Group Build: Dynafet

[Pars]

Well, went ahead and put the output devices (2 pairs) in and fired it up. Seems to work. Had 2 DMMs, one on each J/K source resistors. Started out at 1.05Vdc (500mA), and was dropping. I only left it on for ~30 secs (2 x) and they had dropped into the 800mV range, but that is still 400+ mA. Did not try adjusting VR3 yet, as this seems to be too high. VR3 was set to max resistance, but I will verify this. It is bolted to a heatsink (hifi2000 pesante dissipante). The dynahi angle brackets I have do not go completely under the TO-220 tab (1x1" angle). 1x1.25" would be better. Thoughts?

[Pars]

Update: I tried adjusting VR3... I can go up, but not down. I replaced VR3 with 1 100K pot. Resistance at max is ~34K (it is in parallel with R19 + R20 or 10K total), including R62 of 3.3K. Voltage across the 2 ohm resistors dropped to ~700mV, so 350mA bias current. I measured DC offset and it was around 2mV, and seemed stable (for as long as i dared to leave the amp on, or 30 secs.). What effect does increasing or decreasing R17 and R18 (currently 20K) have? Should I drop these to 10K or go up?

[luvdunhill]

don't raise the pot. Keep it 20K. Lower R17 and R18 to the values I suggested (in reading past posts, it looks like you didn't change these and kept them at 20K, is that right?). If you're only using two devices, go even lower for R17-R18, say 2K. Probably easiest to just parallel another value on the back of the board. What are the two resistors at the alternative feedback point? Did you use the recommended 2K? Perhaps we should find the ref designators for these parts so we don't confuse each other....

Edited January 27, 2011 by luvdunhill

[Pars]

Yes, you told me to leave these at 20K (R17 and R18), so I did. Like I said, I have 4 devices (2 pair) in. R19 and R20 (alternate feedback point) are 5K. Either you or digger told me that was what you used when I asked if it was 2K or 5K.I sent you a pm also Marc.

Edited January 27, 2011 by Pars

[luvdunhill]

Yes, you told me to leave these at 20K (R17 and R18), so I did. Like I said, I have 4 devices (2 pair) in. R19 and R20 (alternate feedback point) are 5K. Either you or digger told me that was what you used when I asked if it was 2K or 5K.I sent you a pm also Marc.

Jacob:

"After some tweaking (and using the values I have around) I've settled on 2k ohm R17-R18 and 1k in series with the pot. I'm using a 20k pot but the usable range wouldn't be much less with 5k."

Marc:

"My recommendation was R17, R18 = 10K and a 20K pot with a 3.3K series resistor. The two 5K resistors at the feedback point are also changed to 2K"

digger945:

"1K pot, 1K for series resistor (R62) and 2.2K for R17-R18"

Kevin:

"One other thing, i forgot to mention is the 5k resistors are now actually 2k on my prototype so as to match the gain between the two different feedback types."

Heh, clear as mud.

[digger945]

What I used was for a single pair of output mosfets, one p and one n, and was an effort to lower the bias as much as possible. The fact is, that these mosfets seem to like higher bias, and the output will distort rather quickly as the bias is lowered. I will try to scare up my notes tomorrow, but I am thinking that the bare minimum bias to maintain a good output would be 150mA+.

[Pars]

I replaced R17/18 with 10K resistors and put the 20K pot back in. Down to 450mV.

Scott: from one of your previous posts (a year ago or so, probably before the boards were even ordered)

For the resistors used in the Vbe multiplier' date=' where originally KG had 20k resist/20k pot/20k resist, I'm using [b']1k for the adjustment pot and 1k for the resistor in series with that pot(R62), and 2.2k for R17 and R18. I recommend that the adjusment pot be set to the middle of the range before installation, or if you know how many turns it is then set to the middle after installation. I have found these values to be good for adjustment down to levels that are acceptable for use in my dynamite should I decide to transplant them, and still be able to exceed 300mA if you want to crank it up a bit. All of my testing lately has been done with only one J76 and one k213, both mounted on 1.5" Aavid's just like a Beta22 would be.

I may drop R17/R18 to 2.2K and see what happens. And also try to decrease the pot size further. I know I have some 2K, but will wait to do this until I can get it well below 150mA. I figured 150mA would be my target, but I want to have adjustment range on either side of that. I assume that if I start mucking with the pot size and series resistor (R62) before I am able to get it down that far that I will be cruising on fail

At 150mA, would I be looking at ~300mA per channel for power draw (4 devices or 2 pairs per board), and should a single Sigma22 handle this OK? Since I already have this case, a single sigma22 is mandatory (no room for a second one).

[Pars]

Also, the gate voltage is at 2.99 Vdc currently (at the 450mV/225mA bias). I may be missing something, but don't the gates set the bias (or apply it I guess would be more accurate)? Since these are in parallel and driven by the corresponding diamond buffers, wouldn't the Vbe multiplier settings be pretty much independent of how may output devices you are actually running?

Edited January 27, 2011 by Pars

[digger945]

I can't find my notes, Chris. I am pretty sure that what I wanted to do, was to see just how low the bias could be set and still sound good and maintain a decent looking output waveform.

You are right, the Vgs of the mosfets will determine the drain current. I don't see how the number of output mosfets would matter, with or without the buffer.

I don't have the proto any longer, so it may be a while before I could do anything to help out much. To my memory, I used the values that you bolded in the quote above to get down around something like ~2v to 2.5v Vgs.

I'm sure I was using just one sigma(2" heatsinks) to power everything, and a 162VA toroid with it. I think the sigma is good for about 1 amp with stock heatsinks.

I think I had much better success when applying a load to the output as I was trying to get everything adjusted up.

[Pars]

Thanks Scott!

I replaced R17/18 with 3.3K resistors (and socketed them this time so easier play time). Min. bias of 160mV (80mA), max of over 500mV (250mA). Dug thru my parts and I don't have any 2.2K resistors . I do have some R60s but I'm not going to try to cram those in. I was going to replace the 5K R19/20 (the alternate feedback point) and see what effect that had. Then drop R62 from 3.3K to maybe 1K, per your findings. So if anyone has any 2.2K PRPs...

If I have to order them, should I get 2K or 2.2K? The latter is probably more useful overall.

I put the Fluke in series with the positive voltage feed. At 150mA bias, the board is drawing ~375mA. Heatsinks on the sigma are pretty cool, so I don't think running a stereo set biased at 150mA is going to be a problem. DC offset looks good, and pretty stable at 3mV. I have not put the servo opamp in yet, and have not tried to adjust offset with the pots which are both set to 500 ohms total for the resistor/pot combo.

[Pars]

I found some 2K resistors I had (NTE ), so went ahead and put sockets in for R19-R20 and put 2K in place of the 5K that I had in there. I also replaced R62 with a 1K (from 3.3K). Bias range (with a 20K pot) is now ~160mV to over 800mV. I also took a stab at using the alternate feedback point. Bias did not change from where I had it set, but interestingly DC offset did. It went from ~5mV to 165mV. I have not done any trimming of VR1/VR2 yet to compensated. I switched back to the std. feedback for the time being.

Kevin (or anyone): how is gain calculated for the alternate feedback point?

Also, I added a tab to the BOM spreadsheet and have a table in there of various Vbe combinations and the result on bias. Thought it might be helpful

At this point it appears that I am going to wind up very close to where Scott did in his tests of the prototype.

[Pars]

Soniccraft order came today I think I arrived at my final values for the Vbe, targeted to give me 150mA or higher (if I want).

R17-R18: 2.2K

R19-R20: 2.0K (I may go back to the 5K i had in here as I don't think I will use the local feedback option)

R62: 1K

VR3: 2K

adjustment range: 241-600+ mV, for 120 - >300mA

Values are almost identical to what Scott (digger945) arrived at. See link for BOM.

My link

EDIT: All testing so far has been at the default gain of 11 (1K/100 ohm). I'll do some testing at lower gains, maybe down to about 5. IIRC, that is about as low as this thing is stable? Someone correct me if I'm wrong

Edited February 1, 2011 by Pars

[luvdunhill]

Nice!

The lower gain just needs more compensation. I tested it down to 3 I believe.

I'd definitely try both feedback points, it's worth it.

[Pars]

DC offset goes up too high on the local feedback point. With no CCS adjustment (VR1/VR2), I'm at ~2mV or less using global feedback vs. 165mV at the local feedback setting. These are both without the servo in. I confirmed with Kevin that the 100 ohm to ground (R56) for the global feedback should remain for the local feedback as well. Not sure how to calculate gain using the local feedback setting. It would appear that the FB resistor (R55) and the combination of R19 || R20 are involved... He did state that the servo would be required for the local feedback to work correctly, but it seems that making a servo compensate for that much offset is the wrong thing to do. He also stated that global feedback was better.

[DigiPete]

Great work Chris! Glad you are figuring this stuff out.

So the board with 4 output devices draws 375ma at 150ma bias;

so that should be x2 for stereo, and x4 for balanced, correct?

Should be no problem for my dual Dynahi PSUs.

A gain of 5 or 6 would be good to try out, as that is what I plan to run.

Look forward to your final parts recommendations.

PS I have output devices matched by Marc,

since I will only use half the devices for my 4 ch amp, I can spare them if you need.

Cheers,

Pete

Edited February 2, 2011 by DigiPete

[Pars]

Pete,

Yes, that would be x4 for balanced, or 2 for stereo. I'll recheck my measurements on that also. Are you using Dynahi PSUs or sigma22s? Can't remember what a sigma is good for, but I think I should be alright with a single sigma22 for 2 channels. If not, I'll just decrease the bias a bit. With one channel, the heatsinks on the sigma don't even get warm after 15-20 minutes. I still have some dynahi PSU PCBs, so I suppose I could always build one up, but I like the sigmas I've used a lot. I'll also be sure to check the board over at a lower gain by changing the 1K to 475 or so. I don't have a signal generator, so I'll see what I can do with my M-Audio Firewire box as far as checking compensation. I think Scott did some testing also, and he kindly is shipping me a pair of K devices to replace the one I stupidly fucked up. I'm finishing the 2nd board now at the posted values.

[DigiPete]

Pete,

Yes, that would be x4 for balanced, or 2 for stereo. I'll recheck my measurements on that also. Are you using Dynahi PSUs or sigma22s? Can't remember what a sigma is good for, but I think I should be alright with a single sigma22 for 2 channels. If not, I'll just decrease the bias a bit. With one channel, the heatsinks on the sigma don't even get warm after 15-20 minutes. I still have some dynahi PSU PCBs, so I suppose I could always build one up, but I like the sigmas I've used a lot. I'll also be sure to check the board over at a lower gain by changing the 1K to 475 or so. I don't have a signal generator, so I'll see what I can do with my M-Audio Firewire box as far as checking compensation. I think Scott did some testing also, and he kindly is shipping me a pair of K devices to replace the one I stupidly fucked up. I'm finishing the 2nd board now at the posted values.

Yes I have two Dynahi PSUs which are powering my Dynamite. I plan to swap out the Dynahi boards with DynaFet boards

I have a spare dynahi psu board if you need one, I think. OF course its hard to find the regulators on that board now a days

Edited February 2, 2011 by DigiPete

[Pars]

Worked on the second board tonight, and then started doing some measurements etc. on the completed one.

I seem to be seeing some oscillation on it. With the input grounded, I am seeing something in the range of 220mVp at around 62.5MHz (scale 0.02us/div, 0.8 divisions... think I calculated the frequency correctly?). This is at the default gain (1K, 100 ohm, so 11?) with a 10pf compensation cap (mica). Also, decreasing bias as far as I could (224mV or 112mA) did not have any effect. I'm currently running at 150mA bias. Varying which feedback method was used made no change.

Any thoughts?

I do have the gain resistor (R55) as well as R17/R18 and R19/R20 socketed.

EDIT: After looking back at some of Jacob's posts, he seems to indicate that the paralleled VAS transistors might be causing some problems? These would be Q9/Q11 and Q10/Q12 on the schematic? I may try pulling one on each side and see what that does...

Edited February 3, 2011 by Pars

[DigiPete]

great stuff. I'm just a board stuffer, so I'll let you guys do the heavy lifting to figure this stuff out.

[Pars]

Pulled Q11 and Q12... no joy. Initially looked like it had done it, but as soon as I connect anything to the input, it oscillates. I'll poke around some more. I would guess that the board would not like running without feedback? I assume feedback will inject oscillation back into the input?

Any other input welcome. I have not looked at the rails yet, etc. Power is from a known good sigma22.

[Pars]

Pulled one pair of output devices. Oscillation seems to be gone as far as I can see on the scope. However, running a sine wave thru it, it is obvious that there is still some oscillation present looking at the waveform instability. From a DC perspective, the bias, etc. seemed stable. Looking at waveforms, it is obvious it is not.

Interestingly, I had pulled the 220pf input cap. This resulted in oscillation only when the volume pot was all the way down (or almost) or all the way up. Once I replaced the cap, this went away.

Summary of changes to this point:

1) R17-R18 2.2K

2) R19-R20 2K

3) R62 1K

4) VR3 2K

5) One set of VAS transistors removed (Q11 and Q12)

6) Removed one pair of output devices (so only one pair installed)

7) Minor deviation from BOM: the bypass caps spec'd as 10nf are implemented as 22nf ERO MKP1837.

Some of these changes are bandaids as the thing should run with all 4 sets of output devices, etc.

I'm running out of ideas...

Edited February 4, 2011 by Pars

[Pars]

Talked to Kevin via PM a couple of times tonight. He sent me a pic of his board, so I've been going though it and have a discrepancy. R31-R34 on his board are 100R. In the Eagle files I have (which match his prototype and not the final group buy boards), the Eagle schematic says 10R, which is what I used. However, if I highlight R34 on the board in Eagle, it says 100R (these are the combo 0805/RN55 resistor packages that Jacob used on the board). So, which is right? I think I will go with 100R and see what happens.

In the pic, for the Vbe multiplier, he has R17-R18 as 10K, I assume a 20K pot, and 4.75K for R19-R20. IIRC, he said in the thread that he changed the 5K resistors to 2K, so he may have replaced these since the pic was taken.

EDIT: Was looking at the wrong resistors. 10R is correct.

Edited February 5, 2011 by Pars

[Pars]

Did some more looking last night, and I am pretty sure the oscillation I was seeing was from the scope probes. It seemed more prevalent on the N-channel source resistors than on the P channel. I was clipping the scope probe to the output side if these and grounding the probe to the PCB ground. If I moved the scope probe perpendicular to the board, the oscillation goes away completely.

I was also testing the board with no load on the output, so I should probably use a load of some sort. I saw what looked to be oscillation on the tops of a sine wave, but I think I may have been overdriving it as it went away when the volume was reduced some from max on the pot. More looking today and will settle on bias current. I am thinking about dropping it down from the 150mA to 75mA or so. I don't think a stereo board pair will run off a single sigma22 at more than 150mA without doing some more serious heatsinking.

I have run some RMAA tests on the one board (in mono). When in mono mode, RMAA does some sort of phase tests (shown on the opposite channel graphs) which I don't know how to interpret. I'll try to post one and see if anyone can explain it to me or tell me if it is a good or bad result

Edited February 5, 2011 by Pars

[DigiPete]

It would be good if you have all 4 output devices on that board going for your RMAA test

If you could check the final configuration for both 150ma and 75ma bias at low gain approx 5-6, that would be great

[Pars]

Did some testing at both x11 and x5.5 gain. Still seeing some oscillation on the tops of the waves at full volume. Used my load box (not sure if it was amb or tangent that had posted that), with loads of none, 33 and 330 ohms. Waveform looks good at all volumes at 330 ohm load. Some oscillation present on 33 ohm load, and no load.

gain 11, volume at ~90%. Top is input, bottom is output

Gain 11, volume 100%. Putting my hand across the angle bracket and caps (above the board, not touching it) can eliminate most if not all the oscillation. Parasitics?

Also, before any of these shots, I removed all sockets (R17-R20), but installed a socket for R55. I also went back to R17-R18 at 3.3K, with a 20K pot. The oscillation seems to be worse the higher the bias. I think I will go for 75-100mA bias on mine. At 75mA and no load, most of the time there is no oscillation present.

At 150mA, gain 5.5, there is oscillation present at pot all the way down at all loads other than 330 ohm.

Edited September 2, 2017 by Pars