Intrigued by how well my cheap 'n cheerful ESP P-06 build worked out, I've been playing around with an even simpler RIAA phono preamp made with an opamp > passive EQ > opamp. The PSU is repurposed from an earlier Hagerman Bugle build (+/-15VDC, using 7815/7915 ICs).
I believe I have a circuit that 'sounds good'. Now I'm playing around with the output capacitor and pulldown/load resistor.
I've looked around at other similar projects, and have found a lot of variation in values chosen for the output DC blocking cap (Cout) and output pulldown/load resistor (Rload). Many traditional 1980s-style designs use a large value capacitor with lower value resistor. For instance, I found one Graham Slee design which uses 10uF for Cout and 47k for Rload. Rod Elliott shows a few example designs that use Cout = 22uF and Rload = 22k ohms. But then the Muffsy and an ELENCO design use Cout = 1uF and Rload = 100k, while the RJM VSPS uses 2.2uF and 100k ohms. On the other hand, the TNT Solidphono goes to the other extreme and uses Cout = 0.33uF and Rload = 330k ohms, like early 1960s tube circuits would have used.
I'm left with the impression that the choice of values is less important than the resulting time constant. Is that correct?
One thing to take into consideration is the load impedance presented by the device to be driven by the phono preamp output.
- If that's going to be a preamp or amp with a 10k ohm volume control, then a 1uF output cap will result in an F3low of 16Hz, which would introduce a noticeable rolloff of low bass. (Perhaps that's desirable, though, as a rough 'n ready rumble filter?)
- Looking at the other extreme, if Cout = 22uF and Rload = 22k, and that feeds a 10k volume control, even though the resulting load (22k//10k) will be only 6875 ohms the F3low will be down at 1Hz. The Graham Slee values of 10uF and 47k results in F3low of only 0.34Hz, but if the load is a 10k volume control, F3low for that would go up to 1.9Hz. Perhaps that was the design goal there?
Let's say I'm not sure what my RIAA preamp will be driving. It could drive a class D amp with a 10k ohm volume pot on its input, or it could drive my living room stereo with an autoformer volume control (AV) which maintains a very high impedance load. Should I try to find a compromise solution that works OK with both? Or should I optimize for one extreme and assume it will be OK for the other?
I did have 1.5uF and 100k and that was working fine into the living room hi-fi w/ AVC. Into a 10k ohm load the F3low would go way up to 16Hz, but it's not a problem into the AVC.
Now I'm trying a 6.8uF 100V (big) film cap with 56k, which is predicted to have F3low of only 0.4Hz into a light load.
Should I change the Rload to 22k, so the F3low is about 1Hz? Would that conform better to best practices?
What are the pros and cons of bigger vs smaller value C and bigger vs smaller value R in this part of the circuit?
I understand that smaller value capacitors have advantages of lower inductance, lower ESR, etc.
Also, the smaller value of R allows the output cap to charge/discharge more quickly, reducing turn on/off thumps.
PS - Forgot to mention... I've read that a lower value of capacitance (e.g. 1uF) for the output DC blocking cap will have a higher reactance at low frequencies, so can make the circuit more susceptible to picking up hum from its interconnect cabling. Is that independent of the value of Rload? If that's true, and a higher value capacitor (e.g., 10uF) will help reject hum pickup, then perhaps that's an important issue for a standalone phono preamp as opposed to one that's built into a full-function preamp or integrated amp?
Are there other issues of importance?
I believe I have a circuit that 'sounds good'. Now I'm playing around with the output capacitor and pulldown/load resistor.
I've looked around at other similar projects, and have found a lot of variation in values chosen for the output DC blocking cap (Cout) and output pulldown/load resistor (Rload). Many traditional 1980s-style designs use a large value capacitor with lower value resistor. For instance, I found one Graham Slee design which uses 10uF for Cout and 47k for Rload. Rod Elliott shows a few example designs that use Cout = 22uF and Rload = 22k ohms. But then the Muffsy and an ELENCO design use Cout = 1uF and Rload = 100k, while the RJM VSPS uses 2.2uF and 100k ohms. On the other hand, the TNT Solidphono goes to the other extreme and uses Cout = 0.33uF and Rload = 330k ohms, like early 1960s tube circuits would have used.
I'm left with the impression that the choice of values is less important than the resulting time constant. Is that correct?
One thing to take into consideration is the load impedance presented by the device to be driven by the phono preamp output.
- If that's going to be a preamp or amp with a 10k ohm volume control, then a 1uF output cap will result in an F3low of 16Hz, which would introduce a noticeable rolloff of low bass. (Perhaps that's desirable, though, as a rough 'n ready rumble filter?)
- Looking at the other extreme, if Cout = 22uF and Rload = 22k, and that feeds a 10k volume control, even though the resulting load (22k//10k) will be only 6875 ohms the F3low will be down at 1Hz. The Graham Slee values of 10uF and 47k results in F3low of only 0.34Hz, but if the load is a 10k volume control, F3low for that would go up to 1.9Hz. Perhaps that was the design goal there?
Let's say I'm not sure what my RIAA preamp will be driving. It could drive a class D amp with a 10k ohm volume pot on its input, or it could drive my living room stereo with an autoformer volume control (AV) which maintains a very high impedance load. Should I try to find a compromise solution that works OK with both? Or should I optimize for one extreme and assume it will be OK for the other?
I did have 1.5uF and 100k and that was working fine into the living room hi-fi w/ AVC. Into a 10k ohm load the F3low would go way up to 16Hz, but it's not a problem into the AVC.
Now I'm trying a 6.8uF 100V (big) film cap with 56k, which is predicted to have F3low of only 0.4Hz into a light load.
Should I change the Rload to 22k, so the F3low is about 1Hz? Would that conform better to best practices?
What are the pros and cons of bigger vs smaller value C and bigger vs smaller value R in this part of the circuit?
I understand that smaller value capacitors have advantages of lower inductance, lower ESR, etc.
Also, the smaller value of R allows the output cap to charge/discharge more quickly, reducing turn on/off thumps.
PS - Forgot to mention... I've read that a lower value of capacitance (e.g. 1uF) for the output DC blocking cap will have a higher reactance at low frequencies, so can make the circuit more susceptible to picking up hum from its interconnect cabling. Is that independent of the value of Rload? If that's true, and a higher value capacitor (e.g., 10uF) will help reject hum pickup, then perhaps that's an important issue for a standalone phono preamp as opposed to one that's built into a full-function preamp or integrated amp?
Are there other issues of importance?
Last edited:
The output of a MM phono preamplifier with 40 dB of midband gain is roughly 700 times less sensitive to hum voltage than the input, so as long as the reactance of the coupling capacitor is much less than 700 times the low-frequency impedance of the cartridge, any hum issues are likely to occur at the input.
Thanks. It certainly makes sense that hum is more likely to be picked up by the input wiring than anywhere else. Am I correct in inferring that — as far as hum pickup is concerned — it doesn't matter whether the output blocking cap is 1uF or 22uF?
Are there any other issues of importance?
Are there any other issues of importance?
Yes, at least for a RIAA preamplifier.
Regarding other issues, I think you already covered most of them.
For a given time constant, large capacitance, small resistance as compared to small capacitance, large resistance:
1. Frequency response less sensitive to the load impedance
2. Capacitor either expensive or not a very good type of capacitor
3. Somewhat heavier load on the amplifier output
Large time constant compared to small time constant:
1. Less loss of bass and less phase shift
2. Longer settling time, so longer mute required or worse thump
3. If the capacitor is non-linear to a non-negligible extent: less capacitor distortion than with a small time constant, because the signal voltage across the capacitor is smaller
4. Error due to other capacitor imperfections, such as dielectric absorption, also smaller
5. Capacitor less effective as subsonic filter
I think you missed point 3 (and 4) of the second list. It's why Douglas Self likes to use large time constants when using electrolytic coupling capacitors. It's really important when you are forced to use class 2 ceramic coupling capacitors for whatever reason.
Regarding other issues, I think you already covered most of them.
For a given time constant, large capacitance, small resistance as compared to small capacitance, large resistance:
1. Frequency response less sensitive to the load impedance
2. Capacitor either expensive or not a very good type of capacitor
3. Somewhat heavier load on the amplifier output
Large time constant compared to small time constant:
1. Less loss of bass and less phase shift
2. Longer settling time, so longer mute required or worse thump
3. If the capacitor is non-linear to a non-negligible extent: less capacitor distortion than with a small time constant, because the signal voltage across the capacitor is smaller
4. Error due to other capacitor imperfections, such as dielectric absorption, also smaller
5. Capacitor less effective as subsonic filter
I think you missed point 3 (and 4) of the second list. It's why Douglas Self likes to use large time constants when using electrolytic coupling capacitors. It's really important when you are forced to use class 2 ceramic coupling capacitors for whatever reason.
Thanks. Good points.
I think that suggests
1. If you're concerned enough about maintaining LF bandwidth and minimizing phase shift that you choose to go with a higher value output capacitor, make sure it's a very high quality one (and watch out for turn-on/off thumps).
I have some metallized polyester caps in 6.8uF 100V, and some 4.7uF 250V metallized polypropylene that should be up to the task and are small enough to fit on the PCB. It's a large PCB.
2. If you're more worried about capacitor imperfections and you know the preamp will be lightly loaded at all times, a smaller value cap with larger value pulldown resistor is the better choice.
Or in other words...
Everything is a compromise. 😀
I think that suggests
1. If you're concerned enough about maintaining LF bandwidth and minimizing phase shift that you choose to go with a higher value output capacitor, make sure it's a very high quality one (and watch out for turn-on/off thumps).
I have some metallized polyester caps in 6.8uF 100V, and some 4.7uF 250V metallized polypropylene that should be up to the task and are small enough to fit on the PCB. It's a large PCB.
2. If you're more worried about capacitor imperfections and you know the preamp will be lightly loaded at all times, a smaller value cap with larger value pulldown resistor is the better choice.
Or in other words...
Everything is a compromise. 😀
Last edited:
Yes, my two usage cases are the preamp driving a class D amp with 10k input resistance (10k ohm volume pot, and I think the Zin of the amp's input must be at least 20k ohms -- or at least I hope it is!) or the preamp driving my main setup with an Intact Audio autoformer volume control, which presents a high impedance (light) load. So, I actually have the two extremes to contend with as far as loading is concerned. But yes, you're correct there, of course.
That last bit just sank in (had to re-read it).
Am I correct that one implication of points 3 and 4 is that a longer time constant allows use of a lesser-quality capacitor with less disadvantages?
That would suggest a longer time constant is generally a better choice, as long as it doesn't cause collateral damage such as subsonic disturbances overloading or causing unwanted intermodulation in the amplifier circuits or speakers, or turn on/off thumps that can damage downstream devices, Yes?
Last edited:
Good point!
On my cheap 'n nasty DMM, I'm seeing -1.3mV at the output of one channel and -3.2mV at the output of the other channel. I could swap around ICs to find the ones with the lower offset. I'm using LM4562, which are supposed to have low offset compared to other opamps (such as OPA2134).
Also, isn't a bigger worry that a larger DC offset might be present for a few seconds at power on, as the capacitors charge up?
Quick power on thumps would pass through output cap anyway, especially with high value caps. However, it's unlikely with your setup. Do you use CRC in the power supply?
Exactly, but when you use a small capacitance, using a reasonably high quality capacitor is less expensive than when you use a high capacitance.
The PSU uses a full-wave bridge made of MBR1100 Schottky diodes, then a 4700uF-10R-4700uF CRC, followed by 78M15 and 79M15 regulator ICs (+/-15VDC).
I've collected quite a few 'audiophile grade' film caps over the years, so I have a few to play with. So far I've tried a pair of Xicon 4.7uF 250V metallized polypropylene, a pair of ERO 2.2uF 250V MKP 'blue box' caps, and a pair of Axon 1.5uF 630V polypropylene film & foil caps. I really hate to say this, but they all sound just a little different from each other. The differences are exceedingly slight, but I believe I'm not deluding myself (haha, that's funny). I actually wanted the cheapest (they were free) ERO MKPs to work out best, or at least to sound as good as anything else. But then...
I took out a couple of clip leads and shorted the output caps. I'm really sorry, but now the high frequency textures sound better to me. A brush on snare drum, cymbals, the hammer of a piano key, breath in a saxophone, those come across more present and 'natural' sounding than with the output caps in circuit (I'm sorry for the audiophool jargon, really, I am).
So far, no cap sounds better than any cap I've tried. The downside is that there's just enough DC offset that now there's a mild 'pop' when I use my selector switch to change source away from Phono. I might just have to deal with it.
Interesting. I really wanted to hear no difference at all between the cap in circuit or shorted out.
--
PS - There's only a slight turn-off "flmp" noise. Nothing alarming.
Last edited:
I don't know how accurate the LTspice model for LM4562 is, but it's showing a trend.
If I take the TI suggested circuit for LM4562, using a passive RIAA EQ in between two of them set up as your basic non-inverting amplifier stages, if you keep the resistor values small and the capacitor values large (as TI shows), then the output DC offset can be really, really low -- like 20uV. BUT...
If I change that to a higher impedance RIAA EQ, with 10X larger value resistors and 10X smaller value capacitors, the output DC offset goes up much higher, to around 2mV -- which strangely enough, is just what I have, and close to what I measured.
