How to fix DC offset with no trim pots available?

[Moving Ahead]

Hi, I'm new this audiokarma and I'm new to repairing vintage amplifiers. I have hit a brick wall and I'm in need of some technical help.
I have several amps (early 70's) without variable trim pots to align the DC offset.
1/ What can I do to the circuit to better align the DC measurements in such circumstances?
2/ Is there a particular element (ie: cap/ resistor/ transistor/diode ) that the manufacturer put in place instead of a trim pot to regulate DC offset?

This question is directed not at one particular model amp but rather as a general means of amplifier repair. I'm also not concerned to get DC offset perfect in these circumstances but if I can get it somewhat closer to zero, I'll be willing to give it a go.

I hope I can get some help with this, as I feel I need to overcome this problem before I can proceed to the next level.
Many thanks for your help with this problem in advance.

 

[EchoWars]

The input stage of most SS power amplifiers is the point where any DC offset present at the output is corrected for. The most popular way of accomplishing this is through an input stage incorporating some form of differential amplifier in which the two transistors are closely matched for gain and other characteristics.

Read: http://www.audiokarma.org/forums/showpost.php?p=44429&postcount=33

 

[mech986]

 

[pchilson]

I'm interested in this as well. I always see the advice to replace the differential pair as a fix. I've done that and yes the DC offset is lowered but not zeroed. There must be more to the story. Why is the DC there to begin with?

 

[EchoWars]

Because nothing is ever perfect. Once mfgr's switched to split-rail DC coupling for SS output stages, the DC blocking capacitor used in early SS single-supply designs went away. Now it's a matter of using high-gain input stages to track the output DC as well as possible (and associated protection circuitry to protect the speakers if something goes wrong).

Mfgr's have gone to great lengths to minimize DC offset in their designs, and for the most part are successful in keeping offset in check (even most cheap SS amps built from about '75 on have the ability to hold offset under 20mV). But bipolar transistors all have base currents, and these currents translate to voltage errors when these currents flow through resistors. So without the ability to zero out these error voltages with an offset trimpot, the ability to minimize offset is limited by the gain of the transistors (more gain = less base current = less error voltage) and the design of the circuit.

 

[pchilson]

Thanks. Thats all well and good but when you have a DC offset that is not reduced to less than 20mV after you have replaced the differential pair with a "matched" pair, there is still more to the story.
Example... I have a Kenwood KR-5030. When I acquired it the DC offset readings were, left 100mV, right 23mv. I replaced the diff pair with a set that were matched at hfe 453 and 455. The left channel came down to 52mV and the right remained at 23mV. So yes, there was improvement but it didn't solve the high mV on the left channel. Is this simply because the hfe is not exactly the same or is there more to the story?

Thanks

 

[blhagstrom]

Extrapolating from what Glen said, beyond a closely matched set of diff pais, its time to look the tolerances on the rest of the parts in the signal path.

For example, check the resistors and match those to the spec'd values.

Some resistors are born with actual value a bit different than the spec and some old resistors drift from the desired value.

 

[dr*audio]

50mV may be as good as it can be. You could try replacing the electrolytic cap in the bottom leg of the feedback divider. If it is leaky it could give you some gain at DC. 50mV isn't going to damage your speakers, though.

 

[fred soop]

One method used on a lot of amplifiers that actually have an adjustment is to connect a relatively large resistor to the base of the transistor on the feedback side of the differential pair, maybe on the order of 56K. The other end of this resistor is connected to the wiper of a trimpot of maybe 100 to 500 ohms. The ends of that trimpot are connected to the +/- supply rails through large resistors, again on the order of 56K. Value will depend on the rail voltage. You should end up with maybe +/- 2 to 5 volts at each end of the trimpot. Both ends of the trimpot may have a capacitor, maybe 10 uf, to ground. More sophisticated designs may also have zener diodes.

The end result is that the trimpot is adjusted to add a compensating voltage to the feedback side of the differential pair to balance the circuit. The high resistance values ensure that there is little effect on the loop gain of the amplifier.

With this addition, you should be able to get the offset down to less than 10 mV. If not, then, as suggested by others, there is probably something out of tolerance.

With a vintage unit, there could be leakage across the input capacitor, which would add a DC bias to the input side of the differential amplifier.

 

[EchoWars]

Most have a DC blocking cap in the feedback loop to limit closed-loop DC gain to '1' (Cm19 and Cm20 on this receiver). A leaking cap here will cause DC at the output to be excessive and unstable.

