Group Build: Dynafet

[Torpedo]

No oscillation at all?

[Pars]

Nope. At least that my scope can see (Tek 150MHz). DC offset is rock solid (another clue to possible oscillation if it is bouncing all over the place). Could run this without the servos if I wanted to.

Edited May 25, 2017 by Pars

[kevin gilmore]

I can certainly do a new board, it would have to be mostly surface mount, and different output fets.

I can definitely see that running the board without the main electrolytics would effect the low frequencies

Edited May 25, 2017 by kevin gilmore

[Pars]

There were still 4x470uf (2 per rail), but these were close to the power input. The lows were there, just not as nice.

I still have a hard time believing that yours didn't oscillate. Mine did as soon as I put more than 1 pair of outputs in, sometimes with only a single pair (Renesas 2SJ77/2SK214). Yours was running 4 pairs from the picture I saw.

Did anyone try the balanced board? Maybe the only run of those was with lil knight and no one ever got them...

[Pars]

By new board, would this be a balanced (and SuSy) board? What changes to the original circuit would be made? The gate zobels? Output? I might be interested in a layout, though I have been making changes to Jacob's original SE layout to beef up trace widths a bit and add the aforementioned. My copy of Eagle is right at the edge of board size with this, so hampered a bit.

After the SuSy Dynalo (which I like, don't get me wrong), I'm not sure I am a balanced believer. I don't hear full, if any, superiority to my original dynalo built on one of justin's headamp boards. Maybe if I had a fully balanced chain. Or better source. Or better phones. Dunno.

Edited May 27, 2017 by Pars

[luvdunhill]

I would be willing to redo a board from scratch if you sent me a list of modifications (removing dual resistor pads, etc.) I should have time this summer and sounds like a fun project.

[kevin gilmore]

it could certainly be balanced susy.

biggest question is what to use for output transistors

[Pars]

This is what I currently am running, schematic-wise (other than ferrites). Extraneous bits remain (I think) at the input in the schematic. Also the rail bypass caps don't accurately reflect what I have currently in place. Mine has 4x470uf 50V by the power input and 2x220uf 63V up by the output devices, modified to run rail->gnd instead of the original rail->rail. I also have 4x22nf Vishay MKP1837(?) in the C11-12-13-14 positions. Maybe should replace these with some Wima 4.7uf? Gain resistors (R55/R56) are 10K/2K in mine (10K/1K also verified). 4.7K/1K for reduced gain caused oscillation (without the output zobels of 47n/22R in place). The original 1K/100R did not work.

I'm not sure if the ferrites should be accommodated or not in the PCB layout. Choosing which ones might take a bit of work as well. The chip SMD beads I had didn't do shit.

The changes I made to mine were band-aids to get it to work. No scientific effort was expended to optimize parts such as the gate stopper values (47R PRP on mine). Cordell always seems to use the 39pf/100R combo for gate zobel or bypass, but that might reflect him always using a particular FET? Not sure. Mine are 33pf/91R, as that is what I had. I'll have to check the 33pf to make sure that is what is in there.

As for FET output devices for a modern SS DynaFet layout, some interesting threads on diyaudio:

http://www.diyaudio.com/forums/headphone-systems/207578-thor-all-discrete-lateral-fet-headphone-amplifier-17.html

http://www.diyaudio.com/forums/solid-state/193730-transistors-recommended-new-designs.html

http://www.diyaudio.com/forums/headphone-systems/263785-replacing-power-mosfets-sds-labs-headphone-amplifier.html

 

 

dynafet_cby.pdf

[Pars]

On 7/12/2009 at 4:45 PM, j4cbo said:

I think this could be the one...

 

http://b.j4cbo.com/temp/gbf2f-top.png

http://b.j4cbo.com/temp/gbf2f-bottom.png

http://b.j4cbo.com/temp/gbf2f-signal.png

 

On 7/12/2009 at 5:31 PM, kevin gilmore said:

SJ2 NOT necessary.

 

EXCEPT you have to flip pins 1 and 2 of the feedback jumper connector.

 

The inverting input needs to be the middle pin.

After re-reading much of this thread (for I don't recall how many times), the feedback isn'w quite what I expected.

It appears by the attached drawing, marked up to correspond to Kevin's post, that the feedback is either:

1. pins 3-2: output to inverting input
2. pins 1-2: midpoint of R19/R20 (2K, not 5K) to inverting input

For some reason, I thought it was either 1 above, or output to midpoint. In looking the board files over, I see that that is wrong.

Also, for the midpoint feedback (#2), there doesn't appear to be a gain resistor in play, unless R19 || R20 acts as the gain resistor? I'm using the output to input FB, and could not get the midpoint feedback even close to working.

And to clarify on your comment on SJ2 (which was not on the boards), you had never intended to ground the midpoint of R19/R20 in any circumstance?

Thanks!

[kevin gilmore]

r19 in parallel with r20 is the gain resistor

 

[Pars]

Thanks Kevin! Duh on my part but I hadn't thought the gain loop architecture thru on the mid point properly. This could explain why it was very unstable the last time I tried it, with an effective gain of 2 (I think?) with 1K and 1K for R56

The feedback is EITHER from the output stage, or stage 2. Not correct as shown
in jacob's schematic above.

Or in multiloop, a resistor from the output stage to input- AND a resistor from
stage 2 to input-. Plus compensation cap on the output stage, and possibly
the second stage.

I never messed with the multi-loop, and the prototypes have feedback from the
outstage only.

.
Earlier in the thread you had described the FB as multiloop (2nd paragraph of quote) and connected to both feedback points, which seems different from how this was actually implemented (1st paragraph). Not that I want to mess with mine anymore


Sent from my iPhone using Tapatalk

[jose]

I follow the thread because it seems an interesting project.

On ‎27‎/‎05‎/‎2017 at 12:04 PM, kevin gilmore said:

it could certainly be balanced susy.

biggest question is what to use for output transistors

Would there be big differences between the Dynahi-bal and Dynafet?

What characteristics should the new Mosfet have?

[kevin gilmore]

ssdynahi has a lower input impedance and would be fully balanced.

 

[Pars]

Kevin had indicated to me that he hadn't done an SS DynaFet layout, and parts would need to be spec'd for it. I personally don't think i have an interest in building one.

 

In a post I made around 5/27, I had put 3 links to diyaudio threads with some currently available mosfets that could be candidates. Back 1 page in the thread.

 

1) Semelab BUZ900 series (BUZ901P/BUZ906P)

 

 

Sent from my iPhone using Tapatalk

 

 

 

[Pars]

Replaced the sigma22 PSU with a GRLV, and listened to this tonight for the first time since. Yes!

[Rossliew]

Pics of the amp, @Pars !

[Pars]

Go back a page (April 11 post)... still haven't finished casing it yet (this winter...). Both amp boards are on the right heatsink now, and a GRLV on the left heatsink.

[Rossliew]

Ok saw it. Nice impressions of the sound..looks like something worth a little extra attention  

[Pars]

Finally got around to casing this (also posted in the What are you building thread.

Needs a bit of wiring cleanup, and the transformer needs to get bolted down, but sounds great.

 

[Pars]

One additional pic showing one of my favorite aspects: I used TE Conn. D5200 Dynamic connectors for the AC on the voltage select/momentary switch board (Mark Johnson's ACPRSS from diyaudio, modified layout for connectors). These are rated at 600V, are 5.00/5.08mm spacing and don't take up much more room than the screw down phoenix connectors. These are very nice, and IMHO would be a really good choice for the stat stuff (Carbon, GRHV, etc.). There are 2 alignment pins on the footprint, so not a drop-in replacement (these do pull out, so you might be able to get away with that).