Thanks Ian. I wonder if the other Ian (Gingertube) would care to flesh those comments out? I think he's a traveler so pops in and out of here on a regular basis. I guess I'm wondering about what kind of fixed bias? Negative grid or positive cathode? How bipassed if not CCS. Oh Lord, a plain old bipass cap is easy to understand. Is the final iteration ( if there is such a thing) of the 6V6 using source followers on the AX7s or in some other fashion? And is the AX7 all that necessary anyway? Maybe with the 6V6 but the EL-84 is so easy to drive you'ld think an AU7 or AT7 would be enough. I'm told slew limiting is a factor with the higher mu triodes, but then we're not dealing with a phase splitter here are we. I got so much to learn but I want something to listen to while I'm doing it. You know? All my experience has been ham transmitters and receivers from the olden days and guitar amps. None of this modern stuff. I'll be driving Altec VOTT stuff so don't need more than a couple watts. I think 17 watts gets us to 130db or so.....
But then again, I've never used SI in anything other than a power supply anyway. What do I know.
But then again, I've never used SI in anything other than a power supply anyway. What do I know.
Back "on air" after having no internet access for a week and a half.
The proposed zobels from anode to screen taps did NOT work out, at least for the basic Baby Huey. The phase and amplitude plots I did quickly showed that they were just limiting the frequency response of the amp. This DID make some CD's sound better but that is the fault of my CD Player not the Amp. I had a long listening session on vinyl and it was quite clear that the amp was best with no zobels fitted - in fact this was the first amp where I was able to set tracking weight and anti-skate on the arm just by listening.
For the "Standard" Baby Huey (not the fixed biased version with mosfet source followers) the frequency limit is not in output trannies. I found that the HF -3dB point was at 38kHz and there is a slight peak at 53kHz.
To do the phase measurements I used my digital CRO and read of the time delay (group delay) from input to output and then converted those times to phase with the trusty calculator. It was INTERESTING looking at it in the time domain this way.
The following data taken at 1V RMS out into 4 Ohms (I.e at only 1/4 of a watt)
1000Hz 1.02VRMS 4.4us delay => 0 degrees phase
2000Hz 1.04VRMS 4.4us delay => 1.5 degrees phase
3000Hz 1.14VRMS 4.4us delay => 4.3 degrees phase
5000Hz 1.10VRMS 4.4us delay => 7.6 degrees phase
10kHz 1.08VRMS 4.4us delay => 15 degrees phase
20kHz 1.02VRMS 4.4 us delay => 32 degrees phase
30kHz 950mVRMS 4.3 us delay => 46 degrees phase
45kHz 600mVRMS 3.6us delay => 58 degrees phase
53kHz 715mVRMS 3.3us delay =>63 degrees phase
100kHz 425mVRMS 3.0us delay => 109 degrees phase
Adding the 820 Ohms + 2n2 zobels between anode and screen taps on each of the push pull sides resulted in high frequency roll off. Output was 820mVRMS (compared to 1.02V) at 20kHz and 680mVRMS (compared to 950mV) at 30kHz. I tried smaller and smaller zobels and ended up with none at all.
Cheers,
Ian
The proposed zobels from anode to screen taps did NOT work out, at least for the basic Baby Huey. The phase and amplitude plots I did quickly showed that they were just limiting the frequency response of the amp. This DID make some CD's sound better but that is the fault of my CD Player not the Amp. I had a long listening session on vinyl and it was quite clear that the amp was best with no zobels fitted - in fact this was the first amp where I was able to set tracking weight and anti-skate on the arm just by listening.
For the "Standard" Baby Huey (not the fixed biased version with mosfet source followers) the frequency limit is not in output trannies. I found that the HF -3dB point was at 38kHz and there is a slight peak at 53kHz.
To do the phase measurements I used my digital CRO and read of the time delay (group delay) from input to output and then converted those times to phase with the trusty calculator. It was INTERESTING looking at it in the time domain this way.
