Simple discrete unity gain buffer

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
My friend Angshu designed this buffer for me, when I wanted a very simple unity-gain discrete buffer. His simulation runs show THD at -107dB with a load impedance of 10K. And PSRR seems to be -49.5dB on the positive rail and -53dB on the negative.

What do you guys think of this circuit? Angshu is quite a veteran of transistor based circuits, so I believe this will work pretty much as the sims show; neither of us has breadboarded it yet. I'll probably be using something like a BC550 instead of the BC337 he's used. I guess any high-gain low noise transistor will do just fine. C3 and C1 are for DC blocking. The output will have a DC offset of about -0.7V without C1. This is not a problem for me, because I always put input caps in the signal paths of my power amps anyway. And yes, I'll put a decent regulated PSU for this buffer; just bridge plus caps may not do, since PSRR is way worse than opamps.

Quite often, line-level preamps don't need any gain. Putting a volume control pot before this buffer should be enough to build an active preamp which can drive fairly long interconnects (10 feet?). Many of us need simple buffer-based preamps for many projects. And I need longish interconnects because I am building active loudspeakers, where each speaker has an active xo plus a set of power amps, in a chassis placed adjacent to each speaker. Thus, my interconnects need to go from my preamp (in a central rack) to each speaker. I am not sure passive preamps are a good idea with such long interconnects.

What do you guys think of this buffer instead of an opamp-based buffer? What should I be careful about?

Thanks,
Tarun

PS: I also posted a query about a remote controlled volume control design in another thread. Comments please??
 

Attachments

  • buffer.png
    buffer.png
    63.8 KB · Views: 5,336
Looks good to me.

Perhaps a 470R resistor between the base of the current sink and the LED, increase the cap to 100uF, but nothing more.

Purists might comment adversely on two coupling caps; you could run it without the input cap if the source was safe, but good caps will see you right.

Should easily produce 25Vpp with a Zout around 26/Ic; not clear what current you are running as the LED voltage is not clear. I would assume around 4-6mA.

Current delivery to load will be high on positive half cycle, but limited by the CCS on negative half cycle. There will be highish H2 at high loading.

Cheers,

Hugh
 
AX tech editor
Joined 2002
Paid Member
I agree with Hugh; since you are running two power supplies, the input coupling cap is superfluous and can be left out. On the other hand, if you think you need it because you may have sources with DC, then you could leave out one of the power supplies without adverse effects.

Beware: simulation THD results very seldom have any relationship to real-world results, they are a trend indicator at best. Noting that this really is an emitter follower, how does the sim results change with say 1000 to 3000pF parallel to the load to take care of the cable cap?

Jan Didden
 
AKSA said:
Perhaps a 470R resistor between the base of the current sink and the LED, increase the cap to 100uF, but nothing more.
Will do. Noted.
Purists might comment adversely on two coupling caps; you could run it without the input cap if the source was safe, but good caps will see you right.
Yes, I'll leave it out, almost certainly. I just kept it in the schematic because that's how my friend had simulated it.
Should easily produce 25Vpp with a Zout around 26/Ic; not clear what current you are running as the LED voltage is not clear. I would assume around 4-6mA.
I'm stuck on this question; Angshu is out of town. Once he's back, I'll ask him to fill in the blanks. But I think the current will probably be around the range you quoted.
Current delivery to load will be high on positive half cycle, but limited by the CCS on negative half cycle. There will be highish H2 at high loading.
What I'm really curious to know is how it'll behave with a load impedance made up of, say, 20K resistive plus the impedance of the 10-feet-long interconnects. I will probably be using RG58 or RG59 coax cable used for Arcnet or 10base5 Ethernet as the interconnect. What will that load be like? Does this qualify as "high loading"?

janneman said:
I agree with Hugh; since you are running two power supplies, the input coupling cap is superfluous and can be left out.[/quote
Yes, I'll do that.
Beware: simulation THD results very seldom have any relationship to real-world results, they are a trend indicator at best. Noting that this really is an emitter follower, how does the sim results change with say 1000 to 3000pF parallel to the load to take care of the cable cap?
Angshu has the simulator. Will try that as soon as he returns, and will post results. What L and C values should one take to model longish interconnects? Will I actually have C as high as 3nF if I use video-grade coax? Wow. If these are the figures, how do they carry Ethernet at 10MHz over 100 metres?
And overall, what do you guys think about the quality of this as an output bufffer, compared to a decent opamp (say, NE5532/OPA2134) used as unity-gain inverting amp? I mean, what, if any, will be the audible differences?

Thanks for the help, Hugh and Jan. :)
Tarun
 
AX tech editor
Joined 2002
Paid Member
tcpip said:
[snip]. What L and C values should one take to model longish interconnects? Will I actually have C as high as 3nF if I use video-grade coax? Wow. If these are the figures, how do they carry Ethernet at 10MHz over 100 metres?
And overall, what do you guys think about the quality of this as an output bufffer, compared to a decent opamp (say, NE5532/OPA2134) used as unity-gain inverting amp? I mean, what, if any, will be the audible differences?Tarun


Tarun,

I had some interconnect that were 300pF/m, not that much really. For 10m you get 3000pF. I don't know about the video stuff, maybe you can measure a length with a cap meter? Or contruct a filter with a series R and the cable, then find the half-voltage freq, R = 1/(2. pie.f.C), and calculate the C.

Difference with an opamp: I don't know about the audilel difference, because that depends on a lot of other factors in your system. The measurable differences will be very large.

Jan Didden
 
Cable capacitance

Capacitance of RG-59 is typically 22pF per foot and general purpose shielded cable ( local brands ) for microphones is about 40 pF per foot.
I think Tarun's cable is 10 feet not 10 meters. So at 220pF or 400pF it is pretty OK.

All RG-59's are not the same. Especially at DC. The core conductor can be single core copper or copper coated steel or multistrand copper.
The braid can be stranded aluminum or copper , single layer or multiple layers or aluminum foil plus strands . Quite different from each other.
Additionally the insulator varies from foam to solid and the insulator material varies with the brand. So one cannot use the term RG-59 by itself. You will have to mention the type and brand name.
The same applies for other shielded cables also. Can lead to A LOT of experimenting.

As a line level interconnect the cables with low capacitance will avoid early rolloff . I am not quite sure how the dc characteristic will affect this. As a line level interconnect this should not affect the low end. But this will surely be important if used as a speaker interconnect - as some fans do. LF performance could drop . I see people mention this effect in their set up's .

With a Zout of 1K ohm ( yes ,that's very high ) and a load of 10 K Ohm ( 20 K volume pot ) and cable capacitance of 400pF , the roll off is about -0.1dB at 70 Khz. With a Z out of 100 ohms it will be -0.1db at 700Khz. With a 10 K volume pot the frequencies will be half of the above.

I think Tarun's interlink is quite OK on paper. BUT he will have to try several types to determine how they 'sound'. That appears to be a grey area and needs practical tests - no (known ?) sim can give the correct answer. Don't forget to use good connectors !
Cheers.
 
Re: Cable capacitance

Thanks, Ashok. Good to see you in this thread. :)

ashok said:
Capacitance of RG-59 is typically 22pF per foot and general purpose shielded cable ( local brands ) for microphones is about 40 pF per foot.
I think Tarun's cable is 10 feet not 10 meters. So at 220pF or 400pF it is pretty OK.
Correct. 15 feet (5 metres) max.

All RG-59's are not the same.
Thanks for the enlightening inputs. :) And I am picking up something which has stranded copper core and braided (no foil) shield. Don't know about the insulator material. Any brand of cable I should look for?

As a line level interconnect the cables with low capacitance will avoid early rolloff.
That's what I was thinking. Call me ignorant, but I was under the impression that cable-to-cable differences in RG58/59 would be insignificant for audio frequencies, and the capacitance issues would be inaudible if I drove them using a low-Zout buffer stage. The volume control and other controls, if at all, would come before the buffer, not after. So, at least in my case, the cable will see a low source impedance, and feed directly to a power amp of, say 10KOhms Zin. In case the inputs of multiple power amps are connected in parallel and fed from the same interconnect (active speakers, one amp per driver for multiple midbass units), then the Zin of the load may be lower.

So, driving a cable of 15 feet from a low-Zout buffer of the type I've drawn, and feeding a load of 10K or so, should be safe at audio frequencies. Or so I thought! :confused:

I think Tarun's interlink is quite OK on paper. BUT he will have to try several types to determine how they 'sound'. That appears to be a grey area and needs practical tests - no (known ?) sim can give the correct answer. Don't forget to use good connectors!
Thanks. Will try. :) Regarding RCA plugs, my current favourite are some connectors with the "Monster" label, hefty and very tight-fitting. I'll continue to use those till someone teaches me better.

Tarun
 
Jan,
janneman said:
I had some interconnect that were 300pF/m, not that much really. For 10m you get 3000pF.
I think Ashok has explained the metre-vs-feet issue. But it'll be interesting to measure a length of cable with a cap meter anyway. I'll do it sometime. My cap meter does not measure smaller than 1nF, so I'll need quite a bit of cable first.

Difference with an opamp: I don't know about the audilel difference, because that depends on a lot of other factors in your system. The measurable differences will be very large.
Can you please elaborate on the measurable differences? I'm totally clueless, so any comments, however basic, will be very useful. :) I can think of only two clear differences so far: the DC offset at the output, and the poor PSRR, of my buffer.

And Peranders:
... diamond buffer...
This is the first time I'm hearing the term. I'll certainly look it up.

Thanks.
Tarun
 
AX tech editor
Joined 2002
Paid Member
tcpip said:
[snip]Can you please elaborate on the measurable differences? I'm totally clueless, so any comments, however basic, will be very useful. :) I can think of only two clear differences so far: the DC offset at the output, and the poor PSRR, of my buffer.[snip].
Tarun

Oh yes, DC offset will be MUCH smaller (you can leave out the output coupling cap), the PSRR will be between 100-1000 times better, also the output impedance of the opamp buffer will be between 1000 and 10.000 times smaller, the distortion will be 10 - 100 times smaller. The opamp will also be more flat in freq response and a some other things. Still, if your buffer sounds OK to you, go for it.
I think it is somewhat misleading to call your circuit a buffer. It is not untrue, but it is better characterised as an emitter follower.

Jan Didden
 
janneman said:
Oh yes, DC offset will be MUCH smaller (you can leave out the output coupling cap), the PSRR will be between 100-1000 times better, also the output impedance of the opamp buffer will be between 1000 and 10.000 times smaller, the distortion will be 10 - 100 times smaller. The opamp will also be more flat in freq response and a some other things.
Thanks! Very interesting. I didn't know there would be such a huge difference in Zout, or the distortion. I thought most opamps (other than the top-end newer ones) have distortion figures worse than the 107dB that Angshu's simulation of this circuit had reported. So the distortion bit too is a surprise to me.

And come to think of it, if the primary purpose of my simple buffer is to drive long interconnects with low Zout, and an opamp will have much lower Zout than this circuit, I guess I may as well revert to an NE5532-based inverting buffer. :)

I'm not trolling for flames here, but if this circuit indeed is beaten so thoroughly by any decent opamp, why don't people simply use opamps to make preamps? Why do people bother to make discrete preamps at all these days? I was under the impression that simple discrete circuits can have inaudible levels of distortion, hence the continuing romance with discrete circuits. Now I'm confused. Am I then to understand that discrete preamps is a "taste" thing, like valve amps with their 2% distortion?

I read Per-Ander's thread about diamond buffers, and some other related threads. I can see discrete circuits are delivering 120dB THD or better. But if I can get similar figures, together with very low Zout and good current driving ability, using opamp chips, then why bother with those discrete buffers? Is the pumping of 2V signals into 150 Ohms the only reason to build discrete buffers? (That's the kind of figure I saw in Walt Jung's discrete-buffer article that Per-Anders referred to.) If yes, then I don't need discrete buffers for my work... I can do with NE5532s.

I think it is somewhat misleading to call your circuit a buffer. It is not untrue, but it is better characterised as an emitter follower.
Sorry, I don't know the difference. I thought a buffer is any circuit which has high Zin and low Zout, and (typically) does not have any voltage gain. Doesn't this circuit qualify? Or is my definition wrong?

Thanks again.
Tarun
 
AX tech editor
Joined 2002
Paid Member
tcpip said:
[snip]I'm not trolling for flames here, but if this circuit indeed is beaten so thoroughly by any decent opamp, why don't people simply use opamps to make preamps? Why do people bother to make discrete preamps at all these days? I was under the impression that simple discrete circuits can have inaudible levels of distortion, hence the continuing romance with discrete circuits. Now I'm confused. Am I then to understand that discrete preamps is a "taste" thing, like valve amps with their 2% distortion?[snip]
Tarun

Well, I don't want to start any war either but you yourself said you wanted a simple buffer. That's what you got. You didn't ask for a high precision, wideband, low distortion buffer.

And you asked for the measurable differences, which is what I gave you.

As I said, I don't know the audible differences, if there are any. Your circuit may well have inaudible distortion, depending on the rest of the system. Unfortunately, there seems to be no agreement on what is "unaudible distortion". Some people report that the simple circuits sound "more musical". Again, I haven't seen any definition as to what that is. You really are on your own here, and that is good, because you are the one who has to listen to it. If you like what you hear, why bother with those 1000 times lower THD of the opamp?
But, you asked the question! Don't ask it if there is a chance that you don't like the answer!

Jan Didden
 
Dear Per-Anders,

peranders said:
You could also consider a diamond buffer of some kind. We have discussed it here quite much lately. A diamond buffer is real easy to get going and you wont need a output cap.
Saw your stuff on diamond buffers. Also saw snaps of your PCB with SMT components. When, oh when, will I be able to solder SMT bits correctly??? Sigh... :D

One basic thing I didn't understand was: what is the extra that I get from these super-low-distortion circuits, compared to, say, one of the input+VAS stage circuits that Randy Slone gives in his audio power amp book? His "very good" topology has a differential input with CCS and current mirror, driving a pair of medium power transistors (2SB649/2SD669). He adjusts output voltage to about 33V RMS, and runs from 50V symmetrical rails. And gets 2nd harmonic at 20Khz of 0.0027%, and 3rd harmonic of 0.0002% at 20KHz, and 0.0012% at 50KHz.

And what he considers his "excellent" topology is fully mirror-image, and the input stage drives Darlington VA stages with cascode loading. (I say "stages" because there are two of them, the design being fully mirror-image.) With a 1KOhm load, he's getting 2nd harmonic distortion of 0.0000003% (six zeros and a three) at 50KHz, and 3rd harmonic distortion of 0.00000006% (seven zeros and a six) at 50KHz. His final-stage devices are the 2SB and 2SD ones, as before. The rail voltages and signal amplitude is as in the previous case.

The only problem with the super-duper topology is the component count. I can see 18 transistors, 23 resistors, seven diodes. But then some members have remarked about the component count of the diamond buffer too.

Does the diamond buffer give me something which topologies like Randy Slone's VAS designs don't give me? You must understand that my questions are very primitive, because I know very little. I am just picking up pieces of information from various places and comparing them, in order to understand stuff better. Please don't take my questions as any veiled comment on one or other approach.

Thanks for the help.
Tarun
 
Discrete and opamp chips and all that...

janneman said:
Well, I don't want to start any war either but you yourself said you wanted a simple buffer. That's what you got. You didn't ask for a high precision, wideband, low distortion buffer.
And you asked for the measurable differences, which is what I gave you.
....But, you asked the question! Don't ask it if there is a chance that you don't like the answer!
Thanks a lot for the insight. Your answers have confirmed many of my suspicions.

However, it appears I may have upset you a bit. Please accept my apologies; that was never my intention. Your answers have, in fact, been exactly what I've asked for, and they were very helpful.

I too have heard opinions along the lines of "less components make a more musical amp." Don't know what it means. I listen to a system which has an NE5532-based preamp, every day. I love it. I guess with time I'll learn about the higher levels of musicality. :) It is possible that I may never understand it, considering that one of my gurus, Randy Slone, probably does not understand it yet either, after spending more than two decades in serious audio electronics R&D and writing a few books.

Thanks a lot. :)
Tarun
 
AX tech editor
Joined 2002
Paid Member
Re: Discrete and opamp chips and all that...

tcpip said:
[snip]However, it appears I may have upset you a bit. Please accept my apologies; that was never my intention. [snip]Tarun

Not at all! I respect you, because you are doing the hard thinking and discovering of things, and you are willing to question your own beliefs. That for me is the only way to progress. You're moving about 40dB faster than the average DIYer on this forum. (now I have upset some people I'm sure).

Jan Didden
 
Re: Discrete and opamp chips and all that...

tcpip said:
"less components make a more musical amp."
This _may_ be true sometimes but more often it's the opposite! Some people count parts rather than elements. A Gaincard doesn't consist of 9 parts (or whatever). Count all those transistors inside the LM3875 IC. I'll guess this IC has 50-100 "transistors.

How good an amp is, is determined by the whole solution, from circuit design down to choice of components and pcb layout wiring etc.
 
Tarun,

Jan has given you reasonable factors of improvement over your simple emitter follower if you use a modern opamp. Of course, these figures are appealing.

Like most of us, you have recourse to PSpice to tell you the operational parameters, and Jan's estimates are readily confirmable.

But nagging away in your mind is the notion that 'simpler is better'.

Why do you think that is, and how might you verify it? Given that audio is based so strongly on subjectives, and there's a lot of folklore, what are your options?

Given that you have made a seminal comment about Randy Slone's two decades of experience, and likely ignorance about a few issues, it should be obvious.

You must build it. Then you must listen to it, preferably AB against your beloved 5534 preamp. Then you should invite in friends, and perform the same test on them.

After a couple of heavy listening tests, confirmed by others who understand roughly what to listen for, your hunches will be confirmed one way or the other. You will KNOW!! And you may find that the vanishingly small distortion figures and high specs of opamps are ambivalent; they don't really complete the picture, and in some ways are inconclusive.

If you do all this, you will be assessing your 'product' as the market does; 'How does it sound?'. It is the answer to this question which sells high end. People mostly insist on an audition, so in their heart of hearts they don't much care for the specs.

Hmmm. Wonder why that is?

Cheers,

Hugh
 
I have tried different kinds of follower circuits, measured and tested sonically. The simple follower shown hereabove is only simple, that's all. Good as a starting project for newbies, but do not expect any miracles. Peranders is right, better give a try to some kind of diamond buffer in case that you want to achieve excellent results.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.