• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Sub 1V SINAD?

You mean the circuit theory? At what level? That is a pretty broad subject, but any college-level (senior or grad) text will have sections discussing noise and how to reduce it.
More on the topologies and things like that. They make stuff and usually that stuff has patents or white papers, but I haven't had any luck finding any. You are right on one thing: maybe I should go crack a book on design theory before anything. Its not as if there aren't piles of literature out there on the subject.
 
More on the topologies and things like that. They make stuff and usually that stuff has patents or white papers, but I haven't had any luck finding any. You are right on one thing: maybe I should go crack a book on design theory before anything. Its not as if there aren't piles of literature out there on the subject.
Low bias current and high gain in BJTs, large area for MOSFETs, topology is probably less a concern for noise except to keep it simple in the front end. Since noise is uncorrelated (I assume you are talking about thermal, shot, and so forth rather than quantization noise) most things you add to the circuit (e.g. for lower distortion) tends to hurt noise. I have several books on noise, but they are buried, and pretty old (though the fundamentals have not changed, but they may be hard to find now). If you are an IEEE member you can search the CAS and JSSC libraries for tutorials.
 
Kind of surprised that no solution has been found for consumer audio

I'd say the solution has been found and then some... As others have noted, if you have 130dB at full output, you really need to screw around with gain just to have a chance of hearing a little noise.

A lot of devices can't even be set to gain that low.

Noise is an unavoidable fact of life, but if nobody hears any noise while they're using the gear except extreme edge cases (which is the case for high performing DACs) then I think that's mission accomplished.
 
Low bias current and high gain in BJTs, large area for MOSFETs, topology is probably less a concern for noise except to keep it simple in the front end. Since noise is uncorrelated (I assume you are talking about thermal, shot, and so forth rather than quantization noise) most things you add to the circuit (e.g. for lower distortion) tends to hurt noise. I have several books on noise, but they are buried, and pretty old (though the fundamentals have not changed, but they may be hard to find now). If you are an IEEE member you can search the CAS and JSSC libraries for tutorials.
Actually this brings up another question: So when I look at the spec sheets for things like OPA161x and OPA891, they boast very good noise figures. But I've also read that you have far more control in discrete, and that inherently with any IC you are left at the mercy of whatever design choices were made in its inception. Is it worth exploring discrete options when op amps are on the table?
I'd say the solution has been found and then some... As others have noted, if you have 130dB at full output, you really need to screw around with gain just to have a chance of hearing a little noise.

Noise is an unavoidable fact of life, but if nobody hears any noise while they're using the gear except extreme edge cases (which is the case for high performing DACs) then I think that's mission accomplished.
True, we have come a very long way. But when I look at so many designs, while I understand that we have in fact come a long way, there are still proven design choices that can be made in order to better optimize for noise and distortion content.
 
I have several books on noise, but they are buried

Are your books buried in noise? You need FFT (Find F#$%ing Textbook) to help dig them out!
 
Are your books buried in noise? You need FFT (Find F#$%ing Textbook) to help dig them out!
LOL, missed you, John! Might take more than that; I think they are in boxes in the black hole masquerading as our basement storage ("exercise") room.
 
Actually this brings up another question: So when I look at the spec sheets for things like OPA161x and OPA891, they boast very good noise figures. But I've also read that you have far more control in discrete, and that inherently with any IC you are left at the mercy of whatever design choices were made in its inception. Is it worth exploring discrete options when op amps are on the table?
What is your technical background?

There are pros and cons of using ICs and discrete transistors, again discussed in various texts and courses. ICs offer the ability to tightly couple devices for matching and thermal, can reduce circuit parasitics and potential noise coupling, and with a practically unlimited (relatively speaking) of devices available you can implement a lot of circuitry difficult to produce on a printed circuit board (PCB). You can add digital processing for control and compensation on the same chip as the analog circuits using a mixed-signal process. Discrete potentially offers more control of devices if you are designing your own circuits, and the ability to mix disparate devices that would require a complex IC process (e.g. CBiCMOS -- complementary bipolar and mosfet devices on the same chip), and it may be easier to get high-voltage devices if needed.

There are a lot of great ICs out there, designed by very competent designers, so chances are their design choices are good ones. If you go discrete, you need to have the design skills to beat the ICs if that is your goal. For low-level signals, with very low noise and distortion, you need solid circuit and board layout skills so it may be hard to beat an op-amp design based on an eval board. Even for power amps modules are available that will likely exceed the performance of most DIY folk.
 
