Again: you could do it with just 1 small IC and a small 5.xV adjusted local reg like MIC5205 or the like.
There is not a problem, there is a nice challenge. To make it even smaller than the pastic surface of a LM317 for instance.
You revert to opamps which do not care as much but we are discussing DAC, Vref etc. where overhoot is killing IC's.
There is not a problem, there is a nice challenge. To make it even smaller than the pastic surface of a LM317 for instance.
You revert to opamps which do not care as much but we are discussing DAC, Vref etc. where overhoot is killing IC's.
Last edited:
LM317 comes in sot-223 package as well. But again, I do not consider LM317 for vref directly. I would consider it (with denoiser circuit) to power the reference for Vref, if it stays within the limits. Absolutely perfect for AVCC as well.
Could be possible that MIC5205 would benefit from the denoiser circuit as well
Could be possible that MIC5205 would benefit from the denoiser circuit as well
We could also go the hardest way, just a MAX6070 with some caps at the output without any added stuff
List extended again:
Ratio of the white noise and the DC voltage for various voltage and current references:
85A2 (glow discharge): 648.8 nV/sqrt(Hz) and 85 V, so 7.633E-9 Vref/sqrt(Hz)
LM723 (Zener / avalanche diode): 86 uV from 10 Hz to 10 kHz at 5 V output voltage, so 172.1E-9 Vref/sqrt(Hz)
LM317 (bandgap): 0.003 % of Vout from 10 Hz to 10 kHz, so 300.2E-9 Vref/sqrt(Hz)
uA7805 (bandgap): 40 uV from 10 Hz to 100 kHz at 5 V, so 25.3E-9 Vref/sqrt(Hz)
LT1236 (buried Zener reference): 2.2 uV from 10 Hz to 1 kHz at 5 V, so 13.98E-9 Vref/sqrt(Hz)
LT3081 (unknown, presumably bandgap): 5.7 nA from 10 Hz to 100 kHz at 50 uA, so 360.5E-9 Iref/sqrt(Hz)
LT3042 (unknown, presumably bandgap): 6 nA from 10 Hz to 100 kHz at 100 uA, so 189.7E-9 Iref/sqrt(Hz)
TL431 (bandgap): 125 nV/sqrt(Hz) at 2.495 V, so 50.1E-9 Vref/sqrt(Hz)
MAX6070 (bandgap): 5 uV from 10 Hz to 10 kHz at 1.25 V, so 40.02E-9 Vref/sqrt(Hz)
Of course it's not entirely fair comparison since the 85A2 uses 467.5 mW in its recommended operating point, far more than the LT1236 or the references of the LM723, LM317 and uA7805. Besides, when the noise is specified from 10 Hz to x Hz, there will be some 1/f noise included.
Well AK4468 is about 10 years newer. You keep saying to use newer parts! Also there's that whole NDA stuff about datasheets and stuff which complicates things. I'm not that much into exclusive clubs. Plus sourcing a legit chip etc.
And the most important thing is that I think that the specs of the AK4468 are more than enough for my needs, and the rest of my gear, for some time to come. I personally don't believe that ES9018 with 120dB DNR is going to make a difference from 117dB DNR of AK4468. Or thd+n of 108dB vs ak4468 104dB is such a difference.
Yes, MIC5205 is pretty nice and small, and small part count. So then MAX6070/1 + MIC5205 should do it.
I see MAX6071 has an extra sense gnd pin, and faster settle time. Is that something that is needed in this application? Gnd sense is irrelevant as the VCC sense as current is nothing and they are as close as possible to the DAC chip.
I can only get a MAX6070A 5V which is stocked where I shop and is also the A part, better tempco.
You put the specs of MAX6071. MAX6070 is 2.5uV from 10hz to 10kHz for 1.25V.
Could be possible that MIC5205 would benefit from the denoiser circuit as well
Thanks for having me spilling my coffee
ONLY when they are indeed better.... You are not a woman are you? Our discussions seem like quarrels with my ex
Ok, I am making your BOM: LT3042, MAX6071, MIC5205. Do you need any more good parts in your/our design? No advice on the DAC chip itself as I detest anything that has more than 2 channels. In general I am playing without a DAC although several ESS Sabre 9018K2M and ES9023 are around.
Last edited:
WRT LT3042 Ref pin, that is exactly right.
