Hitachi HA-330 dc offset - help needed

I was able to bring the offset down to around 80mV on the left and 165mV on the right. After reconnecting R916, now both channels are around 155mV and 160mV. Really weird πŸ˜…
Yeah... this is what I expected.

Did you clean the flux?

You also didn't answer a few of my previous questions:

What is the offset reading with no speakers connected?
Is the DC offset stable with temp change; what is the reading when the amp is cold?


If everything looks DC-stable, there are no oscillations at the output when the amp is driven into a resistive load with a bit of small capacitive load added in parallel and the gain of both channels looks the same (input voltage vs. output voltage ratios are the same... or, the sinewave clips at the same point on both channels with no load connected), my suggestion is to use a 1k trim pot in place of R705. The trim pot should be pre-adjusted to 680 ohms. Then try to tune the DC offset to 0. Use a multi-turn trim pot of good quality. Let us know if it worked okay.
 
Last edited:
Just to confirm my interpretation:

Right Channel output: 165mV with no load, 160 mV with R916 installed.
Left Channel Output: 80mV with no load, 155mV, again with R916 installed.

This is very disconcerting. Connecting R916 applies about 4mA from output to the B supply rail (-40V), yet it results in an additional output 80mV DC offset??? This suggests the DC output impedance is 80mV/4mA = 20 ohms! Further R916 is the Left sense resistor, but R115 is the analogous sense resistor for the right channel. Ideally, both channels should be immune to these sense connections. (I recommend choosing a consistent reference point for these mV resolution measurements to avoid uncertainties, eg. bottom of R703L for left channel.)

[A rant directed at Hitachi: this is an appalling excuse for a schematic. The box depicting/enclosing the right channel is a particular irritation as it conflates with the ground buss. End rant.]

I wonder if the bias current adjustment is working as specified on page 4 of manual? Would you confirm?

Thanks.
No theres no difference with or without R916
As for the bias it seems to work fine as I can adjust it as the manual is telling. All the measurements are done with the bias adjusted to spec
 
Yeah... this is what I expected.

Did you clean the flux?

You also didn't answer a few of my previous questions:

What is the offset reading with no speakers connected?
Is the DC offset stable with temp change; what is the reading when the amp is cold?


If everything looks DC-stable, there are no oscillations at the output when the amp is driven into a resistive load with a bit of small capacitive load added in parallel and the gain of both channels looks the same (input voltage vs. output voltage ratios are the same... or, the sinewave clips at the same point on both channels with no load connected), my suggestion is to use a 1k trim pot in place of R705. The trim pot should be pre-adjusted to 680 ohms. Then try to tune the DC offset to 0. Use a multi-turn trim pot of good quality. Let us know if it worked okay.
I only measure the offset without anything connected to the output. The offset goes down a little bit when the amp gets hotter but it’s not significant
 
I'm reassured to read that the amp output shows no sensitivity to R916 current.

I'm attaching a simplified sketch of the amplifier and have tabulated observed bias voltages (correct interpretation?) and the resulting bias currents. To allow estimating Beta of the input differential pair, I've estimated the emitter tail current to be 13.1V/R706 = 4mA and Q701 collector as 0.6V/R706 = 0.9mA. So Q702 collector current is about 3.1mA. Corresponding Q701 Beta is taken to be 0.9mA/Q701 base current, and Q702 Beta as 3.1mA/Q702 base current.

            Q701 Base   Q702 Base   V out        Q701 Bias   Q701 Beta   Q702 Bias   Q702 Beta

Left:     252mV         229mV         157mV     2.5uA          360               3.3uA        795
Right:   296mV         285mV         159mV     2.9uA          310               5.7uA        574


Examining left channel results, Q701 Beta seems reasonable but Q702 Beta of 795 seems suspicious. Note that C705 has 229mV of reverse bias applied and I still suspect leakage. I recommend swapping polarity as this appears to be a design/drafting error. But it might not make any difference.

I don't believe efforts to balance the differential pair by adjusting R706 resistance will be very fruitful. Let's assume the input pair is perfectly balanced with 0V offset voltage error, equal current and Beta in each transistor, and base current in each is 2.5uA. Then Q701 and Q702 base voltage will be 252mV. Voltage drop across R716 will be 4.5mV and resulting amp output will be about 248mV. The fundamental problem is the differing source resistances, as rayma pointed out in post #8.

If remaining output voltage is still deemed excessive, I suggest trimming residual output voltage by adding the resistor shown in the sketch. For example, let's assume the numbers just above, namely that Q702 base voltage is 252mV and drop across R716 is only 4mV. We'd need to impose an additional 248mV drop across R716 to yield 0V at the output. So:

Voltage across R trim is (14V-Q702 base voltage) = ~13.77V

Additional needed current through R716 is 248mV/22K = 11.3uA

R trim = 13.77V/11.3uA = ~1.2M

I hope I haven't made arithmetic or other errors.
 

Attachments

Last edited:
I'm reassured to read that the amp output shows no sensitivity to R916 current.

I'm attaching a simplified sketch of the amplifier and have tabulated observed bias voltages (correct interpretation?) and the resulting bias currents. To allow estimating Beta of the input differential pair, I've estimated the emitter tail current to be 13.1V/R706 = 4mA and Q701 collector as 0.6V/R706 = 0.9mA. So Q702 collector current is about 3.1mA. Corresponding Q701 Beta is taken to be 0.9mA/Q701 base current, and Q702 Beta as 3.1mA/Q702 base current.

            Q701 Base   Q702 Base   V out        Q701 Bias   Q701 Beta   Q702 Bias   Q702 Beta

Left:     252mV         229mV         157mV     2.5uA          360               3.3uA        795
Right:   296mV         285mV         159mV     2.9uA          310               5.7uA        574


Examining left channel results, Q701 Beta seems reasonable but Q702 Beta of 795 seems suspicious. Note that C705 has 229mV of reverse bias applied and I still suspect leakage. I recommend swapping polarity as this appears to be a design/drafting error. But it might not make any difference.

I don't believe efforts to balance the differential pair by adjusting R706 resistance will be very fruitful. Let's assume the input pair is perfectly balanced with 0V offset voltage error, equal current and Beta in each transistor, and base current in each is 2.5uA. Then Q701 and Q702 base voltage will be 252mV. Voltage drop across R716 will be 4.5mV and resulting amp output will be about 248mV. The fundamental problem is the differing source resistances, as rayma pointed out in post #8.

If remaining output voltage is still deemed excessive, I suggest trimming residual output voltage by adding the resistor shown in the sketch. For example, let's assume the numbers just above, namely that Q702 base voltage is 252mV and drop across R716 is only 4mV. We'd need to impose an additional 248mV drop across R716 to yield 0V at the output. So:

Voltage across R trim is (14V-Q702 base voltage) = ~13.77V

Additional needed current through R716 is 248mV/22K = 11.3uA

R trim = 13.77V/11.3uA = ~1.2M

I hope I haven't made arithmetic or other errors.
I have measured the transistors hfe on my cheap tester and both channels had 280 and 370 in each pairs. I ordered replacement ones in enough quantity so I can make matched pairs and rule them out of the equation. I should have them installed in the amp tomorrow and i’ll keep you informed of the result ! Thank you for your much appreciated help !
 
I'm skeptical that it will lead anywhere that's useful. A problem, at least a pitfall, is that the suggested trim pot and/or decreasing R706 to increase Q701 current in the goal of tweaking the bias dilemma messes with open-loop gain, and maybe even instability (i.e. oscillation). I'm doubtful of a satisfactory result. But maybe this my paranoia.

A fundamental problem remains the almost 5 to1 difference in source resistance and a perfectly matched differential pair wouldn't correct that issue. Metaphorically, purposely unbalancing a differential pair to compensate for mismatched source resistance seems like using a new sin to cure an earlier sin.
πŸ˜€

My suggested remedy has some virtues: it has no effect on amplifier feedback dynamics, it's brute-force-simple, and it's easily installed/tweaked.
 
I replaced Q701/Q702 for two matched pairs of ksa992 with hfe of 430. Now the left channel has 97mV of dc and the right channel has 91mV after about 10min of warmup. Bias was set according to spec after the warmup and before taking measurements.

Do you guys think it’s worth going further or should I just leave it like that ? I could replace the driver pairs in hope of getting closer to zero but not shure about that