Bonjour Yves, c'est un plaisir de te lire!
Yes, I have to say that your design really impressed me for its simplicity and efficacy, and it's what I basically applied to a PP guitar amp to see how it reacted.
I've read the evolution of the project from your very smart and elegant two-valve-only amp, to the actual design. The UL has been integrated to reduce the THD of the EL84s, but with Hammond trafos with 43% UL taps, the ratio is not the right one (see here: UL for EL84. 22-24% is better than the dogma....40% is lowest distortion.... ).
Then it has been modified the CCS on the tail of the PI, to increase its AC impedance and increase the balance of the signal. Also, from your 1 mA, notes tend to prefer 1,2 mA and no more than 1,4 mA (700 uA per side).
Then it has been applied the trick to reduce 3rd harmonic content: EL84 Amp - Baby Huey
Then the output tubes have been switched from cathode bias to fixed bias to improve the recovery when overloaded.
Then powerdrive has been applied, with a simple resistor on the mosfet.
Final developement has been to increase the driving capability of the powerdrive by applying a CCS instead of the simple resistor to load the mosfet of the Powerdrive. The current flowing through it determines the bandwidth of the Powerdrive, that's because the R33 and R38 need to be lower for the EL34 version: the input capacity is almost the same (11 vs 15 pF), but it's needed to swing almost three times to get full power from the latter.
This is the status of the art as far as I know.
To better explain how the current of the powerdrive influences its bandwidth:
To ensure the right slew rate for the Powerdrive:
I need around 25 Vpp up to ten times the bandwidth, so 200 kHz, so slewrate will be:
2 x pi x f x V = 2 x 3,14 x 200 kHz x 25 = 31,4 V/us
One EL84 has an input capacitance of 11 pF (let's say 15 considering wiring?).
C = i x dt / V
dV/dt = 31,4 V/us
then
15 pF = i / slewrate
so
i = 15 pF x 31,4 V/us = 470 uA
So it could be enough to have 1.5 kOhm on the CCS.
I will keep 390 Ohm and 1,7 mA.
Should be correct, but again please feel free to correct any error.
I would also like to link an interesting thread about output transformer comparison for this amp:
Red Light District/Baby Huey (EL84 P-P) Output Iron
To ensure the right slew rate for the Powerdrive:
I need around 25 Vpp up to ten times the bandwidth, so 200 kHz, so slewrate will be:
2 x pi x f x V = 2 x 3,14 x 200 kHz x 25 = 31,4 V/us
One EL84 has an input capacitance of 11 pF (let's say 15 considering wiring?).
C = i x dt / V
dV/dt = 31,4 V/us
then
15 pF = i / slewrate
so
i = 15 pF x 31,4 V/us = 470 uA
So it could be enough to have 1.5 kOhm on the CCS.
I will keep 390 Ohm and 1,7 mA.
Should be correct, but again please feel free to correct any error.
I would also like to link an interesting thread about output transformer comparison for this amp:
Red Light District/Baby Huey (EL84 P-P) Output Iron
I quote myself because I've found this Mullard document ( http://primarywindings.com/wp-content/uploads/2017/04/Mullard-Circuits-for-Audio-Amplifiers.pdf ) that at pages 22-23 says:
This seems not being confirmed by tests done by gingertube on the two similar BH84 amps, as I remember him saying that fixed bias performed better around overdrive, but most probably that was not of interest in the tests performed by Mullard.
If you see the table at page 23, it is interesting to see that the suggested impedance when using 20% taps is lower than in triode and also lower than in pentode, not something in between as I would have expected. So with 20% taps the suggested Raa is 6.6 kOhm instead of the most available 8 kOhm, but differences should be minor.
So... how 43% became a standard for transformer builders?
I'm soldering components on the six pcbs and I would underline few things.
R16 and R17: both on EL84 version (1.6 kOhm and 1 kOhm respectively) and in the EL34 version (15 kOhm and 10 kOhm respectively), R16 is around 1.5 times R17.
Why?
From gingertube's guidelines, led current must be between 2 and 5 mA.
So being the voltage linked to the powerdrive and the PI's CSS the same and around 3 times the bias voltage, we'll have:
Around -36 Vdc on the led, so around -34.3 Vdc to be dropped by R16 and R17.
So for 2 mA it would be needed to have 10 kOhm and 6.8 kOhm.
While for 5 mA it would be needed to have and 3.9 kOhm and 2.7 kOhm.
I would go in between at 3.5 mA with 5.6 kOhm and 3.9 kOhm.
I've noticed some differences from the BOM, the pcb, and what I think should be correct. I will collect them and report, by now I noticed:
- R3 and R4 should be 10 kOhm instead of 1 kOhm as shown
- R11 and R12 are they really 4.7 kOhm? Is there any AB2 with this high value?
- R16 and R17 as shown before
- R18 would be around 1 kOhm instead of 680 Ohm for EL84
- R20 and R21 that are shown as 470 kOhm but they should be 1 MOhm
About components to be mounted on the back side of the pcb:
- C5, C6, C7 (to keep them far from how tubes and mount the pcb closer to chassis)
- C10, C11 (to mount the pcb closer to chassis)
- JP1 (to be able to modify the heaters from 12AX7 to russians equivalents without dismounting the pcbs)
- R5, R41, R42 (to be able to modify the bias and balance and check it from bottom, as doing it from top would bring the high risk of shorting checkpoints to chassis)
Please correct if there's anything wrong in what I wrote, or if anyone has done things differently.
Thanks
R16 and R17: both on EL84 version (1.6 kOhm and 1 kOhm respectively) and in the EL34 version (15 kOhm and 10 kOhm respectively), R16 is around 1.5 times R17.
Why?
From gingertube's guidelines, led current must be between 2 and 5 mA.
So being the voltage linked to the powerdrive and the PI's CSS the same and around 3 times the bias voltage, we'll have:
Around -36 Vdc on the led, so around -34.3 Vdc to be dropped by R16 and R17.
So for 2 mA it would be needed to have 10 kOhm and 6.8 kOhm.
While for 5 mA it would be needed to have and 3.9 kOhm and 2.7 kOhm.
I would go in between at 3.5 mA with 5.6 kOhm and 3.9 kOhm.
I've noticed some differences from the BOM, the pcb, and what I think should be correct. I will collect them and report, by now I noticed:
- R3 and R4 should be 10 kOhm instead of 1 kOhm as shown
- R11 and R12 are they really 4.7 kOhm? Is there any AB2 with this high value?
- R16 and R17 as shown before
- R18 would be around 1 kOhm instead of 680 Ohm for EL84
- R20 and R21 that are shown as 470 kOhm but they should be 1 MOhm
About components to be mounted on the back side of the pcb:
- C5, C6, C7 (to keep them far from how tubes and mount the pcb closer to chassis)
- C10, C11 (to mount the pcb closer to chassis)
- JP1 (to be able to modify the heaters from 12AX7 to russians equivalents without dismounting the pcbs)
- R5, R41, R42 (to be able to modify the bias and balance and check it from bottom, as doing it from top would bring the high risk of shorting checkpoints to chassis)
Please correct if there's anything wrong in what I wrote, or if anyone has done things differently.
Thanks
C1, C2 and C5 I've used 100 nF instead of 220, because with a load of 1MOhm the -3dB point is at 1,6 Hz, and because I already had alot of Vishay Ero 1813 around for my guitar amps.
Another point is R7 and R8. In the BOM they are 600 mW 220 kOhm resistors, but in my experience with guitar amps, plate resistors improve the dynamic of the stage if they are 1 to 2 W instead of smaller ones. Has anyone experience on that?
Off course I noticed that after having soldered all twelve...
Another point is R7 and R8. In the BOM they are 600 mW 220 kOhm resistors, but in my experience with guitar amps, plate resistors improve the dynamic of the stage if they are 1 to 2 W instead of smaller ones. Has anyone experience on that?
Off course I noticed that after having soldered all twelve...
I selected four of them being all 217 kOhm, then soldered in one amp:
Tube-Town Store - Resistor Metaloxide 2 Watts / 220 kOhms
instead of the 600 mW ones, but it is needed to pull them in from the back of the pcb, because they are just same lenght of the holes.
I tried RN65Ds, but they are too long and cannot fit.
I will report the difference in future.
As for the CCS' red led dropping resistors, I went for 5.6 kOhm and 10 kOhm. This makes just something more than 2 mA.
Tube-Town Store - Resistor Metaloxide 2 Watts / 220 kOhms
instead of the 600 mW ones, but it is needed to pull them in from the back of the pcb, because they are just same lenght of the holes.
I tried RN65Ds, but they are too long and cannot fit.
I will report the difference in future.
As for the CCS' red led dropping resistors, I went for 5.6 kOhm and 10 kOhm. This makes just something more than 2 mA.
May I ask if this kind rheostat is used in Hi-Fi too?
Tube-Town Store - Rheostat Alpha 100 Ohm / 5 Watt
In instrument amplification is sometimes used to reduce the hum due to heaters.
I usually keep the heaters in AC and elevate them around 40Vdc.
This solution usually costs less than a central tap for heaters when you ask custom transformers. Two 100 Ohm resistors costs of course even less.
Thanks for your feedback
Tube-Town Store - Rheostat Alpha 100 Ohm / 5 Watt
In instrument amplification is sometimes used to reduce the hum due to heaters.
I usually keep the heaters in AC and elevate them around 40Vdc.
This solution usually costs less than a central tap for heaters when you ask custom transformers. Two 100 Ohm resistors costs of course even less.
Thanks for your feedback
Hey all, going to build a BH EL84 once the next GB by Prasi is shipped.
Been following this thread (and the Wiki) and all the comments by Bas, Francois, Zintolo, Mark and Ian have been enormous. Really appreciated.
I know OPT's have been discussed ad nauseam but can't help but ask again due to a particular circumstance.
First, I'm wrapping up a Tubelab SPP with EL84's (I'll be comparing the BH to the SPP if I get around to
it) and for that one I used all Hammond iron. 270HX, 1650FA and 193H choke.
For BH I want to try something else.
I happen to have a credit at Sowter due to a MC SUT order that I cancelled, so was thinking to use it towards OPT's.
The U064 is their cheapest, but am concerned the rating is on the low-side at 10W, 80mA DC per side...
They offer 20% or 43% UL and still not convinced which is preferred.
So basically looking for a confirmation here if I can confidently go with U064.
To be fair, I think they are a bit underrated.
The 1608 weighs 2.5lbs while the Sowter weighs 3.3 lbs.
Any thoughts here appreciated.
Been following this thread (and the Wiki) and all the comments by Bas, Francois, Zintolo, Mark and Ian have been enormous. Really appreciated.
I know OPT's have been discussed ad nauseam but can't help but ask again due to a particular circumstance.
First, I'm wrapping up a Tubelab SPP with EL84's (I'll be comparing the BH to the SPP if I get around to
it) and for that one I used all Hammond iron. 270HX, 1650FA and 193H choke.
For BH I want to try something else.
I happen to have a credit at Sowter due to a MC SUT order that I cancelled, so was thinking to use it towards OPT's.
The U064 is their cheapest, but am concerned the rating is on the low-side at 10W, 80mA DC per side...
They offer 20% or 43% UL and still not convinced which is preferred.
So basically looking for a confirmation here if I can confidently go with U064.
To be fair, I think they are a bit underrated.
The 1608 weighs 2.5lbs while the Sowter weighs 3.3 lbs.
Any thoughts here appreciated.
If it must be Sowter I would go with these, just to be safe: SOWTER TYPE U082
The reason why I went for 23% DL taps is that old Mullard says there's basically no need to exceed 20% like DL, plus can be squeezed more power, plus shunt feedback should linearize the response towards g1=0 (that is indeed more linear with 43% rather than 23%).
There will be less need to go into AB2 with 23% DL, but screen stoppers will need IMHO to be higher than with 43% DL, because the plates go lower in voltage, so the dV is higher with 23% DL, and so secondary emission, and screens need to be protected more.
Please note that I kept 8k Raa even with 23% DL, but datasheet suggests 6.6k in this case (it goes more on the right of the pentodes' knee, where curves are more equally spaced). Again, I went with 8k because of shunt feedback. Without it I would probably have gone with 6k6.
The reason why I went for 23% DL taps is that old Mullard says there's basically no need to exceed 20% like DL, plus can be squeezed more power, plus shunt feedback should linearize the response towards g1=0 (that is indeed more linear with 43% rather than 23%).
There will be less need to go into AB2 with 23% DL, but screen stoppers will need IMHO to be higher than with 43% DL, because the plates go lower in voltage, so the dV is higher with 23% DL, and so secondary emission, and screens need to be protected more.
Please note that I kept 8k Raa even with 23% DL, but datasheet suggests 6.6k in this case (it goes more on the right of the pentodes' knee, where curves are more equally spaced). Again, I went with 8k because of shunt feedback. Without it I would probably have gone with 6k6.
Thanks zintolo for the explanation.
Some of the concepts there are beyond my scope of knowledge right now.
The U082 is way out of my price range unfortunately.
I asked them about making an 8k, 15W for smaller price/size and looks like it should be good. Since it is custom, I have the choice of 25% or 43% UL taps.. I chose 25% but can change it if recommended otherwise.
Some of the concepts there are beyond my scope of knowledge right now.
The U082 is way out of my price range unfortunately.
I asked them about making an 8k, 15W for smaller price/size and looks like it should be good. Since it is custom, I have the choice of 25% or 43% UL taps.. I chose 25% but can change it if recommended otherwise.
Working on another project (an EL34 amp a friend of mine gave me and that you can find here: improvements on 12AX7 12AT7 EL34 schematic? ), I've had an idea that work very good on simulations (I'll try it during the weekend).
Basically the Baby Huey concept is to apply local feedback to the power tubes in order to linearize them. So far so good.
But then why not applying this concept to the full amp? The PI is actually out of any nfb loop, but there's an easy way to include it in a short loop: take the signal at the source of one mosfet driver and reinject at the nfb input of the phase inverter.
This way the nfb will work on the PI and the Driver only.
On that amp I've applied (actually simulated only) local feedback on PI&Driver + UL + cathode feedback + gnfb and I've got those results on 8 Ohms:
With all the limits of the models and simulations, distortion is very low and basically low order only.
I've not yet built my BH so I cannot test it, but if anyone would like to try, it would be great.
Basically the Baby Huey concept is to apply local feedback to the power tubes in order to linearize them. So far so good.
But then why not applying this concept to the full amp? The PI is actually out of any nfb loop, but there's an easy way to include it in a short loop: take the signal at the source of one mosfet driver and reinject at the nfb input of the phase inverter.
This way the nfb will work on the PI and the Driver only.
On that amp I've applied (actually simulated only) local feedback on PI&Driver + UL + cathode feedback + gnfb and I've got those results on 8 Ohms:
Code:
Harmonic Frequency Fourier Normalized Phase Normalized
Number [Hz] Component Component [degree] Phase [deg]
1 1.000e+03 3.966e+00 1.000e+00 -1.00° 0.00°
2 2.000e+03 2.826e-04 7.125e-05 129.46° 130.47°
3 3.000e+03 1.624e-04 4.096e-05 171.85° 172.85°
4 4.000e+03 9.218e-05 2.324e-05 -173.48° -172.48°
5 5.000e+03 7.195e-05 1.814e-05 179.97° 180.97°
6 6.000e+03 6.114e-05 1.542e-05 -177.77° -176.76°
7 7.000e+03 5.414e-05 1.365e-05 -179.90° -178.90°
8 8.000e+03 4.584e-05 1.156e-05 -179.72° -178.72°
9 9.000e+03 4.191e-05 1.057e-05 -179.88° -178.88°
Total Harmonic Distortion: 0.009107%(0.009551%)
Code:
Harmonic Frequency Fourier Normalized Phase Normalized
Number [Hz] Component Component [degree] Phase [deg]
1 1.000e+03 1.587e+01 1.000e+00 -1.01° 0.00°
2 2.000e+03 4.628e-03 2.916e-04 99.87° 100.88°
3 3.000e+03 3.550e-03 2.237e-04 157.61° 158.62°
4 4.000e+03 4.107e-04 2.588e-05 -157.37° -156.36°
5 5.000e+03 1.015e-03 6.397e-05 178.61° 179.62°
6 6.000e+03 2.768e-04 1.744e-05 155.08° 156.09°
7 7.000e+03 9.420e-05 5.935e-06 5.23° 6.23°
8 8.000e+03 2.356e-04 1.485e-05 -142.37° -141.37°
9 9.000e+03 2.868e-04 1.807e-05 -179.77° -178.76°
Total Harmonic Distortion: 0.037509%(0.037689%)
Code:
Harmonic Frequency Fourier Normalized Phase Normalized
Number [Hz] Component Component [degree] Phase [deg]
1 1.000e+03 2.298e+01 1.000e+00 -1.04° 0.00°
2 2.000e+03 4.777e-03 2.079e-04 115.35° 116.38°
3 3.000e+03 5.535e-01 2.409e-02 0.38° 1.42°
4 4.000e+03 1.883e-02 8.195e-04 90.04° 91.07°
5 5.000e+03 1.789e-01 7.786e-03 179.70° 180.74°
6 6.000e+03 1.252e-02 5.447e-04 -90.95° -89.92°
7 7.000e+03 2.088e-02 9.089e-04 -176.71° -175.68°
8 8.000e+03 3.935e-03 1.712e-04 -98.07° -97.03°
9 9.000e+03 2.828e-02 1.231e-03 1.60° 2.63°
Total Harmonic Distortion: 2.538261%(2.539844%)With all the limits of the models and simulations, distortion is very low and basically low order only.
I've not yet built my BH so I cannot test it, but if anyone would like to try, it would be great.
At what input and output signal levels were these results taken?
With feedback around the driver stage it will reduce the gain available from the driver. Therefore it will take a larger input signal to drive the amp.
Cathode feedback in the output stage reduces the gain of the output stage so increases the demands upon the driver stage. This will further reduce the sensitivity of the whole amp.
gNFB will reduce the gain of the whole amp, requiring higher voltage signal level from the signal source.
Generally, amps are designed to reach full power output with about 1Vrms signal input. Some will aim for 2Vrms signal input (corresponding to 0dBFS from a DAC or CD player) but that doesn't account for sources with lower signal level like FM tuners or many phono stages.
In simulation, what is the voltage sensitivity of this virtual amp circuit?
-
With feedback around the driver stage it will reduce the gain available from the driver. Therefore it will take a larger input signal to drive the amp.
Cathode feedback in the output stage reduces the gain of the output stage so increases the demands upon the driver stage. This will further reduce the sensitivity of the whole amp.
gNFB will reduce the gain of the whole amp, requiring higher voltage signal level from the signal source.
Generally, amps are designed to reach full power output with about 1Vrms signal input. Some will aim for 2Vrms signal input (corresponding to 0dBFS from a DAC or CD player) but that doesn't account for sources with lower signal level like FM tuners or many phono stages.
In simulation, what is the voltage sensitivity of this virtual amp circuit?
-
Hi rongon,
all values are taken at 8 Ohm secondary output.
Last value is obtained with 850 mVrms at the input, first one with 70 mVrms.
But main point is not about that other amp, is about applying local feedback on the PI+Driver on the BHEL84 together with shunt feedback on the EL84s. Both are more demanding for the PI, but EL84s do not need too much swing to give full power.
Same applies for cathode feedback, indeed, but this was not the focus of my curiosity.
all values are taken at 8 Ohm secondary output.
Last value is obtained with 850 mVrms at the input, first one with 70 mVrms.
But main point is not about that other amp, is about applying local feedback on the PI+Driver on the BHEL84 together with shunt feedback on the EL84s. Both are more demanding for the PI, but EL84s do not need too much swing to give full power.
Same applies for cathode feedback, indeed, but this was not the focus of my curiosity.
Apologies for this question which is surely answered in this thread, but I could not find a conclusive answer.
I saw the ‘275VAC’ label on my Baby Huey EL84 boards, and since I had a Toroidy TSTA0250/001 with a secondary with a 275VAC tap, I thought no more of it, until powering up for the first time today. Unfortunately the B+ is just too high with that one - around 380VDC.
Scanning the thread here, it looks like some have plumped for a custom Toroidy PT with a 250VAC secondary.
1) Is there an off the shelf PT for the BH EL84?
2) What secondary should I order? Maybe something for a guitar amp with a 50V fixed bias tap?
I saw the ‘275VAC’ label on my Baby Huey EL84 boards, and since I had a Toroidy TSTA0250/001 with a secondary with a 275VAC tap, I thought no more of it, until powering up for the first time today. Unfortunately the B+ is just too high with that one - around 380VDC.
Scanning the thread here, it looks like some have plumped for a custom Toroidy PT with a 250VAC secondary.
1) Is there an off the shelf PT for the BH EL84?
2) What secondary should I order? Maybe something for a guitar amp with a 50V fixed bias tap?
