Hi Loek,
that's a neat and compact arrangement.
Shame about the shared PSU for the two stereo channels.
Have you ever investigated the splitting of the rectifiers/smoothing caps to independant supplies for each channel?
that's a neat and compact arrangement.
Shame about the shared PSU for the two stereo channels.
Have you ever investigated the splitting of the rectifiers/smoothing caps to independant supplies for each channel?
Shame about the shared PSU
Thanks Andrew for the compliment.
At that time i was happy with the C-core transformer made by hand at a local factory(expensive!) and the 2x 15kuF caps i found isomewhere.
The chassis is as most tube amps are made: big components on top, the wiring and small components on the bottom side.
That is also the reason that it is now more difficult to change to a double PSU so i decided to make a new Leach wicth Jens boards.
But that is future music...
Bye, Loek
Thanks Andrew for the compliment.
At that time i was happy with the C-core transformer made by hand at a local factory(expensive!) and the 2x 15kuF caps i found isomewhere.
The chassis is as most tube amps are made: big components on top, the wiring and small components on the bottom side.
That is also the reason that it is now more difficult to change to a double PSU so i decided to make a new Leach wicth Jens boards.
But that is future music...
Bye, Loek
AndrewT said:
Yes, it should be less than 22k. It should be 20K if the series resistor is 2K.
The series resistor forms a low pass filter with the input capacitor, and helps with interaction with the source. It does not deliberately unbalance. My amp has very low offset with this arrangement. Just because your amp acts better with higher R17 does not mean Leach is deliberately unbalanced. It could mean that your component matching it not ideal.
AndrewT said:
I don't know what you are getting at, here. You are trying to measure DC offset. Therefore, you are measuring DC conditions. At DC, the capacitors in the feedback network are open, and R12+R13 set the base bias current in the negative input transistor. On the input side, R17+R11 do the same, and should add up to the same. If you directly short the input to ground, then you only have R11 on the input side to ground. This will clearly unbalance the base currents.
If you use high-beta input transistors, you can minimize the base currents and the voltages they develop across the input resistors, and therefore minimize voltage differences. If you have Self's book, he has a good explanation of this.
Leach is a DC coupled amp. If you connect a direct coupled preamp to it, the output impedance will effectively short R17. You can add an input cap to prevent this at DC. Leach's preamps had a series cap in the output, as do many preamps. If you want to adjust the DC offset with a preamp connected, it may be impossible by increasing R17, because the parallel combination with the preamp output resistance may never get high enough to balance, no matter how high you raise R17. A capacitor input may be the only way. In my transformer coupled input, I have to raise the series input resistor which is in series with the low output inpedance of the transformer. This, of course, means changing the input capacitor to maintain the input cutoff frequency. But I am also reducing the values of resistors in the feedback network to make this more practical (to reduce the value needed for the series input resistor), which also means raising the value of the feedback cap. The space available on Jens' boards for this cap limits me to a value of 470uf or so, so this sets the conditions for how low I can lower the feedback resistance, and therefore how low I can go with the input series resistance to get balance with the transformer in circuit. I haven't done this yet, so I don't know the results yet.
Hi Pooge,
I realise this posting procedure can lead to confusion.
I will restate my set up conditions.
I am using an AC coupled input with the series DC blocking cap in place.
I am using the AC coupled NFB loop with the DC blocking cap in place which also reduces the DC gain to about one.
When checking the output offset and/or adjusting it, I short out the input RCA to ensure that the open circuit that otherwise be there cannot cause an input instabilities that may cascade down stream (although I think Jens' interpretation of Leach does not suffer from this input instability problem). The AC coupled short also shows a low AC impedance to the input stage and reduces the amplifier noise to a minimum. This is slightly artificial and could mimic the source impedance of the pre-amp for more accuracy in measuring AC performance, but since I am setting up for DC offset the noise/source impedance (AC coupled) is of no consequence.
In the foregoing test situation It is normal for a differential amplifier to have near zero output offset when the input conditions on the inverting and non-inverting inputs are identical.
Leach has not made the two sides of the LTP inputs identical. The series resistor forming the RF input filter is after the grounding resistor. The grounding resistor should tap off the input line after the series resistor. Then the non-inverting input arrangement mimics the inverting input arrangement. That is what I meant when I said Leach has deliberately unmatched the input conditions, which he has done by moving one resistor.
