Is a capacitor so scarce?
I was driving recording head by class D amp, long before "class D" name appeared. I don't remember if I used any coupling cap, most probably not.
...I also stopped using LP, for the same reasons.
So let's skip the 'no_offset-no_adjustment-DC_perfection' requirement from the spec. We simply don't have unsaint gain in the chain between line level and speaker.
The running discussion about offset is like optimizing the air drag coefficient of a combine harvester. Of course it can be done. It can even be done with great perfection and adorable know how, but it does not necessarily lead to a great harvester. It might even contradict. For an optimized air drag coefficient most likely one will end up with a very small ground clearance and narrow rotator....
If you have a solution with optimized air drag coefficient, but without impacting the relevant properties and without unsaint complexity - that's of course cool. Just put the schematic for discussion.
I still don't see Vgs matching between N and P channel tracking over temp, diff pairs make this a common mode error.
Quickly breadborded the design in #451, and did a few tests - did find only 10 2N5458s, so the selection was a little bit poor ; BC550/560 in the gain/driver stage, MJE243/253 running at ~70 mA in the output stage. At 10 dB gain BW is in the MHz range out of the box, and THD @ 1Vrms, 1 kHz into 30 ohm (~30 mW) is about 0.004% (only 2nd and 3rd harmonics), quite load insensitive (THD drops to 0.0038% into ~300 ohm). No compensation needed; DC drift good (for audio, at least) while not spectacular: output stage Iq rock stable, output offset in the +/- 5 mV range en plein air over a couple of hours. Well, it needs some more work, but seems promising - going to get plenty of J113s, do a better selection and see what happens
Ciao,
L.
; BC550/560 in the gain/driver stage, MJE243/253 running at ~70 mA in the output stage. At 10 dB gain BW is in the MHz range out of the box, and THD @ 1Vrms, 1 kHz into 30 ohm (~30 mW) is about 0.004% (only 2nd and 3rd harmonics), quite load insensitive (THD drops to 0.0038% into ~300 ohm). No compensation needed; DC drift good (for audio, at least) while not spectacular: output stage Iq rock stable, output offset in the +/- 5 mV range en plein air over a couple of hours. Well, it needs some more work, but seems promising - going to get plenty of J113s, do a better selection and see what happens Ciao,
L.
Comparing the published datasheets, other than Idss bracketing and max voltage, it appears from the characteristic curves that the 2SK246 has a much smaller transconductance than each half of the 2SK2145.
I'm about to measure the sample I'm testing for that parameter. I also have a lot of 246s I could measure.
2145 gm
Well I just snapped the source lead off of 2145 sample #1, but it served its purposes admirably
Sample two has Idss of 2.80mA, and the follower circuit an offset of -5mV. The transconductance is 13.3mA/V, based on the 6dB attenuation of a low-level 2kHz signal. This corresponds reasonably well with the datasheet curves, which are for a 10V Vds.
This indicates that midband noise should be, for the follower, about 1.29nV/sqrtHz (the current source Idss load entails a 3dB increase in equivalent input noise). Not too shabby, but I will attempt a noise measurement at some point as well. The 3dB additional noise could be eliminated, with a good deal of additional work, if the devices were in fact isolated, but so it goes --- in most applications that noise will be negligible.
Well I just snapped the source lead off of 2145 sample #1, but it served its purposes admirably
Sample two has Idss of 2.80mA, and the follower circuit an offset of -5mV. The transconductance is 13.3mA/V, based on the 6dB attenuation of a low-level 2kHz signal. This corresponds reasonably well with the datasheet curves, which are for a 10V Vds.
This indicates that midband noise should be, for the follower, about 1.29nV/sqrtHz (the current source Idss load entails a 3dB increase in equivalent input noise). Not too shabby, but I will attempt a noise measurement at some point as well. The 3dB additional noise could be eliminated, with a good deal of additional work, if the devices were in fact isolated, but so it goes --- in most applications that noise will be negligible.
Quickly breadborded the design in #451, and did a few tests - did find only 10 2N5458s, so the selection was a little bit poor
.
Great amp, Coluke! One more in ths thread!
2145 seems to be a two-chip dual
Sample #1 having served its country well, I ground down through the encapsulant to uncover the innards. I finally got to a square positioned over a gate lead about 400um on a side. It's plausible that this is one of the chips --- that's about the size of a 4416 chip, which transistor, although lower capacitance than the 2145, does not take up all of the die size.
And I still have fingerprints, although as I rubbed away I switched fingers from time to time. I guess this is what safecrackers were supposed to do, to increase their sensitivity.
Sample #1 having served its country well, I ground down through the encapsulant to uncover the innards. I finally got to a square positioned over a gate lead about 400um on a side. It's plausible that this is one of the chips --- that's about the size of a 4416 chip, which transistor, although lower capacitance than the 2145, does not take up all of the die size.
And I still have fingerprints, although as I rubbed away I switched fingers from time to time. I guess this is what safecrackers were supposed to do, to increase their sensitivity.
Yes, that was a quote about the 246 from EUVL, to which I was taking exception.
The 2145 etc. is about like a half-area SK170, about the same gm/Cin figure-of-merit. 1nV/sqrtHz is also what I got based on gm and the Van der Ziel theory.
It is strange why such big capacitance, because current drive is needed. But Brad was right, power supply transients on on / off could magnetize the head.
I mean pretty high dynamic resistance of the output stage. "Easiness" depends on frequency and resistance. The lower is resistance, the higher capacitance is needed for the same frequency response. And vice verse.
Not an easy task, if one uses NFB at the same time to define gain and/or to reduce overall distortion.
That one should better use NFB by current, not by voltage. To define gain and reduce overall distortion. However, HF bias adds additional requirements for filtering, but it was done.
Wavebourn, do you mean by that circuit similar to Widlar Op-Amp Current Source?
should be better ??
With the large amount of nfb used in Scott's circuit the distortion should be lower -- lower than .001%.
but your numbers are comparable to the little circuit I put up when it also has not well matched devices. When well matched, the distortion at moderate nfb is measured as under .001% (1v/30 Ohms). -RNM
Quickly breadborded the design in #451, and did a few tests - did find only 10 2N5458s, so the selection was a little bit poor
Ciao,
L.
With the large amount of nfb used in Scott's circuit the distortion should be lower -- lower than .001%.
but your numbers are comparable to the little circuit I put up when it also has not well matched devices. When well matched, the distortion at moderate nfb is measured as under .001% (1v/30 Ohms). -RNM
Last edited: