Yamaha CR-420 Offset: - 61mV by design!

First let me say I am a Yamaha lover, and I don't want to offend any other member of "the familly" by this title.
Let me say also that I don't think there is any particular problem with an offset as low as this. A 60mV offset in a 3 Ohms load (typical DC resistance for a 4 Ohms speaker) represents only 1.2 milliwatt of power, and only 0.6 milliwatt in a 6 Ohms load. There is nothing here to justify all the attention a small offset like this received on some threads!!
I don't think Yamaha designed this receiver with a -61mV offset goal, but they certainly figured it was inoffensive. I think so...

As a Yamaha lover, I was curious to find why others CR-420 owners frequently measured an offset in the vicinity of -60mV. I was also curious to see if there is a solution to this.

As an intellectual exercise, here is an analysis of the problem. If you don't want to follow the math section, you may skip to the end of this thread, where you will find a very efficient and easy to implement solution, or you may jump to another thread!

You will find a complete and very readable schematic for the CR-420 here:

http://www.hifiengine.com/manuals/yamaha/cr-420.shtml

For the sake of simplicity, I've reprinted part of this schematic (part of left channel only) and I've added some informations like measurements, computed values and formulas to it. I suggest that you print the first attach file to follow this analysis. (The value for Vb3 and Ve (TR403 TR405) on this file is computed after a change in R407)

In this analysis, I make some reasonable simplifications and assumptions:

Vbe = 600mV for all transistors

Ic = Ie for all transistors

hFE = 300 for TR403 and TR405, corresponding to the mean value for the 2SC1571 F or G, the original transistors.

The first step is to compute the bias current across the Vbe multiplier, to see if the base current from TR407 can be neglected in the actual value of Ic3 (Ic3 is the collector current in TR403).

At standby, the emitter of TR409 is at -1.2V since there is 4X0.6V across the Vbe multiplier. This bias current is set by R421 and R423 and the -30V supply rail at:

I= 28.8V / 4.4kOhms = 6.55 mA

This is the current across TR407. I figured a hFE of 200 for the 2SA659 and to be sure, I measured the voltage across R413 and found 16mV on both channels. This translates to a base current of only 34uA from TR407, not far from 6.55 mA / 200, and I neglected this current in the rest of the analysis since it represents only 3% of the estimated value for Ic3.

Now, we need to compute the exact value of Ic3 because this value sets the base voltage for both transistors (yes, both) in the differential stage.

From TR401, R405 and R407, we have

100 Ie1 + 0.6V = 1200 Ic5

Posing Ie1 = Ic1 = Ic3 since Ic3 and Ic1 >> 34 uA

Ic5 = (100 Ic3 + 0.6 ) / 1200 = (Ic3 / 12) + 0.5 mA

Here is the problem! With this value for R407, Ic3 is about 3 times larger than Ic5 (we will see this later)

Also, Vb3 = - 18270 Ib3 = - 18270 Ic3 / 300 = - 60.9 Ic3

Ve for TR403 and TR405 is given by the supply voltage (-13.7V) and R411

Ve = - 13.7V + 5600 (Ie3 + Ie5).

It is also set by Vb3 and the Vbe voltage

Ve = Vb3 - 0.6V

We have three variables, Ve, Ie3 and Ie5. We have also three equations. Easy to solve!

Ve = - 13.7V + 5600 (Ic3 + Ic5)

Ve = - 13.7V + 5600 (0.5mA + (13 Ic3 / 12))

Ve = - 0.6V - 60.9 Ic3

Solving the two last equations by subtraction to get rid off Ve, we get

Ic3 = 1.68 mA

The rest follows like spring water

Ib3 = 1.68 mA / 300

Vb3 = - 18270 (1.68 Ma / 300 ) = - 0.10V

Ve = - 0.70V

Ie3 + Ie5 = (13.7 - 0.7)V / 5600 = 2.32 mA

This is the Tail current, It

Ic5 = 2.32 mA - 1.68 mA = 0.64 mA

Ib5 = Ic5 / 300 = 0.64 mA / 300 = 2.14 uA

Vo, the output voltage at standby is

Vo = Vb5 + R415 Ib5 = - 0.1V + (18000 X 2.14 uA) = -0.1V + 0.0385V

Vo = - 0.0615V = - 61.5 mV THIS IS THE OFFSET VOLTAGE for perfectly matched transistors in the differential stage of a CR-420.

As you can see, the problem is because Ic3 = 2.63 Ic5


But a solution exist if we can make Ic3 = Ic5, and it is very easy if we replace R407.

Posing Ic3 = Ic5 = It / 2 = 2.32 mA / 2 = 1.16 mA , we have

R407 x 1.16 mA = 0.6 V + (100 x 1.16 mA)

R407 = 620 Ohms

With this value, the base current for each transistor in the differential pair will be 3.87 uA

Vb3 = Vb5 will be - 71 mV

Vo will be

Vo = - 0.071V + (18000 x 3.87 uA) = ZERO !!!


A practical solution:

I've bypassed R407 with a second resistor, soldered on the copper side of the printed circuit board. The exact value for this resistor was selected by experimentation because I did't try to match the transistors in the differential pair.

With needle nose jumpers, I shunted the original resistor with another resistor starting with a 1200 Ohms. See the pictures on the attach files for the Left and Right channel positions for R407, R408. Turn the receiver OFF when you put the jumpers in place, the spacing is very small here.

The advantage of this method is that once the jumpers are in place, you can trim the value of the additional resistor with the amp ON.

My results?

Original Vo R Ch = - 61mV, -4 mV after shunting R407 with 1200 Ohms

Original Vo L Ch = - 51 mV, 0 mV after shunting R408 with 1300 Ohms

Good tweakings in your CR-420 !!

I am quite confident that replacing the original resistor (R407, R408) with a 620 Ohms resistor, if you have well matched transistors in the differential pair, you will have very close to 0V offset voltage in your CR-420.

(edit on 2014-04-12: take a look to this thread http://www.audiokarma.org/forums/showthread.php?t=578021&page=4 )

wow! at least you done your research! Nice addition to the site!

The million-dollar question is, "Can you detect any benefit, other than the lower measured values, having gone though that work?".

Intellectual exercise, that's all!

Great stuff ecluser! My 420 has higher than expected offset readings which I believe are the primary cause of a 'thump' when I turn the unit off. I'll have to give this a try.

-Matt

Hi Matt!
I don't think a small offset like this (near - 60 mV) can be the cause of a "thump" when you turn the unit off, unless you have very sensitive speakers. But when you say your CR-420 has higher than expected offset readings, what do you get exactly?
From your previous thread, which initiated my analysis incidentally, you reported a "normal offset" for a CR-420. That is before you tried to match the transistors in the differential pair. Where are you in this affair now?

I know I suggested to you in the past to match this pair, but from what I found from my analysis, I changed my mind about this. What I think now is that it is much more important to equalize the collector currents by replacing this damned 1.2K resistor (R407).

You may replace it by a 620 Ohms resistor, as I suggested, OR by a series connected 100 Ohms resistor and a forward biassed diode (1N4148 or similar) to make a nice current miror. There is enough room to make this change on the board. If you do this modification, the operation point of the differential stage will be very close to the ideal situation (equal Ic), even if there is a great mismatch in the hFE of the two transistors. The output offset voltage would be very small after this modification.

I would like to insist on the fact that the important point is not to have a low output offset voltage, but to have equal collector currents in the differential pair.

I would be curious to know what you get with a 100 Ohms resistor + a diode.

Have a nice day!

ecluser said:

I would like to insist on the fact that the important point is not to have a low output offset voltage, but to have equal collector currents in the differential pair.

Click to expand...

How about we have both equal collector currents and zero offset, seeing as how both are desirable things...?

No, I don't believe fretting over small amounts of offset is a worthwhile behavior in the large scheme of things, but if it gets guys interested in how their amp works, who's to argue? FWIW, I can easily measure increases in THD with offsets as low as 20mV. Inaudible (as far as I'm concerned), but measurable.

RE: The input stage of the 420. I'm surprised. What was Yamaha thinking?:scratch2: Title of thread ought to be: 61mV by (bad) design.

Every CR-420 that I have ever had on my workbench (at least 6!) has had DC Offset readings EXACTLY as noted above (excellent first post BTW ecluser!).

Indeed, what were they thinking, BUT[!] I do like the CR-420 very much--it's a great little receiver with many of the features of its larger siblings, but a small footprint with plenty of power!

Hi ecluser -

I have been working on two different Yamaha pieces recently. A 620 which measures at ~15mV max DC offset (if I recall correctly) and the 420 which is high at ~78 and 95mV DC on the right and left outputs, respectively.

I haven't done anything additional with the 420 yet, because it really sounds pretty good as is. The only motivation I have for twiddling with the diff pair is to eliminate the small turn off thump that it has. It's entirely possible that the thump on the 420 is from something else though.

I'll definitely have to give your suggestions for R407 a try. When I do, I'll report back and let you know what happens.

-Matt

EchoWars said:

How about we have both equal collector currents and zero offset, seeing as how both are desirable things...?

No, I don't believe fretting over small amounts of offset is a worthwhile behavior in the large scheme of things, but if it gets guys interested in how their amp works, who's to argue? FWIW, I can easily measure increases in THD with offsets as low as 20mV. Inaudible (as far as I'm concerned), but measurable.

RE: The input stage of the 420. I'm surprised. What was Yamaha thinking?:scratch2: Title of thread ought to be: 61mV by (bad) design.

Click to expand...

As for equal collector currents and zero offset voltage, IT IS EXACTLY WHAT I DID. But to achieve this goal, you must first correct the unbalance in the collector current.

In the CR-420, it is impossible to correct the unbalance in the collectors currents if you don't change the damned R407 (or R405), unless you change the tail current by a large amount. This, in turn, would change the transconductance of the differential stage and the open loop gain of the amplifier.

By changing R407, I didn't change the gain of the amp and this modification has no detrimental effect on the stability of the amp.

It has been suggested by some members to replace the transistors in the differential stage by transistors having a much higher hFE. Well, this change would reduce the base current in the transistors, and by the way the base voltage and the offset voltage at the output.

BUT, it changes nothing in the unbalance in the collector currents. This is the unbalance in the collector currents which is responsible for the rise in distortion (not the output offset voltage in itself), mainly a rise of second order distortion if the differential stage is unbalanced. You may measure the same offset voltage in two amplifiers having very different unbalances in the differential stage, as well as you may measure different offset voltages between two amplifiers having the same unbalance in the differential stage. The offset measurement in itself means nothing because it is related to the collector current by the base resistance, and many other parameters.

This thread is not a general discussion, but is dedicated specifically to the CR-420.

I was curious to understand the cause of this offset because it (-60mV or so)was a generalised observation, and when I saw the cause, the solution became trivial to me and easy to implement. I did it, why not? As I said before, after I found the cause I was more concerned by a correction in the unbalance in the collectors currents than I was by the offset, but a correction to the first corrected the second problem.

I don't know what was Yamaha thinking, but I did appreciate this lovely receiver in my bedroom!

Appears we are on two different pages. I understood perfectly well what you were saying and doing in the original post, but my reply has obviously been seriously misconstrued (the quoted sentence of yours that appears in my reply was its basis).

RE: The Yammie. Good detective work. Best of luck to you.

ecluser said:

I would be curious to know what you get with a 100 Ohms resistor + a diode.

Click to expand...

I was too curious to wait and I've builded a replica of the (CR-420 main amp)differential stage on a breadboard to test this suggestion. I was able to observe the X shaped Ic vs Vi curves. It is a delicate measurement because there was no feedback and only 15mV peak-to-peak was sufficient to saturate the amp.

To resume, I prefer the replacement of the original R407 resistor (1.2 kOhms) by a 620 Ohms resistor, because it works very well and is easy to implement.

The problem with the diode version of the current mirror is that it is sensitive to mismatch between the Vbe drop in the transistor vs the voltage drop in the diode (Vd), with the 100 Ohms resistor for R405. The lower the value of the tail current, the greater is the problem.

If you want to go for a current mirror, I suggest that you replace R405 by a 1kOhms resistor.

If you look at the attached file, you will understand why the resistor method is acceptable. For a variation of 0.05V in Vbe, the diode version of the current mirror gives a variation in the collector current of Q1 (TR403) which is about 4 times greater (with the original 100 Ohms R405 resistor) than the resistor method.

It is not unusual for a small transisor to have less than 0.55V for Vbe at Ic close to 1 mA, and a voltage drop higher than 0.6V in a signal diode (1n4148 or 1N914) for the same current. I measured 0.625V on my transistor curve tracer for the diode I used in this test.

Go for the easy way, it works so well!

ecluser said:

To resume, I prefer the replacement of the original R407 resistor (1.2 kOhms) by a 620 Ohms resistor, because it works very well and is easy to implement.

[...]

Go for the easy way, it works so well!

Click to expand...

I just wonder why the Yamaha engineer who developed the circuit didn't go this way 30 years ago... could it be that there are stability problems that aren't obvious? A circuit for mass production has to be stable first and "audiophile" second...

Martin

Resurrection Time

I picked up this CR-420 at the gw for 7.99. It sounds surprising after a de-oxit shower. :thmbsp:

I have a couple questions, though.

1. I adjusted the bias using the tp pins 1-4, see pic. Is this the correct method?(I've searched unsuccessfully for a service manual) I know ecluser adjusted bias using the .47ohm resistors.Here Two screws are frozen on the bottom(read stripped), so i'm unable, at this time, to get to the underside of the receiver. Consequently i'm unable to get at the .47ohm resistors to adjust trim pots.

2. I plan on doing the resistor mod in order to bring the dc-offset down, and am going to replace r407 and r408 with 620 Ohms resistors. So, what should the specs of the replacements be? 1/4watt 5% ok?
Readings are -71mV,-61mV


Thanks for the help, -Mike

For your frozen screws, you may try to pinch the head of the screw with a sharp wire cutter (hold the cutter at strait angle with the back plate), or a wise-grip.

I replied to your PM this morning but it seems that you didn't receive it. The test points are on either sides of one 0.47 Ohm resistor, so the method is the same. Use what is more convenient to you.

For the resistor, I preffer to bypass the original with another resistor from the underside of the board. This is more flexible and more precise. This way you can trim the original 1.2k resistor to something close to 600 Ohms (with 2.5% tolerance) with standard 5% tolerance resistors. In the (-71mV) channel, you will end up with a final value less than 600 Ohms, probably close to 560 Ohms. The power dissipated in the resistor is very low, of the order of 1mW. So, 1/8 Watt is more than you need.

Good luck!

Thanks!

Nope, i didn't get your PM.

I ended up grinding those screw heads off, it took less than 2 mins for both of them.

Thanks for the resistor info, on to making a capacitor list!

Also, have your tried adding a 2.2 250v film cap across the main filter caps?

If so, what were your results?

Ok, so I took more accurate readings of the offset with the bias adjusted, and they are

-67.8mV and -76.5mV:thumbsdn:

Can you hazard a guess as to the resistance needed to bring the offset closer to 0?


I know you posted all the math previously, but i'm an idiot when it comes to that stuff.

Thanks for your help, -Mike

I can't guess the exact new value for that resistor with great precision. It depends on too many parameters that may be different from one amplifier channel to another. You see, I assumed a Vbe of exactly 600mV in my analysis, but it may be 585mV or 610mV in your amp. Consider also the tolerance of the other resistors in the first stages... Don't bother with the maths, try the jumper and shunt resistor. It is fast and easy to find the correct value with this method.

I didn't bypass the main filter caps. I removed two resistors in the loudness circuit (R331 and R332) to keep a flat response in the high frequencies, and I changed the impedance of the phono preamp. But I don't listen to vinyls any more with this receiver since it is now in the bedroom.

Add DC Offset Adjustment

You can Add DC Offset Adjustment. 3 x 22k resistors, 1 x 1kVR, 2 x 4148D and one non-polar 100uf 16volt cap. Come from the main power supply + & - supply through a 22k to pin 1 & 3 of the VR. Now tie the 4148s to pin 1 & 3 and the other ends of the 4148s to ground. Use the last 22k and the 100uf to pin 2 of the VR. Send the other end of cap to ground. Last connect to other end of the 22k from pin 2 to the base of the first transistor in the pair. The 4148 gives regulation. Now you can adjust to 0 dc offset as needed.:yes:

This idea is not new, but why take an helicopter to go from point A to point B when you can go by foot in less than one minute? If you want to have an adjustable offset, all you need to do is to replace the fixed resistor R407 by a trim pot.

If you use your technique, better to use a larger resistor between the trim pot and the input transistor. A 22k here would reduce the input resistance to ~22k