A friend here recently rebuilt his 555 and found a large difference in the DC output offset between the two channels and he asked me to measure my working, original amps for comparison. I will also explain the source of this error as time permits.
Here is my data:
Input open circuited which is the right way to do it for this amp, taken cold:
....... No Load 4 ohm Load
Right 23mV 23mV
Left 15mV 14mV
Here is my data:
Input open circuited which is the right way to do it for this amp, taken cold:
....... No Load 4 ohm Load
Right 23mV 23mV
Left 15mV 14mV
Last edited:
Did you look at the schematic?
Do you realize that you are not making any sense?
Most outputs have a low Z that is true, however they are usually
AC coupled with a cap and therefore have no bearing on the DC offset.
Shorting the inputs to the ADCOM is wrong for this measurement unless your
pre is DC coupled.
Look at the schematics and stop repeating yourself here.
No scales on the curve again.
Mods would you please delete his posts and keep him out of here.
He is thread crapping again.
Didn't your mother teach you that if you do not have anything "nice" to say, then to say nothing at all?
We all know your opinion of the GFA-555 amp...
PB2 - I recently acquired a GFA-555 (early production 64XX serial number) and I have 1.4mV / Right and 22.1mV / Left (no load). I measured mine with the input "open".
Last edited:
Nice work Nick, did you come to the conclusion that the 33.2R on the on diff amp was for better linearity, but they left the other so that the clip indicator would come on strong thus they wanted it more non-linear? I don't see any other reasonable reason ...
When you added the second 33.2 did you get better balance with matched devices?
We really have two simultaneous requirements, one is to have the diff pair split the tail current evenly. This design without a mirror doesn't do this very well - even with matched devices. and we want to bring the output DC offset as close to zero simultaneously. Output DC offset, even for a near perfectly matched diff pair (I came to this conclusion working on an amp with a dual device for the diff pair) is highly dependent on the VAS device matching the device used in the prototype were the value for R7 was chosen. Many designs of this type have R7 variable with a pot. What this does is make .5 of the total Ie provided by the current source and approximately delivered at the collector of the pos diff transistor bring the output to zero volts. It is otherwise dependent on many things, VAS Vbe and beta, and probably the matching of the rest of the devices - mainly Vbe. I would probably socket the VAS and select one to provide this result if the goal was to do it without changing R7 or adding a pot.
I believe that most other solutions such as a pot between the emitters of the diff pair skews the division of current between the two transistors.
I also believe that a servo that injects a small current to either base of the diff pair may drive the output offset very low, but it does not balance the current in the diff pair - in fact it mismatches it such that pos diff transistor provides the correct current to the VAS transistor.
These are all small errors, probably not greatly increasing distortion, but it is interesting, I think, to understand the ideal goals.
When you added the second 33.2 did you get better balance with matched devices?
We really have two simultaneous requirements, one is to have the diff pair split the tail current evenly. This design without a mirror doesn't do this very well - even with matched devices. and we want to bring the output DC offset as close to zero simultaneously. Output DC offset, even for a near perfectly matched diff pair (I came to this conclusion working on an amp with a dual device for the diff pair) is highly dependent on the VAS device matching the device used in the prototype were the value for R7 was chosen. Many designs of this type have R7 variable with a pot. What this does is make .5 of the total Ie provided by the current source and approximately delivered at the collector of the pos diff transistor bring the output to zero volts. It is otherwise dependent on many things, VAS Vbe and beta, and probably the matching of the rest of the devices - mainly Vbe. I would probably socket the VAS and select one to provide this result if the goal was to do it without changing R7 or adding a pot.
I believe that most other solutions such as a pot between the emitters of the diff pair skews the division of current between the two transistors.
I also believe that a servo that injects a small current to either base of the diff pair may drive the output offset very low, but it does not balance the current in the diff pair - in fact it mismatches it such that pos diff transistor provides the correct current to the VAS transistor.
These are all small errors, probably not greatly increasing distortion, but it is interesting, I think, to understand the ideal goals.
Last edited:
It is difficult to say what the designer was thinking (it appears Nelson Pass designed the original GFA-555, but Adcom may have added or revised the circuit after his initial design) regarding the input circuitry. I would guess that a lot of the knowledge that Doug Self has compiled in his books was not known or widely adhered to back in the early '80's and a lot of the designs were based on trial and error and empirical experience. Although not perfect, the input circuitry in the GFA-555 seems to be forgiving and works well given its simplicity. I ran a few AP curves before and after the resistor mod and there is no difference in THD, etc.
Initially, in the original (upon receiving the amp, I opened it and it appeared to be all original, never repaired) configuration, the DC off-set was 78mV left and 1~2mV right. After inserting the 33.2 ohm resistor in the inverting emitter leg, the off-set dropped about 20mV. "Unbalancing" the diff. pair dropped the off-set another 40mV. I used my meters Hfe setting to verify the diff. pair beta delta.
I also tried a near perfectly matched diff. pair and this did not affect the DC off-set in the channel, which leads to the assumption that other mechanisms are at work leading to the off-set.
It would be interesting to note the difference in performance and off-set if current mirrors were placed in the 555 input circuitry, and delete the clip indicator temporarily, and properly degenerate the emitters.
Interesting observations there.
I'm not sure that the clip indicator is hurting anything at all.
The collectors are roughly current sources and therefore don't
care much what you put in their path to the supply within
reason. The clip transistor B-E junction is more like the load
the VAS presents to the other side so it might actually help.
I'm not sure that the clip indicator is hurting anything at all.
The collectors are roughly current sources and therefore don't
care much what you put in their path to the supply within
reason. The clip transistor B-E junction is more like the load
the VAS presents to the other side so it might actually help.
The fellow who's amp I was talking about originally swaped the VAS transistors without it helping the problem. I should have known since he matched many things in his amps. This still does not disprove the claim that the balanced Ic of the diff transistor driving the VAS base with the R7 load has to provide 0 DC offset for it to perfectly balance with 0 DC offset.
A few more measurements for anyone interested should help shed more light on this:
Measure these voltages relative to the amp output pin:
Q17 base
Q13 base
Q12 base
Q11 base
These relative to ground:
Q4 collector
Q7 collector
Relative to the positive rail, these should probably be done with a 10K
resistor at the tip of the pos DVM probe so as to not put a heavy capacitive
load there and upset the amp stability:
Q7 base
Q5 base
A few more measurements for anyone interested should help shed more light on this:
Measure these voltages relative to the amp output pin:
Q17 base
Q13 base
Q12 base
Q11 base
These relative to ground:
Q4 collector
Q7 collector
Relative to the positive rail, these should probably be done with a 10K
resistor at the tip of the pos DVM probe so as to not put a heavy capacitive
load there and upset the amp stability:
Q7 base
Q5 base
Let me add these so that we have the full picture:
Measure these voltages relative to the amp output pin with
no load on the output :
Q17 emitter
Q18 emitter
Q19 emitter
Q20 emitter
Q13 emitter
Q14 emitter
Q15 emitter
Q16 emitter
All the previous measurements should also be with no load on the output.
Measure these voltages relative to the amp output pin with
no load on the output :
Q17 emitter
Q18 emitter
Q19 emitter
Q20 emitter
Q13 emitter
Q14 emitter
Q15 emitter
Q16 emitter
All the previous measurements should also be with no load on the output.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.