Group Build: Dynafet

[DigiPete]

From Digger

I think 10pF will be fine and I've tried it down to 4.3 gain I think. I have had some stability problems with my Wavetek. I'm shopping for another function generator now.

Still I have tried 10pF for the comp cap and used the square wave out of the scope calibrator(50mV @ 1kHz) and it looks perfect from a gain of 4.3 to 11. I'm gonna try lower than 4 this weekend sometime.

For the resistors used in the Vbe multiplier, where originally KG had 20k resist/20k pot/20k resist, I'm using 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 have had no problems at all thus far. The fact is I fully intend to build a 2 channel amp using the Aavid's just as a HP amp to maybe sell later sometime, and to use to tweak values for the rest of the gang as we get started in the build thread when Marc gets ready.

[j4cbo]

Gettin' there.

[digger945]

From Digger

I think 10pF will be fine and I've tried it down to 4.3 gain I think. I have had some stability problems with my Wavetek. I'm shopping for another function generator now.

Still I have tried 10pF for the comp cap and used the square wave out of the scope calibrator(50mV @ 1kHz) and it looks perfect from a gain of 4.3 to 11. I'm gonna try lower than 4 this weekend sometime.

For the resistors used in the Vbe multiplier, where originally KG had 20k resist/20k pot/20k resist, I'm using 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 have had no problems at all thus far. The fact is I fully intend to build a 2 channel amp using the Aavid's just as a HP amp to maybe sell later sometime, and to use to tweak values for the rest of the gang as we get started in the build thread when Marc gets ready.

I have since used a feedback setup with 1k and 681R(gain of 2.5) and 7 or 10 or 15pF(I'll recommend 10p just the same) comp caps. It looks the same on the scope(very good) any way you build it.

I'm still shopping for a function gen. to replace the Wavetek 188.

[Vortex]

This is my first post at this place, thanks for letting me in.

I have been reading through this thread in the last couple of days.

The efforts and collaborations of some DIY-ers here to bring this excellent project to the point we are now is very inspirational.

It’s good to see the first assembled PCBs are alive and kicking. Keep up the good works!

As some of you (Scott & Marc) know, I’m working on a balanced DynaFET too.

In order to properly understand the electronics circuits, I was trying to find the (final) schematic of this amp but I couldn’t find it in this thread.

(Hopefully I have not overlooked something). Much appreciated if someone can help me with this. Thanks in advance !

[j4cbo]

I don't have my laptop with the final schematic/board with me at the moment, but this: http://b.j4cbo.com/temp/gbf2.0rc1-sch.png is very close. I'll upload the one that matches the boards that were ordered when I get home.

I have two boards up and running now, with a single set of output devices each attached to s22-style heatsinks. Bringup was painful. When I first powered it up I got several volts of oscillation at 10 MHz - yikes! This happened on two different boards and with either feedback point, so I figured it wasn't just a build error. There seemed to be a lot of noise/hash/oscillation/something coming from the VAS stage, so I tried putting a 22pf ceramic between base and collector of one of the transistors on each side; that didn't fix things either. Colin suggested removing one of the paralleled transistors. After doing that it was entirely well-behaved. Later this weekend I'll take it in to a lab on campus with good function generators and get some measurements.

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.

The oscillation is worrying me a bit. Did anyone else have trouble with using two non-compensated transistors per side in the VAS stage? The only difference I can think of is that I'm using a different version of the PCB layout, but my SMT-only one should have far fewer parasitics, and the circuit itself is identical. I really want to figure out what the most appropriate value for VAS compensation is, but 22pF was the only low-value C0G chip cap I had on hand; I'll poke around and try to find some other values to test with.

But even now it's hard to tear myself away from listening...

Vortex: are you using the through-hole group buy PCBs, my SMT version, or something else?

[digger945]

I have two boards up and running now, with a single set of output devices each attached to s22-style heatsinks. Bringup was painful. When I first powered it up I got several volts of oscillation at 10 MHz - yikes! This happened on two different boards and with either feedback point, so I figured it wasn't just a build error. There seemed to be a lot of noise/hash/oscillation/something coming from the VAS stage, so I tried putting a 22pf ceramic between base and collector of one of the transistors on each side; that didn't fix things either. Colin suggested removing one of the paralleled transistors. After doing that it was entirely well-behaved. Later this weekend I'll take it in to a lab on campus with good function generators and get some measurements.

I have experienced the same problems with the prototype board. Using only a pair of compimentary output devices has helped a lot. Removing the DMM probes from the output devices was one main culprit of noise and oscillation(trying to measure Vgs + adjusting offset while looking at the output on the scope)

My dummy load is a pair of 75ohm 50W resistors parallel(when I'm not using HP's;))

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.

As mentioned earlier, I'm using almost the same values here. Basically I lifted them from the Borbely Servo 50 which uses similar values for the exact same fets as drivers for the output stage.

The oscillation is worrying me a bit. Did anyone else have trouble with using two non-compensated transistors per side in the VAS stage? The only difference I can think of is that I'm using a different version of the PCB layout, but my SMT-only one should have far fewer parasitics, and the circuit itself is identical. I really want to figure out what the most appropriate value for VAS compensation is, but 22pF was the only low-value C0G chip cap I had on hand; I'll poke around and try to find some other values to test with.

The only deviation from the norm that I have tried thus far is substituting IRF mosfets in the output. All oscillation/noise problems vanished.

I've tried everything from 7pF to 68pF for comp caps, and 10 to 33pF seem to be the best overall. I've had little luck with mid-feedback.

For feedback resistors, I keep the 1k the same and vary the other, but I think I'm gonna try something lower than 1k soon.

Vortex: are you using the through-hole group buy PCBs, my SMT version, or something else?

He's using GB through hole boards.

Lemme know if there is anything you want me to try. I can supply scope shots to give an idea of what's going on if it helps.

A few scope shots I took. These were using 8 outputs(4 N and 4 P on the one proto board) at the time(back in August I think).

[Vortex]

I don't have my laptop with the final schematic/board with me at the moment, but this: http://b.j4cbo.com/temp/gbf2.0rc1-sch.png is very close. I'll upload the one that matches the boards that were ordered when I get home.

...are you using the through-hole group buy PCBs, my SMT version, or something else?

As digger945 said, I'm using the GB through hole boards.

Thanks for the link to the schematic, btw.

[luvdunhill]

I would recommend more current, as the circuit is much more stable with more current. My suspicion is you'll have problems until you get real heat sinks in there, unless you use the IRF devices, which take a lot less voltage. How much bias did you did you have when experiencing oscillation? Also, what are you using for R19-R20 and what gain setting?

Edited November 8, 2009 by luvdunhill

[j4cbo]

It was oscillating at very low bias, about 25mA with no heatsinks. Now that I've got the larger heatsinks it's still running cool at 75mA; I guess I'll recheck the stability with the higher current.

R19-R20 are 2k, the series closed-loop feedback resistor is 1k, and the shunt resistor is 220R, so gain is 5.5 with either feedback point.

[DigiPete]

I'll be working to finish the group parts buy order details over the holiday period,

just too busy to work it right now.

[hitoridekimasu]

All right, many thanks. Please, keep us informed.

[atothex]

bump

[Torpedo]

Good idea atothex, otherwise it'd seem the project has died killed by the Pass one

[DigiPete]

I have an email out to digger and luvdunhill to help me finalize parts values.

Once I get that feedback, I will post the parts list and approximate cost for each board.

Any takers can then contact me and we will finalize the group parts buy costs and collect the money to place the orders.

Sorry for the delay.

Pete

[Torpedo]

no worries, we're just beneficiaries of your work, and can wait as long as needed Thanks for the update and the effort.

[atothex]

x2

My apologies if I came off as naggy.

[DigiPete]

no sweat. Sorry for not getting this done more timely.

[luvdunhill]

Pete:

In your case, I'd recommend building it and socketing a few parts. You're the outlier here. You're pushing it on heat sinking, PSU, and transformer, IIRC.

For everyone else, the recommendations in this thread are solid, but they will depend on how much heat sinking you have. There's no way we can predict that, or plan for that. For compensation, just err on the high side, you'll be fine.

I just don't think anyone is willing here is willing to figure out exactly the best part combination for every unique situation. Just too many variables. Take the first step and get a single board up and running and we can help you from there.

[DigiPete]

Marc, appreciate the offer to help with my particular setup.

As for the group parts buy, I can work it for the next few weeks, then I will have to turn it over to someone else, as I will not have time to work it again for many months.

[DigiPete]

Ok folks. Been working the spreadsheet here is where it stands right now.

Price of parts per conventional board is approximately $26 excluding taxes, fees and shipping,

this includes vishay dales, Nichicon Muze for the big caps, wima for the box caps. The OPA445s are out of stock at Mouser, so this adds $2 per board to get them from DigiKey.

Parts values are the standard values published when the boards were designed.

I have not recently priced the PRP resistor option, but if there is much interest, it can be done, but this adds about $10 per board per my last price check /don't know about availability.

One question is should we order overage on the resistors for better matching, etc?

[Pars]

Soniccraft has a 20% sale going on right now (prp)

[DigiPete]

Soniccraft has a 20% sale going on right now (prp)

hmmm... well that drops the cost to $8 more per board for PRPs

not smart enough, but are there key resistors (signal path) that could be replaced with PRP?

[luvdunhill]

hmmm... well that drops the cost to $8 more per board for PRPs

not smart enough, but are there key resistors (signal path) that could be replaced with PRP?

I'm never sure what signal path means. Shunt resistors are just as important as series resistors, and for whatever reason people like to ignore this fact.

[hitoridekimasu]

Now that we've got here, 2$ or 10$ more per board won't be a problem for me if I'm going to get a bit better sound (even if not). I don't know the opinion of the others, but this is my opinion. Thanks for your effort, Pete. It seems you are short of time.

[Pars]

It appears that Soniccraft's sale is over.