Inspired by the Zen series of A class amplifiers designed by Nelson Pass, LTPZen substitutes the source follower of Zen Variations 4 for a long tailed pair. LTPZen is an amplifier with a bit more open loop gain, a bit more feedback, an extended frequency response and reduced distortion. At first sight, the circuit looks more complicated, but the component count is similar and it is still only a two stage design. The additional complexities lie outside the signal paths. A balanced supply eliminates the loudspeaker decoupling capacitors and the number of remaining signal path capacitors have been reduced to a minimum.
For further details visit www.pmeweb.co.uk/audio
Details are available as a pdf download LTPZen.pdf
Anyone else tried anything like this?
An externally hosted image should be here but it was not working when we last tested it.
For further details visit www.pmeweb.co.uk/audio
Details are available as a pdf download LTPZen.pdf
Anyone else tried anything like this?
nice:
I like variations you made ,especially with current sources here and there;
regarding output CCS modulation,I always think that Papa's way of modulating it from output sense have also something with back EMF and all this stuff............
I already talked several months ago with my handy SS consigliero Oly about that.......and seems that he take my ramblings (about EMF) seriously......
Papa,care to put one liner in answer?
I like variations you made ,especially with current sources here and there;
regarding output CCS modulation,I always think that Papa's way of modulating it from output sense have also something with back EMF and all this stuff............
I already talked several months ago with my handy SS consigliero Oly about that.......and seems that he take my ramblings (about EMF) seriously......
Papa,care to put one liner in answer?
Actually I missed the fact that the current source was constant
on the schematic.
I think the effect of modulating the current source (using output
current to the load as the factor) is not particularly aimed at
back emf, but is intended to reduce the work of the Common-
Source gain stage. To this end it reduces distortion and/or the
bias current requirements in general.
on the schematic.
I think the effect of modulating the current source (using output
current to the load as the factor) is not particularly aimed at
back emf, but is intended to reduce the work of the Common-
Source gain stage. To this end it reduces distortion and/or the
bias current requirements in general.
Thanks to everyone for their comments.
The improvement in the distortion figures for the LTPZen circuit are partly due to the increased open loop gain and partly due to the to configuration of the Aleph current source. With LTPZen, you can choose either output current CCS modulation à la Zen V4, or input voltage modulation as shown.
Having tried both methods with LTPZen, input modulation roughly halves the distortion. This is because the output current will be effected by non linearities of the output transistors, so with output modulation, the signal feeding the CCS transistor is more distorted than the signal feeding it’s driver.
On back EMF? With a theoretical, purely inductive voice coil load Lc, the voltage across the speaker, Vc= Lc.∂Ic/∂t. For a sinusoidal waveform the current and voltage are out of phase. With a real loudspeaker, this I-V phase shift will be frequency dependent.
Zen V4 and LTPZen are both voltage amplifiers and use overall voltage feedback. However, LTPZen uses negative voltage feedback and voltage control for the output stage; the phase of the output current is irrelevant.
Zen V4 uses the output voltage in it’s overall negative feedback arrangement but senses the output current in the CCS. The fact that the output current will not be in phase with the output voltage should not matter. The LRC components of the loudspeaker load are still a linear system and the distortion is still caused by non linearities in the transistors. However, In Zen V4 you do not have a suitable CSS drive signal with the correct phase so input CSS control is not possible.
With LTPZen, input CSS control is possible and has been adopted accordingly.
The improvement in the distortion figures for the LTPZen circuit are partly due to the increased open loop gain and partly due to the to configuration of the Aleph current source. With LTPZen, you can choose either output current CCS modulation à la Zen V4, or input voltage modulation as shown.
Having tried both methods with LTPZen, input modulation roughly halves the distortion. This is because the output current will be effected by non linearities of the output transistors, so with output modulation, the signal feeding the CCS transistor is more distorted than the signal feeding it’s driver.
On back EMF? With a theoretical, purely inductive voice coil load Lc, the voltage across the speaker, Vc= Lc.∂Ic/∂t. For a sinusoidal waveform the current and voltage are out of phase. With a real loudspeaker, this I-V phase shift will be frequency dependent.
Zen V4 and LTPZen are both voltage amplifiers and use overall voltage feedback. However, LTPZen uses negative voltage feedback and voltage control for the output stage; the phase of the output current is irrelevant.
