I think it is indeed G-R noise. With a quantum efficency in the 90 % range, almost each photon creates an electron-hole pair. These have to recombine again. If you consume all available current, which can be set by illumination intensity, recombination takes mainly place at the electrodes. If there are excess electron- hole pairs, they recombine at other places and thus creating noise.
If this is the noise effect, the best operating point then would be to set illumination intensity to a level where the available short circuit current of the cells is only a little bit higher than the required current and the cell voltage just starts to decline
If this is the noise effect, the best operating point then would be to set illumination intensity to a level where the available short circuit current of the cells is only a little bit higher than the required current and the cell voltage just starts to decline
Last edited:
I still have a lot of samples from the days of working on the pulse oximeter and indeed some LED's have large low frequency noise, noise around 1Hz was important for obvious reasons.
I still have a lot of samples from the days of working on the pulse oximeter and indeed some LED's have large low frequency noise, noise around 1Hz was important for obvious reasons.
Yeah, I know those cute devices... However, here I think it is likely a different mechanism. I am not sure if the LED noise reflects in light and then noise in the solar cell, or in this case it is G-R noise from the cell itself. But sure as hell, white LEDs (in fact blue, from a semi perspective) have horrible noise, much worse than the (more or less) monochromatic red LEDs. At least because of the wide bandgap of InGaN.
Scott, I'm not using the transformer to 'terminate' the 200R microphone.
It's used to transform the 200R source to Ropt = rt(Rnv x Rni) the optimum source resistance for the 'OPA'. This is about 7K1 for 5534 and 450R for AD797 (from the datasheet). These are the source resistances where the 'OPA' gives the best NF. The Jensen website has a couple of practical examples of this 'matching' for AD797 and 5534, IIRC. The ultimate NF at Ropt is also a measure of the range of source resistances with 'good' noise performance.
If the 'OPA' sees a source resistance lower than Ropt, its voltage noise 'dominates'. If it sees higher, its current noise 'dominates'. That's why the LN FET i/p OPAs can give very good NF .. provided you can wind a transformer to match their very high Ropt without crummy HF.
No. I was comparing 2 transformer i/p LN amps. From their NFs, my 80s Calrec design is 1.3dB closer to SOTA than Lepaisant et al. It's down to the transformers.
If you want an example closer to their 0R5, it's likely some of the better ribbon transformers could also better Lepaisant et al as they would target better than 3dB NF with 0R1 ribbons.
Richard, I believe you need to align your ducks in a row, on the beach.
http://michael.e.gruchalla.org/WebpagePapers/OptimumSourceResistance.pdf
I'm quoting this only because it directly addresses your [wrong] way of thinking. Otherwise, any half decent book on LF noise (and I mean, no RF) will tell you exactly the same thing as above.
I'm afraid this is off the track and only will serve to confuse folks, these principles do not apply to transformerless circuits that is where this started. A user wants some degree of control of the gain structure, since the sensitivity and resistance of a MC cartridge scale in very different manners you can't optimize everything and the 0.5nV or so FET input looks pretty good no matter what you are using. Virtually any step up ratio will give nearly 0dB "NF".
BTW virtually no one specs the self noise of ribbon mics why is that? The only one I found was 25dBA which is rather ordinary.
The confusion here might be because the best noise performance is seen at occurring at Ropt. In the attached graphic, that would be where the total equivalent noise line most closely approaches that of an ideal resistor wrt thermal noise.
All Ropt is telling you is that below it’s value the amplifier noise dominates and above it the source resistance dominates. In any amplifier system with Ropt < c. 0.75 that of Rgen, Rgen will dominate (cf we can just perceive c 1.5 dB power ie 10 log)
Here’s a link to the tool that will plot these graphs Opamp Noise Visualizer
All Ropt is telling you is that below it’s value the amplifier noise dominates and above it the source resistance dominates. In any amplifier system with Ropt < c. 0.75 that of Rgen, Rgen will dominate (cf we can just perceive c 1.5 dB power ie 10 log)
Here’s a link to the tool that will plot these graphs Opamp Noise Visualizer
Attachments
Last edited:
The Ropt concept gets a little silly for FET amps. If you take a 0.5nV composite FET amp virtually any source impedance works.
That's strike one; strike two would be accepting that the best noise performance indicator for audio would be the Noise Factor, Noise Figure or even SNR. Which, myself, I am having a hard time digesting. Audio is not space communication, where the SNR is one of the critical parameters for correctly demodulating the signal. I now recall modulation 64 and 256 QUAM schemas, Viterbi decoders, trellis code, punctured codes, etc...
. O tempora o mores 
Even for RF, impedance matching for max power transfer isn't universal, although that's often assumed. RF inputs can be and sometimes are "noise matched".
For playing vinyl we'd ideally like a number for noise referred to 5cm/s (and THAT specified as either horizontal or per channel - please!) with RIAA or AES overlaid. With or without A weighting?
But for designing MC preamps this is silly. Volts.
All good fortune,
Chris
I've never claimed this. It applies only to a particular case, the common or garden CB amp which we now know Duraglit isn't.
For best noise, matching is to Ropt = OPAvoltagenoise/OPAcurrentnoise = rt(Rnv * Rni) rather than 'input resistance'.
Guru Wurcer, I've already pointed out the good performance of LN FET i/p OPAs due to their very high Rni (just another way of saying very low current noise) You just need to get the source seen by your 'FET OPA' sufficiently smaller than Rnv.
Trying for Ropt is in most LN FET OPAs impractical cos HF problems with HiZ windings but as Guru Wurcer says, you don't have to go so far for excellent performance.
But to cut a long story short, someone asked whether a SUT would give better results for MC. The answer is unless you are a lot cleverer than me and have access to much better core materials, you won't beat 1.5dB NF even with mucho $$$. I found this circa 1980 when even an unmatched Duraglit beat a large & $$$ matched Denon transformer. Lepaisant et al certainly don't.
It's likely Les Watts' active ribbon (and possibly the active Royers) ... which combine my suggestions on transformer design and Guru Wurcer's insights into LN FET OPAs to the limit, beat Lepaisant et al.
One place where Lepaisant et al do shine, but don't really explain, is their transformer(s) + amp go up to 100kHz. This is still an area of Black Magic to even old fogey LN transformer amp designers. You can get a flavour of this in RDH 4 edition and also old Wireless World articles by GG Baxandall & others.
In theory, a LN FET i/p amp with an appropriate transformer (like Lepaisant et al) should beat my 1.5dB NF but I note Jensen don't show such a design and they have most certainly looked.
____________
BTW, my/BBC/IBA et al NF corresponds directly to the Noise Contours in LN BJT datasheets. In a LN transformer amp, you run the i/p BJTs in the best conditions for best NF and the transformer ensures they see the optimum source resistances for this.
Last edited:
The 'self noise' of a ribbon is usually the Johnson noise of its 'nominal resistance'. Most (all?) ribbons are too inefficient for acoustic resistance noise to be seen. I would love to see a BBC 4038 with enough sensitivity/efficiency such that acoustic resistance noise is a problem
Sensitivity and 'nominal resistance' will give you 'self noise' for practically all passive ribbons.
Last edited:
I think this has been beaten to death, you continue to talk about "NF" like it's a property of a circuit with no regard to source resistance. The Lepaisant circuit will have 0.06nV referred to input noise with your 3 Ohm cartridge as a source that is a lot better than 1.5dB "NF".
What ribbon element has more than a few 10's of milli-Ohms resistance?
EDIT - Sorry I found numbers more like .5 -.7 Ohms very thin foil these days. So Lundahl has a 1:37 transformer with 0.05 Ohms primary resistance, what's the problem?
Last edited:
So WTF is this 1.5dB NF number (I understand you don’t adhere to the standard NF definition) and how did you calculate or measure it? Use some basic arithmetic and real numbers please.
That's the same number Jensen comes up with on some of their applications, it seems maybe adding a constraint on bandwidth increases loading loss. Not sure of the details, not enough info on all the transformer specs.
If you do the plot for a JFET amp it does not perform as well as a bipolar amp at low Rgen (JFET noise voltage dominates) but at higher source resistances, it tracks the pure resistance line very closely to very high values before deviating and going more vertical - see the OPA627 plot in the graph. Of course this all makes sense because current noise in a JFET is very low but noise voltage is higher that bip types.
I am not saying this how to ‘match’ Rsource to Rin - simply if you follow the 1.5 noise guideline you’ll be ok. So any Rsource below Ropt and your amp is then the bigger noise contributor. Importantly, there’s no point in spending money to take equivalent amplifier noise lower than 1.5 x Rgen. Anyway this is audio so everything will be driven to extremes . . .
Last edited:
Err.rrh! NF is ALWAYS with regard to a source resistance.
Lepaisant et al's Fig 5 shows 3dB NF with regard to their 0R5 source. My 1.7dB NF Calrec transformer mike amp is with regard to 200R for a dynamic microphone.
The NOMINAL resistance of the ribbon is what you see at the output of the transformer .. usually 200R - 600R.
It's the ribbon resistance transformed by the square of the transformer ratio. That's why you use that with the nominal sensitivity of the mike to determine 'self noise'.
If you use the resistance of the ribbon, you would have to use the sensitivity AT THE RIBBON to determine 'self noise'.