Intrest Check: 120w @ 8ohm Tube Buffered LM3875 Amplifier

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hi,

I am not completely sure on this because your circuit diagram is very hard to read and my valve electronics is not great. But I don't think your circuit will work correctly without R103.

The negative input on the amp chip is a virtual earth and with a voltage output from the tube and nothing to limit the current the amp is just going to go straight to a rail fighting the current from the valve output. I think you need R103 to limit the current and define the gain.

Also with the plane you have used for the chip amplifers you have shared the bulk decoupling rectifier currents and the signal ground for the amplifer on the same plane. This is likely to make it hum. I would seperate the decoupling caps from the plane and route them seperately either with thick track (my preference) or with a sperate plane.

I am also struggling to see how the ground follows the signal through the tube stage it appears to go via the star earth, surely this can't be right. Shouldn't the ground follow the signal through the low power stage and the onto the high power stage.

As I said I am not a tube designer so I may have completely misunderstood what is going on but it must be worth another check through at least.

Can I ask, have you prototyped this design. If not I would strongly suggest trying it out first. Even if you just do the tube front end and connect it up to an existing chip amp design. Otherwise I think you might spend out on alot of money on PCBs that will need massive modifications before they work properly.

Regards,
Andrew
 
gfiandy said:
Hi,

I am not completely sure on this because your circuit diagram is very hard to read and my valve electronics is not great. But I don't think your circuit will work correctly without R103.

The negative input on the amp chip is a virtual earth and with a voltage output from the tube and nothing to limit the current the amp is just going to go straight to a rail fighting the current from the valve output. I think you need R103 to limit the current and define the gain.

Also with the plane you have used for the chip amplifers you have shared the bulk decoupling rectifier currents and the signal ground for the amplifer on the same plane. This is likely to make it hum. I would seperate the decoupling caps from the plane and route them seperately either with thick track (my preference) or with a sperate plane.

I am also struggling to see how the ground follows the signal through the tube stage it appears to go via the star earth, surely this can't be right. Shouldn't the ground follow the signal through the low power stage and the onto the high power stage.

As I said I am not a tube designer so I may have completely misunderstood what is going on but it must be worth another check through at least.

Can I ask, have you prototyped this design. If not I would strongly suggest trying it out first. Even if you just do the tube front end and connect it up to an existing chip amp design. Otherwise I think you might spend out on alot of money on PCBs that will need massive modifications before they work properly.

Regards,
Andrew

Andrew,

Thankyou for the feedback, i agree with you my schematic diagram is not easy to read, becuase originally i was just planning to use it for myself. I have uploaded a slightly better version, here, that is also slightly updated.

R103
I have had long considerations on R103, and if it is actually needed or not, you are correct, R103 should not actually be omitted, and on the latest board design i have placed R103 back, as i belive there will indeed be stabilty issues without it.

I have also seriously considered the output grounding, and i now plan to run a thick trace, directly from the speaker output terminals to the star ground.

The reason the input ground is dumped directly into the ground plane, is because it is no longer needed, if you look at a pinout for a LM3875 it does not actually contain any kind of ground connection, so i do not actually see a problem with this setup.

I have actually prototyped the tube front end, and it works surprisingly well as a stand alone buffer, and i did not notice any problems with hum or grounding with the setup i used with the input grounding.

I do plan to route to input ground directly to the star ground, instead of just dumping it straight into the plane.

Normally, whilst constructing tube amplfiers, i just dump the signal ground directly into the star ground, and i have yet to see a problem with this.

I hope i have helped to clear things up a bit for you.

Rhys
 
Hi Rhys,

Thanks for the updated sch. It now reads much better especially round the amps where I see you have gone to a positive input version rather than an inverting design.

I agree that a sperate path from the output to the star GND is a good improvement.

There are two points in this design where ground is defined for the amplifer, if either of them have 50Hz on them the amp will hum as noise will be injected. This is the bottom of R150 and the bottom of R152. You may get away with it as with two rectifers for the different rails the currents are quire well defined in the PSU circuit. However I would still seperate them from the input stage to be sure.

Looking at the app note, you have omitted Rb the 1K input protection resistor on the positive input. You can probably get away with this as you are AC coupled into what should be a fairly high impedance circuit. But I would still put in an option for it as the power up characteristices of a tube are poorly defined and could cause a problem, you can always fit a 0R if you find its not required. On the same basis I would put the option in for the RC to ground on the output (often called a Zobel ) and the RC option around the feedback resistor, this way if you do have a stability problem you have options to improve it. All these are all optional as it will probably work ok without but why take the risk. (I normally design comercial stuff where you can't afford to take any risks as you don't know what the load speaker will be)

On the tube front end, if you find it works dumping ground into the star then I can't argue as I haven't built any tube stuff.

I am glad to hear you have prototyped, you never really know what will happen untill you have tried it :)

I hope it works out well.
Andy
 
gfiandy said:
Hi Rhys,

Thanks for the updated sch. It now reads much better especially round the amps where I see you have gone to a positive input version rather than an inverting design.

I agree that a sperate path from the output to the star GND is a good improvement.

There are two points in this design where ground is defined for the amplifer, if either of them have 50Hz on them the amp will hum as noise will be injected. This is the bottom of R150 and the bottom of R152. You may get away with it as with two rectifers for the different rails the currents are quire well defined in the PSU circuit. However I would still seperate them from the input stage to be sure.

Looking at the app note, you have omitted Rb the 1K input protection resistor on the positive input. You can probably get away with this as you are AC coupled into what should be a fairly high impedance circuit. But I would still put in an option for it as the power up characteristices of a tube are poorly defined and could cause a problem, you can always fit a 0R if you find its not required. On the same basis I would put the option in for the RC to ground on the output (often called a Zobel ) and the RC option around the feedback resistor, this way if you do have a stability problem you have options to improve it. All these are all optional as it will probably work ok without but why take the risk. (I normally design comercial stuff where you can't afford to take any risks as you don't know what the load speaker will be)

On the tube front end, if you find it works dumping ground into the star then I can't argue as I haven't built any tube stuff.

I am glad to hear you have prototyped, you never really know what will happen untill you have tried it :)

I hope it works out well.
Andy


Andrew,

Thanks again for the reply, i think i will cut the main amplifier plane in half, one half for the front end itself and one half for the power supply, and just simply link them to star ground with a thick trace. Do you think this would help to remove 50hz hum that could possibly be injected?

I was going to leave the zobel network off the board, and leave the builder to install it on the speaker terminals, i have never found any disadvantages to doing this.

I thought that if i needed to go the RC option around the feedback resistor aswell, i would simply do this on the back side of the board to keep paths as short as possible to help defeat the problem, personally i think this would do the trick but some may disagree.

Thankyou for the hint about the Rb resistor, i will place a spot on the pcb, and as you said just install a 0R if it is not needed.

Rhys
 
Yes splitting the plane should do it. If you think about the returning current for the rectifier circuit it passed through the diode to the capacitor, through the capacitor and into the plane, back to the diodes then back to the connector. So long as this path is solid and not shared with any of the input paths you should be ok. To get the best hum rejection I think you should take two seperate grounds back to the star one from the pos and one from the neg which is why I would track them. But a plane should have low enough impedance for you to get away with it.

I guess for DIY your other solutions are good, although if you put the Zobel on the far end of the speaker cables you introduce the inductance of the speaker cables into the circuit and risk creating a tank oscillator circuit at some undefined frequency.

Regards,
Andy
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.