Let's say you want 3.3V OP which implies using a 33k R on V ref, there is 23nV/rt Hz straight off the bat. But we also have to consider current noise at ref pin since it is essentially a (very accurate) current source. Since the LT304x can work with very low dropout voltages, you can assume V ref current comes through an internal mirror. This will imply a lot of (current) noise gain and I think that is where your white noise is coming from.
It only has low noise due to the sledge hammer cap attenuating that noise on V ref.
I've used similar to John W's circuit for years. The secret is the LPF with schottky diode across the series R to 'boot up' the voltage to within 0.2V of required operating voltage. It can then slowly drift up to actual required
voltage with as slow a LPF as you like.... 0.01Hz LPF -> go for it... measure the noise on the OP of that.
TCD
Thanks for having me spilling my coffee
ONLY when they are indeed better.... You are not a woman are you? Our discussions seem like quarrels with my ex
MIC5205 has a roughly 36uV noise figure. Nonoiser lm317 managed 0.6uV best case I think. That's almost 36dB difference. I have no idea if the circuit is compatible or would help.
I'm ok with the analog parts, for digital power supply I'll probably have to use that dreaded lt304x reg. Dunno yet what other parts I'll need for digital but I should be able to figure it out.
Thanks for the info!
edit: max6070A not max6071B. told you already used the max6070A part. you helped with the mic5205, far cry from "our" design but whatever
Last edited:
Yeah it is only slightly less worse than standard LM317/7805 but with way less board space. I was merely pointing out that reasonable regs exist in that format (there are better ones but that is up to you to find out). Besides that the PSSR of the MAX6070/71 is there as well and some might say maybe the combination is "good enough". The low noise reference is the goal.
Last edited:
Exactly.
I need an 8ch DAC as I want a DSP to replace my active crossover for my biamped speakers, and want to later measure and correct for room etc, and maybe work on the headphones as well. Might add a subwoofer or two at some point et voila! 8 channels. Also all this needs to fit 10cm X 10cm pcb, with full tht resistors/capacitors for all 8 differential outputs, CLC filtering for three different rails, and three full sized LM317 installed, with absolutely no heat issues towards the caps (and ADAU1452 included). Hence some of my space limitations.
Maybe you prefer the phase shift and inefficiency of a passive crossover?
Last edited:
Yes I do. The less stuff I have the better it is. The simpler stuff is, the better it is in general (not always but very very often). The less cabling I have the better it is. The more complex chains are, the more issues can/will occur. The more design choices one has to make, the more circles one will be running in. The more operational choices there can be made, the more user errors can occur. For squeezing out the last 5% out of an already satisfying chain maybe 200% more money and 500% more effort must be invested. So, in general, simplicity has my preference.
No thanks, you are welcome. I am sure the combo MIC5205 or any other small low noise reg and MAX6070/71 will lift performance of the DAC. The predefined limit of 100 x 100 mm board space will benefit from their size as well.
No thanks, you are welcome. I am sure the combo MIC5205 or any other small low noise reg and MAX6070/71 will lift performance of the DAC. The predefined limit of 100 x 100 mm board space will benefit from their size as well.
Last edited:
You are sending mixed signals like some of my exes
That statement is so meaningless. There's situations where it's completely true, others where it's completely false. You just choose to use it when it suits you.
tht reg allows me to move the heat away from the pcb, to the case. lm317+denoiser is a very good use in this case (pun intended). lt3042 would keep the heat in the pcb. Plus I don't need the extra anyway, anywhere between 12-15V is just the same for the opamps. Low drop is really not important.
The digital part is getting a dc/dc converter as I have two 9VAC lines at 0.8A each. 15VAC lines have plenty to spare on the analog part, and overall wasted power is pretty low even with the lm317.
I have no idea who you are, I just interacted with you with the denoiser thing. Have you designed anything, and if so, could you point me to any project? Really curious of seeing some of your stuff you made, and I don't mean in a visual way, I want to see what design choices you made.
i had a chance once to sign an NDA for ess9018 when got into the only audio commercial project i ever worked for ...besides being a very expensive chip, ess part didn't impress me more than a dac found in a 30 years old cd player...There's no real proof that AKM dac's or adc are worse and they made a career in professional recording consoles already.
As far as i can remember without consulting the real project we used lots of cuk and step down smps followed by aADP low noise ldo regs .We had a section for sync-ing the enable signals for all the regulators as they were so many to preserve the unmeasurable noise of ess parts.The really funny part was when we found an audible hum on one channel due to the master pot wires peaking some EMI from the NEONS in the room with the open case ...It setteled down when completely shielded with the metal case, but it was realy fun to see a litteraly 1300 pounds worth of circuits suffering from audible ordinary hum
I can actually do better now with no regulator at all...just very carefull wiring.
Last edited:
Well..of course with modern dacs you need 1.1...3.3v regulators so you need special parts just because the technology shifted on lower and lower power supplies, but that made larger space for smps to do the hard job before the very low noise LDO so the industry switched on using very high frequency PSRR LDO because of these two factors: very low voltage, high current supplies and the additional smps supplies able to handle the heat when going from the regular 110...230 vAC mains to your 1.1...3.3volts. Most probably no modern soundcard company would agree on using lm317 based regulators.Most of them have a 12...24 Vdc external smps from the mains followed by other smaller smps, followed again by multiple ldo's for providing all the needed voltages in a modern digital equipment and that's the main rule for 30 years or so.I still have a MTCHLINE Philips video audio controller center which is full of smps down to the last ldo's ,no series regulator in a 30 years old digital-analogue equipment.
Vref is multiplied by the digital bit weights in dac chip's D/A section. If it varies in time, one might reasonably expect intermodulation distortion to result (which IME, can be audible and objectionable at very low levels).
Is going to a higher Vref let's say 5.5V going to result in better THD+N? Datasheet allows it as a continuous operating voltage. Absolute maximum is 6V, and their specs are rated at 5V and 3.3V for Vref (but 3.3V Vref specs are also with 3.3VACC). There's 5dB degradation for THD+N between 5V and 3.3V for Vref. And 4dB of dynamic range. If this scales linearly to 5.5V then it means 1.5db gain in THD+N and almost 1dB more of dynamic range (which gets degraded by digital volume control).
This is not worth going to the trouble of finding a decent 5.5V reference but I was more curious from an academic point of view.
Last edited:
Higher Vref will result in a little higher dac output voltage. Its a mulitiplier, so in terms of dB a few tenths of a volt may not much in the way of perceived volume level.
Where Vref gets more tricky, IMO only, is in the noise and distortion it produces during music playback. The audio output gets multiplied by those things. Distortion arises because the current draw of the Vref load is not constant. When more current is drawn the regulator must respond by trying to hold the voltage constant. Since its error amplifier has finite gain, the voltage will never be perfectly constant. Also, since the error amplifier will have some nonlinearity, its response in trying to hold the voltage constant will not be perfectly proportional to the current draw. Thus drawing a sine wave of current will result in some very low level voltage fluxuations at harmonics of the current sine waveform. This can be seen using a distortion analyzer, which was done for a few audio voltage regulators by forum member John Walton.
Of course, the whole question about voltage regulator linearity is how nonlinear does it have to be before it can be audible by a human listening to music played by a dac? IME, and no doubt some will differ, with some dacs the Vref regulator sound will be heard if the dac is otherwise well designed enough that Vref imperfections are not masked by bigger problems in the dac and in he rest of the reproduction system. Okay, that's my two cents on it. I would suggest to experiment if you want to learn. Find out for yourself what you can or can't hear.
Where Vref gets more tricky, IMO only, is in the noise and distortion it produces during music playback. The audio output gets multiplied by those things. Distortion arises because the current draw of the Vref load is not constant. When more current is drawn the regulator must respond by trying to hold the voltage constant. Since its error amplifier has finite gain, the voltage will never be perfectly constant. Also, since the error amplifier will have some nonlinearity, its response in trying to hold the voltage constant will not be perfectly proportional to the current draw. Thus drawing a sine wave of current will result in some very low level voltage fluxuations at harmonics of the current sine waveform. This can be seen using a distortion analyzer, which was done for a few audio voltage regulators by forum member John Walton.
Of course, the whole question about voltage regulator linearity is how nonlinear does it have to be before it can be audible by a human listening to music played by a dac? IME, and no doubt some will differ, with some dacs the Vref regulator sound will be heard if the dac is otherwise well designed enough that Vref imperfections are not masked by bigger problems in the dac and in he rest of the reproduction system. Okay, that's my two cents on it. I would suggest to experiment if you want to learn. Find out for yourself what you can or can't hear.
Last edited: