Hello !
I am building an LM317-based CCS board for a tube LTP and I have adjusted the differential track lengths to within a millimeter with serpentines as seen in professional pcbs (see image). This was mainly as an educational exercise but I figured why not put all chances on my side since it is easily done.
Note that the board hosts two independent channels and the lengths are adjusted for the +/- tubes in a channel but the channels don't have the same lengths between them - One is 10mm longer, which is 33% more with respect to the other channel. The adjusted tracks start from the on-board differential balance adjustment trimpot; everything from that pot's wiper to ground is common-mode.
As I understand it, differential routing length trimming is critical when dealing with high-speed circuits, where the tracks being much longer than the signal wavelength they carry behave like transmission lines. Our audio circuits are not high speed but tubes can oscillate at HF. The current project is an input stage using triode-connected pentodes, I read these are prone to oscillate up to the MHz range? Also the LM317 itself can oscillate.
So back to the current question, how critical is track and wiring length for this project and tube differential pairs in general? The board doesn't host the tube sockets and so if perfect balance is needed the wiring to the +/- sockets will of course have same lengths. Now that I'm finished writing I realize I should be talking about impedance matching, but in this project's context tracks and wires have same width/gauges and are far apart, so parasitics are mostly equal and signal cross-coupling should not be an issue... I think 😎
Thanks for any insigths!
Joris
I am building an LM317-based CCS board for a tube LTP and I have adjusted the differential track lengths to within a millimeter with serpentines as seen in professional pcbs (see image). This was mainly as an educational exercise but I figured why not put all chances on my side since it is easily done.
Note that the board hosts two independent channels and the lengths are adjusted for the +/- tubes in a channel but the channels don't have the same lengths between them - One is 10mm longer, which is 33% more with respect to the other channel. The adjusted tracks start from the on-board differential balance adjustment trimpot; everything from that pot's wiper to ground is common-mode.
As I understand it, differential routing length trimming is critical when dealing with high-speed circuits, where the tracks being much longer than the signal wavelength they carry behave like transmission lines. Our audio circuits are not high speed but tubes can oscillate at HF. The current project is an input stage using triode-connected pentodes, I read these are prone to oscillate up to the MHz range? Also the LM317 itself can oscillate.
So back to the current question, how critical is track and wiring length for this project and tube differential pairs in general? The board doesn't host the tube sockets and so if perfect balance is needed the wiring to the +/- sockets will of course have same lengths. Now that I'm finished writing I realize I should be talking about impedance matching, but in this project's context tracks and wires have same width/gauges and are far apart, so parasitics are mostly equal and signal cross-coupling should not be an issue... I think 😎
Thanks for any insigths!
Joris
Last edited:
Triode Wired Pentodes.
Did you put traces and through holes for g1 grid stoppers on your PCB?
And did you put traces and through holes for g2 screen grid stoppers to connect from screens to plates?
You might not need either kind of grid stoppers.
But if you ever have to layout another PCB, put those 4 grid stoppers in there (you can use wire jumpers if no resistor is needed).
I paralleled a 12AT7 dual triode.
The plates were directly tied together.
A 900V IXYS current source was connected from B+ to the plates.
Grids both had grid stoppers, the non-grid ends were connected together (grids separated by 2X the grid stopper resistance).
Each cathode had its own individual self bias resistor, with its own individual bypass capacitor across the resistor. So . . .
The advantage was near equal currents in each cathode, but that did not account for the fact that through the bypass capacitors, the cathodes are effectively connected together at audio frequencies and RF.
All that mess, ended up being an Oscillator.
The solution was to put a 1k Ohm resistor from each plate, and connect the other ends of the resitors to the IXYS current source (plate load).
The RC coupling cap (to the next stage) was connected to the junction of the two 1k resistors and the IXYS current source.
Oscillation stopped.
I was fortunate, I used point to point wiring, not a PCB.
Murphy's Law Correlaries:
If an Oscillation can happen, it will.
If you build an amplifier, you get an oscillator.
If you build an oscillator, you get an amplifier.
:>)
Did you put traces and through holes for g1 grid stoppers on your PCB?
And did you put traces and through holes for g2 screen grid stoppers to connect from screens to plates?
You might not need either kind of grid stoppers.
But if you ever have to layout another PCB, put those 4 grid stoppers in there (you can use wire jumpers if no resistor is needed).
I paralleled a 12AT7 dual triode.
The plates were directly tied together.
A 900V IXYS current source was connected from B+ to the plates.
Grids both had grid stoppers, the non-grid ends were connected together (grids separated by 2X the grid stopper resistance).
Each cathode had its own individual self bias resistor, with its own individual bypass capacitor across the resistor. So . . .
The advantage was near equal currents in each cathode, but that did not account for the fact that through the bypass capacitors, the cathodes are effectively connected together at audio frequencies and RF.
All that mess, ended up being an Oscillator.
The solution was to put a 1k Ohm resistor from each plate, and connect the other ends of the resitors to the IXYS current source (plate load).
The RC coupling cap (to the next stage) was connected to the junction of the two 1k resistors and the IXYS current source.
Oscillation stopped.
I was fortunate, I used point to point wiring, not a PCB.
Murphy's Law Correlaries:
If an Oscillation can happen, it will.
If you build an amplifier, you get an oscillator.
If you build an oscillator, you get an amplifier.
:>)
Thanks for sharing your experience - every bit of info helps. I will be using IT coupling so there are oscillation risks there too I think...
I try to avoid putting signal connexions on pcbs, prefering point-to-point wiring on the socket. So the triode connexion is a very small wire on the relevant socket pins, the g1 stoppers are soldered directly to the socket pins and any g2/anode stoppers will go there as well. The signal input sockets get the input resistors and both signal grounds are tied at the sockets to form the input star ground point. The pcb goes between that ground point and to each cathode.
I try to avoid putting signal connexions on pcbs, prefering point-to-point wiring on the socket. So the triode connexion is a very small wire on the relevant socket pins, the g1 stoppers are soldered directly to the socket pins and any g2/anode stoppers will go there as well. The signal input sockets get the input resistors and both signal grounds are tied at the sockets to form the input star ground point. The pcb goes between that ground point and to each cathode.
The electrical wavelength of a 20kHz signal is 9.3 miles, or 15km.
I would just use short direct paths where possible. The path length matching just raises inductance and capacitance which can make the circuit more prone to HF oscillation.
Path length adjustment for HF involves transmission line calculations including the ground plane or neighboring traces as well as magnetic flux linking to calculate the line Z and loss. Not a hand calculation.
That said, if it makes you feel good to attempt matching, I doubt it will cause any problems.
Path length adjustment for HF involves transmission line calculations including the ground plane or neighboring traces as well as magnetic flux linking to calculate the line Z and loss. Not a hand calculation.
That said, if it makes you feel good to attempt matching, I doubt it will cause any problems.
I'd be more worried about a bear coming in through your window... even in a 5th floor Toronto apartment.
LMAO! For what it's worth, I carry a pepper spray just in case 😉
As long as HF and RF has already been mentioned . . .
Actually, if you use a doublesided PCB, with a ground plane on the back, then there are 2 speed of light velocities.
The top of the trace (air path) is very near to the free space speed of light.
The trace, to the PCB body material, to the ground plane is much slower than the free space speed of light.
When you look at something such as a 3GHz serial data stream which is coming across a PCB (that uses very expensive substrate to make it low loss),
The square wave nature of the serial data Smears, because of the 2 different velocities, that causes 2 different arrival times.
The question is, when did a single transition time from 0 to 1 (or 1 to 0) . . . arrive, arrive? Oh, twice per single transition, I get it, I get it.
High speed serial data implementation is not simple, simple.
Actually, if you use a doublesided PCB, with a ground plane on the back, then there are 2 speed of light velocities.
The top of the trace (air path) is very near to the free space speed of light.
The trace, to the PCB body material, to the ground plane is much slower than the free space speed of light.
When you look at something such as a 3GHz serial data stream which is coming across a PCB (that uses very expensive substrate to make it low loss),
The square wave nature of the serial data Smears, because of the 2 different velocities, that causes 2 different arrival times.
The question is, when did a single transition time from 0 to 1 (or 1 to 0) . . . arrive, arrive? Oh, twice per single transition, I get it, I get it.
High speed serial data implementation is not simple, simple.
Last edited:
The more I learn about parasitics, gain/bandwidth, output loading, etc... the more respect I have for people designing RF transmitters, radar, 5G and so on... I look at my projects and I'm like man, I'm barely scratching the surface. But still having fun !
I had hands-on experience with signal reflection on a home-made SPDIF cable about 30-40 feet long. Strapped a BNC to RCA adapter on some old-style video camera coax cable but for the other end soldered a regular RCA jack. The reflections would cause the output to oscillate (buzz) every other day as the clock timing was lost - until I properly terminated the cable. No problem since then (years).
I had hands-on experience with signal reflection on a home-made SPDIF cable about 30-40 feet long. Strapped a BNC to RCA adapter on some old-style video camera coax cable but for the other end soldered a regular RCA jack. The reflections would cause the output to oscillate (buzz) every other day as the clock timing was lost - until I properly terminated the cable. No problem since then (years).
Last edited:
Them danged radio waves get into everything. I had a 20ft. RCA cable just a bit loose at the mixer input jack and it was picking up an FM radio station about 50 mi. away, because the contact junction was acting like a diode, I guess. I twisted the plug a bit to pull it and it quit, shoved it in better and all was well.
Things only get worse with the switch to digital transmission, wi-fi, cell towers and all those DC/DC converters. Currently wearing tin foil hat.
Absolutely unnecessary.
Your picture does not show any "professional PCB"
It looks very amateurish.
That sounds a little harsh, but it may be somewhat true, since our Joe here actually is an amateur, AFAIK.
From my current state of understanding, I would advise you to aim... well... basically for the opposite from what you were doing: Use short and wide tracks as much as it is feasible. Don't waste the PCB real estate with a "pretty" wiggle, but instead use a single track that is as wide as your wiggle. In your sample above that'd be somewhat overkill already, but assuming your thin traces are 10mil, better use at least 50mil. That will reduce the actual resistance, as well as as the parasitic inductance (which is what may become a problem at some places and thus needs to be as small as possible), and at the same time will be much more robust in Real Life (TM). Just imagine a tiny plated-through hole with a 10mil trace attached - you make a mistake and have to unsolder the component again; maybe even a second time. You'll strip off that little bit of copper in no time.
I was not attacking the poster by any means, of course, but the notion that the design shown is "Professional".
Somebody might take it at face value and design something similar.
My main beef with it is the ultra narrow tracks, which make it very fragile.
The track-pad joint, as shown, is very easy to crack , leading to "mystery failures"
As you mention, with what I strongly agree, is to make tracks as wide as possible for strength.
The added current carrying capacity is a bonus.
Somebody might take it at face value and design something similar.
My main beef with it is the ultra narrow tracks, which make it very fragile.
The track-pad joint, as shown, is very easy to crack , leading to "mystery failures"
As you mention, with what I strongly agree, is to make tracks as wide as possible for strength.
The added current carrying capacity is a bonus.
Amateur here guilty on charges of amateurism 😎
Just to clarify, I didn't imply my pcb was professional, but rather that I added serpentines which are seen in professional pcbs... But that's moot at this point. I am re-designing the board without them and larger traces. Thanks
Just to clarify, I didn't imply my pcb was professional, but rather that I added serpentines which are seen in professional pcbs... But that's moot at this point. I am re-designing the board without them and larger traces. Thanks
Last edited:
All the factors that are compensated for, whether they need to be compensated or not . . .
Are great Fuel for Marketeers Brochures. Tell your customers all those things are why they need to purchase your amplifier.
Just my opinions (I was a Marketeer for 5 years in the T&M industry)
Are great Fuel for Marketeers Brochures. Tell your customers all those things are why they need to purchase your amplifier.
Just my opinions (I was a Marketeer for 5 years in the T&M industry)