What is your technical background?
I've been a test technician for most of my career, but a lot of different things from consumer electronics to some IT stuff to aerospace/defense. Jack all trades, but I do not have a degree so, admittedly, there are gaps in my knowledge base. But that doesn't mean I can't teach myself new things, and use the knowledge I do have to help extrapolate information from the knowledge I don't.
There are pros and cons of using ICs and discrete transistors, again discussed in various texts and courses. ICs offer the ability to tightly couple devices for matching and thermal, can reduce circuit parasitics and potential noise coupling, and with a practically unlimited (relatively speaking) of devices available you can implement a lot of circuitry difficult to produce on a printed circuit board (PCB). You can add digital processing for control and compensation on the same chip as the analog circuits using a mixed-signal process. Discrete potentially offers more control of devices if you are designing your own circuits, and the ability to mix disparate devices that would require a complex IC process (e.g. CBiCMOS -- complementary bipolar and mosfet devices on the same chip), and it may be easier to get high-voltage devices if needed.
Funny you should mention that. I am in fact wanting to make a board. An acquaintance of mine is providing some extra funding and said as long as the design is good, go ham. So currently I'm hunting piggies ;)
There are a lot of great ICs out there, designed by very competent designers, so chances are their design choices are good ones
That's what I figured. It would make sense given the how good some of the figures are on spec sheets. I was just curious as to how much design compromises are made, but if they're that good, then who cares?
 
But when I look at so many designs, while I understand that we have in fact come a long way, there are still proven design choices that can be made in order to better optimize for noise and distortion content.
Well, the steady parade of ever-better SINAD coming from Topping, Fosi, SMSL, and others is proof that you are right. And of course high-end lab gear performs even better.

At the most extreme end of noise reduction, we have LIGO which is used to measure gravity waves, and is so sensitive they keep track of the surf on the coast so they can account for the vibration... 250 miles away.

So perhaps the ideal DAC is somewhere in between a Topping E30 and that. ;)

Any good engineer also asks what they are building the device for. When we're talking music playback through normal speakers or headphones, there is only so much noise reduction you can do before it gets silly.

If the noise level is such that nobody will ever hear it unless they intentionally set out to hear it, "better optimized" is a weird term to apply to changes that ultimately don't benefit the user.
 
Compromises are made no matter what. Silicon area, IC process cost, PCB area and discrete device costs, etc. etc. etc. Noise and distortion has been below audibility for many years, at least for the best designs, so expecting to improve on today's best is a significant challenge. MOSIS offers fairly inexpensive IC processing, but for a DIY'r discrete devices and a simple (2 layer) PCB will probably be cheaper. But I would not expect to beat the state of the art.
 
Any good engineer also asks what they are building the device for. When we're talking music playback through normal speakers or headphones, there is only so much noise reduction you can do before it gets silly.
Eh this one is just a headphone DAC/AMP, so nothing high power. Can definitely go class A and have little to no issues.
but for a DIY'r discrete devices and a simple (2 layer) PCB will probably be cheaper. But I would not expect to beat the state of the art.
Definitely not trying for the heavy weight belt. Respectable, and maybe even a bit novel, is good enough for me :)
 
Yes if the noise increase in the silents parts of actual recordings compared to no signal you're still good . I would test with a file of total silence or something some DAC's sneakily mutes physically with no signal ? :)

I think we are close to physical limits if the signal max level is going to be 2v or 4v for example ( that why no 24bit DAC has 144dB SINAD ,but are 20-21 bit in practice ).

This can be very complicated fast and depends on source and sink impedance etc .

I do wonder what the physical limit is the electrons and atoms in our components are jiggling around at room temperature ? :) so where are for example the noise of a passive resistor at 100Ohm compared to a 2v signal ?

Is it not why both IEM's and MC/MM cartridges are noise limited especially MC carts as they have ridiculous signal levels ?
 
LOL, missed you, John! Might take more than that; I think they are in boxes in the black hole masquerading as our basement storage ("exercise") room.
Ha! We also have two and a half of those "storage" rooms. The half refers to a space filled with electronics equipment and my small device repair area. One room functions as both a woodshop and storage, while the other is mainly a "prepping" room for household supplies. Lesson learned: more space inevitably leads to more stuff...
 
Ha! We also have two and a half of those "storage" rooms. The half refers to a space filled with electronics equipment and my small device repair area. One room functions as both a woodshop and storage, while the other is mainly a "prepping" room for household supplies. Lesson learned: more space inevitably leads to more stuff...
I think George Carlin had a bit about this :p
 
Eh this one is just a headphone DAC/AMP, so nothing high power. Can definitely go class A and have little to no issues.
You can get about two gazillion decent to great ones off the shelf these days for a few hundred bucks. What are you hoping to achieve that one of those won't?
 
You can get about two gazillion decent to great ones off the shelf these days for a few hundred bucks. What are you hoping to achieve that one of those won't?
Reasons why I am doing this:
  • because I'm bored
  • because I can
  • because I feel like it
You may choose any combination you like as an answer :p
 
I do wonder what the physical limit is the electrons and atoms in our components are jiggling around at room temperature ? :) so where are for example the noise of a passive resistor at 100Ohm compared to a 2v signal ?

Formula for noise voltage of a resistor.
1737035606980.png


For 100 Ω and BW=∆f=20kHz that computes to U_N ≈ 0.18µV or -141dB compared to 2V.
 
Also 50mV is 32dB closer to the measurement system noise floor than 2V is.
 
Back
Top Bottom