---

[Quoting amb]

The result is pretty much what I expect. One way to make the MOSFET version faster and have wider bandwidth is to drive them with a complementary emitter follower (which will lower the impedance and therefore lower the RC time constant of the gates. Since the CFP BJT output stage employs two additional transistors, I think it's fair to give the MOSFET version similar luxury...

In the β22, I took a different approach. Cascoding the MOSFETs also reduces the effective Cgd, In conjunction with a high VAS stage current this also makes things faster.

---

LOL. Who said that electronics has to be fair. But, I agree that two more active devices could make the mosfet version faster.

I am curious though about the reduced Cgd. I am familiar with the reduction in Miller capacitance in a cascode with gain. But I am not familiar with a reduction in Cgd in a follower. Can you give me a reference for this? I'd like to read about it.

In the meantime, after the last set of exercises, here is a new schematic. There are Cdom caps on both drivers and an optional feedback compensation cap if it's needed. I've also added some optional rail caps because on the boards the input stages are some distance from the power inlets.

And here are two possible boards. One with the tube on the topside and one with the tube on the bottom side. Someone will have to decide which one to make.

Topside:

Bottomside:

Any of you guys interesting in building this amp? It would be pretty easy, if boards were made and if you ordered an s22 from amb (but don't tell him it would be for this amp or he might not sell you one ). It may not be a cheap amp to build, but it won't break the bank either.

[Edited by runeight on 01-02-2007 at 11:10 PM.]

---

[Quoting runeight]

I am curious though about the reduced Cgd. I am familiar with the reduction in Miller capacitance in a cascode with gain. But I am not familiar with a reduction in Cgd in a follower. Can you give me a reference for this? I'd like to read about it.

---

[Edited by amb on 01-03-2007 at 03:34 AM.]

---

[Quoting runeight]

And here are two possible boards. One with the tube on the topside and one with the tube on the bottom side. Someone will have to decide which one to make.

---

I liked your layout where the output transistor's rear sides were both facing right, so they could be mounted to the chassis sidepanel or heatsinks, directly or with angle brackets. *duck* Just a suggestion, I'm not a slavedriver, if I were one I'd have ordered changes to be done with deadline yesterday and punish my myrmidon for not keeping the deadline

[Edited by Steinchen on 01-03-2007 at 03:27 AM.]

However, if you don't mind, I still have some questions about the Power Supply. Is that four sets of inputs per amp - because each rail is referenced to a common ground? Not that I have anything against the S22, but that's a very high-level supply. Are there other options that might allow someone interim use while building up to the best performance level?

Do you have a ball park or relative estimate since you mentioned, "It may not be a cheap amp"? There doesn't seem to be too much unless it's in the power supply.

I am still curious about the heater supply (6.3VDC) - how is that provided?

On a different issue, is the idea is to adjust offset between sets of plates in each tube, rather than the cathode-bias-adjust method? If so, how is the adjustment made between the tubes of two channels? Does that put a retail-stereo premium on tube matching? Or is that only relative to the bias adjustment on the output trannies?

Finally, what case options did you envision for these boards? Will two of them slide into one of the Hammond 1455's?

Sorry for all the questions - probably just more of my lack of understanding. I am so glad that you are still here and pushing the envelope on DIY tube-hybrids!!

[Edited by tomb on 01-03-2007 at 06:49 AM.]

---

[Quoting amb]

---

[Quoting runeight]

I am curious though about the reduced Cgd. I am familiar with the reduction in Miller capacitance in a cascode with gain. But I am not familiar with a reduction in Cgd in a follower. Can you give me a reference for this? I'd like to read about it.

---

I think you misunderstood me. Cascoding reduces the effective Cgd, an emitter follower doesn't. But, the emitter follower has lower output impedance than a common emitter amp (which the VAS stage is). This means that the RC time constant with the MOSFET gate capacitance is much lower (thus extending bandwidth and increasing speed). This has essentially the same effect as lowering Cgd.

---

Yes, I must have. I thought you were saying that in the b22 the cascoding of the output fets reduced the Cgd. Since they are operating as followers I would not understand this. But there are many things that I don't understand.

Steinchen, topside it is. I tried the "facing right" layout first. It was a little hard to do with those big heatsinks on the demo board limitations. But I'll try again. I'm pretty facile with this circuit at the moment and it won't take too long. I'll get to it right away.

Tomb, the connector on the right side has five ports. These are:

-30V, PS GND, OUT, OUT GND, +30V.

The s22 is a very high level supply and it will probably give the best results. But we can try to create a less complex supply. The problem will be ripple is we are drawing 200mA to 300mA at idle. Either lots of big caps or active devices. If we need active devices, then the s22 becomes the better choice.

Actually, the amp itself may not be too expensive now that you mention it. The board components are mostly cheap. It will, as usual, be the PS and other parts that cost the money.

The heaters can be 6.3V AC. Just buy a suitable transformer with enough current capacity and hook it up to the heater headers. If you want DC we can design a simple DC supply. But I think that AC will work alright. A nice 12.6V toroid would reduce EM. You can run the 6.3V heaters in series. If it has two 6.3V secondaries, you can put one heater per secondary.

No tube matching is required so long as the tubes are operating properly. The amp has a built-in servo, but it needs just a little bit of help because it doesn't have enough gain. Since these are monoblocks, the offset adj will set the offset to zero for each channel separately. You may have to adjust it if you change tubes and as the amp ages, but the built in servo should do most of the work.

I haven't thought about the enclosure at all. If you guys have some board dimensions that you want me to stick with, I'll do that if the demo Eagle will let me.

I like working with the members of this forum to figure out new ways to spend our money, get you guys in trouble with girlfriends/wives, and make interesting new headphone amps in the process.

---

[Quoting runeight]

I like working with the members of this forum to figure out new ways to spend our money, get you guys in trouble with girlfriends/wives, and make interesting new headphone amps in the process.

---

aaaah, the Devil's advocate...you would have thought I would have seen this sooner ..dB

---

[Quoting runeight]

I thought you were saying that in the b22 the cascoding of the output fets reduced the Cgd. Since they are operating as followers I would not understand this. But there are many things that I don't understand. :)

---

For the rest of you guys, here's a new board with the power transistors facing right so Steinchen can attach them to brackets.

There is one connection to the star ground at the input jack. The headphone jack should also be connected to the star ground, wherever that is, not to the board. So should the PS ground. There are separate input grounds so you can connect your shields and jacks at the input.

Is this the final one?

And here is the Eagle schematic as a link.

[Edited by runeight on 01-03-2007 at 05:36 PM.]

[Edited by j4cbo on 01-03-2007 at 07:09 PM.]

[Edited by runeight on 01-03-2007 at 08:38 PM.]

If you use a 60VCT (or equivalent whatever)to make the split supply, then it seems to me that the easiest thing to do is to put about 10000u/50V as the filter caps followed by lm317/337 regulators set for 30V. And the usual set of components to get the best performance from the regs. This is a pretty standard split supply and will maintain 30V if the bias is adjusted.

This seems like the simplest way to go to get things up and running.

I am sure that you can get a small 6.3V transformer. For example Hammond 166g6 will do if you stick with 6gm8s or 6922s.

Triad makes cheapo transformer for this and there are probably a handful of toroids. Hopefull this won't add too much to the space requirements.

---

[Quoting tomb]

Sounds OK, but the heaters on two 6GM8's will pull 660 ma. You suggested a separate xfmr earlier, but is there a way to incorporate it into the PS design?

---

My inclination would be to actually plan for AC on the heaters, especially if a second transformer is to be used. However, the terminal block should add a place for the center tap to be grounded, or, even better, to hook to a voltage divider to bias the heater up above the cathode few volts.

Regarding the CT, since these are monoblock boards, the CT should go directly to the star ground.

Biasing the heaters is not a bad idea. Even though the cathodes are only a volt or two above ground, a little bias on the heaters wouldn't hurt. However, we still have the mono board issue. I guess we could put a bias chain on the board and then the builder would install components for just one of them. OTOH, my inclination is to leave this off and do it p2p if someone want to have heater biasing. But, I'm happy to add this if the rest of group here wants them.

One other thing to note about the boards is that pin 9 of tube is grounded. This is assuming a tube with 9 as the shield.

If you want to tube roll with tubes where pin 9 is the CT of the heater, then we need to have a further discussion about what to do with pin 9 on the board and how to work the heater supply to be flexible for a wide variety of tubes. MHO is, however, that between the 6gm8 and the 6922 you'll have enough variety to choose from.

Here's a new schematic showing pins on the tube.

[Edited by runeight on 01-04-2007 at 01:29 AM.]

As to heater biasing, I think PTP is fine. I just had to do this for the first time, and it was with a 5842 which is much more volatile than anything being used here. It is unlikely that it will be necessary here, and if it is, a couple of PTP resistors is good for you.

As for the PSU, if you don't want to build a sigma22 a dual TREAD should be fine too. I agree also on a separate transformer for the heater.

just a final wish for the pcb layout: may you move R18, R19, R22 and R23 close to the output transistors and provide drills for horizontal mounting of the trannies to angle brackets please

Can we get the dual TREAD-like PS and heater bias adjustment on a board, then - something to keep the parts organized? The basic TREAD is really good, but it could probably use bigger/dual power caps for this stuff - as Runeight suggested with the 10,000uF cap above. If you left the xfmrs off, that would leave plenty of flexibility for those of us with creative xfmr wishes.

About potential cases - I would just pick one of the width dimensions in the Hammond 1455 series that fits one of the dimesions of your boards - maybe sliding them in sideways, along with the PS board (hint, hint). Even if we end up with something where the boards have to be supported in the middle, but the slots still held them on the sides, that would be of benefit.

Pin 9 as a heater - is that on that Russian Magnum tube?

[Edited by tomb on 01-04-2007 at 06:36 AM.]

Steinchen, I am not sure what you mean about the drills and moving the resistors. That is, I understand how you want to heastink the transistors, but I don't understand exactly what you want to do with the board. I would like to send you the eagle files and have you make these mods. This ok with you?

I don't know about a PCB for the PS. It depends on what you guys decide to use. Also, someone will have to get boards made. And we don't know who that is yet. So we should discuss this too.

As for cases, the board is currently 3.8" x 3.6". Are there any 1455 cases that are tall enough to hold the sk129 heatsinks?

Tomb, on many tubes pin 9 is a center tap on the heaters and sits electrically between 4 and 5. Using a capacitor/jumper here permits us to leave pin 9 floating for tubes that are connected like this (or ac coupled to ground).

---

[Quoting runeight]

Steinchen, I am not sure what you mean about the drills and moving the resistors. That is, I understand how you want to heastink the transistors, but I don't understand exactly what you want to do with the board. I would like to send you the eagle files and have you make these mods. This ok with you?

---

---

[Quoting runeight]

Are there any 1455 cases that are tall enough to hold the sk129 heatsinks?

---

[Edited by Steinchen on 01-04-2007 at 09:31 AM.]