jacco vermeulen said:
Ilimzn,
why is the BS107 preferred, not BS170?
You can still use BS170, no problem. IIRC The BS170 is a higher voltage device, it has slightly higher capacitances and slightly lower gm, at the same Id, while it's increased voltage handling is not required, the increased gm and decreased C's of the other two might be preferable. Still, differences are small - and I am pulling the relative comparisno of BS107 and 170 out of my failing memory 🙂 so please check this out before you decide.
You mean those folded springmetal high Napoleon hats ?
Yes, something like that - with the required dab of silicone grease too 🙂
Hi ilimzn,
You are thinking along the same lines as I am.
My understanding is that the input signal is converted into a current through the series resistance and base resistance (bulk emitter resistance being much lower, not as relevant. Am I right?) Any non-linearities in the B-E circuit will distort the signal unless it's driven by a low impedance. This distortion would then be outside the amplifier and corrective feedback. I do not know the magnitude of this, just trying to rationalize this in my mind. How close am I here?
With Fet's and tubes, the signal is presented a voltage with only variable capacitance and gate charge (with a mosfet) issues. Am I still on track? The voltage to current conversion is therefore inside the input device rather than before it. I don't know how to put this in math terms. Math is my weakest point.
Now, wouldn't the input signal be a much higher amplitude at the device control terminal using a fet or tube comparing to a BJT?
-Chris
You are thinking along the same lines as I am.
My understanding is that the input signal is converted into a current through the series resistance and base resistance (bulk emitter resistance being much lower, not as relevant. Am I right?) Any non-linearities in the B-E circuit will distort the signal unless it's driven by a low impedance. This distortion would then be outside the amplifier and corrective feedback. I do not know the magnitude of this, just trying to rationalize this in my mind. How close am I here?
With Fet's and tubes, the signal is presented a voltage with only variable capacitance and gate charge (with a mosfet) issues. Am I still on track? The voltage to current conversion is therefore inside the input device rather than before it. I don't know how to put this in math terms. Math is my weakest point.
Now, wouldn't the input signal be a much higher amplitude at the device control terminal using a fet or tube comparing to a BJT?
-Chris
anatech said:
I'd say more or less right on track. I would not be prepared to compare magnitudes between BJT and JFET/MOS, though, but read on.
Without going too deep into it, a good example would be to connect a, say, 5k series resistor from ideal source to the input of an amp in a simulator, and look at the THD at the input. This is not a bad example for an amp that needs no preamp - a 10k volume pot mid-way would give a 5k series resistance to the amp input. Given 47k as the usual input impedance, that is actually a rather optimistic approximation of the real world.
In a LTP non-inverting configuration used by most amps, any input nonlinearity of one active element, that is not cancelled by the other in the pair, remains outside of the feedback loop.
In a BJTs, there are several, hie modulation due to Ic imbalance (as LTP actually has to source current to thenext stage, especially at HF) - this is combatted by degeneration, hfe modulation with Ic and Vce, Cbc capacitance modulation with Vce. Any difference in dynamic impedances on the two inputs of the LTP that 'syphens' input current away from the actual active element control terminals, produces with the source impedance, a different potential divider in 'front' of said inputs, and some form of this difference is essentially out of the NFB loop.
The major difference with FETs is a MUCH higher input impedance, and no equivalent to hie modulation, so series impedance is essentially negligible - BUT the advantage is lost as we go up in frequency due to generally higher input capacitance, and further, if the inclreased and nonlinear Cgd comes into play. Given the lower gm of the devices, it is also plausible that the voltage swing across this nonlinear C would be higher than in the case of the BJT.
In other words, I would not venture into debating magnitudes of the top of my head, and possibly not at all - but I think I would not be far off in saying that as long as Vd/Vc are sufficiently higher than the peak input (to minimize the influence of Cgd or Cbc), and we use a high gm FET, there is probably a breakeven point, with FETs leading at LF (of course, the 'L' in LF begs for a definition).
As for the nature of the distortions, I can't say anything but that they would definitely be different in nature (different harmonic spread) owing to different Vin to Iout laws.
It should be noted that JFETs also have a form of 'gate charge'.
Exactly, and this is because of the lower gm. This higher voltage, or rather, voltage change, ipacts on the higher and potentially more nonlienar Cgd of FETs. Because of this, the nonlinear impedance of this C at HF becomes the dominant part in input impedance. Keeping it's influence as low as possible (forinstance by clever cascoding), and keeping the gm as high as possible (for instance by using a MOSFET and high tail current), are keys to giving the FET an advantage over the BJTs, or rather, extending up the frequency range in which the FETs can have an advantage.
ilimzn said:
Rather strange to quote myself but I forgot to add that this is probably where the merits of the Lavardin constant current constant voltage configuration are. By keeping the operating environment of theinput pair constant, these differences are minimized.
Hi ilimzn,
Thank you for your take on things, they make sense to me.
-Chris
That makes sense, it must be a lower magnitude than other factors because I haven't heard of this before. I suspected there must be something of this nature going on.
That is something I almost always do, since 1990 at any rate, in my own designs. I'm not familiar with the "Lavardin constant current constant voltage configuration". That is unless you mean the constant power arrangement for diff pairs. I usually find the cascode is enough to achieve rather good performance.
Thank you for your take on things, they make sense to me.
-Chris
I made some measurings with the jfet version... But i can't really tell the difference, with my first measuring i forgot the dummy load 🙄 , so the load was only ~38ohms, giving thd of ~0.002%.
Having 8ohms as dummy load, thd increased to 0.01%, showing that i have severe grounding problems in my measuring setup. Previous measurings with mpsa18 had 0.005%, doubling THD is what i expected as the nfb was lowered. I have the feeling that the THD rise with frequency lowered, i had 0.01% at 1khz, and 0.014% at 10khz.
http://www.lf-pro.net/mbittner/rmaa/Symasym_jfet.htm
Mike
Having 8ohms as dummy load, thd increased to 0.01%, showing that i have severe grounding problems in my measuring setup. Previous measurings with mpsa18 had 0.005%, doubling THD is what i expected as the nfb was lowered. I have the feeling that the THD rise with frequency lowered, i had 0.01% at 1khz, and 0.014% at 10khz.
http://www.lf-pro.net/mbittner/rmaa/Symasym_jfet.htm
Mike
Hi Mike,
Yes, you have a ton of noise there. Try to remove the common from the output into your sound card. The signal is already ground referenced.
-Chris
Yes, you have a ton of noise there. Try to remove the common from the output into your sound card. The signal is already ground referenced.
-Chris
It may be, but at the wrong point!
The problem with anything PC related is that the signal ground is directly coupled, but at an unknown point, to the earthing line of the mains net. If you have anything else grounded that has ground connecte to earth (like your scope, for instance), you get a ground loop. My vote would be to separate the earthing line of the PC from the actual earth by two antiparallel diodes (5A types would do well here), in parallel with several 10s of ohms of resistance (or even more). Either that, or use an isolation transformer.
The problem with anything PC related is that the signal ground is directly coupled, but at an unknown point, to the earthing line of the mains net. If you have anything else grounded that has ground connecte to earth (like your scope, for instance), you get a ground loop. My vote would be to separate the earthing line of the PC from the actual earth by two antiparallel diodes (5A types would do well here), in parallel with several 10s of ohms of resistance (or even more). Either that, or use an isolation transformer.
Michael,
your measurement does not help. Download the following file and burn it to CD:
http://web.telecom.cz/macura/1102.zip
your measurement does not help. Download the following file and burn it to CD:
http://web.telecom.cz/macura/1102.zip
anatech said:
PMA said:
mlloyd1 said:
anatech said:
As always,
it is interesting to visit this topic about MikeB's major project SymaSym.
Good discussion and good advices.
Think both parts selection, mechanical construction and wiring will be what it takes to make a real good performance amplifier.
Just take a schematic, with however good and interesting circuit solution, wont get you there.
An amplifier built using transistors and discrete components is a very complex system.
What works on patented papers, wont always work within limits of real life limtiations of parts you can find around at a given time in history. Sometimes you just have to take what you can get, if not in your neighbourhood, so at least in your home country.
Then there is money.
All are not made out of money and can afford what others can afford. Some may even have family and some may need to buy some food and clothing. All have not free lunch!
One thing mentioned here is wiring.
My own personal opinion is, that a DIY audio amplifier project offered in public
should have detailed wiring instructions. Preferably with wiring diagram.
I find this is too many times overlooked. People often can come around repeatedly asking about wiring / grounding problems.
So for 'my own SymaSym', my first published project which is under construction
I have already booked a page: 7. Wiring Suggestion
You will find it in my 😉 nice 😉 new website for audio projects
http://lineup.awardspace.com/
Regards everybody
from lineup
PMA said:
Hi Pavel, thanks, i already had burnt a CD with all kind of test signals, one of the tracks was the rmaa test signal, sine waves with different freqs (1,2,6,10khz) and IMs (10+11khz,19+20khz).
I don't know where the hum garbage is coming from, i already hunted it, i tried isolating earth of the PC, then i found that the hum starts when i connect the players mains cable. If only amp is switched, on the noise level looks quite clear.
But using the soundcard as signal source gives the "horrible" measurings. 😉
Mike
Mike,
I can't find silver mica 10pf capacitor. Maybe 10pf gimmick capacitor will be a good substitute ... or maybe a 3-10pf air trimmer ? 😕
I can't find silver mica 10pf capacitor. Maybe 10pf gimmick capacitor will be a good substitute ... or maybe a 3-10pf air trimmer ? 😕
A very simple and a good way to make small caps in RF projects
http://web.telia.com/~u85920178/begin/gimmik-0.htm
http://web.telia.com/~u85920178/begin/gimmik-0.htm
Roender,
Mail me your adress and i'll bootkick a letter pidgeon with a pair of 10pF silver micas your way.
Mail me your adress and i'll bootkick a letter pidgeon with a pair of 10pF silver micas your way.
MikeB said:
Hi Pavel, thanks, i already had burnt a CD with all kind of test signals, one of the tracks was the rmaa test signal, sine waves with different freqs (1,2,6,10khz) and IMs (10+11khz,19+20khz).
I don't know where the hum garbage is coming from, i already hunted it, i tried isolating earth of the PC, then i found that the hum starts when i connect the players mains cable. If only amp is switched, on the noise level looks quite clear.
But using the soundcard as signal source gives the "horrible" measurings. 😉
Mike
Hi Michael,
1) the 1102.5Hz signal I have provided is exact 1/40 of the Fs(44.1kHz) - think about it and probably try it.
2) in case of hum/buzz garbage I have had working solution - make a diff amp with INA217 and place it in front of the soundcard. You will have no hum anymore and its distortion is far below that of the soundcard.
Caps, transistor legs, pcb traces are antennas anyhow. Nothing wrong with twisting your own. Besides, you can avoid the mica dielectric issues. 😎
PMA said:
Ah, you mean that the permanently changing aliasing creates very high order "harmonics" ? I believed that they would be averaged out...
Yes, balanced input will really help, my last electronic order is a while ago, i'm a bit out of money these days.
Mike