I think in the video you posted he's talking about distortion, on the output of an audio stage? I was not interested in distortion, rather in PSRR and noisefloor of power supplies.
I have measured BC337 and BC817 and with BC817 the noise measured lower (on discrete + denoiser supply). And this was predicted by the LTSpice sim.
With BC550 instead of BC337 the LTSpice sim shows higher noise. But for BC550 I used Cordell's models, as I don't think there's a native model for BC550.
You are correct, the noise floor of the system is determined mainly by the bit depth. 16 bit will be around -120dB and 24 bit will be around -140dB.
To measure noise below these values you must amplify the noise a known amount to get it up into the range the system can measure.
Most people use a +60dB low noise amplifier (LNA) which will multiply the noise 1000 times. Very handy because if it reads uV, then the real noise is 1000 times less than that, nV.
Some people use a +80dB LNA which would move the "real " measurement decimal point to the left 4 places, because it multiplies everything by 10,000 times.
This procedure does not work with distortion measurements and is not needed because the distortion is usually high enough to read directly for the equipment diy builders play with.
I have found the BC817's, in -16 and -25 selections.
I compared them to BC337-16 and -25 in my transistor noise tester:
The results are consistent: the 337's produce a noise output of 120mV, whatever that means (I don't remember exactly how the output relates to the noise, I would need to browse my documentation to know the exact relationship) and the 817's produce 90mV, thus significantly less.
I used the tester rather than a denoiser, to eliminate the effects of the 317, protection resistor, etc.
That's pure transistor noise performance, and depending on the context, it might make a difference, or not.
40 or 50 years ago, the 817 began its life as a pure electrical clone of the 337, but at some point they diverged.
Both my samples are from Philips, made in the mid-2000's, just before the NXP conversion.
Attachments
They are extremely cheap, and completely generic which means there is no guaranteed consistency from one lot to another.
I purchased mine 1~2years ago from banggood, but if I place a new order now, I am not sure they will show the same performance.
There is a constant though: processes have improved, even for dirt-cheap components, and this means that recently produced parts will generally perform better than historic, vintage types; for cheap, generic parts at least.
If you try to buy a good power BJT or LatFet from a cheap oriental source, chances are you are going to be disappointed.
I purchased mine 1~2years ago from banggood, but if I place a new order now, I am not sure they will show the same performance.
There is a constant though: processes have improved, even for dirt-cheap components, and this means that recently produced parts will generally perform better than historic, vintage types; for cheap, generic parts at least.
If you try to buy a good power BJT or LatFet from a cheap oriental source, chances are you are going to be disappointed.
I am a latecomer to this thread.
Elvee may I express my thanks for the work you have done.
I have built a basic denoiser onto a 317 and it performs as your description.
I also had 3 LT 3083 regulators on hand to do some comparative measurements and the denoised 317 noise level came in at about 6 db better than the LT3083. I used a PECO scope spectrum analyser and an old fashioned Tektronics Linear scope.
The real proof to me is that Pk to Pk noise in the scope is 1.2mv for the LT and 0.8mv for the 317.
( both set at 12v with a 25 ohm load)
The 317 does exhibit a voltage overshoot of about 5% on startup which takes a few seconds to stabilise. ( as mentioned in an earlier post )
Most of my "in use" supplies use a floating pre regulator. I have order sufficient components to allow me to delve a little deeper and try the other versions of denoiser.
I use LT 80XX series regulators a lot as their TO220 5&7 pin cases are easy for DIY but they are becoming increasingly difficult to obtain in the UK and cost up to £10 each.
Ill report back on future progress.
Elvee may I express my thanks for the work you have done.
I have built a basic denoiser onto a 317 and it performs as your description.
I also had 3 LT 3083 regulators on hand to do some comparative measurements and the denoised 317 noise level came in at about 6 db better than the LT3083. I used a PECO scope spectrum analyser and an old fashioned Tektronics Linear scope.
The real proof to me is that Pk to Pk noise in the scope is 1.2mv for the LT and 0.8mv for the 317.
( both set at 12v with a 25 ohm load)
The 317 does exhibit a voltage overshoot of about 5% on startup which takes a few seconds to stabilise. ( as mentioned in an earlier post )
Most of my "in use" supplies use a floating pre regulator. I have order sufficient components to allow me to delve a little deeper and try the other versions of denoiser.
I use LT 80XX series regulators a lot as their TO220 5&7 pin cases are easy for DIY but they are becoming increasingly difficult to obtain in the UK and cost up to £10 each.
Ill report back on future progress.
Thanks for your feedback; always good to know
For noise and other low-level annoyances, I wouldn't rely too much on a scope: the level is too low, and it is easy to misjudge the actual value.
A spectrum analyser with a sufficiently low floor is better, but ideally you need some kind of audio analyser, AC microvoltmeter or soundcard with a LNA frontend. Look at the other thread for details
The overshoot should normally not be an issue, but in case it is, you can increase the 220µ coupling (startup will take longer though) or decrease the 22µ (it will reduce the VLF correction)
You can also consider discretes alternatives: Trileru and I have described some variants
Has anyone come up with a stable LM337 DieNoiser (cfp version)?
I assembled LM317 and LM337 boards by Tombo56, who was very helpful with debugging. We weren't able to figure out a way to get the LM337 board to work (it was oscillating badly) without disabling the second transistor (conversion to DeNoiser). I used STMicro chips.
My other issue is that the LM337 overshoot is much higher than LM317 (about 3V vs 1V). Is this more typical of DeNoiser or just the LM337 chip? I am somewhat concerned about startup overvoltage as I set the output voltage to 16.5V (we calculated for 17V, but it ended up being a little lower in reality).
I experimented with different resistor values with no luck to change this. I see that Elvee recommends increasing 220u coupling or decreasing 22u. Just to confirm, either of should reduce the overshoot magnitude?
I assembled LM317 and LM337 boards by Tombo56, who was very helpful with debugging. We weren't able to figure out a way to get the LM337 board to work (it was oscillating badly) without disabling the second transistor (conversion to DeNoiser). I used STMicro chips.
My other issue is that the LM337 overshoot is much higher than LM317 (about 3V vs 1V). Is this more typical of DeNoiser or just the LM337 chip? I am somewhat concerned about startup overvoltage as I set the output voltage to 16.5V (we calculated for 17V, but it ended up being a little lower in reality).
I experimented with different resistor values with no luck to change this. I see that Elvee recommends increasing 220u coupling or decreasing 22u. Just to confirm, either of should reduce the overshoot magnitude?
Hi Trileru, thanks for the reply.
The compensation network is 33 ohm in series with 22uF.
Output capacitor - do you mean the 220uF capacitor or the 100uF capacitor to ground on the output side of the regulator?
I installed these recommended capacitors:
220u capacitor is EEU-FM1E221B (low esr)
100u capacitor is EEU-FC1J101 (high esr I believe, for stablity/capacitive loading)
The compensation network is 33 ohm in series with 22uF.
Output capacitor - do you mean the 220uF capacitor or the 100uF capacitor to ground on the output side of the regulator?
I installed these recommended capacitors:
220u capacitor is EEU-FM1E221B (low esr)
100u capacitor is EEU-FC1J101 (high esr I believe, for stablity/capacitive loading)
Last edited:
Thanks.
I am a little confused about the ESR - it seems like I was incorrect to say that the output capacitor is high ESR.
The EEU-FC1J101 capacitor has dissipation factor (DF) = 0.08 for this 63V version.
I see that ESR is calculated by:
ESR = DF*Xc = DF / (2 * pi * f * C)
So my output capacitor has:
ESR @ 120 Hz = 1.061 Ohms
ESR @ 100 kHz = 0.001273 Ohms
Looking up a 100uF 50V capacitor with ESR actually specified as "61 mOhm" shows DF = 0.10.
I can try 2x22nF and 1-3R for compensation.
I am a little confused about the ESR - it seems like I was incorrect to say that the output capacitor is high ESR.
The EEU-FC1J101 capacitor has dissipation factor (DF) = 0.08 for this 63V version.
I see that ESR is calculated by:
ESR = DF*Xc = DF / (2 * pi * f * C)
So my output capacitor has:
ESR @ 120 Hz = 1.061 Ohms
ESR @ 100 kHz = 0.001273 Ohms
Looking up a 100uF 50V capacitor with ESR actually specified as "61 mOhm" shows DF = 0.10.
I can try 2x22nF and 1-3R for compensation.
Panasonic FC 100uF/63V is perfect for LM317 +de/dienoiser output.
For LM337 you need something like Panasonic FR 470uF/25V (20mm long version). Model number EEUFR1E471L. I've always had good results with this particular capacitor and LM337 + de/dienoiser, but you could try others that have the same low ESR.
Also the compensation network is important here, 22R is too much. 1R-3.3R should be fine, and 22nF to 47nF for the cap. This combo should work.
edit: EEU-FC1J101 datasheet has ESR stated as 0.256R at 100kHz. This is fine for LM317, but too high for LM337.
For LM337 you need something like Panasonic FR 470uF/25V (20mm long version). Model number EEUFR1E471L. I've always had good results with this particular capacitor and LM337 + de/dienoiser, but you could try others that have the same low ESR.
Also the compensation network is important here, 22R is too much. 1R-3.3R should be fine, and 22nF to 47nF for the cap. This combo should work.
edit: EEU-FC1J101 datasheet has ESR stated as 0.256R at 100kHz. This is fine for LM317, but too high for LM337.
Last edited:
Thanks, I will have to try the new capacitor. Right now I am trying 1.8R for the compensation network.
What's your recommendation for the CFP feedback resistor? (TBH I am a little confused on the nomenclature, since the output is taken from the top rather than the bottom of the "CFP") .I see 180R to 560R on different designs.
What's your recommendation for the CFP feedback resistor? (TBH I am a little confused on the nomenclature, since the output is taken from the top rather than the bottom of the "CFP") .I see 180R to 560R on different designs.
Ok, I actually just tried 220R. I was getting way less oscillation (only VLF sawtooth). Then I tried various capacitors in my box and arrived at the 220uf EEU-FM1E221B that I also used for coupling. It's only 25V but it will never see more than 20V. And voila! It works!
If there is interest I can draw up a diagram to show the exact parts I used.
EDIT: I spoke too soon, after a long warm up it began showing spurious peaks and random deep spikes... VLF is back, just VVLF.
If there is interest I can draw up a diagram to show the exact parts I used.
EDIT: I spoke too soon, after a long warm up it began showing spurious peaks and random deep spikes... VLF is back, just VVLF.
Last edited:
560R seems like a good all-around value for that resistor.
For the LM337 output cap I had short ones that don't work even if low ESR. My best guess is that they need to have some ESL as well, hence why I recommend the 20mm long one from the FR series.
I discovered the extra inductance requirement while trying ceramic caps for the output of the regulator. Adding a small 20nH diy aircore inductor allowed for small ceramic output caps.
For the LM337 output cap I had short ones that don't work even if low ESR. My best guess is that they need to have some ESL as well, hence why I recommend the 20mm long one from the FR series.
I discovered the extra inductance requirement while trying ceramic caps for the output of the regulator. Adding a small 20nH diy aircore inductor allowed for small ceramic output caps.