The following data taken at 1V RMS out into 4 Ohms (I.e at only 1/4 of a watt)
1000Hz 1.02VRMS 4.4us delay => 0 degrees phase
2000Hz 1.04VRMS 4.4us delay => 1.5 degrees phase
3000Hz 1.14VRMS 4.4us delay => 4.3 degrees phase
5000Hz 1.10VRMS 4.4us delay => 7.6 degrees phase
10kHz 1.08VRMS 4.4us delay => 15 degrees phase
20kHz 1.02VRMS 4.4 us delay => 32 degrees phase
30kHz 950mVRMS 4.3 us delay => 46 degrees phase
45kHz 600mVRMS 3.6us delay => 58 degrees phase
53kHz 715mVRMS 3.3us delay =>63 degrees phase
100kHz 425mVRMS 3.0us delay => 109 degrees phase
Adding the 820 Ohms + 2n2 zobels between anode and screen taps on each of the push pull sides resulted in high frequency roll off. Output was 820mVRMS (compared to 1.02V) at 20kHz and 680mVRMS (compared to 950mV) at 30kHz. I tried smaller and smaller zobels and ended up with none at all.
Cheers,
Ian
I finally got the BH circuit into a 6L6 (6P3S) "test-bed" amp. It sounds very good. Nice bass and good detail, good on rock'n'roll too. Due to the grids being at -28V to achieve a 20W Pd with 370V B+, UL, and fixed bias, I don't know if full power will be achievable due to available voltage swing from the diff amp, but the amp has plenty of power for my little FR speakers.
First up I tried 10K feedback resistors, then 22K, then 32K (10K + 22K). It could do with a little bit more, it is not sounding clinical yet, far from it, but I'm losing overall gain with each increase, not sure how far to go. There is no GFB at the moment.
I am tempted to buy some 6V6 output tubes but I am concerned the OPT primary anode to anode impedance may be a bit low at 5300 ohms, or is that nothing to worry about?
Is it possible that smaller tubes like EL84, 6AQ5, 6P1P, 6V6, require less overall feedback than larger output tubes, like 6L6, EL34? Just wondering since I have two small output tube amps that get away with very little feedback.
Any comments/answers/guidance appreciated, I'm all ears.
First up I tried 10K feedback resistors, then 22K, then 32K (10K + 22K). It could do with a little bit more, it is not sounding clinical yet, far from it, but I'm losing overall gain with each increase, not sure how far to go. There is no GFB at the moment.
I am tempted to buy some 6V6 output tubes but I am concerned the OPT primary anode to anode impedance may be a bit low at 5300 ohms, or is that nothing to worry about?
Is it possible that smaller tubes like EL84, 6AQ5, 6P1P, 6V6, require less overall feedback than larger output tubes, like 6L6, EL34? Just wondering since I have two small output tube amps that get away with very little feedback.
Any comments/answers/guidance appreciated, I'm all ears.
The size and/or output power has nothing to do with it. What's important is the nature of the harmonic distortion. The 6L6 or its RF versions, 807 and 1625, produce less overall THD. What they do produce, however, has a high harmonic content above h3. Running pure open loop pentode mode with no NFB applied can sound really nasty with some program material. These types require a bit more NFB help than other types. (Local feedback + gNFB)
Others like the 6V6, 6AQ5, etc., can sound better since these produce mainly h3. The 6BQ6 is another one, capable of putting out nearly 40W while operating quite conservatively, its distortion is mainly h3. Open loop, these sound a bit "edgy", but without the same sort of pentode nastiness. Mainly add just enough gNFB to take the edge off.
Thanks for that Miles. I also found this post here and think I am heading in the right direction. Also found where some of the gain went, silly me left the 6 ohm resistors across the outputs after it was on the bench...a bit more fiddling and I will have a nice amp, the circuit is great, this one is a keeper.
Ian.
Ian.
Hi Ian444,
My faith is that this schemo does not need UL (personnaly, I hate UL
).
I know I disagree with Ian (gingertube) on this point, but local feedback already scales down internal tube's Rp.
Feeding your 6L6's screens with some "clean" source around 250v to 300v and adjusting bias accordingly will somewhat increase the gain and remove almost all Miller effect back to the driver.
Yves.
My faith is that this schemo does not need UL (personnaly, I hate UL
).I know I disagree with Ian (gingertube) on this point, but local feedback already scales down internal tube's Rp.
Feeding your 6L6's screens with some "clean" source around 250v to 300v and adjusting bias accordingly will somewhat increase the gain and remove almost all Miller effect back to the driver.
