Gilmore DC offset problems

[Garbz]

Hi all

I just fired up my Gilmore for the first time and now i'm working through the niggling problems. I just managed to tame the -ve opa548 by adding a 0.12uf capacitor between pin1 and 2 (-vin +vin). This brought it down to a manageable temperature and i can keep my hand firmly on the psus now without anything cooking. Generally it's still hot but then again that's what the heatsinks are for in the first place right

Anyway back to the problem. I've confirmed and noticed the following:
1. The voltage output is +16.25 -16.20 from the psu. Which seems acceptable enough.
2. The output transistors of the amp itself are hot to the touch, but not burning hot, as is the servo for which i've used opa227.
3. All LEDs light up

4. When various LEDs are in the output DC swings around a bit between +- 1.5v. The lowest i've managed to get it was 0.8v

I assume that when i remove the LEDs that the servo stops working as the voltage stablises. The right side has -1.147v and the left side has +0.988v all relative to ground.

The main problem i think was the transistor matching, I kinda had no idea what i was doing. The resistors were all matched very closely (i'm using 0.1% generics but they seemed generally pretty close to begin with). LEDs were also very closely matched but i found that i got much better results with LEDs which were not matched to each other. I don't know why.

I didn't understand how to match the transistors. The NPNs all gave me rougly 11.6v some swinging out to 11.9 and some as low as 11.4. This was apparently just as expected and i noticed that the the higher LEDs measured higher using the hFE measure on my DMM. The PNPs however were strange. FOllowing the diagram they all gave me 2.4 - 2.7v instead of the 5-6 that kevin said they should measure.

Anyway i can allways pull them off and remeasure. THe main problem i'm still having regards how to match them. The PNPs and NPNs give different voltage measures and have different hFE measures too. So how do i match them?
What i ended up doing was grouping all the transistors with similar hFE ratings together and grabbing the middle most set as this by all logic would mean they are pretty closely matched.

Anyway facts:
about -1vDC on right side,
about 1vDC on left side,
voltages swing all directions when LEDs are in place.
i'm lost for ideas

Any help would be greatly appreciated, i've been working on this amp for a good 3 months now in the RnD and i can't wait to hear the final product.

 

[Glassman]

the LEDs needs to be in their place, they are not just a servo components, they form CCS needed for the whole amp.. remove the OPA227, that way the offset should stabilise, then try different LEDs, you should find some with which the offset will be in the order of tens of milivolts, if you put the servo opamp back it will null such offset..

oh and one important note: are you testing with input shorted to ground??

 

[Garbz]

thanks a lot for the response glassman. For once the headwize forums were faster then head-fi

Btw what's the CCS?

Ok so far i've made a few changes:
I've added the pot which loaded the input to ground.
I removed the 1Mohm resistor from the output to the opa227.
From there i matched LEDs.
Then i put the resistor back in.

These are my results so far. Left/Right

1Mohm Disconnected - ~50mV/80mV
1Mohm Reconnected - ~20mV/25mV

This is as low as the offset gets. I assume there's still some components which are disagreeing with the servo unit as it didn't get all the way down to or close to 0. It also took about 40 seconds for the offset to get so low, it started around the 60mV mark.

So what can i change to make the servo work properly? Should i check LEDs again? Or is 50/80mV too much for the servo unit to handle?

 

[drewd]

Quote:

Originally Posted by Garbz Btw what's the CCS?

Constant Current Source?

-Drew

 

[dip16amp]

Yea, the LED is part of the constant current source.
I've seen the J109 and K389 cause the 50mV to 80 mV offset because the JFETs within them are not matched very well. Swapping them may not be easy in your case, so swapping LEDs that differ by .1 or .2 volts may do it or swapping 500 ohm resistors with higher or lower values like 10-20 ohms difference.

 

[Garbz]

all good now thanks everyone

0.9mV / 8.1mV or thereabouts after it settles for about 30 seconds

Btw i'm looking at an output protection circuit but since i don't want a relay in the signal path i was wondering how the Gilmore amp will handle a full short for about 20 seconds (output to GND)

On the one side everything on the output is behind a resistor mostly 25ohms, however the output stage is hot enough as it is.

If i short the output to GND for about 20seconds it should nullify all the DC offset untill it's stabalised. Path of least resistence and all that. Why take a 300ohm path when a hard short is availubale.

would that work?

/EDIT: Followup question does it matter that both LEDs are differnt brightnesses? The ones connected to the -ve rails are much brighter than the +ve however the offset is small. I assume it's all part of how it works? Does it effect the class-A biasing?

 

[steel]

Quote:

Originally Posted by Garbz 2. The output transistors of the amp itself are hot to the touch, but not burning hot, as is the servo for which i've used opa227.

I have a same problem on Gilmore Amp
the out transistor (4 1015&1815) are very very HOT near to more 90 C
How to solve this problem.
thanks

 

[00940]

Garbz : did you check the grades of your transistors ? Sounds like you have the same problem than I've here, 2sc1815 are y grade and 2sa1015 are gr grade (or the contrary). Transistors being mismatched creates offset and are (very) hot. I've got measures of roughly 50mv of offset/channel without servo and 4mv with servo.

 

[Garbz]

Hi steel your post looks familiar. I replied in the headwize forum but i'll add it here too:
Quote:

I re-etched the pcb, i had a shorted trace on the last one. But it happened again and after very close inspection one of the transistors was the wrong way around. Check the output DC, chances are that it's huge, like 1V or similar. In my case i had npns where pnps should have been. And in my case it was the transistors connected to the LEDs which were the wrong way around. I just missread the schematic when i put it together. I recomend that for all ur holes u put in machined sockets to allow you to easily swap transistors. It drove me nuts trying to match them. But after getting it to work and careful mathcing i got the DC down to 9.4mV and -8.3mV which zeros after about 40seconds. /EDIT: ALso check orientation against the schematic and not the board layout. THere was a problem in the original board layout.

00940: In my batch i had about 3 GR of each type and 22 Y grade. However i found that my pnp transistors typically had a hFE of 20 to 30 higher then my npns. In the end i decided to put it all in matched with the hFE reader on my multimetre, and then swap a few of the GR components in to level off the offset. It started at around 30-50mV which after mixing and matching a few GR parts, and then adding a heatspreader accross the top, ended up with the above measurments. Typically i get 0.2 - 0.5mV once the servo kicks in which takes about 30sec to 1 min to fully work.

pic:

 

[steel]

Quote:

Originally Posted by Garbz Hi steel your post looks familiar. I replied in the headwize forum but i'll add it here too:

yes, I search your post in headwize...

thanks
My pcb bought from "Headamp.com"
the offset with servo
1.x/2.x mV is fine and working fine too

LED 1.87V and 25R is 0.68V
means the idling current is 27mA per transistor
maybe
I must try to find solution to reduce the current