DIYHiFi.org

Hello (not) again !
I'm starting this new build thread to cover a project that I've been planning for a while, two "monoblock" fully differential in-out power amps.

And why BEG? Well by the time I'll finish it will pretty much reflect my state of affairs...The plan is to make two nice cases with big radiators and the whole kit: modular construction so it can host various future projects, good quality supplies with higher than needed ratings (at least for this first try), protection modules, connectors and so on...
So overall BEG will not be a cheap chip amp and name tries to confirm it: Balanced Expensive Gainclone. I guess we need to name amplifiers like hurricanes - to distinguish between failures

A discrete design is tempting and there are plenty to choose from, however I need to test this implementation first and after that... whatever the ‘flavour of the day’ will be.
There must be a few similar projects around (and at least one in kit form from the Twisted Pear crew )... but quite different.
The first minor step is already undertaken; I created some new symbols and footprints. Soon the real fun will begin: making a smoke free prototype. The schematic is below, lots of head spinning feedback with the usual note on values (especially the * ones) not yet being defined.

of course I will need to (re)build a suitable preamp in due time...probably a ‘combo’ discrete+OpAmps.

Hey, something different, "head spinning feedback".
2K input impedance is somewhat low, no?
You have lower input impedance on the input chip than on the LM4780.
Also, you don't need the servo and IMO you don't even need the input chip, if you are going to use a preamp.
Simple(r) usually sounds better but then, it's your project, you may have reasons to do it like this and I don't want to spoil the fun.

_________________
Carlos Filipe

"Any intelligent fool can make things bigger, more complex, and more violent. It takes a touch of genius - and a lot of courage - to move in the opposite direction." Albert Einstein

Yep , proudly confusing feedback since 1970 (or was it '73? - if we look at the big thread elsewhere)


The preamp will be able to drive ~2k without major problems (and I don't plan to use another source for the PA, except the EMU card when doing the tests - itself able to drive that impedance). As for the OPA it has to drive ~2.5K of combined diff. load so going lower or higher than that will pose different sets of issues.



I'll start with R41=0, R40=open, the servo may be 'pressed' into service later only if really needed. The LMs being differential will partly take care of offset problems.
The challenge with this project is to tame the high(ish) open loop gain while aiming for extended freq. response and of course stability.
And in the quest for the perfect oscillator I may even try the LME49724 which will give about 145dB OLG.

Btw, I have an open question: is there a spice model for the LM3886?
I know that some use the OPA544 (or LM12?) instead but that doesn't help my case....

And the supply for the power section, in the 'nice to have' version. For this project I have dual independent secondaries but I'll plan some jumpers in the layout that will also permit future use of a center-tap transformer. Not nice, of course a center tap transformer will cancel the advantage of 4 pole 'lytics.

A whooping ~180mF for both channels, perhaps even more to allow for some extra local decoupling.
Looks pretty easy, ...convincing the 'Secretary of the Treasury' that those Mundorfs are really essential, not remotely so...

I would put 0.22 Ohms series resistors after the rectifier diodes (before the first caps), because those big 22,000uF will generate some nasty charging spikes.

_________________
Carlos Filipe

"Any intelligent fool can make things bigger, more complex, and more violent. It takes a touch of genius - and a lot of courage - to move in the opposite direction." Albert Einstein

Thanks for the suggestion Carlos, I will consider it.
Right now I'm thinking how to implement a little trick on this supply....(if it works) I'll post the details.

OK, it's not alright
It involved using the MAX4373 and a power mos switch to implement a sort of latching protection....too much trouble with eliminating (filtering) the ripple and load dependent variations.

However for now I'll post some additional stuff that will discretely sit on the amplifier board....let's call it stand-by and power up/down sequencing.
Vcc1 and Vee1 will be +/-15V regulated.
VCC and VEE will be about +/-24V
All zeners are 15V except D12 which is 24V.

I got some explaining to do. What's with the red line?
I posted the idea for the section at the right side of the red line in a thread here...I mean there. It shows that I've been cooking this amp for a while... Basically it can be used for any number of clones if R42 is tuned accordingly (or even for discrete power-amps should their topology allow)

It is extremely simple, just a resume: Q1 and Q2 realize a wired AND, at power-on Q1 switches on last allowing for some delay and at power-off Q2 will switch off first when VCC-VEE is less than about 26Vdc. The chips will be unmuted only when both mosfets are on, enabling the current source to draw a total of approx. 4mA from all chips. R25,...,28 from the first schematic have migrated here and ensure that each chip supplies his equal share of current.

At the left there's the STAND-BY section that disables the OPA and the clones. And there is an additional precaution: the chips are also muted while the regulated +/-15V is bellow approx. +/-12V. Implemented with Q6 (poor man's comparator) and the optocoupler. This makes sure that the global feedback loop is always in control only after cycling the power correctly and that it doesn't push the outputs into the rails if there are timing differences between all those supplies.

First I salute roibm's work - attachements are visible again

Some updates on this project. I'll start with the side-dish, since I need a balanced preamp. For the volume control section there are several solutions and I have yet to decide what to use:
1. directly the circuit attached, here the seed value for the input resistor is 2.49k
2. use a differential FET follower similar to the one proposed by Bob Cordell in his book, with the attenuator placed at the input (seed value for the input res ~5k)
3. same as above followed by the circuit attached but with a seed value of input R=1k

The switch will be one of those easy to find, blueish 23 positions attenuators.
I'm using already one in my current preamp and I'm quite happy with their reliability.
The idea is to put what little available CMRR to good use and provide a lowish input Z (but perfectly 'driveable' by a normal source) to avoid bandwith limiting due to parasitics. The input Z seen differentially varies with attenuation, but not as bad as it may seem - see table. Of course the schematic is simplifed and assumes an ideal voltage source....

There are also several ways to include a SE to BAL conversion, I'll try to post a complete schematic soon...

Referring to the first schematic of this thread there is a major change. I decided to reduce the gain of the power op-amps down to 2. Less troubles when closing the global feedback-loop. One advantage of this approach is that more feedback (~14dB) is available at low-mid frequencies. Also this compensation method conserves the open loop gain in the passband and the slew-rate. However a couple of dB's of loop gain are lost towards 20kHz (I mean less than 3dB)

I'm actually experimenting with this circuit and I will post some measurements as soon as I'll be happy(er) with the end- result. The values for the lead-lag compensation network ensure only 45deg phase margin and I get some overshoots (5%)...so it's still work in progress...

AN-1604 is a good resource

In my experimental set-up actually both LM3886 are contained in one LM4780, I only used the LM3886 symbol for clarity.
OPA1632 merely converts my signal generator to a balanced source. I will add another OPA1632 to emulate the whole circuit after/if this phase ends succesfully. Loading is still in SE mode and quite light (due to whimpy radiators)