Yves.
Well guys I decided on the Baby Huey and am in full swing, just need a little info on the PSU schematic.
For a stereo setup do I need two seperate input transformers so pretty much just making two of the PSU schematics, or can I just use one tranny?
I've been slightly confusing myself with my noobness and am trying to ask the right question I think if someone ticks a box from my thoughts below it may help,
Re: Gingertube PSU schematic
1)Schematic as it is powering both Left and Right Channels?
2)Two of the schematics as they are with two Trannies, one for each channel?
3)One tranny with two seperate PSU's from the schematic paralelled from the secondary's?
or 4) you tell me?
a pic would be a help as well or links to pics.
I've got the bias blocks all drawn up so at least they're out of the way, any help would be golden as well as the odd noob stab thrown in as well.
For a stereo setup do I need two seperate input transformers so pretty much just making two of the PSU schematics, or can I just use one tranny?
I've been slightly confusing myself with my noobness and am trying to ask the right question I think if someone ticks a box from my thoughts below it may help,
Re: Gingertube PSU schematic
1)Schematic as it is powering both Left and Right Channels?
2)Two of the schematics as they are with two Trannies, one for each channel?
3)One tranny with two seperate PSU's from the schematic paralelled from the secondary's?
or 4) you tell me?
a pic would be a help as well or links to pics.
I've got the bias blocks all drawn up so at least they're out of the way, any help would be golden as well as the odd noob stab thrown in as well.
The short answer is that 1), 2), and 3) above are all viable. If you are building mono blocks then two of everything are needed, with each power transformer current rating about 1/2 of what you would need for the stereo version. The component values would change a bit to get the target B+ voltage you need.
The way it's drawn is 1) schematic as it is powering both channels (stereo version).
Read through the thread, there are lots of posts about the PS design, like using a choke and/or using a CLC design to be able to adjust your B+ voltage. Lots of info about power transformer selection also.
You may also want to download Duncan Amps free PSUDII software to model the PS once you have chosen a power transformer; it will allow you to tweak various components to look at the effect on ripple voltage and B+ voltage. It's quiet easy to use and very educational. One hint-first thing, set the default 5K load to constant current and use the tube circuits current draw as the load.
PSUD2
Just an update on the orginal CCS biased BH (the one from page 1 circuit).
I have confirmed that the frequency response high limit is due to the Miller capacitance of the output tubes. I found that a worthwhile improvement could be gained in imaging and top end detail by rolling off the shunt feedback at higher frequencies to precompensate those losses. I have 15K as the shunt feedback set resistor. I fitted a 1K5 + 2n2 zobel across that 15K. This does not apply to the later circuit with mosfet source folowers (because its high frequency limit is not due to output tube Miller capacitance).
Cheers,
Ian
I have confirmed that the frequency response high limit is due to the Miller capacitance of the output tubes. I found that a worthwhile improvement could be gained in imaging and top end detail by rolling off the shunt feedback at higher frequencies to precompensate those losses. I have 15K as the shunt feedback set resistor. I fitted a 1K5 + 2n2 zobel across that 15K. This does not apply to the later circuit with mosfet source folowers (because its high frequency limit is not due to output tube Miller capacitance).
Cheers,
Ian
Hi Gingertube, many thanks for a great design.
1) I'm using 1609's (10k Raa) instead of the 1608's. Would this make any difference to the values of the shunt feedback resistor + zobel network to be used. Currently I'm just using 16K but liking the idea of a tweek
2) I've aquired a nice pair of 2.5H 140ma inductors. I was wondering if they could be utilised to improve channel separation. I was thinking of replacing the pair of 47r resistors in the power supply currently doing channel separation duty with these inductors. Perhaps any benefit to channel separation would be marginal.
After the fast diode rectifier I've got 100uf-1uf-10H-100uf-1uf leading to the 47r resistors per channel and finally 100uf + 0.22uf Auricap low esr at the centre tap of each OT.
(CCS biased)
Best Regards
Bill
1) I'm using 1609's (10k Raa) instead of the 1608's. Would this make any difference to the values of the shunt feedback resistor + zobel network to be used. Currently I'm just using 16K but liking the idea of a tweek
2) I've aquired a nice pair of 2.5H 140ma inductors. I was wondering if they could be utilised to improve channel separation. I was thinking of replacing the pair of 47r resistors in the power supply currently doing channel separation duty with these inductors. Perhaps any benefit to channel separation would be marginal.
After the fast diode rectifier I've got 100uf-1uf-10H-100uf-1uf leading to the 47r resistors per channel and finally 100uf + 0.22uf Auricap low esr at the centre tap of each OT.
(CCS biased)
Best Regards
Bill
Last edited:
Try 1K6 + 1500pF or 1800pF as the zobel across your 16K .
Replacing the 47R with the 2.5H 140mA inductors MAY help - it is an easy mod to try and hence easy to back out if you don't like it.
When tweaking that shunt feedback zobel - the resistor should be 1/10the the value of the shunt feedback set resistor. The capacitor should be sized to give a RxC approx 3.3X10^-6. You will know when you get it right. If you stand well back from the speakers (say 4 m) and walk forward to within about 1m in front of the speakers you should note no real change in imaging. That tells you you have got the PHASE response correct which in turn tells you that the zobel is right. Carefull there is one setting past optimum (1 capacitor size too big) which is superficially attractive but doesn't stand the long term listening test. I have actually decided to back my 1K5 + 2n2 off to 1K5 + 1800pF, NO 1800pF so I used 1K8 + 2n2 instead - almost as good.
Cheers,
Ian
BTW - this represents the last tweak I'm going to do to the CCS biased version of the Baby Huey. It sounds great BUT only up to a certain level (loud), past which the Achilles Heel of the CCS bias arrangement shows up - Its mediocre overload characteristic. To fix that you need to go to the 2nd version with the fixed bias and the Mosfet Source Followers, which incidentally also fixes the problem of driving the output tube Miller Capacitance which is why this tweak is not suitable for that version.
Replacing the 47R with the 2.5H 140mA inductors MAY help - it is an easy mod to try and hence easy to back out if you don't like it.
When tweaking that shunt feedback zobel - the resistor should be 1/10the the value of the shunt feedback set resistor. The capacitor should be sized to give a RxC approx 3.3X10^-6. You will know when you get it right. If you stand well back from the speakers (say 4 m) and walk forward to within about 1m in front of the speakers you should note no real change in imaging. That tells you you have got the PHASE response correct which in turn tells you that the zobel is right. Carefull there is one setting past optimum (1 capacitor size too big) which is superficially attractive but doesn't stand the long term listening test. I have actually decided to back my 1K5 + 2n2 off to 1K5 + 1800pF, NO 1800pF so I used 1K8 + 2n2 instead - almost as good.
Cheers,
Ian
BTW - this represents the last tweak I'm going to do to the CCS biased version of the Baby Huey. It sounds great BUT only up to a certain level (loud), past which the Achilles Heel of the CCS bias arrangement shows up - Its mediocre overload characteristic. To fix that you need to go to the 2nd version with the fixed bias and the Mosfet Source Followers, which incidentally also fixes the problem of driving the output tube Miller Capacitance which is why this tweak is not suitable for that version.
Last edited:
Ian,
Many thanks for the extensive explanation, will try this out , in the meantime will order the parts for the fixed bias version - been too happy with the ccs version. Compared to my ss amp the baby huey has an uncanny ability to reveal studio effects such as vibrato and reverb, backing harmonies, and odd dissonant notes, and ...........
Many thanks for the extensive explanation, will try this out , in the meantime will order the parts for the fixed bias version - been too happy with the ccs version. Compared to my ss amp the baby huey has an uncanny ability to reveal studio effects such as vibrato and reverb, backing harmonies, and odd dissonant notes, and ...........
CCS
Thanks Adamus, and as Monty Python once said "explain the logic underlying that statement" ;-)
I am sure the Baby Huey is great in terms of sound quality, I was just trying to get a sense on the CCS 'block'
My understanding is that fixed bias's objective is to ensure delviery of a constant current, taking into account changes in valve dynamics. As such, solid state offers a great choice, as in the BH.
However, as the bias is not part of the signal path, I am not sure I understand how it effects sounds quality, apart from keeping all tubes drawing the same current.
