Thanks for this aboos.
This was what I was looking for.
Smaller capacitors can sometimes give more noise cos noise current.
In the IC mike amps like the excellent AD2019, THAT1512, TI163, there is an optimum size of input capacitor. You want to make them small cos big ones are more prone to zap the input devices which can't have much protection else voltage noise suffers. (Wayne has a couple of AES papers on the subject under "EVIL Phantom" or similar
) Many people think the min. size is determined by LF response but it's actually the noise current which sets the limit. Optimum size has the 'current noise' rising only below 20Hz.I think LTspice does this right if the model is correct. But you are right that the effect of the caps on 1/f noise may be the main consideration.
__________________
Do we have any more practical experience with ZTX851/951 1/f or greater noise? So far ...
GOOD : Wayne, H&H
BAD : syn08 .. selection required
Absolute nonsense on all counts Syn08.
Again you blunder in swinging the baseball bat without having checked anything.
1. If you took the trouble to look at the pdf in the first post you will see I have been publishing the amplifier only noise from the very first update (tables and graphs near the front)
2. Regulators or what ever you wish to call them. The current draw on these head amps is low so making them from parallel low noise opamps is quite feasible along with output RC filtering. The sim says 30 pV/ rt Hz. I guess it’s as fanciful then as all the amplifier sims?
Again you blunder in swinging the baseball bat without having checked anything.
1. If you took the trouble to look at the pdf in the first post you will see I have been publishing the amplifier only noise from the very first update (tables and graphs near the front)
2. Regulators or what ever you wish to call them. The current draw on these head amps is low so making them from parallel low noise opamps is quite feasible along with output RC filtering. The sim says 30 pV/ rt Hz. I guess it’s as fanciful then as all the amplifier sims?
An audiophile gets the cartridge they deserve.
3 Ohms 0.1mV is an SUT job in my book!
Agree, I have a 1 ohm cart and it definitely needs a SUT (o/p 0.04mV / 5cm / sec).
Du..uh! Wat iz an SUT??3 Ohms 0.1mV is an SUT job in my book!
Bonsai, I think you will find (like syn08) when you build an actual circuit that the LN PSUs rapidly becomes more complex than everything else.
I tried to integrate this as the CE input stage of a complete Virtual Earth RIAA amp but after superhuman PSU efforts, only managed to equal but not beat the 3mA Duraglit special. I was expecting at least the 1dB that Gerhad's supa dupa amp has over what I thought was Common Base Duraglit, in theory.
syn08 & Guru Wurcer's LTspice sims .. and especially Guru Han's insight that Duraglit was equivalent to a wonky Virtual Earth .. explains it after 20 yrs. It looks like Duraglit never relinquished its title of quietest practical MC head amp
The commercial amp went back to 0.5mA on 'Newton' cos that was less demanding of quiet PSUs and this was still quieter than anything else on the market. The PSUs to take full advantage of a 3mA 'Newton' were as complex as the rest of the circuit (cheapo 2x50W integrated amp)
From original users, I would have said 3mths for 3mA Duraglit and discerning users. More than 1 yr for the 0.5mA versions. That's allowing for the odd night when you forget to switch off the device.Hans Polak said:
The first thing you notice when the cell runs down is the noise increases. But Duraglit is so much quieter than lesser devices that many users didn't even notice that.
Other stuff, like gain, changes too but this is what you notice first when the cell is nearly dead.I think 3mA is a sensible max current to run for a C-cell and will continue to recommend that except for Golden Pinnae .. I mean specially discerning users like Bill.
I also think 'matching' is still worthwhile. After all, the 'matched' condition is actually BETTER than 3dB NF ie better than I thought. Running more current does reduce THD but matched Duraglit at 'overload' is already several orders of magnitude less THD than the cartridge.You really only need to get noise & THD sufficiently below vinyl noise to obtain the improvements in World Peace and a Better Sex Life I describe in my document
For a powered commercial amp, I think the Solar Cell + LEDs has the advantage. What's the difference in cost?
Bill, why don't you build a proper Duraglit special? Unlike our friends in da rebel colonies, you have access to pukka British Army issue Duraglitbillshurv said:
I see a business opportunity for you. Sell Duraglit tins to da rebel colonials for 100 Euros a pop + p&p. Justify the price by claiming they are hand-carved from solid Unobtainium by Shropshire (or even rarer, Manchester) Virgins. A bargain!!
Last edited:
Bill, why don't you build a proper Duraglit special? Unlike our friends in da rebel colonies, you have access to pukka British Army issue Duraglit
I see a business opportunity for you. Sell Duraglit tins to da rebel colonials for 100 Euros a pop + p&p. Justify the price by claiming they are hand-carved from solid Unobtainium by Shropshire (or even rarer, Manchester) Virgins. A bargain!!
I'd swap for Australian eze-start. Down there you can buy the same stuff but its called 'start you bastard' which is exactly what you are screaming when you squirt it into a recalitrant engine.
No problem, I'm not the one to stop you living your simulation fantasies
. So you think there is no further need for low noise regulators like the LT3045 (2nV/rtHz), simple electrolytic filtering buffers are good enough, right?Hints: capacitors ESR, ESL, collector current noise (for transistors) and the resulting noise voltage drop at the output, output current noise voltage density for op amps, input noise "regulation" (virtually zero dB outside electrolytics).
Last edited:
You made my day
. We have some similar **** here, used to start the snowblowers at -35 centigrades.
Last edited:
Why are power supplies with extremely low output noise levels required at all?
A Maxwell/Duraglit type Headamp with a 2 R cart and x30 gain has
- input noise level of 290 pV/sqrt Hz (incl. 2 R cart)
- output noise level of 9 nV/sqrt Hz
The power supply feeds the collectors of the transistors and thus feeds into the output. So, even in the case of zero PSRR, a power supply noise density of 2nV/sqrt Hz is already completely transparent and gives no noise penalty. Everything below in my opinion is useless overkill.
In a non floating power supply concept, it is much more diffficult to get a really noisefree ground and keep all the rectifier and mains crap out that easily can spoil the good looking LTSpice numbers.
Same holds true for electrically or magnetically coupled 50/60 Hz hum with all the harmonics. Keeping them at or below the noise is a real challenge.
A Maxwell/Duraglit type Headamp with a 2 R cart and x30 gain has
- input noise level of 290 pV/sqrt Hz (incl. 2 R cart)
- output noise level of 9 nV/sqrt Hz
The power supply feeds the collectors of the transistors and thus feeds into the output. So, even in the case of zero PSRR, a power supply noise density of 2nV/sqrt Hz is already completely transparent and gives no noise penalty. Everything below in my opinion is useless overkill.
In a non floating power supply concept, it is much more diffficult to get a really noisefree ground and keep all the rectifier and mains crap out that easily can spoil the good looking LTSpice numbers.
Same holds true for electrically or magnetically coupled 50/60 Hz hum with all the harmonics. Keeping them at or below the noise is a real challenge.
No problem, I'm not the one to stop you living your simulation fantasies
Hints: capacitors ESR, ESL, collector current noise (for transistors) and the resulting noise voltage drop at the output, output current noise voltage density for op amps, input noise "regulation" (virtually zero dB outside electrolytics).
Do the sims. The files are in the zip folder.
No claims of perfection, but a route (of probably many) to low noise regulation/filtering is in there with a bit of development.
How can you use a 2 nV/rt Hz reg on a circuit with 0 dB PSRR?
You will still degrade the noise performance aboos, but I agree its not as bad as it first seems in relation to the output noise. With the feedback designs it also arises in the first stage.
I am getting 5.4nV/rt Hz output noise on the powered Duraglit c.f. 242 pV/rt Hz input referred noise.
I am getting 5.4nV/rt Hz output noise on the powered Duraglit c.f. 242 pV/rt Hz input referred noise.
Last edited:
The headamp feed from a noiseless PS will have 9nV/sqrt Hz noise density at its output translating into 1.26µVrms noise over 20 kHz bandwith (leaving out some potential additional real world 1/f noise). With a PS having 2nV/sqrt Hz noise density and a PSRR of the amp of 0dB, output noise will degrade to 1.29µV rms.
This 30 nVrms penalty in my view is purely academic.
If the headamp is not feedback free from output to input for AC signals, then the story might be different - but most of the lowest noise headamps we are discussing here are CB BJT circuits without AC feedback from output to input.
Furthermore, if using a single floating supply (e.g. Maxwell/Duraglit), PSRR is not 0 dB but 40dB or higher while it is in the order of only 6 dB per supply rail for grounded dual supplies (e.g. Newton).
The next amplifier stage does not care where the input noise it sees is coming from. Yes, you feed the PS noise directly into the second stage but noise from the headamp is already much higher.
This 30 nVrms penalty in my view is purely academic.
If the headamp is not feedback free from output to input for AC signals, then the story might be different - but most of the lowest noise headamps we are discussing here are CB BJT circuits without AC feedback from output to input.
Furthermore, if using a single floating supply (e.g. Maxwell/Duraglit), PSRR is not 0 dB but 40dB or higher while it is in the order of only 6 dB per supply rail for grounded dual supplies (e.g. Newton).
The next amplifier stage does not care where the input noise it sees is coming from. Yes, you feed the PS noise directly into the second stage but noise from the headamp is already much higher.
Sorry, saw it to late.
With 5.4 nV/sqrt Hz, total output noise over 20 kHz bandwith will be 0.756µVrms which gets degraded to 0.806 µVrms by a 2nV/sqrt Hz PS if PSRR is 0dB. So noise performance is degraded by 0.55 dB / 50 nVrms
Thanks, but simulation is part of your problem, too many effects are not modelled (check again the hint list). As a rule, the simulator is a brain extension, not a brain replacement as you trend to see it.
The PSRR of the gain stage is a different story. In case it is low (but not 0 dB), a low noise true regulator (for its line regulation) followed by a low noise buffer can be used. One of the solutions I used is to make the buffer out of a CFP cap multiplier. The local current feedback loop effectively cancels the series power transistor Rbb and replaces it, as the CFP equivalent, with the Rbb of the driver transistor, which is one of those ultra low noise devices. It is also important to realize that the PS output noise current could be the limiting factor, not necessary the output noise voltage.
For the common base stage, the input referred noise PSRR is approximatively equal with the stage gain, so ultra low noise regulators/buffers are indeed not really required.
Last edited:
"As a rule, the simulator is a brain extension, not a brain replacement as you trend to see it."
More insults. To be expected of course.
"The PSRR of the gain stage is a different story. In case it is low (but not 0 dB), a low noise true regulator (for its line regulation) followed by a low noise buffer can be used."
That's what I've done.
"It is also important to realize that the PS output noise current could be the limiting factor, not necessary the output noise voltage."
No disagreement there.
More insults. To be expected of course.
"The PSRR of the gain stage is a different story. In case it is low (but not 0 dB), a low noise true regulator (for its line regulation) followed by a low noise buffer can be used."
That's what I've done.
"It is also important to realize that the PS output noise current could be the limiting factor, not necessary the output noise voltage."
No disagreement there.
Andreas,
I tend to agree with your arguments concerning the power supply noise, but for a lower gain like 20dB, I would prefer a higher PSRR than zero,
That's why for the sake of completeness I have attached an Image showing the PSRR for the version with 100uF+500uF and for 5mF+10uf.
Results speak for themselves.
Hans
Attachments
Hans,
the PSRR of the 'Duraglit' like amps is not bad. I investigated this only at 1kHz and found also values in the 40+ dB range.
+1 for your suggestion to use the large caps at the base as this obviously significantly improves the PSRR for low frequencies and brings it to 40+dB over the entire frequency range.
the PSRR of the 'Duraglit' like amps is not bad. I investigated this only at 1kHz and found also values in the 40+ dB range.
+1 for your suggestion to use the large caps at the base as this obviously significantly improves the PSRR for low frequencies and brings it to 40+dB over the entire frequency range.