What is offset ?? 🙄 I don't have them problems. My "worst amps" are 5mv . The "supers" are under .5mV. Wherever the "CCS ADJ." is adjusted to has absolutely NO effect on offset. Theoretically.. any offset , if it existed , would be an interaction between OPS and speaker (warm voicecoil ??)
OS
My listening tests show that output offset does have a sonic impact; the closer it is to zero volts and without drift, the better the resolution and a quieter background; however, when this is achieved, the effect of a balanced LTP on low level resolution and speed is more pronounced and the musicality goes up a couple of notches.
Interesting. Ostripper, will your boards work with matched dual input transistors? I would use MAT0X, THATxxx, LS32xx...
- keantoken
- keantoken
Very interesting question keantoken... something I wondered over too.
Obviously only the LF unit would be effected in most set ups.
http://www.diyaudio.com/forums/solid-state/116731-can-d-c-offset-cause-non-linear-distortion.html
I've heard that DC offset can decrease distortion of come caps, which may effect crossover'd speakers, but I don't know if it's anything that would be audible.
- keantoken
- keantoken
I just use my DMM B-E test on a lot of 10- ksc1845's ,find at least 2 pairs that are very close (.717 / .719 for example) put them in with my offset trimmer centered.. end up at 6-12 mv .. a slight turn of the offset ADJ. , easy to get that down to under 2mv.
Just tested my "baby's" after 16 months .. still just 4-5mV offset. I don't like to use any proprietary duals , just 7 cent mouser fairchild's. I am CHEAP! 😀
OS
keantoken
This soundstage effect was first discussed, then verified by Os back in the Current Mirror thread well over a year ago. Luckily , G.K. isn't around any more to poo poo the subjective findings.
The LS313 and LS352 duals are ideal for use in preamps and lower supply rail poweramps. Sometimes a little hard to obtain though.
Alex
Last edited:
I thought they had newer ones, like LS3250? But never mind, I remember now there's no place on the planet you can get those. Vapor-state electronics!
I don't doubt Os's findings, but I want to make sure.
- keantoken
I don't doubt Os's findings, but I want to make sure.
- keantoken
kt
I first reported these findings about VERY close balance of the LTP to Electronics Australia in 1987 , then again to Silicon Chip magazine in 1989. As is so often the case , the EE's came out in force, much like they do here quite often, to dismiss subjective findings.I also exchanged emails with Hugh on this subject quite some time back.With a fully differential topology you also need very good balance of the 2nd pair,as it affects the close balance of the input LTP due to thermal considerations.This was very evident in the ETI5000 by David Tilbrook from much earlier. Suzy J tested an old ETI 5000
(>25 years old !) that I had modified to overcome this, and had been recovered from a friend's garage. She remarked on how good the distortion figures were on the channel where the electros hadn't dried out, and asked if I had access to distortion measuring gear. I hadn't.
Alex
Hello keantoken
It's not related to the offset voltage but with the transistors linearity, the linear region of a transistor are very narrow, wen the LTP are unbalanced; one transistor will pass more current and there will be a higher Vbe, but the other transistor will pass less current and the Vbe will go down, so the global negative feedback will be unbalanced and less precise, so an LTP unbalance will cause more distortions.
I can say that matching VBE and gain of the LTP pair and balancing them give a real boost of the sonic and soundstage quality of any amps.
Bye
Gaetan
Last edited:
Agree with Gaetan. On all my amps, kits included, I match the input stages wherever I use an LTP. It does make a big difference.
Interesting (as usual),guys!
Would someone like to propose a mechanism (yes, I'll play the objektivist for a bit) for the the sound stage to be influenced by the static balance and suggested linearity range of the input pair?
This isn't just a mouthful of a question, I think there's a book, manual and novel in here too, if I'm not mistaken! After all, it's one thing to make a claim, another to concur for more than aesthetic or social principles and further, to make product changes to conform with what may only appear to be so.
The fact that that you may be considering linearity as an important part of high-end or perhaps "desirable" Audio is surely problematic. Most - well, many including me -consider selective non-linearity as highly beneficial.
bon apetit.
Would someone like to propose a mechanism (yes, I'll play the objektivist for a bit) for the the sound stage to be influenced by the static balance and suggested linearity range of the input pair?
This isn't just a mouthful of a question, I think there's a book, manual and novel in here too, if I'm not mistaken! After all, it's one thing to make a claim, another to concur for more than aesthetic or social principles and further, to make product changes to conform with what may only appear to be so.
The fact that that you may be considering linearity as an important part of high-end or perhaps "desirable" Audio is surely problematic. Most - well, many including me -consider selective non-linearity as highly beneficial.
bon apetit.
Ian Finch
Some time back, in a personal communication, Hugh gave me what appeared to me , to be a very plausible explanation. Unfortunately, my PC was hit by a virus a while back, and I no longer have Hugh's reply.
Perhaps Hugh can recall his explanation ?
Alex
Perhaps balance can be seen as an impedance match between the two transistors. Impedance also considers frequency response.
Also consider a notch filter. Say that the notch is where the transistors are balanced, and the offset and signal control where we fall on the X axis. Signals only briefly pass through the balanced section. But when balanced properly, all signal start at the bottom of the notch, where there is the best linearity (and also the best speed). Consider then that all waveforms start at 0 offset. If our spatial sense is tuned to the time difference of abruptly starting sounds, a balanced LTP makes sense because all these sounds will start in the most linear region. Also consider that reverberated sounds are never as loud as the music, they are lower volume and perhaps can fit better into the linear region.
Now consider imbalance. The signal starts low, out of the spike. As it grows, one half of the wave sees decreasing sensitivity and the other half sees an increasing sensitivity towards the notch - the sensitivities cancel out sort of. Then the signal grows past the notch, and then sees a more quickly decreasing sensitivity. So with an unbalanced LTP, the sensitivity or feedback factor shows not a smooth curve but a slow curve then a knee.
Those are my thoughts for the moment.
- keantoken
Also consider a notch filter. Say that the notch is where the transistors are balanced, and the offset and signal control where we fall on the X axis. Signals only briefly pass through the balanced section. But when balanced properly, all signal start at the bottom of the notch, where there is the best linearity (and also the best speed). Consider then that all waveforms start at 0 offset. If our spatial sense is tuned to the time difference of abruptly starting sounds, a balanced LTP makes sense because all these sounds will start in the most linear region. Also consider that reverberated sounds are never as loud as the music, they are lower volume and perhaps can fit better into the linear region.
Now consider imbalance. The signal starts low, out of the spike. As it grows, one half of the wave sees decreasing sensitivity and the other half sees an increasing sensitivity towards the notch - the sensitivities cancel out sort of. Then the signal grows past the notch, and then sees a more quickly decreasing sensitivity. So with an unbalanced LTP, the sensitivity or feedback factor shows not a smooth curve but a slow curve then a knee.
Those are my thoughts for the moment.
- keantoken
O 'ell, Sandy, this could be difficult. Kean's explanation sounds pretty good.
I think it went like this: Any non-linearity in the input stage, which extracts the error signal from the feedback signal, will affect the amp profoundly. As the amplitude of the input increases (this applies particularly to bass signals, which are BIG), the voltage differential across the inverting and non-inverting input grows, as a function of the output amplitude divided by the open loop gain. It might be, for example, 1mV.
This of course means one LTP device is on harder than the other. The problem is that the reference for these two signals is not taken from the base at all, but from their conjoined emitters (assume no degeneration). Consequently, the Vbe of each device is important. It doesn't take a genius to see that if one device passes 1.2mA, for example, and the other 0.8mA (for 2mA stage current) then the Vbes will vary by an amount on the order of the differential voltage across the bases. This means that the output of the LTP, the error signal if you will, will not be particularly accurate.
In essence, this means that the transfer function of the LTP is S shaped, flattening out with increasing differential input. This is a serious threat to linearity, particularly at high amplitude. The solution is to either find a way to prevent Vbe from varying on the LTP devices, OR increase OLG to a point where the differential input across the LTP is very, very small, keeping the LTP close to balanced even at high inputs. You could use a CFP front end, as I do in one of my amps, OR you could use and additional gain stage, but this tends to have serious implications for stability since you must bring the OLG to heel by the HF pole, and this is tricky.
Hope this is what you meant, Sandy?
Hugh
I think it went like this: Any non-linearity in the input stage, which extracts the error signal from the feedback signal, will affect the amp profoundly. As the amplitude of the input increases (this applies particularly to bass signals, which are BIG), the voltage differential across the inverting and non-inverting input grows, as a function of the output amplitude divided by the open loop gain. It might be, for example, 1mV.
This of course means one LTP device is on harder than the other. The problem is that the reference for these two signals is not taken from the base at all, but from their conjoined emitters (assume no degeneration). Consequently, the Vbe of each device is important. It doesn't take a genius to see that if one device passes 1.2mA, for example, and the other 0.8mA (for 2mA stage current) then the Vbes will vary by an amount on the order of the differential voltage across the bases. This means that the output of the LTP, the error signal if you will, will not be particularly accurate.
In essence, this means that the transfer function of the LTP is S shaped, flattening out with increasing differential input. This is a serious threat to linearity, particularly at high amplitude. The solution is to either find a way to prevent Vbe from varying on the LTP devices, OR increase OLG to a point where the differential input across the LTP is very, very small, keeping the LTP close to balanced even at high inputs. You could use a CFP front end, as I do in one of my amps, OR you could use and additional gain stage, but this tends to have serious implications for stability since you must bring the OLG to heel by the HF pole, and this is tricky.
Hope this is what you meant, Sandy?
Hugh
Hi Hugh
Yes, you did give an explanation about the S shape previously, but not in as much depth as now.
Regards
Alex
Good idea, Hugh. I will see what this gets me. Thanks.
I think Carlos does this on his Blame ST (bootstrapped EF VAS) , he uses a "hard CCS" and a CM on the tail , which alters harmonics in a bad way (lower THD , but too much H3)
OS
Pete,
I really don't like the idea of more gain. I could well be wrong, but each gain stage brings phase shift, and I try to get phase shift to less than 3 degrees at 100KHz. I find less phase shift makes compensation easier, and somehow helps imaging.
It was done in the eighties by Professor David Tilbrook, who did the ETI5000 and ETI6000amps. Suzy Jackson has done a variant of the later amp which has extremely good specs. BUT, I'm not convinced that very low THD correlates well to good sound, chiefly because of the harmonic content, which with lots of gnfb extends out to deep space.
Good luck!
Hugh
I really don't like the idea of more gain. I could well be wrong, but each gain stage brings phase shift, and I try to get phase shift to less than 3 degrees at 100KHz. I find less phase shift makes compensation easier, and somehow helps imaging.
It was done in the eighties by Professor David Tilbrook, who did the ETI5000 and ETI6000amps. Suzy Jackson has done a variant of the later amp which has extremely good specs. BUT, I'm not convinced that very low THD correlates well to good sound, chiefly because of the harmonic content, which with lots of gnfb extends out to deep space.
Good luck!
Hugh
You are right , I should be happy with my 48 DB . Deviate too far away from the "baby" and it no longer is the "baby". As it is now, it mimics the harmonic content nearly exactly with lower distortion (still ,way more than the "blameless") and has a little more gain. It is what it is... 🙂
OS
Yes, and we might marvel at the sensitivity of the ear which, with a bit of training, can pick up these subtleties...... Even very low thd amps have nasties in them.
I gravitate more and more to Earle Geddes approach about distortion.
Cheers,
Hugh
I gravitate more and more to Earle Geddes approach about distortion.
Cheers,
Hugh