Blue).   However it is very wise to place a 1k resistor from the ground side of that cap to the noisy blue line to prevent the amp going nuts when you disconnect the "new" wire that goes to the quite ground.

Actually might not need that 1k the LM380's internal resistors might be enough to handle the disconnection case.

[Bunkey]

Try connecting the ground side of the cap on the - input of the LM380 to the quite ground (light Blue).   However it is very wise to place a 1k resistor from the ground side of that cap to the noisy blue line to prevent the amp going nuts when you disconnect the "new" wire that goes to the quite ground.

Actually might not need that 1k the LM380's internal resistors might be enough to handle the disconnection case.

Well spotted!

Definitely performing better. It didn't solve the oscillation but it has improved the bandwidth/clarity significantly - I have pinch harmonics again! Man I thought I'd just got sloppy 
Thanks that's a great quality improvement.

AFAIK the unused input on the LM380 can be left open if you like to add a bit of radio broadcast to your tone; or straight grounded but doing so causes it to clip asymmetrically with the slight bias offset, so capacitively coupling allows a little more headroom. Could be a useful characteristic to know if one ever felt compelled to overdrive the chip!

my small diy guitar amp looks a lot more janky, being a cigar box with a big black heat sink sporting a good loking TO-3 transistor, a huge DC choke and a modern industrial looking switching power supply mounted on top. it really looks quite dangerous

* i chose it for its looks, its a DELCO DTS-410, and the red lettering and the font and the shiny case, i really like its looks.

** its gangster looking. gold plated legs and heat-sink plate in a grey plastic body.

Thanks Iain. I love the idea of this thing. 4.5A is pretty nuts though 
The ebay img seemed to get nerfed on my browser btw so I've redone the link

I actually managed to salvage a big old 10,000uF LCR elecrolytic from the A60 (the other had spilled its contents across the board); I'll dig this out and and give it a try if it tests good.

Do you post any of your builds?

[iainpunk]

I love the sound of this thing.

that's great to hear!

4.5A is pretty nuts though

yes, it is, but its class A, and provides 12-ish watts into 16 ohm, before break-up
EDIT: it should have been 1.5A, the 4.5A quiescent current was with another inductor, which only had 1.25 Ohm, but only 2H, this one has 30H and 3.8 Ohm. i was a bit troubled with the temps the choke reached, burning almost 30W of heat... with the new choke, i have higher efficiency and more volume level with my 16 ohm cabinet, but a lower volume from my 4 ohm cabinet, so its a bit of a trade off.

Do you post any of your builds?

rarely, i rather focus on new and unfinished ideas when it comes to posts rather than things from the (recent) past.
the rest is helping people here.
another reason is that my taste in circuit design and guitar tones is quite... extravagant. i love lofi sounds and "trick-based" circuit design. most people seem to like more conventional circuit design and sounds.
EDIT: i love the designs i'm working on as long as i am in the process, i keep about 1 in 20 designs in the box they are in, the rest goes on to the ''i dont like it anymore'' pile and i re-use the box

cheers

[Rob Strand]

Definitely performing better. It didn't solve the oscillation but it has improved the bandwidth/clarity significantly

Maybe you should add the Zobel Network recommended in the datasheet in parallel with the network you have.   Worth a try.

Something to try, at least temporarily, is a 100uF cap soldered right on the power pins of the LM380.   A few years back Paul Marossy had no end of problems with modern LM386's (not LM380) oscillating  and that solved it.   The design worked fine without the caps on older chips.

[iainpunk]

i believe he already has Zobel-Network on there, looking at his earlier schematic.
i do echo your suggestion of the Zobel-network to anyone else reading this and planning on building an amp in the future!
it did solve some ringing-when-clipping problems in my sketchy DIY amp.

i think i want to build a configurable-zobel-box. basically a 12 position switch for capacitors and a vintage ''notched'' rheostat. would be a nice tool to try out zobel network values on future diy amp builds. if you want to build multiple amps.

cheers

[Rob Strand]

i believe he already has Zobel-Network on there, looking at his earlier schematic.

Yes, but he's got a large-ish cap which was chosen based on the sound.  Maybe the large cap isn't working so well at high frequencies. That's why I suggested putting the network from the datasheet in parallel.   Really, I'm not expecting it to fix it but when things act weird is pays to try the "right" things.

[Bunkey]

Definitely performing better. It didn't solve the oscillation but it has improved the bandwidth/clarity significantly

Maybe you should add the Zobel Network recommended in the datasheet in parallel with the network you have.   Worth a try.

Something to try, at least temporarily, is a 100uF cap soldered right on the power pins of the LM380.   A few years back Paul Marossy had no end of problems with modern LM386's (not LM380) oscillating  and that solved it.   The design worked fine without the caps on older chips.

Ah yes I'm familiar with the thread. Well I have a 100nF 'right across the pins' with a 10uF in very close proximity - a few mm's - swapping this to a 100uF has no useful effect other than changing the tonal emphasis of the amp.

This is very much an 'additive feedback being coupled into the pre-amp' issue, as opposed to instability of the LM380 itself.

Likewise changing the zobel for the recommended value (2.7ohm / 0.1uf) doesn't fix this. I've tried a lot of values here, again more for their tone shaping, and arrived at the 10uF based on calculations for what actually gives the flattest impedance curve with the speaker I'm using.


With the better grounding on the -Ve input and the 10,000uF reservoir things are moving in the right direction anyway. Maybe a 22,000uf would be an improvement still.

These values look huge compared to whats generally accepted in terms of filtering though - Why the supply & AC signal ground ripple has to be so miniscule I think is a product of the less-than-ideal DC power source with the floating ground.
The solution right now seems to be to limit the pre-amp gain so that risidual ripple isn't amplified beyond the threshold that's going to couple back into the early stages. With the inclusion of that tube stage before the power amp, the total gain far exceeds the threshold; so a tiny bit of risidual audio ripple at the pre-amp ground (caused by the constant heater hum) is turning into the additive feedback you can hear in that clip.

I feel if there were a way to better isolate the audio clipping ground in the pre-amp from the output stage and supply grounds then I might be able to get away with more gain, which it kind of needs.

i love the designs i'm working on as long as i am in the process, i keep about 1 in 20 designs in the box they are in, the rest goes on to the ''i dont like it anymore'' pile and i re-use the box

'Perfectionism' This is exactly why I still haven't boxed this thing up!

I'll not bother with the firewood permit this year - Just employ one of your amp circuits instead and it can heat my house 

[Rob Strand]

I feel if there were a way to better isolate the audio clipping ground in the pre-amp from the output stage and supply grounds then I might be able to get away with more gain, which it kind of needs.

One possibility is the preamp is oscillating or interacting with power amp to cause oscillation.   A few possibilities:
- Interaction through the power rail: An RC filter on the power to the preamp.    Something like 100 ohms and 100uF.
- If the preamp is interacting with the power due to wiring then adding cap to roll-off frequencies outside of audio is helpful. 
   Such caps can help grounding related oscillations as well.
- If the preamp alone is oscillating then that's a whole different story.   That's a design issue.  Normally fixed by careful placement
  and choice of cap values on the preamp.  However,  you would only expect this if the preamp has a feedback loop.

[Bunkey]

One possibility is the preamp is oscillating or interacting with power amp to cause oscillation.   A few possibilities:
- Interaction through the power rail: An RC filter on the power to the preamp.    Something like 100 ohms and 100uF.
- If the preamp is interacting with the power due to wiring then adding cap to roll-off frequencies outside of audio is helpful. 
   Such caps can help grounding related oscillations as well.
- If the preamp alone is oscillating then that's a whole different story.   That's a design issue.  Normally fixed by careful placement
  and choice of cap values on the preamp.  However,  you would only expect this if the preamp has a feedback loop.