Zen V4 uses the output voltage in it’s overall negative feedback arrangement but senses the output current in the CCS. The fact that the output current will not be in phase with the output voltage should not matter. The LRC components of the loudspeaker load are still a linear system and the distortion is still caused by non linearities in the transistors. However, In Zen V4 you do not have a suitable CSS drive signal with the correct phase so input CSS control is not possible.
With LTPZen, input CSS control is possible and has been adopted accordingly.
Thanks fig.
I think a jfet input stage is a fine idea. With no signal and the volume fully up, you can just about hear a small amount of background noise. It is detectable on a 'scope too.
With the existing mosfets you can vary the current over quite a range. The gain - bandwidth product is highest with a current of ~ 30mA and tail / current source resistors of 270 ohms. However, I have opted for a slightly lower open loop gain and push the long tailed pair quite hard, at around 300mA.
Can you suggest a suitable part?
I think a jfet input stage is a fine idea. With no signal and the volume fully up, you can just about hear a small amount of background noise. It is detectable on a 'scope too.
With the existing mosfets you can vary the current over quite a range. The gain - bandwidth product is highest with a current of ~ 30mA and tail / current source resistors of 270 ohms. However, I have opted for a slightly lower open loop gain and push the long tailed pair quite hard, at around 300mA.
Can you suggest a suitable part?
Great work, congrats on a inspiring article! Going to print it out and will diserve a place among the Zen articles....
Personally I found the outtro (the table where THD, FB, OpenL Gain en are compared) of Nelson's PLH article an eye opener! Could you extend your artcile with a similar comparison?
Personally I found the outtro (the table where THD, FB, OpenL Gain en are compared) of Nelson's PLH article an eye opener! Could you extend your artcile with a similar comparison?
Dank u wel tschrama, Nelson!,
It would be interesting to compare Zen V4, LTPZen with PLH. However, you yould need to make a comparison with with like for like components and matched gain and there is quite a lot of work involved here. As Nelson says, we don't want to get too serious! There are some other designs, on speaker protection, power supplies, water cooling etc, which I would like to publish first.
Anyone come up with a good jfet part for the long tailed pair yet?
It would be interesting to compare Zen V4, LTPZen with PLH. However, you yould need to make a comparison with with like for like components and matched gain and there is quite a lot of work involved here. As Nelson says, we don't want to get too serious! There are some other designs, on speaker protection, power supplies, water cooling etc, which I would like to publish first.
Anyone come up with a good jfet part for the long tailed pair yet?
Well, the 'ol standby 2sj109 were avialable at BK in Pa. USA a few day's ago??? I don't know about those high currents in the front end though. 5mA should just about do well...
Allen, you suggested your output transistor choice was lower C. I don't see that in the datasheet? Might an IRFP044N do better?
Also, without the cap to ground under R10 you might see more offset than with it. I don't know, maybe your R5 and associated comps can do the job very well?
Can you tell me how you determine the R11 and R16 values's? Asside from trying to match the diff pair loading, I beleive that will be key to low output distortion design...
And, Can you actually get as good positive voltage swing on the output (T8 & T9 Driving T7) campared to the boostaraped design of the Zen's?
Allen, you suggested your output transistor choice was lower C. I don't see that in the datasheet? Might an IRFP044N do better?
Also, without the cap to ground under R10 you might see more offset than with it. I don't know, maybe your R5 and associated comps can do the job very well?
Can you tell me how you determine the R11 and R16 values's? Asside from trying to match the diff pair loading, I beleive that will be key to low output distortion design...
And, Can you actually get as good positive voltage swing on the output (T8 & T9 Driving T7) campared to the boostaraped design of the Zen's?
Thanks again fig. This peer review business keeps us on our toes!
You are quite right about the higher gate capacitance of IRFP3710: 3000pf vs 2400pf for IRFP044. In choosing a higher voltage part, I remember comparing Qg and Qgd and concluded that the higher gate capacitance would probably be offset by the lower capacacitance associated with the drain-body junction. This will be more diffuse to give a higher breakdown voltage. My comparison was simply made on the basis of the -3dB points, and I neglected to compare the frequency response plots properly. As you will see from the attached image. In substituting the parts I neglected to adjust the high frequency compensation capacitor, (the two blue plots). If you don't intend raising the voltages then IRFP044 is better!
On DC stability, You could put in a 10µF film capacitor in series with R10, but I found that it wasn't really necessary and overall negative feedback can be directly coupled. I guess it is a question of the amount of thermally induced drift you can accept. My system is water cooled which restricts the temperature rise to <30°C and the output stabilises fairly quickly.
I used a BC639 in the CCS. The value of R16 depends upon the hfe of the part used. Going back to the original Zen V4 configuration with output modulation; BC639 has a slightly higher gain than ZTX450, so I had to reduce the transistor drive current by raising the value of R16 from 1.5k to 1.8K. Using the two parallel 0.47 ohm sense resistors into an 8 ohm load, the output sensing arrangement senses 0.235/(8+0.235) ~ 1/35th of the output voltage. With LTPZen you are sensing all of the potential across tail resistor R4, so the feed resistor R11 will need to be increased by a factor of 35. Hence 62k. If you are using a ZTX450 in the CSS, R11 will need to about 52K. The optimum value will depend on hfe, so measure it first.
The led based current source T8 does make a difference to output swing. The voltage across T9 is clamped to 0.66 +VgsT7 ~ 4.7V. This restricts the positive output swing to about 15 Volts. The original bootstrap arrangements connects the output to the mid point of a potential divider so you can get a bit closer to the positive rail. I wanted to get the very low lLF performance as good as I could. You could use a higher or seperate positive supply voltage for T8, but that is maybe getting to complicated. It comes down to personal preference.
You are quite right about the higher gate capacitance of IRFP3710: 3000pf vs 2400pf for IRFP044. In choosing a higher voltage part, I remember comparing Qg and Qgd and concluded that the higher gate capacitance would probably be offset by the lower capacacitance associated with the drain-body junction. This will be more diffuse to give a higher breakdown voltage. My comparison was simply made on the basis of the -3dB points, and I neglected to compare the frequency response plots properly. As you will see from the attached image. In substituting the parts I neglected to adjust the high frequency compensation capacitor, (the two blue plots). If you don't intend raising the voltages then IRFP044 is better!
An externally hosted image should be here but it was not working when we last tested it.
On DC stability, You could put in a 10µF film capacitor in series with R10, but I found that it wasn't really necessary and overall negative feedback can be directly coupled. I guess it is a question of the amount of thermally induced drift you can accept. My system is water cooled which restricts the temperature rise to <30°C and the output stabilises fairly quickly.
I used a BC639 in the CCS. The value of R16 depends upon the hfe of the part used. Going back to the original Zen V4 configuration with output modulation; BC639 has a slightly higher gain than ZTX450, so I had to reduce the transistor drive current by raising the value of R16 from 1.5k to 1.8K. Using the two parallel 0.47 ohm sense resistors into an 8 ohm load, the output sensing arrangement senses 0.235/(8+0.235) ~ 1/35th of the output voltage. With LTPZen you are sensing all of the potential across tail resistor R4, so the feed resistor R11 will need to be increased by a factor of 35. Hence 62k. If you are using a ZTX450 in the CSS, R11 will need to about 52K. The optimum value will depend on hfe, so measure it first.
The led based current source T8 does make a difference to output swing. The voltage across T9 is clamped to 0.66 +VgsT7 ~ 4.7V. This restricts the positive output swing to about 15 Volts. The original bootstrap arrangements connects the output to the mid point of a potential divider so you can get a bit closer to the positive rail. I wanted to get the very low lLF performance as good as I could. You could use a higher or seperate positive supply voltage for T8, but that is maybe getting to complicated. It comes down to personal preference.
Allen, thanks. I mostly thought that R16's value might be a killer because it looks like 180 on the schematic??? If you say 1.5-1.8K, yes, that is much more in the range I would expect it needed to be.
The IRFP044(N) part has even better C (and charge) numbers, while still maintaining high gm... Smaller die I beleive and less thermal capabilities though. But, with your water cooling, that could work well...
Good work
The IRFP044(N) part has even better C (and charge) numbers, while still maintaining high gm... Smaller die I beleive and less thermal capabilities though. But, with your water cooling, that could work well...
Good work
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.