The Multi Amp aka Dynalo Mk2

[Pars]

^ probably. At 1% tolerance, a 390R could be as high as 394 and a 402R as low as 398.

A better question is where are you sourcing the MPSW06/56 transistors from, since they were discontinued awhile ago. Most of these that you might find on eBay/alibaba, etc. are most likely fakes.

[thuytn]

Thanks, Pars.

There is a local shop here who still has a couple thousands of each. All in original boxes. 

[Pars]

Just a follow up on my SS Mini build where I used the MPSW thru-hole transistors instead of the PZTA SMD version. Listening tonight, after 1 hour or so, the chassis top measured a max of 46°C, and scanning inside with the IR gun showed a max of 66°C. I did not remove the top to scan individual devices, but placing my hand on the top ( after 1 hour and 2 hours), it seems like it runs considerably cooler than I remember the SMD versions that I built. Sounds great as well, though I think the same as the SMD version.

 

 

 

[MASantos]

So I finally got to finish my dynalo mini build.

It is playing nice with 2SK170/2SK79 input jfets, 420ohm led resistors and 255 bias resistors. Temperature with cover on is 103C  on left side (- channels) and 81C  on right side  (+ channels) transistors, is this sort of difference normal?  I checked bias and L- is around 340mV while R+ is around 290mV. could this have anything to do with the offset trimpots?

All the bias resistors 255ohm and led resistors (420ohm) measure within spec. Voltage drop on the leds around 1.65V

When I remove the lid the temperatures drop quite a bit to around 60C-75C which is understandable.

Sounds great though with my sennheiser HD650 recabled with 4 pin XLR. I would just like to understand this diferrence between plus and minus channels

 

[Pars]

340mV is rather high for a mini (17mA). What does your DC offset look like without the servo opamps in?

I would check the bias across every 20R resistor for starters. Reset all trimpots to midpoint (with an ohmeter) and go from there. The 2 pots by the 255R bias resistors can cause some wonkiness if you go too far in the adjustment of them (same with the full size amps).

[MASantos]

I measured the bias across all the 20R resistors and they're all flowing between 270-340 after a few hours with the top plate on. there are some variations in each channel between the 20R resistors, some might go to 340 and the one next to it only to 310. could this be transistor variation?

I had the DC offset dialed to about +-5mV before installing the jumpers on the back side of the board. 

I'll remove the jumpers tomorrow and reset the trimpots. 

[Pars]

That is channel to gnd DC offset? The +/- offset pretty much takes care of itself. I wouldn't mess around too much with the 2 trimpots by the 255R's. From what I found, you can chase your tail trying to get it closer and closer and then wind up with one of them at a near dead short and things go south quickly from there. The differences you are seeing are probably transistor variances. The resistors from the BOM are all 0.1%, so should be fairly close. The ones I built, I got offset dialed into +/-2mV without the servos in.

[MASantos]

Channel to GND DC yes.

So you just play with the 2 trimpots close to the LED's to adjust offset?

I'll reset the 4 trimpots close to the input transistors to midpoint and go from there. I'll need to remove the 0 ohm resistors first. 

[Pars]

 

On 10/20/2016 at 4:52 PM, mwl168 said:

The 3 trim pots on each channel adjust the output DC offset. You want to get the offset as close to 0V as possible reference to ground. The adjustment needs to be made without the servo engaged.

If you are using the TO92 version boards, the pot next to the LED by the power connector affects the output of both phases while the 2 pots in the middle of the board each adjust one phase of the output. I first adjusted the pot next to the LED so that the inverted and non-inverted output offset are fairly evenly spread in reference to ground, I then adjusted the pot on each output phase to bring the offset close to 0V. The goal, when you done adjusting, is to get the DC offset between the two output phases as close to 0V as possible. This is very tricky work due to the constant thermo drift so experiments and patience will be needed.

There are quite a few related discussions in this thread if you have not already read them.       

 

On 11/30/2018 at 4:01 PM, Pars said:

Looking at the amp from the front, the channel layout is
L- L+
R- R +

The two pots by the LEDs balance the +/- for that respective channel. The other 4 pots adjust the channel closest to it. There is some cross action with the other phase but not a lot. Normally you don’t need to do much with these. You can also get yourself in trouble with these if you go too far. If that happens, reset to midpoint (using a DMM on ohms) and start over.


Sent from my iPhone using Tapatalk

 

[MASantos]

This might be a dumb question, should I short the inputs?

[Pars]

Yes. If you have the volume pot in, just turn it down all the way and that is good enough.

[MASantos]

Success! I managed to get offset to within +-1.5mV on both channels. Installed the servo jumpers and music is great!

One question: I let the amp cool down and then measured offset during startup. It got very high, above 150mV and descended in about 20-30 seconds to +-1mV, is this behaviour normal even with the servos in the circuit? MIght this damage the headphones if plugged during startup?

[Pars]

Congrats!

When you saw the offset spike, were you checking the single-ended output (L+ or R+ to gnd)? If you look at the balanced offset (L+ to L- or same for R channel), you won't see this. It is something I noticed awhile back and is caused by the Meanwell bricks not coming up at the same time on power-on. Balanced phones should be fine (common mode or balanced offset won't spike). I would not leave single-ended phones connected on either power up or power down.

On 5/22/2019 at 9:33 PM, Pars said:

Re: Mini Dynalo

I thought I had posted this here, but apparently I didn't (had some PMs going at the time).

When I was getting ready to ship the last mini out that I built, I noticed an odd problem in that occasionally DC offset on the SE would jump to ~500mV at startup. To make a long story short, I determined via scope that the Meanwell bricks don't always start up at the same time. When the + comes up first, the servo opamps freak out briefly, causing the offset. Without the 0R jumpers in you'll never see this. And it only affects SE output; the balanced output offset is always fine.

A month or so ago, while perusing diyaudio, I ran across a comparator circuit used for power management in, of all things, the O2 amp. This circuit, placed before or after the regs, with proper parts selection for voltages, would cure this "problem". I realize virtually no one would build one of these for SE output use, but one thing to be aware of. If you are using SE phones with these, wait about 30 secs. before plugging them in.

I did PM this info, and this circuit to Kerry, but he has a lot of fires burning and redoing the mini probably isn't even on his radar

[Kerry]

1 hour ago, Pars said:

Congrats!

When you saw the offset spike, were you checking the single-ended output (L+ or R+ to gnd)? If you look at the balanced offset (L+ to L- or same for R channel), you won't see this. It is something I noticed awhile back and is caused by the Meanwell bricks not coming up at the same time on power-on. Balanced phones should be fine (common mode or balanced offset won't spike). I would not leave single-ended phones connected on either power up or power down.

Yeah, I don't have time at the moment, but If I were going to fix this, I'd just use the enable function of the TPS7A regulators with a timer.

PS  I did these boards in DipTrace and have not converted them over the Eagle yet (which I use exclusively these days), so It's a complete redo. 

[Pars]

Thanks Kerry. I'm not interested in redoing mine. As long as I am aware of it, its fine

 

[MASantos]

12 hours ago, Pars said:

Congrats!

When you saw the offset spike, were you checking the single-ended output (L+ or R+ to gnd)? If you look at the balanced offset (L+ to L- or same for R channel), you won't see this. It is something I noticed awhile back and is caused by the Meanwell bricks not coming up at the same time on power-on. Balanced phones should be fine (common mode or balanced offset won't spike). I would not leave single-ended phones connected on either power up or power down.

I was measuring L+ L- to GND. I didn't measure between L+ and L- I'll check that quickly. 

No single ended phones atm so I'm safe there. 

[MASantos]

No offset between positive and negative sides on the same channel!! Cool!!

[Pars]

14 hours ago, Kerry said:

Yeah, I don't have time at the moment, but If I were going to fix this, I'd just use the enable function of the TPS7A regulators with a timer.

PS  I did these boards in DipTrace and have not converted them over the Eagle yet (which I use exclusively these days), so It's a complete redo. 

Kerry,

Since many (most?) used Amb external regulators instead of the onboard, most would not be able to access the enable pins/function of the regs. Just a thought.

[judo]

I did not see Amb's board but some of the available little tps boards have enable pin.  The fix can use them(or the mini GRLVs ? ) too paralell with the onboard regs maybe .  You can replace the Amb's and use them for other purpose.  Brainstorming for the cost of somebody else I know.

 

[Kerry]

11 hours ago, Pars said:

Kerry,

Since many (most?) used Amb external regulators instead of the onboard, most would not be able to access the enable pins/function of the regs. Just a thought.

My evil plan is to get people used to smaller and smaller parts until those regs look big 🙃

[Pars]

Curious... did you ever successfully install a couple of these regs on those 3oz boards? I know a couple of people did, but I never did get one to work. Ground planes with 1/3 to 1/2 of the pins tied to ground on these little parts seems to be to be somewhat challenging.

[Kerry]

I've done this a number of times, but on 1oz copper.  I used a stencil with an oven, which really helps.  I've also de-soldered and re-soldered them.  Another way is if you let the paste melt on the board first without the chip and then apply flux and heat, you'll be sure to have the right amount of solder on the pads.  

[judo]

To the heat issue:
I have a "standard" mini from the first group buys. Case, the unfortunate 3 oz board, 402 led resistor, 255 bias resistor, 15 mA bias. I used it with offboard regulators and without the top cover mainly for a time. Now I squeezed in the onboard regs second times a bit better with the same heatsinks but only with open lid. I would not like to reduce the bias so I experimented with @Skooby's resolution(see in this thread) a little. In my version there is almost 10 mm the distance from the transistors's heatsinks to the lid so I made a "heat bridge" from two 5 mm thick aluminium pieces as you see in the picture. The result is more than I hoped. Measuring the temp on the heatsink hotter side:  with the top cover 85-86C; without the cover 70C. With the "heat bridge" and the cover 63C. The absolute temperature is depends on the ambient, the music, the headphones a bit but the more than 20C gain is stable. This was a surprise to me, the mod is easy maybe helps somebody. You can secure it with screws, widen it etc.. Maybe later. It is stabil mechanically I am happy with it as is. Thanks for Skooby!

The temperatures were measured after cc. 90-120 minutes when they were stable.

Edited April 14, 2020 by judo
a remark

[Pars]

Interesting. Where you have the temp probe was always the hottest area when I was using an IR gun. Maybe a bit further left. Any reason you didn't populate the rail caps (above the right heatsink)?

[judo]

1 minute ago, Pars said:

Interesting. Where you have the temp probe was always the hottest area when I was using an IR gun. Maybe a bit further left. Any reason you didn't populate the rail caps (above the right heatsink)?

a long story They were popped and  I always  forget to order them.