Every gain stage has its own 10uF decoupling cap (4 in the pre-amp) and there's a 100nf wima to ground located between pre and power for RF filtering (not sure how effective this is). I did audition an RC filter between stages at one point but it didn't sound good. There's every possibility that actually some of the interaction is softening things up and benefitting the sound but there comes a point where it's doing more harm than good.

I'll put together another clip just now without the tube stage comparing low and high volumes so you can hear what I mean with the NFB issue. I don't really get oscillation without that extra stage - just NFB - It's like the ground is acting as that global feedback loop.

[Bunkey]

Actually it's 3am and I just realised how frigging loud I'm being 

I'm probably asking a bit too much for you to listen to a clip and troubleshoot my amp based on the sound you hear too - this is wishful thinking bordering on procrastination tbh.

I need to set it all out on a fresh breadboard as I intend to build it, using shorter runs of better jumper wire, and take it from there.

I really appreciate your help though Rob, thank you.

[iainpunk]

i love the designs i'm working on as long as i am in the process, i keep about 1 in 20 designs in the box they are in, the rest goes on to the ''i dont like it anymore'' pile and i re-use the box

'Perfectionism' This is exactly why I still haven't boxed this thing up!

I'll not bother with the firewood permit this year - Just employ one of your amp circuits instead and it can heat my house 

its not perfectionism, its more the effect of a short 'attention interest span'.
i might be designing a 100W amp soon, with about 10 to 15% efficiency, basically a room heater with speaker driving capabilities. ill send you the design files when im done drawing them!

have you tried isolating you pre amp like a pedal, and send it to a amp you know works?
this demonstrates if the "feedback" is in the pre amp or between the pre and power amp.

cheers, Iain

[Bunkey]

have you tried isolating you pre amp like a pedal, and send it to a amp you know works?
this demonstrates if the "feedback" is in the pre amp or between the pre and power amp.

Yeah the pre-amp doesn't oscillate plugged direct into the usb interface with headphone monitoring.
This is litereally just the power amplifier output coupling back into the pre-amp in a feedback loop, likely via ground.
This loop is quite stable too, seems to have a ceiling probably at the clipping diodes, so the volume doesn't run away on itself...

I just replaced the tube stage with a single fetzer of low gain and got an entirely different flavour of feedback, based on the harmonic spectrum generated by the fetzer valve. I can see the spectrum in ableton's EQ display, which is probably entirely useless information but pretty cool anyway.

[Bunkey]

[iainpunk]

that oscillation sounds freaggin amazing in that context, now i want a oscillating fuzz too!!

[Bunkey]

That was actually two completely unrelated tracks I accidentally triggered together then mixed to taste 

The non-oscillating rhythm part (if you can call it that) is an example of the amp loud without the extra stage - you can hear it's got some pretty hard clipping going on, not very smooth at all - a bit broken tbh - I can't accept that, so that's what I'm trying to problem solve here.

The oscillating with the extra stage is just... yeah  it's a bit nuts. Could definitely use the added wamth and body from either JFETS or tubes though so I guess I've got some head scratching to do here.

[Rob Strand]

Are the sockets for the input jack and the speaker jack (I'm assuming you have one) attached to the same metal plate so the grounds are looping back the through the sockets?    That could cause problems.  From your pics it looks like the speaker is connected directly to the power amp.

On a different line of thinking you could be getting some sort of motorboating.   When the amp is driven hard it pulls the supply rail down due to the regulator current limit.   That then upsets the power to the preamp and it starts motorboating.

A test would be to put a diode in the positive rail between the regulator and the preamp.   You might need to put a 100uF cap on the preamp power rails as well - that's on the preamp side of the diode.   The existing 10uF's might work they might not.

[marcelomd]

That actually sounds great. Would be right at home in an indie record, or movie soundtrack, with a bit of production.