The 2sc5200/sa1943 -
http://www.toshiba.com/taec/components2/Datasheet_Sync//66/7890.pdf
are right on the money with Hfe (55), SOA (1.4A@70V) , voltage... as compared to the NJW0281/0302 's I use.
http://www.onsemi.com/pub_link/Collateral/NJW0281-D.PDF
The MJW/NJW's have a better SOA (1.6a@70v) so they will do better driving 4R loads. Gain wise, both will do the job.
I have the toshiba models , and notice the same simulation results with them as compared to the ON-semi devices.
Resistor paralleled with output inductor?? , Never had a problem without it. For the hell of it .. I ran this amp with nothing (no zobel or inductor) for hours with no issue.
OS
ostripper said:
I ran this amp with nothing (no zobel or inductor) for hours with no issue.
OS
If your load is not capacitive, really do not need the inducer.
The bass were better?
Strange! I used another model for the LED red, the fall of voltage, not was the same standart Fairchild model (QTLP690C).
My model:
You measured with multimeter, value simulated is equal to your value measured in real amplifier?
I did consider the MJL21193/4 , lower bandwidth and higher SOA did not scare me. What bothered me was the lower Hfe .. with 4 pairs and just a type 2 EF , I might get some "droop" at higher currents. A triple would be necessary
. they cost a lot ,too $4+ vs $1.40 for the NJW's.Simulation - LED 1.72v, real life 1.77V
resulting CCS current..simulation 3.84mA, real life 3.90mA
close enough for me. In fact, the real amps voltage and current readings rarely differ more than 1% from the simulated values.
OS
Hi Raphael.luc,
---I was talking to measure in simulator---
To measure output resistance in a simulator, I measure the output voltage with no load and then with a load the value of which is step by step diminished until the output voltage is precisely halved.
The found value of resistance is equal to the amp output resistance.
In real life, I own no less than five (used to 8 !) multimeters of 0.1% accuracy I sometimes find that true RMS and even digital readings can be nuisances. I measure damping factors only very rarely by now.
---I was talking to measure in simulator---
To measure output resistance in a simulator, I measure the output voltage with no load and then with a load the value of which is step by step diminished until the output voltage is precisely halved.
The found value of resistance is equal to the amp output resistance.
In real life, I own no less than five (used to 8 !) multimeters of 0.1% accuracy I sometimes find that true RMS and even digital readings can be nuisances. I measure damping factors only very rarely by now.
Re: Suggestion for C7
Very low voltage swing at collector of Q6 because of Q7 connected as a diode.
However, I was wondering if C7 does any good job. Despite my preliminary reservations, it may.
I just remember a discussion here about the power supply rejection at high frequencies of the standard blameless amplifier. I think Edmond Stuart proposed to improve it by connecting a capacitor of same value as Cdom in series with a resistor about 100 Ohm, from the diode side of the input current mirror to ground.
So here, C7 may have some influence on the design by providing a better PSRR at high frequencies and then should not be removed. This is only a supposition and should be checked.
mjurban said:
Very low voltage swing at collector of Q6 because of Q7 connected as a diode.
However, I was wondering if C7 does any good job. Despite my preliminary reservations, it may.
I just remember a discussion here about the power supply rejection at high frequencies of the standard blameless amplifier. I think Edmond Stuart proposed to improve it by connecting a capacitor of same value as Cdom in series with a resistor about 100 Ohm, from the diode side of the input current mirror to ground.
So here, C7 may have some influence on the design by providing a better PSRR at high frequencies and then should not be removed. This is only a supposition and should be checked.
Most of the discussion we have done so far is correctly referenced to the thread front page schema. (1 schematic)
Any others are someone else's or "clips" to share ideas.
I will present the final cascoded one here.
The base amp's device numbers have not changed , the new additional cascode's device numbers are added on after the fact.
I will update the "page 1" schema to this one soon.
OS
Any others are someone else's or "clips" to share ideas.
I will present the final cascoded one here.
An externally hosted image should be here but it was not working when we last tested it.
The base amp's device numbers have not changed , the new additional cascode's device numbers are added on after the fact.
I will update the "page 1" schema to this one soon.
OS
Hi ostripper,
The Pioneer is too simple. But that is the concept. You can use an IC shunt or low noise zener (as opposed to the normal types). The resistor from the top supply could be a current source, why not? Might reduce noise in practice.
Your second diagram looks good. You should not need R5, I'd leave it out. R48 = CCS, R49=shunt reg. You only need to add the current injected by the CCS at R48 to the total and it will mostly cancel out, leaving you with your original tail current.
You're looking good there! BTW, the cascode reduces gate leakage currents if you keep it to 8 or 9 VDC abs. max. Running anywhere around 6~7 VDC would be good, the major consideration being a low noise shunt.
This is what I have been using for about 20 years now. It sounds really nice, even if the transconductance is lower than a BJT. There seems to be something else going on there that the numbers really are not showing, or another distortion effect. My feeling is that the signal gets distorted before ever entering the amplifier due to varying input impedance that a BJT stage may have. Just one of my musings trying to figure out the whys of input device sound differences. If you buffer the signal before a BJT diff pair, the effect seems to go away.
-Chris
The Pioneer is too simple. But that is the concept. You can use an IC shunt or low noise zener (as opposed to the normal types). The resistor from the top supply could be a current source, why not? Might reduce noise in practice.
Your second diagram looks good. You should not need R5, I'd leave it out. R48 = CCS, R49=shunt reg. You only need to add the current injected by the CCS at R48 to the total and it will mostly cancel out, leaving you with your original tail current.
You're looking good there! BTW, the cascode reduces gate leakage currents if you keep it to 8 or 9 VDC abs. max. Running anywhere around 6~7 VDC would be good, the major consideration being a low noise shunt.
This is what I have been using for about 20 years now. It sounds really nice, even if the transconductance is lower than a BJT. There seems to be something else going on there that the numbers really are not showing, or another distortion effect. My feeling is that the signal gets distorted before ever entering the amplifier due to varying input impedance that a BJT stage may have. Just one of my musings trying to figure out the whys of input device sound differences. If you buffer the signal before a BJT diff pair, the effect seems to go away.
-Chris
R5 is gone , that was tested on the "Guinea pig" (real amp) , the zener shunt (replacing r49) is an improvement. The use of a second CCS instead of 220k (r48) is redundant , only 1DB more psrr. The tail reference is rock solid at only .001ma "slop", and the cascode is locked to that through the zener. If we are going to allow for the use of a separate VAS/LTP power supply anyway, it would be definite overkill.
OS
Attachments
Hi ostripper,
But then again, maybe not. In real life the CCS may actually be noisier than the resistor (I doubt it, but could happen), depending on the resistance of your shunt regulator. If it goes up with frequency (common for IC's), then your PSRR might go down the tubes at higher frequencies. With a regulated supply for the voltage amp stage, I don't expect to see much in the way of noise or ripple on the rails. Possibly some audio though.
How much current do you have running through your 220K (R48) resistor? I am just trying to make sure that you have enough current flowing so that base current draw from Q20 and Q21 doesn't cause too much variation. Those devices normally have a beta of around 68 ~ 80, don't they?
In the event that someone (probably many) do not go for the separate voltage amp regulated supply, can you simulate this with and without the CCS for R48. I just love little red LEDs glowing away in the chassis! Just put an AC voltage source in each rail of your sim, on the output transistor side of R44 and R45. Make it a sawtooth of a couple volts. I am being nice about the amount of ripple to expect under heavier loads. For those who decide to go for a voltage amp reg supply, just make it so the CCS can be replaced with a resistor if desired.
Sorry for putting you to work this way ostripper. The answer is of value for any design, so while you are at it ....
Thank you, Chris
Edit: When I tried the On Semi MJW0281A/MJW0302A, I was blown away by their performance. One important characteristic is that they are tightly matched for beta and VBE. I was able to get lower THD in the original (rev 5.?) SymAsym, but at a lower bias current than everyone else was using. I set the bias using a THD meter and scope for the residual. Set using a 4R load, 10 KHz at about 1 Vrms. The THD followed a "valley" with increasing bias current. That mean that bias above the happy area (not critical) increased THD again. That was posted in the original thread. I would recommend using these transistors if possible. Getting them as "real parts" should be easier for as well. For two, yes there are higher rated parts, but you are using 4 pairs here, another pair wouldn't break the bank if really needed. The heat will be nicely spread out across the heat sink. Japanese driver transistors might be a good idea, I have more faith in those. I guess I haven't had a good chance to run the MJE parts through their paces yet.
Completely!
But then again, maybe not. In real life the CCS may actually be noisier than the resistor (I doubt it, but could happen), depending on the resistance of your shunt regulator. If it goes up with frequency (common for IC's), then your PSRR might go down the tubes at higher frequencies. With a regulated supply for the voltage amp stage, I don't expect to see much in the way of noise or ripple on the rails. Possibly some audio though.
How much current do you have running through your 220K (R48) resistor? I am just trying to make sure that you have enough current flowing so that base current draw from Q20 and Q21 doesn't cause too much variation. Those devices normally have a beta of around 68 ~ 80, don't they?
In the event that someone (probably many) do not go for the separate voltage amp regulated supply, can you simulate this with and without the CCS for R48. I just love little red LEDs glowing away in the chassis! Just put an AC voltage source in each rail of your sim, on the output transistor side of R44 and R45. Make it a sawtooth of a couple volts. I am being nice about the amount of ripple to expect under heavier loads. For those who decide to go for a voltage amp reg supply, just make it so the CCS can be replaced with a resistor if desired.
Sorry for putting you to work this way ostripper. The answer is of value for any design, so while you are at it ....
Thank you, Chris
Edit: When I tried the On Semi MJW0281A/MJW0302A, I was blown away by their performance. One important characteristic is that they are tightly matched for beta and VBE. I was able to get lower THD in the original (rev 5.?) SymAsym, but at a lower bias current than everyone else was using. I set the bias using a THD meter and scope for the residual. Set using a 4R load, 10 KHz at about 1 Vrms. The THD followed a "valley" with increasing bias current. That mean that bias above the happy area (not critical) increased THD again. That was posted in the original thread. I would recommend using these transistors if possible. Getting them as "real parts" should be easier for as well. For two, yes there are higher rated parts, but you are using 4 pairs here, another pair wouldn't break the bank if really needed. The heat will be nicely spread out across the heat sink. Japanese driver transistors might be a good idea, I have more faith in those. I guess I haven't had a good chance to run the MJE parts through their paces yet.
68 - 80 ???
21st century man ! more like 400 ! the U gain group is 600+ , I use the "F" 's . My .3ma of cascode current would have no problem in the world doing the 3.2 microamps drawn by the bases of the KSC's (real low , huh).Attached is the PSRR plot , both rails with a 1V out of phase swept sine (I made it so the phase changed with the sweep). also tried it with fixed square "filth" on the rails. 85 - 88 DB was the result over
most of the audio range. This is absolutely confirmed in real life ... NO crosstalk , at least not audible.
Glad you like the MJW/NJW's . For $1.40 ea .. WOW. 100 beta with 20% more SOA than a 2sc5200. 5 pair(1.6A @70V SOA) could match 4 pair of MJL21193's (2A @70V SOA)
OS
Attachments
Hi Holger,
I am running this amp on 52V rails. Confirmed by others, the sonics are really fine compared to an older Phase Lineair amp, a YamahaXM4080 and a P.L. Aleph4. But ofcourse as always it is a matter of taste but this amplifier is performing really well.
Especially under load conditions this thing is performing pretty good.
Some real life measurement THD results:
load: 7,5 Ohm
1kHz 0,0088@25W/ 0,0121@75W
10kHz 0,0148@25W/ 0,0158@75W
20kHz 0,0306@25W/ 0,0332@75W
Equipment:
Tektronix SG505/AA501
I am running this amp on 52V rails. Confirmed by others, the sonics are really fine compared to an older Phase Lineair amp, a YamahaXM4080 and a P.L. Aleph4. But ofcourse as always it is a matter of taste but this amplifier is performing really well.
Especially under load conditions this thing is performing pretty good.
Some real life measurement THD results:
load: 7,5 Ohm
1kHz 0,0088@25W/ 0,0121@75W
10kHz 0,0148@25W/ 0,0158@75W
20kHz 0,0306@25W/ 0,0332@75W
Equipment:
Tektronix SG505/AA501
In real life , with 77v rails , I have not even approached clipping. On
the "Guinea pig" (real one) , I even have the pads there to add diodes. While letting the rails collapse (turn off) , I have ACTUALLY heard it distort , while this did not sound good , nothing blew
.What I have seen is the KSA3503's "sticking" at saturation when grossly overloaded.
OS
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.