992 opamps application, doubt

[Flatpicker]

[quote author="Bauman"]...I´d say that in the beginning I´ve prototyped with OEP transformers + Melcor and I also like the sound of it very much, it have its space in my taste...[/quote]Yes, and this combination is on my list to try when I finish my 992 pres. Thanks for the tip!

 

[tk@halmi]

Fabio,

I just found the power MOSFETs to drive your API output. The 2SK216/2SJ79 MOSFETS (also used by Borbely) are very fast, have low Vgs (0.5 to 1.5) so there is little headroom lost, and have low input capacitance (~100pf). They can handle 0.6 amp and dissipate 30W.
I just ordered a few of them to breadboard a prototype.

Cheers,
Tamas

 

[FredForssell]

I just found the power MOSFETs to drive your API output. The 2SK216/2SJ79 MOSFETS

Yes, I've used that pair as well. As I remember, they worked really well in a headphone driver that I once built using the orginal JFET circuit.

cheers,

 

[tk@halmi]

[quote author="FredForssell"]
Yes, I've used that pair as well. As I remember, they worked really well in a headphone driver that I once built using the orginal JFET circuit.
cheers,[/quote]

Thank You Fred! Here is the schematics for my "MojoFet" opamp:
http://mysite.verizon.net/res75okq/sitebuildercontent/sitebuilderfiles/mojofet_v1.jpg

To set the idle current in the output MOSFETs twist R7 while monitoring current consumption. One should start with a low R7 setting whe powering up, like all the way down to a few ohms. You can monitor the output on the scope using a sine wave. You can see the crossover distortion going away as you increase the value of R7.

Edit:
Oops, I forgot the gate stopper resistors (100 ohm). I will add them this weekend.

Tamas

 

[Bauman]

Great stuff Tamas, thanks!
Yet exercising my patience stuffing the opamps... :?

cheers!
Fabio

 

[Bear]

So where's everyone on their Project 2 homework? I've got pretty intense actual homework, so I've been absentee lately.

I've got some parts on hand: four 992's, four 15v 990A's (cheap on the 'bay), and ten of the profile OT's, and haven't ordered any input iron yet as I wanted the dust to settle and see what people were doing. (Oh, and I might gut my Yamaha Pm board for their discrete opamps. Or maybe not.)

So is there no good reason to do the 992 w/ output iron? No problem, I can use it elsewhere. What about the Hardy 990's?

Is there any point in matching up the 992 with a "character" input transformer, or am I looking for a big clean one?

Bear

 

[Samuel Groner]

So what's the function of R12 (10k)? It looks to my ignorant mind that it causes the impedances with respect to ground to not be equal...

That's true. But it's not even a 1% error. However, it's not a pain to add another 10k and we are very happy now...

Hm, just thinking about it: the opamp will not have true zero impedance at it's output, maybe without the added resistor it will be more precise..?

Samuel

 

[rafafredd]

Tamas, didn´t you had to use a gate stopper resistor on this one? How does it compares with the ZETEX parts?

 

[PRR]

> Oops, I forgot the gate stopper resistors (100 ohm).

And an output. (It is obvious to me, but maybe not to everybody.)

I would draw Q6 "upside-down" from the way I see your drawing. Or better: put S G D labels; FETs can be so confusing.

Is Q5 really a cathode follower? I must be reading it wrong, because that would not let the output swing negative.

Isn't there any better (safer) way to bias FETs? Maybe I've worked with BJTs too much, but I'd be inclined to use an identical MOSFET (another 2SK216) instead of R7. With a couple of resistors, its Drain-Source voltage would be twice the Gate-Source voltage, giving about the right bias on the outputs. Neglecting R5 R6, then Q7 current would equal Q5 current. With R5 R6 and a small adjustment on the bias divider, you could run the outputs a little richer and still (I think) very thermally stable.

> worked really well in a headphone driver that I once built using the orginal JFET circuit.

Headphone drivers designed to 32Ω cans should make lovely low-Z line-amps.

 

[tk@halmi]

[quote author="PRR"]> Oops, I forgot the gate stopper resistors (100 ohm).
And an output. (It is obvious to me, but maybe not to everybody.)
I would draw Q6 "upside-down" from the way I see your drawing. Or better: put S G D labels; FETs can be so confusing.
Is Q5 really a cathode follower? I must be reading it wrong, because that would not let the output swing negative. [/quote]

Thanks for catching these. Sorry everyone for the sloppy drawing. This was my first time using Eagle, in a hurry. I didn't pay attention to pin labels. Q5 should not be a source follower. I will redraw FETs and MOSFETs this weekend and post.

[quote author="PRR"]Isn't there any better (safer) way to bias FETs? Maybe I've worked with BJTs too much, but I'd be inclined to use an identical MOSFET (another 2SK216) instead of R7. With a couple of resistors, its Drain-Source voltage would be twice the Gate-Source voltage, giving about the right bias on the outputs. Neglecting R5 R6, then Q7 current would equal Q5 current. With R5 R6 and a small adjustment on the bias divider, you could run the outputs a little richer and still (I think) very thermally stable.
[/quote]

I have read a couple of places that enhancement mode MOSFETs have a positive temperature coefficient. So as the device draws current and heats up its internal resistance increases that in turn reduces the channel current. Is this an incorrect assumption?
I know in BJT land we have to worry about thermal runaway and a voltage multiplier closely coupled with the output transistors is a safe way to bias.

[quote author="PRR"]Headphone drivers designed to 32Ω cans should make lovely low-Z line-amps.[/quote]

Thansk PRR!

 

[FredForssell]

I have read a couple of places that enhancement mode MOSFETs have a positive temperature coefficient. So as the device draws current and heats up its internal resistance increases that in turn reduces the channel current. Is this an incorrect assumption?

Yes and no. It is a correct assumption when the operating point is set high. At low currents most (all?) MOSFET still have a negative tempco. However, in the type of application that we have be discussing here, typically the output stage would be biased at somewhere around 100 ma. The 2SK216 has a drain current rating of something like 1 amp and I would guess that with a 100 milliamp operating point, it will still have a negative tempco. It will be pretty obvious when you first build the circuit and set the bias.

Cheers,

 

[synthetic]

My next project is a API-based mic pre as well. I was planning to use a design by Pat Morford, who designed a (very slight) variation of the API 312 circuit using a Lundahl 1636 on the input and no output transformer. Actually, he lists an optional 1:1 output transformer, but I was planning to build the circuit without it to start. His design allows you to switch between transformer/non outputs or have two seperate outs on your mic pre. His transformerless output has a 40uF polypropylene cap after the op amp. A $16 capacitor, sure why not.

The design specifies a Millennia MM990 as the op amp, although I was planning to use a Forssell 992 instead. However, the Millennia spec sheet he supplies lists distortion specs with a 600 ohm load, so perhaps this circuit is designed to handle that? As I understand it, it's one of Fred's earlier designs. Perhaps this would be a better choice if one were to use an output transformer.

His spec includes a parts list with plenty of expensive caps (see above), a suggested bread board layout and a power supply design. He used to sell this design for $500 -- not the kit, just the design. Let's just say I didn't pay that, it accidentally showed up in my PM inbox one day. He no longer seems to have this design available for sale, as far as I can tell. My old links are dead, and several google searches on his name and the text of the PDF turn up no leads.

I would be happy to host this PDF on my site, but I wouldn't want to screw him over if he's still selling this design. (Hopefully for less than $500! I would be a little upset if I paid several hundred dollars for this for this.) Does anyone know if he's still around?

Thanks, -jl

 

[Bear]

Aw, hell. Someone used the P-word.

Well, I guess that's an improvement in the state of the universe that he isn't ripping anyone else off for that "kit". That would be, what, at least $350 of profit each "kit"?

So, anyone have a Project 2 board with a non-OT output option? I'd hate to use a JLM board for less than the full shebang.

Bear

 

[tk@halmi]

[quote author="FredForssell"]
Yes and no... It will be pretty obvious when you first build the circuit and set the bias.
Cheers,[/quote]

OK, I will look into PRR's recommendation to use a MOSFET to bias the output.

Thanks for the Help!
Tamas

 

[tk@halmi]

[quote author="synthetic"]Pat Morford. Does anyone know if he's still around?
Thanks, -jl[/quote]

P*a*t seems to be out of business. I wonder whom he pissed off this time...

 

[synthetic]

Well then. if anyone's interested:

[He's back in business so I pulled the link]

I'll pull it if I hear that he's still in business. Until then, enjoy it while it lasts. Are there any problems you see with this circuit, other than the extravagent capacitor budget?

 

[rafafredd]

40uF is needed at the output if you are willing to drive 600ohms unbalanced.

But an EDCOR 600:600 transformer costs half the price :thumb:

 

[rafafredd]

interesting this 20uF cap he uses at the input transformer primary. Nver seen it. What would be the effect? A low cut?

 

[PRR]

> interesting this 20uF cap he uses at the input transformer primary

Explained far back in the Notes. He says you need it when using Phantom. Yes, if you have really badly-balanced Phantom, you get DC unbalance current in the transformer primary, which is bad. But I never heard of it being an actual problem. Maybe he was confusing Phantom with the alternative AB powering scheme, which DOES need a blocking cap here (actually, you normally insert the power voltage here).

I love this slip: "40uf polypropylene, ( can be 4 x 5uf, for example)". Even my $0.69 cardboard sliderule gives better calculations than that.

 

[synthetic]

I've been wondering about that cap, too. It's only switched in when phantom power is on. I assume it's there to keep +48V from getting into the input transformer, but I haven't seen it in other designs.