Has anybody measured Q5 Vce when schematic from post #1 is used and R1 has about 9V voltage drop?
ZTX851 has if Ic is below 0.1A typical 10mV Vce (sat) (max. 50 mv). So if Vce is like 100mV it is not saturated?
A shame I did not measure that.......just for interest. I could open up the box to measure it.
ZTX851 has if Ic is below 0.1A typical 10mV Vce (sat) (max. 50 mv). So if Vce is like 100mV it is not saturated?
A shame I did not measure that.......just for interest. I could open up the box to measure it.
Saturation is really a whole region of low Vce where the beta collapses. See attached curves** which show Ic (Y axis) vs. Vce (X axis) as Ib is stepped for each trace. In this low-Vce region, the slope starts to deviate from the flat lines at higher Vce, where the slope is dominated by the Early voltage. The data sheet "saturation voltage" spec is just one data point from within this region, usually selected as a standardized point of reference for easy comparison between transistor types. Each transistor's saturation curves look a little different, but generally speaking Vce below ~0.5V-ish is in the saturation region... Some have a sharper 'shoulder' than others. Some have a 'quasi-saturation' zone that gives an interesting two-slope saturation characteristic (e.g. 2SC2240).
Anyway, if R1 has more than roughly 7.5V across it, then Q5 collector is going to start dipping into the saturation region. As it does so, beta will fall and base current will rise, pulling the base voltage down. Exactly where a given part will land on the steep part of that curve will vary a lot with production variance, temperature and so on.
Best practice for a cascode circuit is to keep the Vce above the saturation zone with some margin to guard band for effects like thermal drift and aging causing shifts in the JFET drain current, etc. In this circuit, Vce will generally be low, so a part like BC550 or 2N4401 would be a better choice than 2SC2240. In the non-saturation region, the Q5 base and emitter voltages will be largely insensitive to the voltage drop across R1.
Alas, I don't have curves for the ZTX851 at hand. I would guess they are pretty close to the 2N4401. Maybe Mark has them, or I can trace a part later and post it if anyone's interested.
** Credit to @Mark Johnson for posting these elsewhere on the forum. I encourage everyone to further indulge their curves fetish here.
Anyway, if R1 has more than roughly 7.5V across it, then Q5 collector is going to start dipping into the saturation region. As it does so, beta will fall and base current will rise, pulling the base voltage down. Exactly where a given part will land on the steep part of that curve will vary a lot with production variance, temperature and so on.
Best practice for a cascode circuit is to keep the Vce above the saturation zone with some margin to guard band for effects like thermal drift and aging causing shifts in the JFET drain current, etc. In this circuit, Vce will generally be low, so a part like BC550 or 2N4401 would be a better choice than 2SC2240. In the non-saturation region, the Q5 base and emitter voltages will be largely insensitive to the voltage drop across R1.
Alas, I don't have curves for the ZTX851 at hand. I would guess they are pretty close to the 2N4401. Maybe Mark has them, or I can trace a part later and post it if anyone's interested.
** Credit to @Mark Johnson for posting these elsewhere on the forum. I encourage everyone to further indulge their curves fetish here.
Attachments
Last edited:
Q5 should not act as a switch,
so we must ensure that its collector potential is greater than its base potential. If its "operating point" lies nicely on one of the horizontal branches of the characteristic curve ICE(VCE), then the cascode works - the decisive parameter here is IBE, or VBE, static. Ultimately, we are interested in a difference, a delta, i.e. dynamic. The jFets imprint the current flow, current imprinting. They determine the voltage VR1, among other things, via their preset -VGS. If we now set the base potential to Vcc/2 using a fairly low-impedance voltage divider R2 & R3 and determine that Vcc minus VR1 is < than VBQ5 (=Vcc/2), then we have ensured that Q5 can also be replaced by a wire bridge between C and E.
so we must ensure that its collector potential is greater than its base potential. If its "operating point" lies nicely on one of the horizontal branches of the characteristic curve ICE(VCE), then the cascode works - the decisive parameter here is IBE, or VBE, static. Ultimately, we are interested in a difference, a delta, i.e. dynamic. The jFets imprint the current flow, current imprinting. They determine the voltage VR1, among other things, via their preset -VGS. If we now set the base potential to Vcc/2 using a fairly low-impedance voltage divider R2 & R3 and determine that Vcc minus VR1 is < than VBQ5 (=Vcc/2), then we have ensured that Q5 can also be replaced by a wire bridge between C and E.
Last edited:
Sure enough, Mark has posted ZTX851 curves:
https://www.diyaudio.com/community/...-saturation-measured-data.293308/post-4756366
https://www.diyaudio.com/community/...-saturation-measured-data.293308/post-4756366
In Wayne's first quick test setup for the burning amplifier Fest 23, there are four sk170BL boys injecting their total current into resistor R1 (statically we consider it constant). Wayne has confirmed a voltage drop across R1=750Ohms of greater than 9V, i.e. the collector potential of Q5 is 5.3V, while our voltage divider formed by R2&R3 tries to keep the base potential at 7.15V. VC(Q5) is thus considerably smaller than VB(Q5) -> of an NPN transistor, a cascode operation is no longer present.
I know that, and here we can read about 200mV ... the crucial question (of this very special transistor) is: does the base voltage divider potential also drop sufficiently?
Our first intention is a constant drain potential of Q1_4 !
Of course, why not? There's absolutely nothing wrong with the layout.
Of course, why not? There's absolutely nothing wrong with the layout.
wrong!
The current flow is determined (and set) in the first instance by Q1,2,3,4 and their ohmic source resistances (R4,5,6,7 & R8).
Plus the AC voltage that will appear collecter-emitter in operation ?
the delta, yes "Auricle"
But why don't you ask Wayne Colburn directly how all this is connected and how it works, he will certainly be happy to explain it in a generally understandable way. I suspect.
Kind regards,
HBt.
But why don't you ask Wayne Colburn directly how all this is connected and how it works, he will certainly be happy to explain it in a generally understandable way. I suspect.
Kind regards,
HBt.
all you need is to ask Wayne politely to edit schematic in post #1, to have clearly marked prescribed Iq for battery of input JFets
quite possible that said info is repeated several times through thread, but this way it'll be there, for good, in easy way
quite possible that said info is repeated several times through thread, but this way it'll be there, for good, in easy way
I just did a little opamp rolling with surprising results. This was just one channel, completely in the open, so I was not going to concentrate on hum. Cartridge was a Denon DL-110, loaded at 47k. P3 set to 49 dB gain.
OPA1656: LOTS of noise, both high and low frequency. It did play music, but this noise level is uncacceptable.
OPA2134: LOTS of noise, both high and low frequency. Bit less than OPA1656 to my ear. Also unusable.
OPA627 (dual): low noise, nice sound, but one of the two had some low level pulsing noise, didn't notice that with the other one. Will try these again when the amp is built into an enclosure and properly grounded.
LM4562: pulsing noise, started making an audible frequency sweep (like some of those test signals). Did not seem really stable.
NJM2068: best overall. Low noise (but not as low as the OPA627), no obvious audible signs of instability. But remarkable: a lot more gain than the others. Against the OPA627 the music was audibly louder. It would not surprise me if the difference turned out to be 6dB or so.
Remark: the smd capacitors around the opamp in my P3 are 0.12 uF (as the data sheet of several opamps like the OPA1612 that I figured as my go-to device mention 0.1 uF). 220 uF electrolytics are also in (Panasonic FM).
Some Brown Dog adapters are on their way for the OPA1612 and 2210 that I also have.
Could the Denon cartridge be a source of trouble (Wayne mentioned some HOMC's were causing oscillation, that's why the 1k+1nF loading option is in)? Did not test this myself yet as those positions are open "user" positions on my board.
The default preamp for the turntable I used for this test is a DIY Bugle (mostly Bugle MC) with stock LM4562's, linear power supply and dual regulation (7x15's followed by a low noise +/-12V regulator by LDOVR). Compared to that one (set at 50dB gain) the Pearl 3 is somewhat noisier (hiss). I'm using the PSU of the Bugle for this test of the P3, feeding a super reg.
OPA1656: LOTS of noise, both high and low frequency. It did play music, but this noise level is uncacceptable.
OPA2134: LOTS of noise, both high and low frequency. Bit less than OPA1656 to my ear. Also unusable.
OPA627 (dual): low noise, nice sound, but one of the two had some low level pulsing noise, didn't notice that with the other one. Will try these again when the amp is built into an enclosure and properly grounded.
LM4562: pulsing noise, started making an audible frequency sweep (like some of those test signals). Did not seem really stable.
NJM2068: best overall. Low noise (but not as low as the OPA627), no obvious audible signs of instability. But remarkable: a lot more gain than the others. Against the OPA627 the music was audibly louder. It would not surprise me if the difference turned out to be 6dB or so.
Remark: the smd capacitors around the opamp in my P3 are 0.12 uF (as the data sheet of several opamps like the OPA1612 that I figured as my go-to device mention 0.1 uF). 220 uF electrolytics are also in (Panasonic FM).
Some Brown Dog adapters are on their way for the OPA1612 and 2210 that I also have.
Could the Denon cartridge be a source of trouble (Wayne mentioned some HOMC's were causing oscillation, that's why the 1k+1nF loading option is in)? Did not test this myself yet as those positions are open "user" positions on my board.
The default preamp for the turntable I used for this test is a DIY Bugle (mostly Bugle MC) with stock LM4562's, linear power supply and dual regulation (7x15's followed by a low noise +/-12V regulator by LDOVR). Compared to that one (set at 50dB gain) the Pearl 3 is somewhat noisier (hiss). I'm using the PSU of the Bugle for this test of the P3, feeding a super reg.
My quick and dirty voltages. Both channels have similar voltages, which probably isn't surprising, since I matched everything in the circuit
From a post I cannot find in this thread, I realized I saved a schematic with someone else's voltages that had Q5 voltages of Base 6.9V, emitter 6.3V, and collector 8.5V, but they only had 2 mA flowing through one of the Q1-4 source resistors to my 3 mA.
I welcome your thoughts.
From a post I cannot find in this thread, I realized I saved a schematic with someone else's voltages that had Q5 voltages of Base 6.9V, emitter 6.3V, and collector 8.5V, but they only had 2 mA flowing through one of the Q1-4 source resistors to my 3 mA.
I welcome your thoughts.
Seems like the cascode is saturated with Vce only 0.03V? I would try the trick with soldering 2k parallell to R1. I bet Vce wille be much better then. My 9.3 mA Idss parts work fine (as far as the voltages are concerned) with R1 = 383R.
Instead of blaming the opamps, remove the cartridge and substitute a 150 ohm resistor (about the same impedance of the DL-110 but purely resistive) and re-run the tests.
If you have a scope, you would use a long sweep time 10s to examine the output in the same manner that Ti, ADI and others use. It would be good to have some data to support your assertion.
I found some 2.2k, which would give 559 ohms. It sounds like I should try those and if that does not work, 1k (=428 ohms) and 750R (=375). I won’t be able to do that until tomorrow, but I will report back. Seems better than pulling an input FET or redoing the whole front end. Thanks for the tip. So I want at least 1 volt Vce, or would more be better?