OK so I finished and measured two more channels, with the BOM values I originally intended, which is AAK's with minor changes such as PMA's recommendations for C14 when using the Toshiba devices.
No difference in measurements with the channel using Mike's values, and from playing with the connections I suspect only the 2nd HD is for real in the graph of the previous post.
What is likely happening is that the return current of the measurement probe, due to the way I have to use my soundcard because of the way it is wired, goes straight to the input ground. And on AAK's PCB this is where the feedback ground connects to the signal ground plane. So no mysteries as to why there should be artefacts when using a measurement setup such as mine, real HD is likely much lower.
No difference either whether using an earthed transformer, or one with diode lift from earth, not even in noise floor with my equipment.
The original problem was due to having heatsinks not grounded.
Values I tried for C3, 4, (220 to 330p) R8 and R9 (20 to 100 Ohms) and C14 (3.3p 10p) lead to indistinguishible results with my measurement capabilities, but since those are mediocre, it's best to refer to Mike's, AAK's and PMA's more extensive and better performed analyses.
In other words, to any others building AAK's model, forget any speculations I may have made / data presented (don't believe my graphs, not even the best ones). It all looks like it is working fine.
Just ground your heatsinks.
Many thanks for all your inputs, and special thanks to Chris 😀 😀 😀
No difference in measurements with the channel using Mike's values, and from playing with the connections I suspect only the 2nd HD is for real in the graph of the previous post.
What is likely happening is that the return current of the measurement probe, due to the way I have to use my soundcard because of the way it is wired, goes straight to the input ground. And on AAK's PCB this is where the feedback ground connects to the signal ground plane. So no mysteries as to why there should be artefacts when using a measurement setup such as mine, real HD is likely much lower.
No difference either whether using an earthed transformer, or one with diode lift from earth, not even in noise floor with my equipment.
The original problem was due to having heatsinks not grounded.
Values I tried for C3, 4, (220 to 330p) R8 and R9 (20 to 100 Ohms) and C14 (3.3p 10p) lead to indistinguishible results with my measurement capabilities, but since those are mediocre, it's best to refer to Mike's, AAK's and PMA's more extensive and better performed analyses.
In other words, to any others building AAK's model, forget any speculations I may have made / data presented (don't believe my graphs, not even the best ones). It all looks like it is working fine.
Just ground your heatsinks.
Many thanks for all your inputs, and special thanks to Chris 😀 😀 😀
Hi MBK,
I'm just happy you found the problem.
Buy the web site a beer! 🙂 Thanks for the offer though.
-Chris
I'm just happy you found the problem.
Buy the web site a beer! 🙂 Thanks for the offer though.
-Chris
Hmm.soon to try my "modules"(the amp mounted on the heatsink),so I should ground the heatsinks to the amps ground?But when I put togehter the chassis,I will connect the mains earth to the chassies.(Then the amps ground will be connected to the mains earth,which it should not be..
Or should I ONLY gound the heatsinks to the chassis/mains earth,not the amps ground?
Hi Ryssen,
-Chris
Yes, once it's going into a permanent case. On the test bench we often forget to ground the heat sink to anything.
-Chris
Hi MBK,
I'm currently checking THD, and it makes no difference with or without the heatsink grounded. It's never been a problem with the amps I've built.
But, I'm glad to here that's all it was. Now you can get on with building the rest of them.
Al
I'm currently checking THD, and it makes no difference with or without the heatsink grounded. It's never been a problem with the amps I've built.
But, I'm glad to here that's all it was. Now you can get on with building the rest of them.
Al
Hi AAK,
to be honest I didn't even think of it at first, in previous amps (chip, and Mini Aleph) I always managed to use smaller heatsinks per chip or transistor, mount the active device conductively, and isolate the mounting hardware. Never had a problem with it... Here I wanted to do it properly 🙄 but on the testbench didn't think of it, I didn't know it could make a difference... I have much more experience in line level / op amp circuits than discrete power amps, and never had that kind of issues there.
The THD I use it mainly as a proxy btw, since I don't have a scope it's my only means to check for basic "health". I'm not a THD gazer by itself.
I am really glad I did check though, as much grief as it was it would have been even worse had I not figured it out at this stage. Until the final case is done the amps will work for a while in a wooden enclosure and for sure I would *not* have grounded the heatsinks, I just would have kept wondering about the sound.
to be honest I didn't even think of it at first, in previous amps (chip, and Mini Aleph) I always managed to use smaller heatsinks per chip or transistor, mount the active device conductively, and isolate the mounting hardware. Never had a problem with it... Here I wanted to do it properly 🙄 but on the testbench didn't think of it, I didn't know it could make a difference... I have much more experience in line level / op amp circuits than discrete power amps, and never had that kind of issues there.
The THD I use it mainly as a proxy btw, since I don't have a scope it's my only means to check for basic "health". I'm not a THD gazer by itself.
I am really glad I did check though, as much grief as it was it would have been even worse had I not figured it out at this stage. Until the final case is done the amps will work for a while in a wooden enclosure and for sure I would *not* have grounded the heatsinks, I just would have kept wondering about the sound.
Decided to test one chanel tonight....:
And...it´s working!
This is what I got:
Supply:+-44v dc
Offset:142mv is this high or?
I had a 175 ohms resitor as R24,wich at lowest setting gave me 86mv at R27/28,and the sink got cind of HOT after a while..
Changed R24 to 200 ohm and the lowest setting is 59mv.(R27/28).
And it´s about 40-45 d/C,just have to find the maximum setting now....😎
The driver are getting pretty hot to,I have ordered some small sinks for them.
Now that it has been playing for a while,the dc offset is down to 50mv..
I have matched(bought them matched) all the transistors that neded matching except for the drivers and output transistors.
I aint going to judge the sound quality with the speakers on the foto....
Will try the other channel some day.
And...it´s working!
This is what I got:
Supply:+-44v dc
Offset:142mv is this high or?
I had a 175 ohms resitor as R24,wich at lowest setting gave me 86mv at R27/28,and the sink got cind of HOT after a while..
Changed R24 to 200 ohm and the lowest setting is 59mv.(R27/28).
And it´s about 40-45 d/C,just have to find the maximum setting now....😎
The driver are getting pretty hot to,I have ordered some small sinks for them.
Now that it has been playing for a while,the dc offset is down to 50mv..
I have matched(bought them matched) all the transistors that neded matching except for the drivers and output transistors.
I aint going to judge the sound quality with the speakers on the foto....
Will try the other channel some day.
Hi Ryssen,
Your DC offset is high. Your transistors are not properly matched - sorry. The DC value also drifts a lot! This is a sure sing of mismatched parts.
Buy a bunch and match them yourself. Your matches will be much, much closer. What are you using for a diff pair?
Isn't your supply voltage rather high? It is clearly too high for a normal SymAsym, I don't know about the AAK model.
-Chris
Your DC offset is high. Your transistors are not properly matched - sorry. The DC value also drifts a lot! This is a sure sing of mismatched parts.
Buy a bunch and match them yourself. Your matches will be much, much closer. What are you using for a diff pair?
Isn't your supply voltage rather high? It is clearly too high for a normal SymAsym, I don't know about the AAK model.
-Chris
Agree on the DC offset: with the grounded heatsinks I now got 0.1 and 0.2 mV (!) on the latest pair I built. I thought the DMM was broken... Also from my troubles I recommend to keep the input ground absolutely clean, it really is very sensitive because the feedback ground trace on AAK's PCB directly connects to it. So to ground the heatsink I would use the power ground, not the input ground as I see on your pic.
FWIW when I bought the small signal transistors (unmatched) they came on a tape and were remarkably close out of the box so to speak, out of 40 only half a dozen were outside a 10% bracket for both Ic and hfe. So it was no problem to match them, I did 2 sets of measurements of Ic and hfe for the 2 required currents, 1.5 and 2.5mA, and chose the pairs that matched either current the best.
The output transistors were more variable, but still out of 12 pairs I got 8 pairs that were excellent matches, 5 or 6% off only. I matched the output pairs for Ic, hfe and Vbe at 150 mA.
(didn't match the drivers though).
FWIW when I bought the small signal transistors (unmatched) they came on a tape and were remarkably close out of the box so to speak, out of 40 only half a dozen were outside a 10% bracket for both Ic and hfe. So it was no problem to match them, I did 2 sets of measurements of Ic and hfe for the 2 required currents, 1.5 and 2.5mA, and chose the pairs that matched either current the best.
The output transistors were more variable, but still out of 12 pairs I got 8 pairs that were excellent matches, 5 or 6% off only. I matched the output pairs for Ic, hfe and Vbe at 150 mA.
(didn't match the drivers though).
Hi MBK,
Also, you used FET transistors for your diff pair, didn't you? They will also reduce the normal DC offset voltage due to extremely low gate current. Normal BJT's have a "designed in" DC offset that you can calculate. Generally you will be withing a few mV from a simple calculation.
-Chris
That can matter. We've already seen that.
Also, you used FET transistors for your diff pair, didn't you? They will also reduce the normal DC offset voltage due to extremely low gate current. Normal BJT's have a "designed in" DC offset that you can calculate. Generally you will be withing a few mV from a simple calculation.
-Chris
Hi Chris,
The drivers, I forgot about them and then was too placid to pull out the rig again whilst in the middle of stuffing the boards 🙄
No FETs here, I am using 2N5551 as input devices. I am so surprised by the DC offset: the initial 3-5 mV seemed very good already, now this 0.1 mV reading I got really stunned me.
The drivers, I forgot about them and then was too placid to pull out the rig again whilst in the middle of stuffing the boards 🙄
No FETs here, I am using 2N5551 as input devices. I am so surprised by the DC offset: the initial 3-5 mV seemed very good already, now this 0.1 mV reading I got really stunned me.
Tried that now..And the result was with shorted input offste:160mv,
with open input offste:15mv...!? (this was with cold amp)
Why is that?
No FET´s yet...I´m using 2n5551,I have a bunch of them maybe 20-25,I will try to match some,will it be enough to use the e,b,c, "thing" on my DMM?Or else how?
Edit:If matching the smallsignals is not enough,(I will "probably" not match the drivers and output,would the use of a DC servo be a solution?Or would it degre the sound?
Öhumm,Yes I don´t like them..but I added one and..now with open/closed input,dcoffset ~10mv (warm amp).Never seen that before,could you explain why?
Guess I´ll be satisfied with that low offset then..
The AAK verision is said to work all the way to +- 50v.
Ryssen said:
???
You can ONLY skip inputcap for topologies using fet-input or fully symetrical input, all other topologies have an offset at their input by nature (base current must flow somewhere). NEVER skip inputcap for these amps. Never. Really, never.
Mike
Hi,
the input has an input offset current.
This current must flow to somewhere. Most go to ground through a resistor.
Adding that resistor generates an input offset voltage.
This generates the output offset voltage.
From the above the input offset current and the resistor to ground determine the output offset voltage.
Now go and change the resistor value. What do you think the effect will be?
Now look at the input to the amplifier.
There are two connections to the base of the non-inverting input transistor. One input is connected to ground via that resistor we were talking about.
The other input is connected to the RCA via a resistor and capacitor.
The capacitor blocks DC coming from the source. It also isolates that extra resistor from source. If you short the RCA input to ground with the cap in place the extra resistor has no effect on the DC input conditions that the transistor base sees.
But if you short out the cap and short the input to ground you have added another route to ground. The original resistor and the extra resistor are now in parallel. What happens when you change the value of the resistor to ground?
If your source has a DC blocking cap in it's output then that source cap isolates the extra resistor from the source ground. You do not need a second DC blocking capacitor. But if you ever change the source or add a shorting plug to the input RCA then you will change the output offset.
Take care. If necessary play safe and use double DC blocking, but make changes on the basis of an informed decision, not through lack of knowledge/guesswork/hearsay/other un-informed do gooders.
the input has an input offset current.
This current must flow to somewhere. Most go to ground through a resistor.
Adding that resistor generates an input offset voltage.
This generates the output offset voltage.
From the above the input offset current and the resistor to ground determine the output offset voltage.
Now go and change the resistor value. What do you think the effect will be?
Now look at the input to the amplifier.
There are two connections to the base of the non-inverting input transistor. One input is connected to ground via that resistor we were talking about.
The other input is connected to the RCA via a resistor and capacitor.
The capacitor blocks DC coming from the source. It also isolates that extra resistor from source. If you short the RCA input to ground with the cap in place the extra resistor has no effect on the DC input conditions that the transistor base sees.
But if you short out the cap and short the input to ground you have added another route to ground. The original resistor and the extra resistor are now in parallel. What happens when you change the value of the resistor to ground?
If your source has a DC blocking cap in it's output then that source cap isolates the extra resistor from the source ground. You do not need a second DC blocking capacitor. But if you ever change the source or add a shorting plug to the input RCA then you will change the output offset.
Take care. If necessary play safe and use double DC blocking, but make changes on the basis of an informed decision, not through lack of knowledge/guesswork/hearsay/other un-informed do gooders.
Hmm,there´s 40mv dc on the output of my iAudio porttable that I use for testing,so if no cap on input (if I forget about the base current)would be 1,6v at the output (speakers)?(If Symasym aplification factor=40.)Just trying to learn..
Edit:I am going tu use an input cap,just teoretical..
Edit:I am going tu use an input cap,just teoretical..