Melcor Deconstructed

[bcarso]

Hmmmm.

In sim this amp has some crossover distortion of a disturbing character---it could be nominated for an award for Most Improvement Under Output Current Loading. Without something pulling down (or up) the spectrum has predominately odd-order components that go out almost indefinitely, peaking somewhere around the 23rd (!). This is for the conditions Av = 2, Rf/Rdiv each 1k, 100mV peak in at 1kHz. This is with no loading other than the feedback network itself.

And yet the net amount of all this is still fairly low at about 0.01%.

Meanwhile I'm still not sure of the compensation components, but with C3 600pF and C1 1nF there is a big peak at 340kHz.

If the output is loaded to the neg rail with about 5mA things clean up distortion-wise, but then the peak sharpens and reaches almost delta 40dB. Ouch.

I guess the aim of the designer was to avoid any quiescent dissipation in the output devices---it's the only advantage that I can see to the output configuration.

Brad

 

[clintrubber]

Interesting new info ! Hmm, so the 1731 actually qualifies more as a stompbox than an amplification device :sad: If I have a jig made up I'd look at how nasty the THD-residual is. Jeff, do your sims show alike things ?

This is for the conditions Av = 2, Rf/Rdiv each 1k

I expect little from scaling both these resistors to say 20k each, but how about a feedback-cap ? It'll soften the nasty stuff of course, not cure it.

Regards,

Peter

 

[bcarso]

With a 600 ohm load you can see the notches in the time domain pretty easily. At this point the THD going out to the 39th is .062%. You can also see the even components beginning to rise as you go out in harmonic number. Fortunately the ear cuts off somewhere (or Does it muahahaha...).

I'll try some other configurations.

PS: With the load the freq response loses most of the peaking.

 

[Flatpicker]

I made a typo in my last post - should have been C1, not C2.

 

[Flatpicker]

[quote author="Bauman"]yea yea, I have the correct connections but not the correct value for C1, that need to be 1000pF :green:[/quote]I thought we decided that 630pF worked better?? Or was it originally 630pF and we thought 1000pF worked better? :?

 

[bcarso]

OK.

With C1 1000pF, C2 10pF, and C3 22pF (helped distortion a bit compared to 10pF), with 1k feedback and feedback divider R's and 100pF across the feedback R (~peaking-free response), driving 600 ohms, 100mV peak in...

The crossover distortion is a little better at 1kHz, about .0068% going out to the 39th harmonic. It falls off as you go out, unlike the previous behavior, although not very fast. If you go out just to 23kHz it's still .006%.

At 10kHz it's about .097%, again going out to high harmonics.

How will it sound? Well, since it is so low overall I doubt that it would be an instant yuck response (and if this is really representative of what has been built, obviously it sounds ok). But how audible, say, the 7th order stuff would be after you got used to it, is at least a concern.

Again, pulling some quiescent output current to one rail or the other will help a bunch.

 

[tk@halmi]

Yeah, the Melcor is poopie. Pehaps its redeeming quality is that it can drive low impedance loads.

 

[PRR]

> I think there may be some transcription errors in this schematic, in the vicinity of the output stage.

Agreed. And I sure recall doing some serious thrashing here on something very like this, which was at first obviously a can't-work tracing.

I strongly suspect "D2" is not a single diode, but one of those once-popular multi-diode packs that look like a single diode. A dual here makes sense. And my dim memory is that someone looked up the part and indeed it was an obsolete dual-diode.

That then makes the current-limit about 150mA, which is about right if this thing were driving 150 ohm loads.

And of course the crossover artifacts are now down in the mud instead of tickling every little squiggle of audio.

The 360 ohm resistor still bothers me. It protects Q5 from trying to pull a shorted load all by itself, but what protects Q6? Just the fact that R11 will turn-off Q6's driver? Maybe.

C2 is, at first sight, Positive Feedback. Probably it trims the phase response at the top of the GBW (Brad has it in a simulator, I think-- let him ponder it). This would be very dependant on devices and layout. On a recreation, I'd be real inclined to leave it out for initial testing.

> it could be nominated for an award for Most Improvement Under Output Current Loading.

Nah. Look at the output of one early delay-box: 741 plus two UN-biased transistors. Without even a Base resistor to leak light loads.

> so the 1731 actually qualifies more as a stompbox

Oh, no! Back in the day, a LOT of stuff had crossover distortion, and we even praised the "crisp" sound.

Do we all know how OLD this is? TIS98 is a first-generation mass-produced Silicon transistor. At the time, if you weren't AT&T or IBM, you got TI's TIS transistors or you got nothing. 2N5088 is also a very old part.

 

[clintrubber]

from TK:

Yeah, the Melcor is poopie.

I'm wondering, is this based on listening to the original 1731 ?

If so then it all seem to come down to 'use it if you like the taste, otherwise skip'.
OK, that's a statement that holds for a lot of things, but since I don't know much about the use of the 1731 so far I'm wondering about that 1731 and its use & how it is regarded. Since it was dissected & we're able to clone it using those nice boards I initially assumed it was a desirable DOA, but who knows, there are some 'considerations'...

Regards,

Peter

 

[clintrubber]

from PRR:

I strongly suspect "D2" is not a single diode, but one of those once-popular multi-diode packs that look like a single diode. A dual here makes sense. And my dim memory is that someone looked up the part and indeed it was an obsolete dual-diode.

Thanks, that dual for D2 is good info !
Who did the dissecting of the 1731 ? If its remains are still available a quick DMM-diode check could confirm this.
I've a few of Jays boards, could make a few variations.

The internal variables as I understood them so far:
* skipping C2
* C1 630 or 1000 pF (we might just go for say 820 pF here)
* C3 increasing in value
* D2 a single or double diode

And 'outside' the 1731 there's the feedback-cap to toy around with, ~200pF been suggested.

Regards,

Peter

 

[tony dB]

strange conclusion from TK... :?

Most of the folks around here are very happy with their 1731's... :?: :!:
I'm completing 24 1731's myself but don't have any 312's ready to check them out.
Doesn't the circuit where these get put in count for a lot of the sound too?
In what did you put them TK? Did you compare them to 2520's or anything else too?

Didn't Fabio made already very promising progress in his API's?
I might be a little of my trolley here, as I'm still in recovery fase of surgery :sad:

Cheers,

Tony.

 

[clintrubber]

from Tony:

I might be a little of my trolley here, as I'm still in recovery fase of surgery

Hope all is well or is going to be well soon ! :thumb:

 

[clintrubber]

from PRR:

> so the 1731 actually qualifies more as a stompbox

Oh, no! Back in the day, a LOT of stuff had crossover distortion, and we even praised the "crisp" sound.

There you say something. A clean opamp would be without signature ('990 qualifies ?) and the colour-makers all come down to a few this or thats that makes their colour - like here now for the 1731 its x-over profile perhaps.

Unsurprisingly, I would expect that if this 1731 would be 'technically improved' then its colour (whatever it is) would have been changed.

But has anyone compared the original & cloned 1731 soundwise ?
Maybe because of the discussed potential differences their signatures differ. As long as the clone it sounding good then its OK, but who knows, the clone isn't yet what it could be.

Regards,

Peter

 

[Flatpicker]

[quote author="clintrubber"]Thanks, that dual for D2 is good info !
Who did the dissecting of the 1731 ? [/quote]Peter C. (Where is he, anyway?) I'd put my money on the schematic being correct - several have built Peter's version and IIRC, those with originals said they sounded the same.

 

[clintrubber]

from Flatpicker:

Peter C. (Where is he, anyway?) I'd put my money on the schematic being correct - several have built Peter's version and IIRC, those with originals said they sounded the same.

OK, that sound encouraging/good. Maybe the difference in sound with/without the additional diode is just subtle or less, despite the sim-observed crossover-profile ?

A test from a different angle could perhaps be by looking at the current consumption - I assume the additional diode would increase the quiesc. DC-current draw significantly.

Regards,

Peter

 

[Flatpicker]

[quote author="clintrubber"]I assume the additional diode would increase the quiesc. DC-current draw significantly.[/quote]It might - I'm certainly not saying that it wouldn't be an improvement.

 

[pmroz]

.

I strongly suspect "D2" is not a single diode, but one of those once-popular multi-diode packs that look like a single diode. A dual here makes sense. And my dim memory is that someone looked up the part and indeed it was an obsolete dual-diode.

Hmmm... interesting. I have two diodes mounted for D1, not D2. On all 6 1731s. Have to dig through papers etc. to see where i got that 'information'.

Great that PRR has posted some insights here. I bet a post from Peter C could clear some things up.

 

[tk@halmi]

Excuse my over-generalization. I have the DIY version of the Melcor only. I have never had the original one and most likely I never will. My comparison is based on a BTI OA400 discrete opamp (2520 clone) and several chip opamps. I would take the NE5534 anytime over the current DIY Melcor for moderate gain applications. The additional diode for biasing the output should make a significant difference in the performance of the opamp. Perhaps it is time to give it another try.

 

[rafafredd]

I was the one that originaly worked on the 1731 reversing. By the time I had almost no electronics skills at all (and I still know little) and I just accepted it because no one had the time to do it. Maybe it was a mistake, but I took my chances and tried it. I knew it was a only one shot, so I´ve made a very minucious work, drawing the original board layout and component placing. I have to tottally destruct the original opamp to get into the parts. The original layout doens´t looks very clever, with lots of pats just floating and hanging arround, from one side to the other of the board...

In he beginning I got very confused because of diferent pinouts on those old transistors comparing with the specsheets I had. Strangely, the transistors on the melcor had diferent pinouts comparing to the vintage datasheets of the exact same part numbers I´ve dowloaded on the net. But it wasn´t a big deal at all. Once I had the first schem, with all the transistors wired erroniously (I was a dumb ), PRR had a look at this and redrawn it with the right position of the transistors and we ended with a working schematric. Fabio simulated it, and it worked. I don´t know who was the first to try it, but I think it was PeterP.

As I still have the notes and the drawing that I did, I had a look at this today, comparing with the schematics we are using now.

So, there really is a few conflicting things still. I have noted people about this before but it seems that everyboy was very happy because it was working the way it was and people didn´t really cared by the time.

Let´s use the last schem posted by Fabio (in this very same thread, page 1) for values:

C1=680pF, and not 630pF

C2, C3 = 10pF. Those are the rightvalues.

BUT, on the board that I had, C2 was REALLY connected to the negative input pin, or the base of the negative input transistor. Like it or not it was there... Maybe it is strange, but I´ve checked my work many many many times and I did it meticulously and slowly, and that´s what was in there for sure. I don´t know if it will make things better or worse, but I hope better...

D2 = a dual diode. It´s a dual diode for sure. I have the original parts and just measured it.

DECOUPLING CAPS:

Also, There was an important thing that maybe can make the output stage runs smoother. There was NO DECOUPLING CAPS DIRECTLY ON THE RAILS inside the opamps, BUT THERE WAS 10nF decoupling caps from the 10R resistors and the collectors of the output devices junction to ground. Maybe it will handle the output stage more stable.

Those were the only changes. Let me know if you need me to make another schematic with the changes I´ve listed.

 

[bcarso]

Thanks rafafredd for that information.

So---somebody please at least correct the schematic to show two diodes for D2! That will make ALL the difference in crossover distortion. Another 1N4148 should be just fine.

I don't think the decoupling details being before or after R13-14 will have a huge effect on anything but may have been helpful for a given layout. In some cases a straight R in the collectors is more effective for suppressing parasitic oscillations. Generally, the more decoupling caps the better. This design should have fairly good power supply rejection though, even with the absence of an active current source for the input pair.

The connection of C2 you describe is not that peculiar, as it now is negative feedback and just taken ahead of the final output Q's. Its optimal vlaue will have to do with the choice of impedances for the external feedback components, but it may be about right for the 22k-ish values I've seen mentioned.

Brad