When I matched the LTP transistors I found that using the higher gain bc550/560 I could match the NPN to NPN and the PNP to PNP, but Leach suggests best matching is when NPN matches PNP. I could not achieve that. The gains of all my 560swere higher than the gains of all my 550s so I settled for NPN to NPN matching as Leach calls this up as next best alternative. This will result in a slightly higher, but importantly much more consistent input bias current. The bias will always flow in the same direction.
The slight complication with the Leach style feedback comes from the two route high & low frequency paths. This may be having an effect on the DC conditions but I cannot see that at the moment. If this is a cause of DC imbalance could someone explain the mechanism to me?
I have found great consistency in output offset between my two Leach amplifiers using the input resistors Bommed by Jens. But the DC offset is not nor near zero. Both amplifiers need r17 to be adjusted upwards to about 33k to reduce DC output offset to near zero. This offset remains very close to zero over long term testing. It moves slightly but holds there if the PSU supply voltage is changed, even though the Iq changes dramatically causing significantly different temperatures in the output stage.
33k is some 50% higher than 22k.
Why should this be necessary?
Now back to your other point.
The Leach amp schematics I have seen are not DC coupled. All have a DC block in the NFB path.
If one wants to operate the Leach as a DC coupled amplifier, then one needs to short out the DC blocker on the input HIpass filter and short out the BC blocking cap in the NFB path (lower leg). When you do both then the output offset should remain low or near zero. I have not measured either of my amplifiers for this condition. What I have done in my last posting was to point out that if one partially converts the Leach to DC coupled by only shorting out one of the two DC blocking capacitors one will find it very difficult to arrange for the input impedances to be set to minimise the output offset to an acceptably low level and still retain sensible impedances for the two inputs.
This becomes apparent if you do a fully DC coupled Leach and then attach a DC blocked pre-amp (most are) to the input RCA. DC output offset will go high.
The solution is to use Jens' alternative input which he has on board. DC or AC coupled input with two RCAs at the chassis for connecting the appropriate type of pre-amp.
Sorry this took so long, but it may prove useful to one or two.
I realise this posting procedure can lead to confusion.
I will restate my set up conditions.
I am using an AC coupled input with the series DC blocking cap in place.
I am using the AC coupled NFB loop with the DC blocking cap in place which also reduces the DC gain to about one.
When checking the output offset and/or adjusting it, I short out the input RCA to ensure that the open circuit that otherwise be there cannot cause an input instabilities that may cascade down stream (although I think Jens' interpretation of Leach does not suffer from this input instability problem). The AC coupled short also shows a low AC impedance to the input stage and reduces the amplifier noise to a minimum. This is slightly artificial and could mimic the source impedance of the pre-amp for more accuracy in measuring AC performance, but since I am setting up for DC offset the noise/source impedance (AC coupled) is of no consequence.
In the foregoing test situation It is normal for a differential amplifier to have near zero output offset when the input conditions on the inverting and non-inverting inputs are identical.
Leach has not made the two sides of the LTP inputs identical. The series resistor forming the RF input filter is after the grounding resistor. The grounding resistor should tap off the input line after the series resistor. Then the non-inverting input arrangement mimics the inverting input arrangement. That is what I meant when I said Leach has deliberately unmatched the input conditions, which he has done by moving one resistor.
When I matched the LTP transistors I found that using the higher gain bc550/560 I could match the NPN to NPN and the PNP to PNP, but Leach suggests best matching is when NPN matches PNP. I could not achieve that. The gains of all my 560swere higher than the gains of all my 550s so I settled for NPN to NPN matching as Leach calls this up as next best alternative. This will result in a slightly higher, but importantly much more consistent input bias current. The bias will always flow in the same direction.
The slight complication with the Leach style feedback comes from the two route high & low frequency paths. This may be having an effect on the DC conditions but I cannot see that at the moment. If this is a cause of DC imbalance could someone explain the mechanism to me?
I have found great consistency in output offset between my two Leach amplifiers using the input resistors Bommed by Jens. But the DC offset is not nor near zero. Both amplifiers need r17 to be adjusted upwards to about 33k to reduce DC output offset to near zero. This offset remains very close to zero over long term testing. It moves slightly but holds there if the PSU supply voltage is changed, even though the Iq changes dramatically causing significantly different temperatures in the output stage.
