Question re: RC values for blocking DC at output of simple opamp RIAA preamp

1665271135515.png


R3 is in series with C10, which connects to ground (blocking DC from the IN- pin).
R1 shunts IN+ to 0V ground.
R4 is in parallel from IN- to OUT.

Does the value of R4 need to be similar or equal to the value of R1 in order to minimize DC offset at the OUT pin?
 
Yes, until you connect the cartridge, which will have a DC resistance on the order of 1k.

R4 = R2 + ( R1 // Rcartridge )

But here there is unity DC gain, so it isn't as critical as when C10 is shorted and the DC gain is x41
and the resistor balance would be harder.
 
Last edited:
The cartridge I'm using is supposed to have coil DCR of about 550 ohms.

OK, so I take R1 (47k) and add that to the cartridge coil R (550) and R2 (100) to get about 47.65k ohms to IN+.
Now I want to juggle the values of R3 and R4 so that I get the gain I want (about 40X) with R3 + R4 = 47.65k ohms. The closer the match, the better.

Did I get that right?
 
The cartridge is in parallel with the 47k, see post #22.

In this case it would result in a rather low resistance for R4. The op amp would not be able to drive that.
Sometimes there are too many constraints. You could short C10 and adjust R3 and R4, but that may amplify
any remaining DC too much.
 
Last edited:
  • Like
Reactions: rongon
R4 = R2 + ( R1 // Rcartridge )

Ah, yes. Sorry I didn't catch on to that in plain words. And thank you for the coaching, This is great stuff!

I was busily simulating this and found that if I made R3 = 510R and R4 = 20k that I got a very low DC offset at the output of that stage.
Then for the second stage, since I have a large series R of 220k in the EQ network (high impedance), I made (for that 2nd stage) R3 = 5.1k and R4 = 220k.
The simulation predicts DC offset at the output of the 2nd stage at about 250uV, which would be fantastic if I could get that in real life.

As previously noted, in real life I'm seeing DC offset = 2mV (approximately).
LTspice is predicting 890uV.

Right now, in the first stage I have R3 = 1k and R4 = 33k, for about 32X gain. (The idea is to maximize headroom by not slamming the second stage with too hot a signal at low frequencies).

I tried changing R4 of the 2nd stage to 220k, to match the EQ series R, but the resulting offset was still nearly 400uV.

However, if I change the 2nd stage NFB resistors to R3 = 3k and R4 = 200k, LTspice predicts the DC offset will be only 207uV at the output.
If its predictions are off by an equal amount to what they were from what I've actually built, the real life DC offset should be about 2.5X the predicted, so about 500uV. That would be 1/4 of what I'm getting now. That would be pretty good!

I think I'll try that. I may not get to it tonight, but definitely tomorrow. This poor PCB is getting a real workout...
 
I'm not sure it matters that much to balance perfectly in this case. Looking at the 4562 data sheet, input bias current max is 72nA x 6.2K is only .4mV assuming the other input is 0 ohms. So quite small given your DC gain is 1 in this configuration. Vos is spec'ed at .7mV so summing the worst case at the output should be 1.1mV And all this stuff is temperature dependent, so it is going to drift around.
 
Well, this is a great opportunity to learn something, and again, thank you for sharing your knowledge about this with me (and the community). This is great stuff.

I know SPICE predictions are usually overly optimistic, but it does seem that the formula works. The closer I get R4 to the series R loading IN+, the lower the DC offset at OUT.

Of course, with a phono cartridge with only 500 ohms as the series load, there's no way the NFB resistor can be 500R, so a compromise must be struck. About how low in value should I dare to go for R4 in the input stage? I'm new to designing with opamps, and looking at others' designs I don't see the NFB parallel resistor being any lower than about 4.7k ohms, and usually more like 100k ohms. I'm noticing that the lower the value, the higher the predicted THD at 1kHz, but it's always really low, like < 0.003%. THD goes more toward 0.001% as the parallel NFB resistor gets up to 100k ohms in value. I figure that's because the opamp distorts a bit more as it is more heavily loaded. But how heavy is too heavy? When do things like slew rate limiting start occurring? We're talking tiny signal levels here, so perhaps the opamp will perform fine even heavily loaded by its NFB loop.

I also know that the larger the value of resistor, the higher its excess noise. In the second stage, I'm finding that a parallel NFB resistor of 200k is resulting in very low DC offset at OUT (around 100uV). I guess that will be a little noisier than the 47k I have in there now. The preamp is nearly silent now, and I can take a little hiss (my high freq hearing is not very good at my age). It looks like there's a compromise one needs to make between lowest noise and lowest DC offset. Capacitors don't make noise, so I guess that's why you see all the opamp preamps with caps on their outputs.

Anyhow, I'm armed with some new knowledge, and some experiments to perform to see if things work as they look in simulation. This will be interesting. I hope to have solid answers by tomorrow afternoon. Thanks again!
 
Thanks again, again. 🙂

The 4562 data sheet doesn't specify a minimum load, but it does say "dynamic range is maximized by an output stage that drives 2kΩ loads to within 1V of either power supply voltage and to within 1.4V when driving 600Ω loads."

I figure a conservative figure would be a minimum load of 2k ohms.

The old rule of thumb about impedance is that the load impedance should be 10x greater than the source impedance. So I figure a value of 20k would be safe for the feedback resistor, so for my target of 32x gain from the first stage, that would 20k/32 = 625, or perhaps 620R.

With the first stage opamp's NFB resistors at 620R and 20k, and the second stage's at 3.9k and 200k, the predicted DC offset is 114uV.
If I get <0.5mV offset, I think I'll be happy.
 
Hmmm... So the opamp would easily meet its specs driving a 3.9k ohm load.
But would there be any difference in the sound from it?
I could reduce the parallel NFB resistor in the first stage to 10k ohms, and see what I get.
With R3 = 300R and R4 = 10k the DC offset is predicted to be 43uV.
I'll be happy as can be to get anywhere near that low.
 
Balancing the resistances helps to keep the offset small when you have a bipolar op-amp without base current compensation that has matching input bias currents for the + and - inputs.

However, the LM4562 has base current compensation, which is why the input offset current is of the same order of magnitude as the input bias current rather than much smaller than the input bias current. You can keep the resistances driving both inputs small if you want low offset, but there is not much point in making them equal, as the DC currents into the two op-amp inputs don't match anyway.
 
Before I start in with the solder sucker, is an Rload (pulldown resistor) at the output of the preamp necessary if I'm shorting out the output capacitor? I have 56k installed now, but should I simply remove that and leave the preamp's output unloaded, knowing it will be loaded by the downstream volume control and amplifier inputs?

1665324699763.png


There is a 1 meter cable run to the selector switch, then a 1/2 meter cable from the switchbox to the autoformer volume control, then another 2 meter run of cable from the volume control to each powered speaker. The cables are nothing special, but not dollar-store junk either.
 
You can keep the resistances driving both inputs small if you want low offset, but there is not much point in making them equal, as the DC currents into the two op-amp inputs don't match anyway.

I can't keep the resistance driving the 2nd stage input small unless I redesign the passive RIAA EQ network, which I don't want to do. (The series R of the network is >200k ohms.)

However, the first stage input is driven only by the phono cartridge, so I can keep its NFB network to low resistance values, which will be this morning's project.

Everything is a compromise.

I'm going to follow the guidance of LTspice and see if its predictions are accurate. So far it's been pretty close, on a number of fronts.
 
Before I start in with the solder sucker, is an Rload (pulldown resistor) at the output of the preamp necessary if I'm shorting out the output capacitor? I have 56k installed now, but should I simply remove that and leave the preamp's output unloaded, knowing it will be loaded by the downstream volume control and amplifier inputs?
I don't think it matters much. I could think of an advantage of keeping the resistor if the input of the circuitry downstream were capacitively coupled, but that's not applicable.
 
OK, the new resistor values are in the board. The result?
It looks like Marcel was right -- the DC offset at the outputs hardly changed at all.
One channel is now at about -2.5mV, the other is at about -1.1mV.

It also looks like the LM4562 model is off regarding DC offset predictions. Not too bad, but with the values I've chosen LTspice predicts the DC offset would be a minuscule 10uV. Unfortunately, it's more like -1.5mV, 150x worse than predicted.

Is < 3mV of DC offset a problem, in general?
 
Last edited:
Assuming a DC coupled main amplifier with a gain of about 30, it would be < 90 mV at the loudspeaker at maximum volume. The power dissipated by a loudspeaker with 6 ohm DC resistance is then 1.35 mW. It certainly won't do any damage, I also don't expect it to have a noticeable effect on even-order distortion of the loudspeaker.
 
  • Like
Reactions: rongon