Amplifier Design - A Documentation

Started by Bunkey, February 20, 2021, 11:10:28 PM

Previous topic - Next topic

Bunkey

#80
Quote from: marcelomd on April 19, 2021, 08:39:27 PM
That actually sounds great. Would be right at home in an indie record, or movie soundtrack, with a bit of production.
Heatstroked descent into delirious jazz abstractation; the sound of extracurricular brain melt, with fuzz
- I'm not sure the genre is widely recognised.
...just riffing.

Bunkey

#81
Quote from: Rob Strand on April 19, 2021, 07:16:41 PM
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.

What would you expect this to do Rob?

I imagined nothing (all things being well) besides dropping the diode fv from the pre-amp supply but adding it as shown caused the amp to oscillate independent of the volume control and made an almighty racket until the power was cut.

...just riffing.

iainpunk

#82
split power lines like ground, and add some extra filtering to the pre amp's Vcc


hope this helps a bit,
cheers.

edit/ot:
Quotei 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!
the 100 W class A transistor amp is going to be a bridged configuration.


cheers
friendly reminder: all holes are positive and have negative weight, despite not being there.

cheers

Rob Strand

#83
QuoteWhat would you expect this to do Rob?

I imagined nothing (all things being well) besides dropping the diode fv from the pre-amp supply but adding it as shown caused the amp to oscillate independent of the volume control and made an almighty racket until the power was cut.
You might need to add the 100uF across the power rail on the preamp side.   You can also change the diode to a 100ohm resistor (no diode at all).   The resistor is better as you can keep increasing the value to reduce the preamp supply changes.

The idea is to prevent or at least reduce motorboating.

Suppose you have a preamp with a 12V supply and the transistor collectors a biased mid-supply to 6V.  Now you change the supply to 10V.   The collectors will bias to a lower voltage say 5V (just for arguments sake).  Now imagine continually and rapidly adjusting the power back and forth between 10V and 12V  the change in the bias point on the transistor collectors looks like signal.   That signal will be amplified by the power amp.

When the power amp clips the regulator current-limits and that causes the output voltage of the regulator to drop.    The drop in supply changes the biasing of the transistor preamp.   That drop in supply looks like a signal.  The power amp amplifies the signal but the signal causes the supply to drop or rises.   So you end up with a feedback loop where the power amp feeds back signal to the preamp via the power supply and it oscillates.

By putting a diode and cap, or a resistor and cap on the power supply the power supply changes are reduced.  Hopefully to the point where it cannot sustain oscillations.

The ultimate decoupling of the power supplies is to have a separate regulator for the preamp.    *But*  that's only true if the input supply voltage does not drop too low when the power amp clips.   If the input power rail drops too much the preamp regulator will not regulate.   You will get back the same problem you had before where the preamp power rail is dropping because of what is happening at the power amp.

Anyway, the fact you are seeing a change in behaviour is a good indication we are on the right track.

Send:     . .- .-. - .... / - --- / --. --- .-. -

Bunkey

#84
Quote from: Rob Strand on April 20, 2021, 06:56:50 PM
Suppose you have a preamp with a 12V supply...

Thank you, this is immensely helpful.

Google images was far less adept at helping me visualise the concept...

It makes a bunch of sense and goes some way to explain how the amp sounds the way it does - When it's in the right window it has this really nice juicy spongy quality that I never imagined getting from a solid state amp; I'm gonna have to see about introducing this back in a little more intentionally once I've stabilised the supply.

I guess the sag got a bit too baggy in this case.


Quote from: iainpunk on April 20, 2021, 04:47:36 PM
the 100 W class A transistor amp is going to be a bridged configuration.


Love the idea Iain. Some good articles on that site too.
I'd like to partner this pre-amp with a discreet power amp at some point - maybe about 1/10th of the power! - Have some TIP3055's here but not quite sure how to use them yet. Keep me posted.
...just riffing.

Rob Strand

#85
QuoteThank you, this is immensely helpful.

Google images was far less adept at helping me visualise the concept...

It makes a bunch of sense and goes some way to explain how the amp sounds the way it does - When it's in the right window it has this really nice juicy spongy quality that I never imagined getting from a solid state amp; I'm gonna have to see about introducing this back in a little more intentionally once I've stabilised the supply.

I guess the sag got a bit too baggy in this case.

I guess the risk is fixing the motorboating may affect the sound.

A nasty trick occurred to me.     The LM380 has a + input and a - input.   You are currently grounding the - input and putting the signal on the + input.   That's the common method.   So the trick is to drive the - input with the signal and ground the + input (via the input) cap.   Basically swapping the roles of the + and - inputs.    The idea is it reverses the phase relationship between the feedback on the power supply and the preamp.  Maybe just maybe the reversal will turn the positive feedback into a negative feedback.   It might just be enough.     I'm not 100% in favour of doing this.    It would be *much* better to clean-up the supply to the preamp.    However, it might be something which can be used to *if* fixing the preamp supply affects the tone.  See how you go with the supply first.

Something else I noticed is the LM380 has a quiet ground pin which is separate from the noisy ground pin(s) that are used for the output stage.   Technically the LM380 quiet ground should connect to you light-blue quiet ground.   The quiet ground pin number depends on which  LM380 package you have.    If you split the grounds you have to be careful that both grounds are connecting at all times.   If one is disconnected and the other powered is could blow up the chip.  A safety precaution on the breadboard might be to add a 10 ohm resistor between the noisy and quiet grounds near the LM380.
Send:     . .- .-. - .... / - --- / --. --- .-. -

Bunkey

#86
Quote from: Rob Strand on April 21, 2021, 03:02:22 AM
I guess the risk is fixing the motorboating may affect the sound.

A nasty trick occurred to me.     The LM380 has a + input and a - input.   You are currently grounding the - input and putting the signal on the + input.   That's the common method.   So the trick is to drive the - input with the signal and ground the + input (via the input) cap.   Basically swapping the roles of the + and - inputs.

Most certainly! At least now we know how it's interacting, I can figure out how to recreate that in a controlled way - well, attempt to anyway. I have a few ideas so I'm gonna try and muddle through this one myself. No spoilers please  :icon_lol:

I wondered about phase inversion at the opamp input too but the single inverting fetzer stage and the tube stage (using both inverting triodes) both induced feedback at the same sort of volume with their opposite outputs, albeit at different frequenies and slightly different in behaviour. There are two inverting stages in the pre-amp doing different things so I guess these are both reacting in opposite phase to each other and combining in their own way - swapping the phase of the oscillation would be like reversing the order of two pedals in series for a slightly different effect, perhaps.

I will take you up on the quiet ground pin as I had it grounded like that originally then changed it to the drawn schematic to totally distance the chip from pre and rule out possibilities. I shall change that back. All the quiet pins are arranged on one side of the 8-DIP; with the power, bypass and output stage on the other - which is quite handy.
...just riffing.

Bunkey




It's funny looking back through this thread. A lot of energy, limited knowledge. A little embarrassing at times.
I've learned a lot just over the course of this project and the amp has come a long way since the beginning of the year - It's been more or less a full time endeavour.
Ironically, the more it grows the less detail I want to publish online - I have stabilised the amplifier though; I actually used zeners to decouple the preamp and then incorporated an element of feedback to get a similar effect. Thanks again Rob for explaining what was happening,

The clip above still uses the LM380 power stage and I'm pleased to say the amplifier as a whole is stable and relatively consistent through the volume range now.


Since the amp has come this far, and actually sounds good, I've been inspired to build a discrete power amp section instead of the chip before I begin working on the enclosure - thank you Iain.

I've been doing a lot of studying over the past month to this end and have a basic topology in mind using a CFB input stage and quasi-complementary output.
All the reference material approaches the design process with an output in mind, works out the power requirements and then builds a supply around that..

I need to do things the other way round

I'm limited to the power supply I'm already using, partly because it's all that's readily available and partly after picking up a few pearls of R.G. wisdom from the infamous Banana Amp thread..
The max voltage is 25v which drops to 18v at full load, unfortunately the max current of these wall-warts is a less-than-ideal 600mA but I've thus far not found a higher-rated alternative.

...This means I need to design my discrete power amplifier around the limitations of this 'safe' power supply and include some sort of current limiting in the design - I've not yet worked out how to do this. I guess 5W output would be achievable in class B though and tbh I'd be fine with anything around there down the 2W the LM380 is currently putting out.
I've found this to be a great output level for recording paired with this 6" speaker cabinet and hopefully the sample above demonstrates that.

Anyway, that's where we're at. I'll get a power amp knocked up on the breadboard once I've figured out some suitable component values.
Any word on current limiting the output stage would be appreciated  :)

Thanks
...just riffing.

kaycee

The fuzzy sound reminds me of Sungrazer, which is no bad thing to my ears. It's also woolly, like a low gain fuzz face, also nice. Good work.

iainpunk

that sounds great so far! i love the resonant character of the light crunch.

QuoteIronically, the more it grows the less detail I want to publish online
i know that feeling.

QuoteI've been inspired to build a discrete power amp section instead of the chip before I begin working on the enclosure - thank you Iain.
you're welcome!

Quotehave a basic topology in mind using a CFB input stage and quasi-complementary output.
nice, you made me really curious now, do i have to think 'zen amp' or 'death of zen' style?

QuoteI'm limited to the power supply I'm already using, partly because it's all that's readily available and partly after picking up a few pearls of R.G. wisdom from the infamous Banana Amp thread..
The max voltage is 25v which drops to 18v at full load, unfortunately the max current of these wall-warts is a less-than-ideal 600mA but I've thus far not found a higher-rated alternative.
have you looked in to laptop power supplies/chargers? they are often 19v and 2 or more ampere.

QuoteAnyway, that's where we're at. I'll get a power amp knocked up on the breadboard once I've figured out some suitable component values.
Any word on current limiting the output stage would be appreciated
LM317 based current limiter.
> page 11 Figure 9
https://www.st.com/resource/en/datasheet/lm317.pdf
the formula for the current limiter is 1.25/Rset=Imax
1.25 ohm give you a 1A current limiter.
keep in mind the Power rating of that resistor!

cheers
friendly reminder: all holes are positive and have negative weight, despite not being there.

cheers

Bunkey

#90
Quote from: kaycee on May 15, 2021, 04:26:15 PM
The fuzzy sound reminds me of Sungrazer, which is no bad thing to my ears. It's also woolly, like a low gain fuzz face, also nice. Good work.

Hey thanks and thanks again for the new music. Enjoying their self-titled album right now. I love that acid-tinged Kyuss sound.

The goal is sparkly warm clean that breaks into a pretty crunch; somewhere between reedy jazz, desert rock and 60's psychadelic; this is at the upper end of the amp's gain. I love electronic too so there's an element of that; the harmonic percolator thing can be quite synthy.
If you put a boost in front of this you get a similar thick fuzz - The idea is that kinda 'not high gain but still heavy sounding' textured thing that can be layered onto with pedals. The low-end to the eq is probably responsible for the wool and heft. A bigger speaker might be an interesting proposition.
...just riffing.

Bunkey

#91
Quote from: iainpunk on May 15, 2021, 04:39:57 PM

nice, you made me really curious now, do i have to think 'zen amp' or 'death of zen' style?

have you looked in to laptop power supplies/chargers? they are often 19v and 2 or more ampere.

cheers

Very similar yes but the drivers are an NPN & PNP baxandall compensated pair, with a preceding VAS - so one extra BJT.

That Death of Zen might be a good starting point just to get my eye in actually.

I've a few printer supplies here from 15v to 25v but they all tend to be switch-mode, which is a big nope from my experimentation. I'd use an on-board toroidal if I could - and do have one lined up - but after reading about the stringent regulations for anything carrying more than 30v, I'm inclined to get the most out of a low-power linear wall wart; at least for this design concept which will make sense when you see it.

I'll test that regulator setup in the supply line first. The other route is just emitter resistance on the output trannies but apparently that causes all sorts of distortion. I did come across a topology that combines these two ideas though; some external current regulator clamping the lower output transistor with no effect on THD. I think it was a Douglas Self proposal - will see if I can find it again..
...just riffing.

Bunkey

...just riffing.

Bunkey

...I mean, I might just ditch the TIP3055's and use the BD139/140 complementary pair driven by the VAS but idk.

It's gonna be <5W output so...


With 10W total dissipation from the supply I think class A is out of the question whatever I do.

It's early days.
...just riffing.

kaycee

Quote from: Bunkey on May 16, 2021, 03:21:13 AM
Quote from: kaycee on May 15, 2021, 04:26:15 PM
The fuzzy sound reminds me of Sungrazer, which is no bad thing to my ears. It's also woolly, like a low gain fuzz face, also nice. Good work.

Hey thanks and thanks again for the new music. Enjoying their self-titled album right now. I love that acid-tinged Kyuss sound.

The goal is sparkly warm clean that breaks into a pretty crunch; somewhere between reedy jazz, desert rock and 60's psychadelic;

Enjoyed your playing, that's where it took me.if you want Jazz in the mix,try Causa Sui, Mythic Sunship and  Papir all current Euro bands and label mates to each other.

Bunkey

Quote from: kaycee on May 16, 2021, 08:00:36 AM
Quote from: Bunkey on May 16, 2021, 03:21:13 AM
Quote from: kaycee on May 15, 2021, 04:26:15 PM
The fuzzy sound reminds me of Sungrazer, which is no bad thing to my ears. It's also woolly, like a low gain fuzz face, also nice. Good work.

Hey thanks and thanks again for the new music. Enjoying their self-titled album right now. I love that acid-tinged Kyuss sound.

The goal is sparkly warm clean that breaks into a pretty crunch; somewhere between reedy jazz, desert rock and 60's psychadelic;

Enjoyed your playing, that's where it took me.if you want Jazz in the mix,try Causa Sui, Mythic Sunship and  Papir all current Euro bands and label mates to each other.
Excellent :)

Allow me to return the favour...

...just riffing.

iainpunk

Quote from: Bunkey on May 16, 2021, 04:52:07 AM
...I mean, I might just ditch the TIP3055's and use the BD139/140 complementary pair driven by the VAS but idk.

It's gonna be <5W output so...


With 10W total dissipation from the supply I think class A is out of the question whatever I do.

It's early days.
yeah, if your power supply is limited in the output power, a Class A amp would be extremely impractical. the only reason i chose a Class A power amp is that my supply could put out 5A. if my supply was limited at a lower current, i'd go directly for a class D design.

cheers
friendly reminder: all holes are positive and have negative weight, despite not being there.

cheers

Bunkey

Quote from: iainpunk on May 16, 2021, 08:59:02 AM
Quote from: Bunkey on May 16, 2021, 04:52:07 AM
With 10W total dissipation from the supply I think class A is out of the question whatever I do.

It's early days.
a Class A amp would be extremely impractical.

if my supply was limited at a lower current, i'd go directly for a class D design.

cheers
It would be nice though. I've decided this AB LM380 sounds pretty bad once it's in full B territory. Sounds awesome at low volumes and then... yeah... small and mid-band restricted and a bit naff.
I'm hoping the straight class B discrete will outperform it in that regard.

I think you're just partial to a bit of cerebral peac0cking Iain  :icon_lol:
Would you really design a class D amplifier for guitar, over class B?
I can imagine it appealling to robe-clad Sunn worshipers somewhat.. (not sure about much else!).
...just riffing.

iainpunk

#98
Quote from: Bunkey on May 16, 2021, 12:04:41 PM
Quote from: iainpunk on May 16, 2021, 08:59:02 AM
Quote from: Bunkey on May 16, 2021, 04:52:07 AM
With 10W total dissipation from the supply I think class A is out of the question whatever I do.

It's early days.
a Class A amp would be extremely impractical.

if my supply was limited at a lower current, i'd go directly for a class D design.

cheers
It would be nice though. I've decided this AB LM380 sounds pretty bad once it's in full B territory. Sounds awesome at low volumes and then... yeah... small and mid-band restricted and a bit naff.
I'm hoping the straight class B discrete will outperform it in that regard.

that could be the small speaker, they tend to lose more low end the higher the volume goes, basically the bass stays the same volume while the rest gets louder
Quote

I think you're just partial to a bit of cerebral peac0cking Iain  :icon_lol:
yes, i am, but its always 60% serious and 40% ''just because its cool''
Quote
Would you really design a class D amplifier for guitar, over class B?
any day of the week, pure class B is kinda shitty, class AB is a lot better, class A is amazing, class C is just silly (or its an over driven class A amp) and class D is exactly mediocre, but has efficiency on its side, plus you can impart non-linear behavior in the modulator to recreate a more ''analog'' (read non-linear) sound, which can elevate it to super amazing!
Quote
I can imagine it appealling to robe-clad Sunn worshipers somewhat.. (not sure about much else!).
i'm 66% bass player and 33% guitar player, class D is already something im used to for gigging.
small form factor, low in weight, barely gets hot, doesn't waste much power at all.
the only real downside of class D is the output filtering required to not use the speaker wire and voice-coil as a radio transmitter.

cheers
friendly reminder: all holes are positive and have negative weight, despite not being there.

cheers

Bunkey

The speaker is mighty capable. The amp just sounds much harder at higher volumes.  The eq did open up a bit after cranking it just then (I hadn't played the speaker loud much from new) but yeah I'm hoping the CFB input and a bit of control over biasing will have a slightly softer feel than the chip.

I'm sure I'll be able to settle on some grossly underperforming topology with obscene THD figures that sounds wonderful in this application..
...just riffing.