33k is some 50% higher than 22k.
Why should this be necessary?
Now back to your other point.
The Leach amp schematics I have seen are not DC coupled. All have a DC block in the NFB path.
If one wants to operate the Leach as a DC coupled amplifier, then one needs to short out the DC blocker on the input HIpass filter and short out the BC blocking cap in the NFB path (lower leg). When you do both then the output offset should remain low or near zero. I have not measured either of my amplifiers for this condition. What I have done in my last posting was to point out that if one partially converts the Leach to DC coupled by only shorting out one of the two DC blocking capacitors one will find it very difficult to arrange for the input impedances to be set to minimise the output offset to an acceptably low level and still retain sensible impedances for the two inputs.
This becomes apparent if you do a fully DC coupled Leach and then attach a DC blocked pre-amp (most are) to the input RCA. DC output offset will go high.
The solution is to use Jens' alternative input which he has on board. DC or AC coupled input with two RCAs at the chassis for connecting the appropriate type of pre-amp.
Sorry this took so long, but it may prove useful to one or two.
AndrewT said:
The input cap in Leach's amp is not a DC block. Nor is it a high pass. It is a low pass filter tuned around 200K or so. Maybe you are talking about Jens' version?
If you have a series input cap, then the DC base current will travel through the series resistor and the input resistor to ground, so the fact that it is in a different position than others should be of no consequence for DC. The DC path in the feedback is through the "normal" feedback resistors to the "virtual" ground in the output impedance of the amplifier. These paths are virtually identical, and with matched transistors of the ones Leach specified, I had offsets less than 10mv. I also matched resistors around the input stage, just to be anal. If you reversed your unmatched transistors, you may find an offset in the other direction. But your offset is nothing to be concerned with.
I don't see how the split feedback can change the DC operation. That path doesn't turn on until at least 150kHz or more.
Out of curiosity, why not try to measure the offset with the input open instead of grounded to eliminate the possibility of DC leakage through the input cap. Maybe it could have been damaged when soldered.
offset
Maybe an idea to measure the differential voltages through the circuit with a accurate digital voltmeter and find the offset-source to satisfy your feelings?
Loek
Maybe an idea to measure the differential voltages through the circuit with a accurate digital voltmeter and find the offset-source to satisfy your feelings?
Loek
Hi Pooge,
you made me look back to the original Leach texts, yes you're right, Leach shows only the low pass filter on the input.
He has used the mixed AC/DC coupling that I do not recommend.
The extra DC block is provided on the Jens' board.
Out of interest I tried a few more output offset measurements and sure enough the mixed coupling causes excessive output offset.
Here are my measurments with r17=33k, all else as Jens' Bom.
a. input of DC blocking cap to ground, offset=-2.6mV (AC coupled at input and NFB, measurement done cold and will drop to about -1.4mV when fully warmed up)
b. output of DC blocking cap to ground, offset=72.7mV (Leach original connections)
c. output of DC blocking cap thro' 51r to ground, offset= 72.1mV (mimics a 51r source impedance)
If the pre-amp has a DC blocking cap on it's output the offset will drop to -2.6mV and if it is DC coupled the offset rises to about 72mV. A servo in the pre-amp WILL NOT ZERO the offset.
If I were to put back the 22k for r17 the offsets will become 25mV (AC coupled) and about 100mV (mixed coupling, by calculation, I have not measured this last one)
Finally, it's not leakage through the PES input blocking cap.
Could it be leakage through the NFB bi-polar electrolytic?
What direction would the offset move if there is leakage here?
you made me look back to the original Leach texts, yes you're right, Leach shows only the low pass filter on the input.
He has used the mixed AC/DC coupling that I do not recommend.
The extra DC block is provided on the Jens' board.
Out of interest I tried a few more output offset measurements and sure enough the mixed coupling causes excessive output offset.
Here are my measurments with r17=33k, all else as Jens' Bom.
a. input of DC blocking cap to ground, offset=-2.6mV (AC coupled at input and NFB, measurement done cold and will drop to about -1.4mV when fully warmed up)
b. output of DC blocking cap to ground, offset=72.7mV (Leach original connections)
c. output of DC blocking cap thro' 51r to ground, offset= 72.1mV (mimics a 51r source impedance)
If the pre-amp has a DC blocking cap on it's output the offset will drop to -2.6mV and if it is DC coupled the offset rises to about 72mV. A servo in the pre-amp WILL NOT ZERO the offset.
If I were to put back the 22k for r17 the offsets will become 25mV (AC coupled) and about 100mV (mixed coupling, by calculation, I have not measured this last one)
Finally, it's not leakage through the PES input blocking cap.
Could it be leakage through the NFB bi-polar electrolytic?
What direction would the offset move if there is leakage here?
hi,
any ouput offset voltage less than 100mV is acceptable from what i have read, and in my experience having built around 30 of these amps, offset voltage less the +/-100mV did not present any problems for me.😀
any ouput offset voltage less than 100mV is acceptable from what i have read, and in my experience having built around 30 of these amps, offset voltage less the +/-100mV did not present any problems for me.😀
AndrewT said:
But you haven't tried output of DC blocking cap left alone, i.e., without connecting it to ground or another resistor. (Actually, this will probably give you the same offset as "a".) I don't see how you ruled out admittedly unlikely DC leakage throug the DC blocking capacitor. Measure offset with and without the input of the DC blocking capacitor connected to ground. Leave the other side alone.
AndrewT said:
Actually, your DC offset was fine for unmatched transistors.
The drift you see may be more likely related to temperature drifts in the input pair. Put a finger on one of them and notice the drift! Keeping the input pair to zero temperature differential is very desirable for preventing drift. This is why I was looking for a suitable matched pair dual. Alternatively, bonding the pair together to keep them in thermal equalibrium is a suitable option. It would be nice to have a board layout that made this easy. I'm wondering if there's a specialized heat sink out there that will clamp the pair together for this purpose without requiring Krazy Glue.
AndrewT said:
I don't know why you keep worrying about "impedances", which is important under AC conditions, when you are trying to analyze DC conditions, where resistances are what matters. I outlined the DC resistance paths in a previous post, and you keep insisting on shorting these out, adding another one in parallel, or whatever. DC balance is just that. AC is not involved.
If you short the input of an amp with a series input cap, it shouldn't affect DC offset. If you short the amp with no blocking cap, then you essentially short the input grounding capacitor and the balance will be upset. This is why you should measure offset without the input shorted, unless you have a blocking capacitor.
If you have a pre-amp with a DC blocking capacitor in the output, and your amp offset changes when connected to the preamp, then the preamp likely has a resistor to ground at the output after the cap that is placed in parallel with the input grounded resistor in the amp, thus upsetting balance.
So a designer can't design the amp to be perfectly matched with every component. That's why some use an offset trim pot in the input stage. You can reduce the offset by changing the resistor as you have, but there is also the noise factor you have to juggle, as the input transistors have their lowest noise with a particular source impedance, which may not be the lowest impedance.
Self found the highest sensitivity of offset was from base currents. These base currents generate a voltage in the input resistor path and the feedback resistor path. It is these voltages arising out of the base currents which causes the major offset sensitivity. This sensitivity is reduced by lowering the resistance values to lower the magnitude of the base current induced voltages, to thereby reduce the magnitude of variations, and/or use high beta transistors, to reduce the magnitude of the base currents. It is a matched beta that matches the base currents. Of course, matching of the input degeneration resistors helps reduce offset from this source, which Self indicates is not likely as high an error.
So it is not so much that the preamp is DC blocked. It is the paralleling of other resistors to the input resistors (or the shorting of these resistors) that upsets the voltage balance of even equal base currents from a matched transistor pair.
a specialized heat sink
Hi Pooge, if you watch the TO-5 transistors in my last picture you sea the star-type heatsinks.(spring effect)
They also fit on two tied TO-92 type transistors.
Don't know if they still are available. I will search for them.
You can also use heatshrinkable tube.
Greetings, Loek
Hi Pooge, if you watch the TO-5 transistors in my last picture you sea the star-type heatsinks.(spring effect)
They also fit on two tied TO-92 type transistors.
Don't know if they still are available. I will search for them.
You can also use heatshrinkable tube.
Greetings, Loek
Hi Pooge,
actually I did way back to check what effect an open input would have
A basic check, but it does not provide much useful evidence on DC conditions nor noise performance, since a low source impedance is usually connected.
Re: a specialized heat sink
Wow! That picture reminds me of my first Leach with the ground plane. I made the board myself. What a pain that was, to line up everything on both sides of the board without all the modern tools. I couldn't believe it when it worked the first time. It worked for many years until I did something stupid by pulling RCAs while it was on. I was ready to upgrade, anyway.
I have some of the star heat sinks. I'll have to look into that possibility. Don't know if I'd use heat shrink, though. Even though they are not hot transistors, I wouldn't choose to insulate around them. I think it would be better to bond the flat sides together with adhesive, preferably thermally conductive. (Anyone know of any thermally conductive adhesive in a small tube?) Haven't studied the lead twisting issue on this one, because I haven't settled on transistor selection yet.
loek said:
Wow! That picture reminds me of my first Leach with the ground plane. I made the board myself. What a pain that was, to line up everything on both sides of the board without all the modern tools. I couldn't believe it when it worked the first time. It worked for many years until I did something stupid by pulling RCAs while it was on. I was ready to upgrade, anyway.
I have some of the star heat sinks. I'll have to look into that possibility. Don't know if I'd use heat shrink, though. Even though they are not hot transistors, I wouldn't choose to insulate around them. I think it would be better to bond the flat sides together with adhesive, preferably thermally conductive. (Anyone know of any thermally conductive adhesive in a small tube?) Haven't studied the lead twisting issue on this one, because I haven't settled on transistor selection yet.
in addition to using the mentioned "star heat sink" for keepinjg TO92 parts somewhat in thermal equilibrium, i have also wrapped both parts together in a loop or two of thick copper wire with a VERY small amount of heat sink grease between them.
good luck!
mlloyd1
good luck!
mlloyd1
Hi,
sometimes you can turn the pair of To92 through 90 degrees to face flat sides together. Then bond them with instant glue. Very thin=low thermal resistance.
I have not tried this on the Jen's board but it worked on the GB150 SKA pcb. Leave the To92 legs a bit longer and add very thin insulation (stripped from 0.6mm wire) may give sufficient flexibilty to achieve this.
Four turns of thick copper wire around the pair with longish tails gives a bit of extra dissipation capacity.
sometimes you can turn the pair of To92 through 90 degrees to face flat sides together. Then bond them with instant glue. Very thin=low thermal resistance.
I have not tried this on the Jen's board but it worked on the GB150 SKA pcb. Leave the To92 legs a bit longer and add very thin insulation (stripped from 0.6mm wire) may give sufficient flexibilty to achieve this.
Four turns of thick copper wire around the pair with longish tails gives a bit of extra dissipation capacity.
Been re-reading old amplifier papers since Cordell's entry into this forum, in preparation for building this amp, and I'm have a major brain dump. Leach says his amplifier has a GBP of 8.5 MHz, while his specified output transistors have an ft of 2MHz. I thought that ft was the maximum bandwidth, i.e., at a gain of unity. Therefore, how you can get a GBP of 8.5 is not registering with me right now. What am I missing?
Hi,
too good a question.
I could be wrong, so we'll be waiting for a correct explanation, but here's mine.
The fT of the output device is for when it is producing gain, however Leach uses it as a follower (no gain, well just less than 1.0).
The preceding circuit produces all the gain.
The combination of the follower and the voltage amplifier has a GBW product of 8.5MHz, so it's this combination that has a gain of 1 @ 8.5MHz and rises @ 20db/decade as frequency falls. i.e. gain =100 @ 85kHz.
Finally, the open loop gain is corrected by wrapping the NFB loop around the whole thing to reduce the closed loop gain to 23times (22k/1k + 1=23).
too good a question.
I could be wrong, so we'll be waiting for a correct explanation, but here's mine.
The fT of the output device is for when it is producing gain, however Leach uses it as a follower (no gain, well just less than 1.0).
The preceding circuit produces all the gain.
The combination of the follower and the voltage amplifier has a GBW product of 8.5MHz, so it's this combination that has a gain of 1 @ 8.5MHz and rises @ 20db/decade as frequency falls. i.e. gain =100 @ 85kHz.
Finally, the open loop gain is corrected by wrapping the NFB loop around the whole thing to reduce the closed loop gain to 23times (22k/1k + 1=23).
- Status
- Not open for further replies.