Hi Shelah,
Okay, that is acceptable - both channels. Channel 2 may either have transistors that happen to match, or the offset in each pair compensates for the other. Channel 1 needs the transistors matched. Either way, you are good to go for music. Recheck your bias settings and enjoy!
-Chris
Okay, that is acceptable - both channels. Channel 2 may either have transistors that happen to match, or the offset in each pair compensates for the other. Channel 1 needs the transistors matched. Either way, you are good to go for music. Recheck your bias settings and enjoy!
-Chris
I agree that you don't really need matched pairs, the DC offset pot should move the output enough to achieve close to zero offset. The schematic is distorted on my screen but it appears that the DC gain of the amplifier is 2. (1+R2/R1) The two 27k feedback resistors define this, in a non-inverting configuration. So if the input transistors are mis-matched by 20mV then the output will sit at 40mV.
In my book, if the outputs are < 20mV, I call it go enough. This offset likely varies over temperature. I wonder what this amp does 30 sec after power up vs 20 or 30 min. Now, match pairs would certainly help with this aspect. How about adding a servo (integrator)?
Your posted amp is nearly the same as my "spookyamp" or "Symetri".
I just did a basic match on my DMM , not really close.( beta 380 p /400+N)
A better way to adjust offset is by trimming the current of one gender's
red led CCS. Both my designs have a trimmer on both CCS's.
Your design trims the Re of one of the input pairs to achieve offset null.
Kind of a second rate method ! I've never seen that on a "leach" type
symmetrical design ... especially a high power class H OPS one !!
With the trimmed CCS's and 380-400Hfe , about 30+mv with the CCS's equal.
Usually just under a .1ma "tweak" to one CCS can get me <5mv.
Then , I plug my servo IC in ..... .1mv !!
The non-servo'ed design has been reported with @ 10mv offset following
that slight adjustment to the current source.
OS
Just ran my similar amp with the ground referenced cascode and
current sources.
Those mistakes each knock a full 20db PSRR off the design <500hz.
VERY bad ripple rejection with both referenced to ground.
Another thing I noticed was the use of a cascode / EF2. The higher Z
cascode performs best with an EF3.
I actually built a cascode/EF2 , sounded wonderful at low-medium levels.
High volumes , the cascode could not supply the EF2 with enough current.
"Beta droop".
Also , 4 green or red leds instead of the zeners (not ground referenced,either)
To get my 18V cascode in my blameless , I use a zener in series with
a blue LED .... the cascode references the LED's cathode , not the noisy
zener. A full hawksford cascode is better yet.
I don't mean to criticize your design, shelah ... but I can't believe they posted that design.
A better reference design is the leach or the more modern HK990. Same
design , but done correctly.
OS
The use of LEDs to set the input stage bias current is a bit curious. I see that a red LED is specified. Certainly a green or blue LED will likely give you too much drop, putting the bias at a level much too high. Even red LEDs have more the 1.2 to 1.3V drop (which is desired). I'm guessing the 100k resistors (tied to GND) and 680 ohm resistors set the current of each input pair. A matched npn or pnp pair would provide a better current source, where one of the transistors has its collector-base shorted, and probably wouldnt need the 680 resistors.
Alan, Shelah has proved my point I think. Beta match is critical. Tail current match is less of a problem.
Hi Shelah,
Good first steps. Yes, those offset numbers are fine to use your amplifier.
Look into ways of matching those transistors more closely. This will improve the sound quality and drop your DC offset numbers. I'm generally not happy until DC Offset is less than 20 mV. By close matching you can attain 5 ~ 10 mV DC offset and much improved sound quality. This last bit is optional for you, but now you can see how important matching the input diff pairs is. Remember, you now have similar transistors installed. They are not matched, so imagine what matching can do.
Anatech, this is something that I have wondered about a long time but felt too stupid to ask. Why would you have trouble to match the input trannies when you can use a variable resistor to adjust or balance the offset close to zero ?
Hi Shelah,
Good first steps. Yes, those offset numbers are fine to use your amplifier.