 

[sregor]

A few years ago, someone over in DIYaudio.com started a thread for a yamaha receiver with no DC adjust. Analysis found that with the original value resistors, the offset should have been about 50 mV, which it was. The offset was caused by a difference in the collector currents in the input pair transistors, and the poster was able to get the offset to less than 5 mV by changing the resistor supplying current to the input pair to a value which gave equal current to both transistors. So sometimes it is a design problem and can be improved, but it does require understanding of the circuit.

 

[Moving Ahead]

That's Gold. - It's been 24 hours since I started the thread and it looks like a hot topic indeed. I want to thank you guys for getting my grey matter started again.
I'm no more than a amateur buff at this game however I'm picking up some leads to investigate through your comments. I'll get some of this and put it onto the workbench to try and nut it out. Hopefully it won't get the better of me.
Thanks again.

 

[fred soop]

Most have a DC blocking cap in the feedback loop to limit closed-loop DC gain to '1' (Cm19 and Cm20 on this receiver). A leaking cap here will cause DC at the output to be excessive and unstable.

What actually happens is not that the bad or missing cap causes offset but without it, the DC gain is now greater than 1 and the small offset that may be present at the input is then amplified by the same gain as the normal audio frequencies. In this situation, the OP has no offset adjustment and this capacitor could very well be a problem.

Some amplifiers, (Harman Kardon Citation 22 and 24 for example) do not have a capacitor in the feedback loop. However, this design absolutely requires the trimpot adjustment. Low frequency response is limited by the input capacitor. But, using an electrolytic capacitor to determine frequency response should not be done because of the high tolerance (usually 20%) and electrolytic capacitors introduce distortion unless the cutoff frequency is well outside the audible range.

 

[fred soop]

A few years ago, someone over in DIYaudio.com started a thread for a yamaha receiver with no DC adjust. Analysis found that with the original value resistors, the offset should have been about 50 mV, which it was. The offset was caused by a difference in the collector currents in the input pair transistors, and the poster was able to get the offset to less than 5 mV by changing the resistor supplying current to the input pair to a value which gave equal current to both transistors. So sometimes it is a design problem and can be improved, but it does require understanding of the circuit.

Better amplifiers will use a current mirror to enforce this current to be balanced. Bob Cordell in Designing Audio Power Amplifiers has one chapter where he starts with a common conventional circuit very similar to that found in the Heathkit AR-1500 receiver and proceeds to add current sources, current mirrors, emitter follower VAS, and other relatively minor changes. The end result is a circuit where the original distortion was reduced by almost 3 orders of magnitude.

The starting point had relatively low distortion at 1 kHz into an 8 ohm load (0.01%) but was up to 4.1% at 20 kHz and 2 ohms (yes, many 8 ohm speakers will drop to very low values at some frequency). The final result for the modified configuration was 0.00006% and 0.014% for the same 2 conditions.

 

[Moving Ahead]

I had a fiddle on one of my amps. Here are some of my observations. Note: all electroylic caps have been replaced.
DC offset: left- 6.5mV, right 16mV. (I know its already low but I'm more interested in learning how these things work.)
From here I had a look at the resistor(s) that the input differential emitter's share. The original value was 3K9 +/- 5%. The actual measured values were 3K92 and 3K98. So I changed them for new matched ones. Now: DC offset: left- 3.6mV, right 14.3mV. Hmmmmm Not bad for 2 cents.
Next I went in to change input diff transistors. They are:
origi: 2sa640, -50v, -50v, -5v, Ic-50mA, Pd 250mW, 50min fT, 320hfe, Nf4dB
new: 2n5087, -50v, -50v, -3v, Ic-50mA, Pd 625mW, 40min fT, 360hfe, Nf2dB
Now: DC offset: left- 23.8mV, right 21.5mV. There's some gain but now they are closely balanced.
This now brings up a question. Which transistor characteristic (ceo,pd,hfe,ft....) would I look at if I wanted to reduce DC output? Any answers on that one?
So far so good, I'll experiment more in the coming days to see what can be done.

 

[dr*audio]

I had a fiddle on one of my amps. Here are some of my observations. Note: all electroylic caps have been replaced.
DC offset: left- 6.5mV, right 16mV. (I know its already low but I'm more interested in learning how these things work.)
From here I had a look at the resistor(s) that the input differential emitter's share. The original value was 3K9 +/- 5%. The actual measured values were 3K92 and 3K98. So I changed them for new matched ones. Now: DC offset: left- 3.6mV, right 14.3mV. Hmmmmm Not bad for 2 cents.
Next I went in to change input diff transistors. They are:
origi: 2sa640, -50v, -50v, -5v, Ic-50mA, Pd 250mW, 50min fT, 320hfe, Nf4dB
new: 2n5087, -50v, -50v, -3v, Ic-50mA, Pd 625mW, 40min fT, 360hfe, Nf2dB
Now: DC offset: left- 23.8mV, right 21.5mV. There's some gain but now they are closely balanced.
This now brings up a question. Which transistor characteristic (ceo,pd,hfe,ft....) would I look at if I wanted to reduce DC output? Any answers on that one?
So far so good, I'll experiment more in the coming days to see what can be done.

You have to measure HFE and match them as closely as possible. I don't think you can get the offset any better than it was before you changed the transistors.

 

[pchilson]

Kind of thinking out loud here...

If the hfe matching is critical to achieving a zero dc offset, were the assembly lines matching these pairs or just "next one up" installing? If they weren't matching was the theory that the protection circuits were there to handle the mismatch so to speak?

In my Kenwood KR-5030 when I pulled and measured the diff pair I saw hfe readings of 539 and 399 on the pair that design would say should be matched. Is this component drift or just the way it is?

On said receiver there is also a second set of "pairs", Qm5 and Qm7. Should these also be matched?

When I replaced the Qm1 and Qm3 pair with a hfe 455 and 453 pair the offset on that channel halved what it was from 100mV to 52mV. I find no resistors in or around the circuit section that are out of spec value wise. I spot replaced Cm19 to eliminate any possible dc leakage and there was no change.

I realize 50mV is not going to blow up the speakers but just trying to understand the design theory and and resultant tolerance the circuit designers had intended.

Sorry to grab the thread. Just trying to further the discussion.

Thanks

 

[dr*audio]

Kind of thinking out loud here...

If the hfe matching is critical to achieving a zero dc offset, were the assembly lines matching these pairs or just "next one up" installing? If they weren't matching was the theory that the protection circuits were there to handle the mismatch so to speak?

In my Kenwood KR-5030 when I pulled and measured the diff pair I saw hfe readings of 539 and 399 on the pair that design would say should be matched. Is this component drift or just the way it is?

On said receiver there is also a second set of "pairs", Qm5 and Qm7. Should these also be matched?

When I replaced the Qm1 and Qm3 pair with a hfe 455 and 453 pair the offset on that channel halved what it was from 100mV to 52mV. I find no resistors in or around the circuit section that are out of spec value wise. I spot replaced Cm19 to eliminate any possible dc leakage and there was no change.

I realize 50mV is not going to blow up the speakers but just trying to understand the design theory and and resultant tolerance the circuit designers had intended.

Sorry to grab the thread. Just trying to further the discussion.

Thanks

Manufacturers would either buy transistors already sorted by HFE, or have the Inspection Department match them, not on the line. The transistors age and may no longer be matched, that's why yours are so different.

 

[fred soop]

.....
Next I went in to change input diff transistors. They are:
origi: 2sa640, -50v, -50v, -5v, Ic-50mA, Pd 250mW, 50min fT, 320hfe, Nf4dB
new: 2n5087, -50v, -50v, -3v, Ic-50mA, Pd 625mW, 40min fT, 360hfe, Nf2dB
Now: DC offset: left- 23.8mV, right 21.5mV. There's some gain but now they are closely balanced.
This now brings up a question. Which transistor characteristic (ceo,pd,hfe,ft....) would I look at if I wanted to reduce DC output? Any answers on that one?

Matching hfe.

But any amplifier that is "below" the class of having current sources and current mirrors is probably not going to do much better than the approximately 20 mV that you are measuring other than by chance.

To learn more than you want to know about power amplifier design, look for Audio Power Amplifier Design by Douglas Self and/or Designing Audio Power Amplifiers by Bob Cordell. Both cover a lot of ground without completely burying you in complex math. Cordell relies a lot on SPICE simulations. Self has been in the broadcast and recording industries and has designed consoles for both. He has actually built and tested almost all of the variants that he discusses.

 

[jukebox4698]

If I can break in and ask a question please. I understand that a very high DC OFFSET can cause a receiver to stay in protection mode. So how would you check the DC OFFSET when in protection mode? Pioneer SX-950 in protection mode DC OFFSET reads OL. however I do not feel this is correct. if the protection does not open then there should be nothing coming out my speaker connections. Correct?? = OL

Thanks for the help