How good is good enough when designing a line stage?

[markusA]

I'm toying around with LTspice and a tube buffer.
When you design a line stage and run sims, what design parameters are you looking for?

When is distortion low enough, what's your target numbers for the bandwidth, how much hum is good (when is it low enough) and so on.

My current sim show that I'm down -0.00032dB at 20Hz, I doubt I can hear a difference if it's 1dB?
Hum is in the 20-30uV region?

Simulations are just an indicator and the finished product can end up being quite different but when you do run simulations what are your target numbers and what are you looking for?

[mashaffer]

I would not trust simulations for distortion figures personally. It might show if you have something grossly out of whack but I don't trust the models well enough to put much faith in the resulting numbers.

[DF96]

It depends on why you want a line stage. Some people want 'tube sound', so you design for, say, 0.5% distortion and an HF droop. Others want to drive a long interconnect so you design for low distortion and low output impedance.

[markusA]

mashaffer> Like I said, I don't regart sims as gospel truth. I see it as an indication where I'm heading...

DF96> I'm still learning LTspice and distortion is still something I don't know how to simulate. Bandwidth I can do and my current data is mentined above, I don't know what I'm supposed to be aiming for though? Straight as an arrow in HF range and a very small (-0.00032dB) drop in the LF.

I'm not entirely sure how to simulate hum/background noise? Doing it the easy (possibly wrong) way by turning off the signal show noise in the 20-30uV area. Still I don't know if I'm doing it right or wrong or even if the numbers are awsome or sucky?
To me all the numbers seem to good to be true and I figure this will change when I build the actual circuit irl.

Thus the question, how good is good? I don't even know how much ripple I should aim for on the B+?

I figure -60dBu noise on the output should be pretty much inaudiable? Beyond that I'm undecided? Looking for info from you guys who've done this longer than I.

[DF96]

Simulating hum is difficult, because some major sources of hum do not appear in the circuit diagram. You can simulate supply rail ripple, but not induction from the heater circuit.

My guess is that you want noise and hum to be at least -80dB below the normal peak output (which you need to determine). Don't worry too much about distortion, unless it is high (above 1%) - this would indicate clipping or a bad bias point. However, the first stage of design is to decide what your requirements are. If you ask us what our requirements are you will get a lot of different answers. The 'standard engineering' answer might be 20-20kHz at better than -3dB, less than 0.1% distortion. Some will tell you that this is nowhere near good enough, you need 10-30kHz at -1dB and 0.0001% distortion; others will say that these are irrelevant specs - such people might be happy with a 'tuned' frequency response and 5% distortion as long as it is low order.

Just as you need to decide what your favourite food is, you need to decide what 'audio school of thought' you will follow. Some ideas are supported by scientific evidence, but even this is disputed - some think better tests are needed, others reject the idea of testing. To put it bluntly, some design with their calculator while others design with their ears. I am not trying to start a debate here, just alerting you to the background to what seems like a simple question. You can find the debate in other threads.

[Cassiel]

. Some will tell you that this is nowhere near good enough, you need 10-30kHz at -1dB and 0.0001% distortion; others will say that these are irrelevant specs - such people might be happy with a 'tuned' frequency response and 5% distortion as long as it is low order.

Just as you need to decide what your favourite food is, you need to decide what 'audio school of thought' you will follow. Some ideas are supported by scientific evidence, but even this is disputed - some think better tests are needed, others reject the idea of testing. To put it bluntly, some design with their calculator while others design with their ears.

Great summary of what's going on here. Yeah, it doesn't take long to know what people are about. Golden eared (mystical visionaries) camp vs scientific camp whose motto "I trust my distortion analyzer, not my ears" is all over the place. Unfortunately, one has to read it every day. It may be excellent in its objectivity but it really gets tiresome.

[markusA]

I've been around for a while so I know exactly what you're referring to.
I would like to think of myself somewhere in the middle.
Yes, the ears are the final judges but good engineering and measurments are tools that help you achieve hifi-bliss.

What I'm looking for are minimum requirements.
Stuff like
- 2mV B+ ripple is generally ok
- +/- 1dB @ 20-20kHz is good enough that you'll probably not notice it
- -80dBu ripple/noise at the output will probably percieved as a "silent" preamp

Minimum requirements and rules of thumb that tell you when it's time to stop simulating and start building.
The final performance will always be determined in the lab, not in front of a computer screen and a spice simulation.

[analog_sa]

Originally Posted by DF96

such people might be happy with a 'tuned' frequency response and 5% distortion as long as it is low order.

Care to name any such hypothetical people? Or any preamps which have been designed for 5% distortion and 'tuned' response? Or is this simply your fictional idea of what a demented audiophile should be?

[chrish]

asa, I don't think DF96 was 'pointing the bone' at anyone, just using a hypothetical spread of differing views to highlight his point that you need to first determine your design goals. Work out is what it is that makes you happy, then you can design for it.

[DF96]

5% distortion, mainly second order, will be common at high volumes for a SET amp. There are those, both DIY and commercial, who admit to 'tuning by ear' an RIAA network. I did not use the word "demented". I am not trying to start a debate, but explain why answering the OP's question is not straightforward.

Allowed levels of supply ripple depend on allowed levels of hum in the output and the PSRR of the circuit. You start from the output, and work your way back to what supply ripple you need, then include a safety factor. Acceptable hum/noise levels depend partly on whether you are before or after the volume pot - you might want 20-30dB better after the pot so you can turn down the music without hearing noise.