Therefore, my question was with regard to the merits of using the BH design over something like a LM317, which has less components, and seems to be used in other amps with great effect.
As a newbie, just trying to get my head around the various schools of thought on solid state fixed bais and am keen to learn from people on this thread.
Thanking you in anticpation.
Malcolm
Thanks Adamus, and as Monty Python once said "explain the logic underlying that statement" ;-)
I am sure the Baby Huey is great in terms of sound quality, I was just trying to get a sense on the CCS 'block'
My understanding is that fixed bias's objective is to ensure delviery of a constant current, taking into account changes in valve dynamics. As such, solid state offers a great choice, as in the BH.
However, as the bias is not part of the signal path, I am not sure I understand how it effects sounds quality, apart from keeping all tubes drawing the same current.
Therefore, my question was with regard to the merits of using the BH design over something like a LM317, which has less components, and seems to be used in other amps with great effect.
As a newbie, just trying to get my head around the various schools of thought on solid state fixed bais and am keen to learn from people on this thread.
Thanking you in anticpation.
Malcolm
CCS and all that,
The CCS bias of the output tubes is there to guarantee output tube idle current balance (in a no adjustment ever required way). This is the way to get the best out of your output tranny in terms of bottom end and fine detail. The actual audio path for the output tubes is via the CCS bypass capacitors which is where you want to spend your cash as far as good components go.
"Looking into" the cathodes of the output tubes you see an impedance of approximately 1/gm which for an EL84 means about 90 to 100 Ohms. The CCS should therefore offer an impedance of say 1000 times that or 100K minimum. A LM317 based CCS WILL manage that up to about 100kHz above which the device capacitance becomes an issue. I used the "Ring of Two" current source instead which will give marginally better performance particularly at higher frequencies. Using a MJE340 as the pass transistor is not absolutely required but its 300V rating gives it half a chance of surviving a tube short. The LM317 based CCS would be toast after a tube short.
So for output tube bias CCSes the LM317 would be good enough - The "Ring of Two" using the MJE340 is just an attempt to make the thing a bit more "bullet proof" (or test probe slip proof). The LM317 device is also known to have stability issues (ringing) in some circuits. I simply did'nt want to tackle those issues.
Further:
Note however that the diffamp CCS requires a much better CCS and that is why I used a cascode bipolar transistor CCS in that spot. I think I mentioned in a much earlier post that I originally used a single transistor + LED reference CCS there and noted a significant improvement in smootheness and top end detail when I changed it to the cascode transistor arrangement. In my view the cascode bipolar transistor CCS is the best performing "simple" CCS.
Hope this is what you were after as an explanation of what was intended.
Cheers,
Ian
The CCS bias of the output tubes is there to guarantee output tube idle current balance (in a no adjustment ever required way). This is the way to get the best out of your output tranny in terms of bottom end and fine detail. The actual audio path for the output tubes is via the CCS bypass capacitors which is where you want to spend your cash as far as good components go.
"Looking into" the cathodes of the output tubes you see an impedance of approximately 1/gm which for an EL84 means about 90 to 100 Ohms. The CCS should therefore offer an impedance of say 1000 times that or 100K minimum. A LM317 based CCS WILL manage that up to about 100kHz above which the device capacitance becomes an issue. I used the "Ring of Two" current source instead which will give marginally better performance particularly at higher frequencies. Using a MJE340 as the pass transistor is not absolutely required but its 300V rating gives it half a chance of surviving a tube short. The LM317 based CCS would be toast after a tube short.
So for output tube bias CCSes the LM317 would be good enough - The "Ring of Two" using the MJE340 is just an attempt to make the thing a bit more "bullet proof" (or test probe slip proof). The LM317 device is also known to have stability issues (ringing) in some circuits. I simply did'nt want to tackle those issues.
Further:
Note however that the diffamp CCS requires a much better CCS and that is why I used a cascode bipolar transistor CCS in that spot. I think I mentioned in a much earlier post that I originally used a single transistor + LED reference CCS there and noted a significant improvement in smootheness and top end detail when I changed it to the cascode transistor arrangement. In my view the cascode bipolar transistor CCS is the best performing "simple" CCS.
Hope this is what you were after as an explanation of what was intended.
Cheers,
Ian
Last edited: