[palchiu]

Thanks everyone!

 

[SnoopyRocks]

I just finished two amps. A lot more debug time than I anticipated was necessary because I mistakenly ordered 24.9k instead of 24.9 Ohm resistors for the output stage. What really irks me is that I measured every single one by hand and still missed the little 'k' on the DMM each time. Sometimes you only see what you want to see.

After swapping in resistors with the closest values available at hand (20 Ohm) and adjusting R19 to get the output current right, everything works.

Here are some of my observations:

• They sound great - better than a pimeta (OPA627s, dual stacked buffers, tread, class A). Critical listening was not necessary to discern the distinctions, but I did have to AB to tell. The biggest difference is the in the bass, which is deeper, faster and clearer. Clarity has improved and I hear detail (such as smoother decay, lusher reverb, etc) that I had not before, which is really encouraging because this is apparent even with low bit rate mp3s with an audigy 2 zs upsampled to 24/96 in software with kernal streaming as a source. What was most surprising was that even when it was not working properly (high output impedance due to large output stage emitter degeneration resistors and very low output current), it still sounded better than an iPod Mini by itself, albeit with low volume.
  
  
• The pimeta measures MUCH better - linearity in particular (~20dB better SFDR). MCM sent pnps and npns from different gain classes, making matching problematic since the pnp Betas measured around 175 +-15, while the npn Betas measured around 250+-10. I'd like to test again with better matching if I can get some higher gain pnps or lower gain npns. Socketing the transistors sure was a great idea.
• The open loop gain was lower than expected but made sense after thinking about it. Here is where the thermal tracking diodes hurt. I measured an actual closed-loop gain of 2.8 for a feedback factor of (1/4). This translates to an open-loop gain of about 10. I wonder if the JFETs' gm is lower than expected because of low IDss as well...
• LED matching matching was not as big a deal as I anticipated, especially with the trimpots.
• Offset was not hard to get reduce. The AD820 works well as a substitute for the OP27 with lower power consumption. Typical measured offsets were less than 1mV after settling. Using the servo makes thermal bonding diodes superfluous (IMO) and considering the reduced open-loop gain, I would skip the diodes next if I were to do it over again.
• Casework is not particually fun and is difficult to get just right. It's off just enough to bother me, but would take a sharp eye to catch. Ughh...I miss having access to a propper machine shop.

Has anyone else, measured the performance (THD, SNR, etc) of their build or observed the low open-loop gain that I have?

Quote:

Originally Posted by bg4533 Has anyone else built a Dynalo in the Hammond case listed in the BOM? When mine has been in use for many hours it gets pretty warm. Maybe too warm. I need to take some temperature measurements. I just wanted to warn people of this.

I did use the Hammond case from the BOM but don't think that it gets hot. IMO, this is a nonissue.

 

[amb]

Quote:

Originally Posted by SnoopyRocks The open loop gain was lower than expected but made sense after thinking about it. Here is where the thermal tracking diodes hurt. I measured an actual closed-loop gain of 2.8 for a feedback factor of (1/4). This translates to an open-loop gain of about 10.

The gain of the Vas stage is actually less than unity (according to Kevin Gilmore, it's 0.5, and my calculations agree). The gain of that stage could be increased slightly by upping the value of the two 3.3K resistors going to ground at the collectors of the Vas transistors (or even remove those two resistors altogether, which only increases the gain of that stage to a little more than 0.6 by my calculations). Kevin states that the first stage has a gain of 50, which seems higher than what you actually measured. Decreasing the value of the 200 ohm resistors at the source pins of the input JFETs should give that stage a boost. While I agree with Kevin that too much open loop gain is not desirable, as it stands the open loop gain is probably too low. Changing those resistors to 100 ohms looks like a good mod to try.

 

[nikongod]

here are a couple quickies:

1: are the 2 heatsinks necessary? i saw a build that lacks them on the "post your builds page"

2: how rigidly set are the capacitor selections? im assuming not very, and hoping this to be correct. obviously type should be same (dont put an electrolytic where a film belongs...) but what about sizes?

thank you much

 

[palchiu]

I think if your case have venting hole, there no need these heatsinks.

I've burn the dynalo for 72 hours, they are not hot. My finger still can touch they and feel warm.

But the best way is put them on, maybe will extends they life.

 

[robzy]

20 - CAP - Monolithic Ceramic - 0.33uF, 25V or greater7.5mm - P4669-ND - C5, C18

Hey there, in the BOM it mentions the above capacitor, problem is the part number it mentions is not a Monolithic Ceramic but is actually a polyster film, should i go with poly film? or ceramic?

Rob.

 

[dgardner]

Quote:

Originally Posted by robzy 20 - CAP - Monolithic Ceramic - 0.33uF, 25V or greater7.5mm - P4669-ND - C5, C18 Hey there, in the BOM it mentions the above capacitor, problem is the part number it mentions is not a Monolithic Ceramic but is actually a polyster film, should i go with poly film? or ceramic? Rob.

This cap is the integrator for the DC offset servo. The type is not especially critical. Film caps are fine. Try to stay close to the 0.33 value or slightly larger.

 

[sft]

Quote:

Originally Posted by palchiu I've burn the dynalo for 72 hours, they are not hot. My finger still can touch they and feel warm.

had better make sure the bias current is near 16mA. The output transistors are hot after 1 hour.

 

[palchiu]

Thanks SFT!!! I'll check it again.

 

[robzy]

R65 & R66? What should i do if im using a 50k pot? Im totally confused.

[edit]: And for the 100uf electrolytics, all the BOMS i have state them as 25V or greater, but the Detailed Construction Guide i am using says 120uf 35v minimum.... so which is it?

Rob.

 

[amb]

Quote:

Originally Posted by robzy R65 & R66? What should i do if im using a 50k pot? Im totally confused.

Use a resistor value about 10x or more of the pot, so that the resistor doesn't alter the pot's taper adversely. If you're using a 50K pot, then use 470K to 1M ohm for these resistors. The purpose of these resistors is to prevent the JFETs' gate from floating (and losing bias) if the pot should ever become intermittent. That could cause headphone-damaging DC offset to develop.

 

[robzy]

Thanks for that

Hopefully i have one lying around.

[edit]: And just to reiterate my other question (now that its on the previous page: And for the 100uf electrolytics, all the BOMS i have state them as 25V or greater, but the Detailed Construction Guide i am using says 120uf 35v minimum.... so which is it?

Rob.

 

[amb]

Quote:

Originally Posted by robzy for the 100uf electrolytics, all the BOMS i have state them as 25V or greater, but the Detailed Construction Guide i am using says 120uf 35v minimum.... so which is it?

These are on the supply rails which don't normally get more than +/- 15V or 16V, so 25V caps are adequate. But, going to 35V caps ensures that if, for whatever reason (a weird failure or short circuit), the whole difference in voltage between the positive and negative rail were to become present on one of these caps (which would be 30V or 32V), then it would still survive. Since electrolytic caps often fail by exploding into a big mess, some voltage headroom is a good idea.

 

[robzy]

That makes sense, thanks a lot

So now that im getting new caps i decide taht getting ones with the right pitch diamater would be a good idea.

Problem is that the 100v Pana FMs have 5mm pitch... but a minimum order of 10 @ $2 each aint cheap... i guess ill just go for the 35v ones.

Rob.

 

[Garbz]

I don't think the value of the caps is too much of an issue. The output of the power supply if leads are kept short should provide appropriate current reasonably quickly.

Personally I've gone overbaord putting 6x 470uf 16v caps on each board.