Look into ways of matching those transistors more closely. This will improve the sound quality and drop your DC offset numbers. I'm generally not happy until DC Offset is less than 20 mV. By close matching you can attain 5 ~ 10 mV DC offset and much improved sound quality. This last bit is optional for you, but now you can see how important matching the input diff pairs is. Remember, you now have similar transistors installed. They are not matched, so imagine what matching can do.
Anatech, this is something that I have wondered about a long time but felt too stupid to ask. Why would you have trouble to match the input trannies when you can use a variable resistor to adjust or balance the offset close to zero ?
Last edited:
the high offset channel can be improved.
Anatech has told you where to look and what to test for. Do it.
Anatech has told you where to look and what to test for. Do it.
Leaving unmatched input transistors in place and adjusting another circuit parameter simply adds another error that reduces the apparent output offset. BUT that leaves the imbalance in the input currents and leaves the unmatched pair and leaves the unmatched pair with different transconductance in the two sides.
This leads to excessive drift in output offset as temperatures change.
Sort the unmatched pair. Sort the imbalance of the LTP currents. Sort the base resistances. Then the output offset is least and just as importantly the drift in output offset is ALSO least.
Last edited:
the high offset channel can be improved.
Anatech has told you where to look and what to test for. Do it.
If after sorting the LTP transistors and you still have the high offset, then you need to look elsewhere for the error. Come back and confirm your result.
Anatech has told you where to look and what to test for. Do it.
If after sorting the LTP transistors and you still have the high offset, then you need to look elsewhere for the error. Come back and confirm your result.
Hi Andrew,
Many thanks ...
Hi Jan,
The differential pair is where the deviation from the input signal is subtracted from the desired signal. The balance between the two transistors directly affects how accurate the distortion signal will be subtracted so it can be inverted and therefore corrected. So balance here is critical to sound quality. If you have a DC offset problem the last place to correct for it is by unbalancing the diff pair! You are then trading DC offset for distortion. So, where should the DC correction take place? Not the differential pairs, that's for sure. You can servo the DC offset out in the Vas stage, leaving the diff pair(s) in balance.
A balanced diff pair will correct for some DC offset on it's own, but the offset will then be what is designed into the circuit. You can often figure out what the expected DC offset will be for a design assuming transistors with perfect characteristics (like the designer did).
So why would a trimmer control be a problem with mismatched parts? Well, you already know that the circuit isn't going to correct for distortion because it isn't in balance. Knowing this, why proceed at all? Fix the unbalance problem (match the damn transistors), then correct for any remaining DC offset by injecting a correction current or voltage after the diff pair. It it easier to do this in the Vas stage, but you could even do this in the pre-driver stage. I'd recommend playing with the Vas current, or use an active current source against the Vas transistors. Just make sure you can control the current.
-Chris
Many thanks ...
Hi Jan,
The differential pair is where the deviation from the input signal is subtracted from the desired signal. The balance between the two transistors directly affects how accurate the distortion signal will be subtracted so it can be inverted and therefore corrected. So balance here is critical to sound quality. If you have a DC offset problem the last place to correct for it is by unbalancing the diff pair! You are then trading DC offset for distortion. So, where should the DC correction take place? Not the differential pairs, that's for sure. You can servo the DC offset out in the Vas stage, leaving the diff pair(s) in balance.
A balanced diff pair will correct for some DC offset on it's own, but the offset will then be what is designed into the circuit. You can often figure out what the expected DC offset will be for a design assuming transistors with perfect characteristics (like the designer did).
So why would a trimmer control be a problem with mismatched parts? Well, you already know that the circuit isn't going to correct for distortion because it isn't in balance. Knowing this, why proceed at all? Fix the unbalance problem (match the damn transistors), then correct for any remaining DC offset by injecting a correction current or voltage after the diff pair. It it easier to do this in the Vas stage, but you could even do this in the pre-driver stage. I'd recommend playing with the Vas current, or use an active current source against the Vas transistors. Just make sure you can control the current.
-Chris
- Status
- Not open for further replies.