On the other hand, throw in two AD823 opamps (at almost 10X the price) and it looks like the output DC offset will be entirely negligible.
That's in LTspice, not in real life, but maybe its prediction is not far off. If that's the case, then I guess I'll be buying some $12 opamps soon....
If I take the TI suggested circuit for LM4562, using a passive RIAA EQ in between two of them set up as your basic non-inverting amplifier stages, if you keep the resistor values small and the capacitor values large (as TI shows), then the output DC offset can be really, really low -- like 20uV. BUT...
If I change that to a higher impedance RIAA EQ, with 10X larger value resistors and 10X smaller value capacitors, the output DC offset goes up much higher, to around 2mV -- which strangely enough, is just what I have, and close to what I measured.
On the other hand, throw in two AD823 opamps (at almost 10X the price) and it looks like the output DC offset will be entirely negligible.
That's in LTspice, not in real life, but maybe its prediction is not far off. If that's the case, then I guess I'll be buying some $12 opamps soon....
Bipolar opamps have lower DC offset in general. But in circuits with high value resistors can end up with increased DC offset if not properly balanced.
On the other hand opamps with JFET input have higher intrinsic DC offset but are much less affected by resistor imbalance.
On the other hand opamps with JFET input have higher intrinsic DC offset but are much less affected by resistor imbalance.
Thanks, but which resistors need to be balanced?
Again in simulation, I see that higher resistance values in the NFB loops also result in higher DC offset voltage.
In the TI example RIAA preamp circuit, increasing the NFB resistor values by 10X (resulting in the same gain, but less load on the opamps) results in much higher DC offset at the output.
I had 150 ohms and 6.2k ohms for the NFB loop voltage divider. There was a tiny 12uV DC offset at the output.
Changing the resistor values to 1.5k and 62k ohms resulted in a big increase to 680uV DC offset at the output.
Obviously, for DC precision resistor values need to be kept small. Unfortunately, that means capacitor values have to be made large.
To get F3 = 1.59Hz using a 150 ohm resistor, the capacitor has to be 1000uF.
Using 1.5k ohms the cap could be only 100uF.
I guess one way to slice it is to make the NFB resistors low in value, with large value decoupling caps, so you can dispense with the output blocking cap.
--
Again in simulation, I see that higher resistance values in the NFB loops also result in higher DC offset voltage.
In the TI example RIAA preamp circuit, increasing the NFB resistor values by 10X (resulting in the same gain, but less load on the opamps) results in much higher DC offset at the output.
I had 150 ohms and 6.2k ohms for the NFB loop voltage divider. There was a tiny 12uV DC offset at the output.
Changing the resistor values to 1.5k and 62k ohms resulted in a big increase to 680uV DC offset at the output.
Obviously, for DC precision resistor values need to be kept small. Unfortunately, that means capacitor values have to be made large.
To get F3 = 1.59Hz using a 150 ohm resistor, the capacitor has to be 1000uF.
Using 1.5k ohms the cap could be only 100uF.
I guess one way to slice it is to make the NFB resistors low in value, with large value decoupling caps, so you can dispense with the output blocking cap.
--
Both inputs should see the same Thevenin equivalent DC resistance.
Ignore any resistors that are blocked by capacitors, they don't count.
Feedback resistors do count, as the output node is at DC ground potential.
By both inputs, I think you mean both inputs of the opamp, so that would be IN+ and IN-.
In the circuits I've seen, the NFB bottom resistor (to ground) has a DC blocking cap (electrolytic) lifting the NFB loop from DC.
If the IN+ is DC coupled (no blocking cap), does that mean the two inputs (IN+ and IN-) will be imbalanced?
Or does that mean if the IN+ has 47k ohms from that pin to ground, the Thevenin equivalent DC resistance from IN- to ground should also be 47k ohms?
I'm sorry for my lack of understanding here... I'm not trained in electronics, etc.
For each op amp input, ignore any resistor in series with a capacitor, and parallel the rest.
Those two values should be equal, so input current voltage drops across them are equal,
and so the differential input voltage (and output voltage) is low.
That series nfb capacitor is intended to force the DC gain to unity, so the DC input offset voltage is not amplified.
It's used almost universally in phono preamps since it is an easy way to have low DC output offset.
Those two values should be equal, so input current voltage drops across them are equal,
and so the differential input voltage (and output voltage) is low.
That series nfb capacitor is intended to force the DC gain to unity, so the DC input offset voltage is not amplified.
It's used almost universally in phono preamps since it is an easy way to have low DC output offset.
Last edited: