I again hate to say this, and I will probably be censored for it, but the specific value of Re, as Bob Cordell and I both know is very important to give a smooth transition from CLASS A to CLASS B which happens at about 25W into 8 ohms in the JC-1. The optimum value TOTAL is between 1/2Gm and Gm of each output device used at its idle current. Therefore more idle current used, less Re is optimum.
This was first pointed out in 1971 in a HP journal by Bernard Oliver then VP of HP. He solved the problem with a computer simulation AND noted that without extra care taken, an optimum Re might not be very stable.
I personally have used Re's of .05 ohms and maintained thermal stability with my thermal tracking system since the early 70's. 0.12 ohms would have been ideal for the JC-1.
We experimented with 0.1 ohms in HCA 3500 prototypes with success. However, we quickly moved up to 0.15 ohms in the JC-1 after a few potential problems were noted.
In an ideal world, 0.1 ohms would have been better for the overall transfer function due to the added base spreading resistance denoted by Bob Cordell and especially because of the 10 ohm base resistance added by me. Bob is right. A lower base resistance would be better, BUT how low can I make it without blowing up a whole series of amps due to difficult loading conditions?
5ohms, 3 ohms? I really don't know and neither Bob or anyone else has shown me how to derive the exact value. This is my dilemma and I am sticking to 10 ohms until further notice. I sure would like to use 5 ohms, however.
This was first pointed out in 1971 in a HP journal by Bernard Oliver then VP of HP. He solved the problem with a computer simulation AND noted that without extra care taken, an optimum Re might not be very stable.
I personally have used Re's of .05 ohms and maintained thermal stability with my thermal tracking system since the early 70's. 0.12 ohms would have been ideal for the JC-1.
We experimented with 0.1 ohms in HCA 3500 prototypes with success. However, we quickly moved up to 0.15 ohms in the JC-1 after a few potential problems were noted.
In an ideal world, 0.1 ohms would have been better for the overall transfer function due to the added base spreading resistance denoted by Bob Cordell and especially because of the 10 ohm base resistance added by me. Bob is right. A lower base resistance would be better, BUT how low can I make it without blowing up a whole series of amps due to difficult loading conditions?
5ohms, 3 ohms? I really don't know and neither Bob or anyone else has shown me how to derive the exact value. This is my dilemma and I am sticking to 10 ohms until further notice. I sure would like to use 5 ohms, however.
Hi John,
I see no problems with using low value emitter resistors as long as the bias control circuit is up to the challenge. Current sharing would seem to be your biggest problem in a production environment, unless you specify that these parts are pre-screened by the manufacturer. At this point, you can even specify a different package marking to differentiate them from other, similar parts. Of course, you do know that this was a very common practice 30 to 40 years ago. Almost every manufacturer used their own special part numbers to protect what they saw as intellectual property.
The well known Carver brand often used emitter resistors of 0.05 ohms. They were extremely reliable as well, proving that bias control was possible at these values. He even pulled this off with a couple amps that used two pairs of outputs in parallel. Not too shabby!
So, have you not tried going to 4R7? Others have done this, it's not that uncommon in the industry. I've also seen 2R2 as well. 10 R base resistors are not common anymore, I can't think of any amp I've seen in the last 15 years with that high a value.
Your remark is not offensive, nor is it pushing people to attempt something that is dangerous.
You want to see dangerous? Look at what they did at Counterpoint! 2 to 4 pairs of selected mosfets running in parallel with zero additional source resistance of any kind. Of course, this bit him (Michael) in the rear end eventually. These amps were not stable for bias, and the outputs were not selected tightly enough to do this. I know because I have had to select sets by hand, really tightly. Takes me about two days using jigs I made up. Anyway, the result was company failure. This wasn't the only reason, but it sure didn't help any.
-Chris
I see no problems with using low value emitter resistors as long as the bias control circuit is up to the challenge. Current sharing would seem to be your biggest problem in a production environment, unless you specify that these parts are pre-screened by the manufacturer. At this point, you can even specify a different package marking to differentiate them from other, similar parts. Of course, you do know that this was a very common practice 30 to 40 years ago. Almost every manufacturer used their own special part numbers to protect what they saw as intellectual property.
The well known Carver brand often used emitter resistors of 0.05 ohms. They were extremely reliable as well, proving that bias control was possible at these values. He even pulled this off with a couple amps that used two pairs of outputs in parallel. Not too shabby!
So, have you not tried going to 4R7? Others have done this, it's not that uncommon in the industry. I've also seen 2R2 as well. 10 R base resistors are not common anymore, I can't think of any amp I've seen in the last 15 years with that high a value.
Why?
Your remark is not offensive, nor is it pushing people to attempt something that is dangerous.
You want to see dangerous? Look at what they did at Counterpoint! 2 to 4 pairs of selected mosfets running in parallel with zero additional source resistance of any kind. Of course, this bit him (Michael) in the rear end eventually. These amps were not stable for bias, and the outputs were not selected tightly enough to do this. I know because I have had to select sets by hand, really tightly. Takes me about two days using jigs I made up. Anyway, the result was company failure. This wasn't the only reason, but it sure didn't help any.
-Chris
anatech said:
Methinks the IEEE barrier to information is much worse.
That's not an organisation by the engineers, for the engineers.
Even if you are a member, but not of the right sub-sub-group,
2 pages can cost you a lot of money.
With books, the authors don't get much royalties unless the number of prints
is pretty high. Could be even negative for a few 100.
My brother is head of a publishing company and I avoid that topic.
Gerhard
I find that with too low of an Re on a multi-output (pick a number over 2) design that the current sharing (as issues can go) as a barely measurable 'artifact' comes out as slurred/smeared transients and loss of micro-detail. (detail that is buried under a loss of transient definition, a transient that takes far more time to occur, with loss of definition of said transient) This ends up sounding/creating a more 'lively' amplifier, but ultimately fatiguing, congested and lacking inner detail. Mo' better bass..but that's about it, IMHO. The rest goes bad. There is obviously some gain to the lower Re, yes, and that can be heard --but it's seemingly a trade-off over other concerns..
Hi Gerhard, Scott,
That is exactly my point. I'm all for a reasonable markup, just don't gouge "us". I can also accept that the volume is low, so the price will be higher than your average book of the same size. There must be a scale of price vs. sales volume.
They just force many to look around for the same information published in other locations on the 'net. Too bad patents were not easier to read!
I'm sure that app notes are a good place to look too. That is how I keep up with the devices and techniques I use.
Hi Ken,
Ken, I have an idea for the Mission amp that cooked itself. I might be able to rebuild it slightly differently in a way that excessive heat is minimized. That output stack must go for one.
-Chris
That is exactly my point. I'm all for a reasonable markup, just don't gouge "us". I can also accept that the volume is low, so the price will be higher than your average book of the same size. There must be a scale of price vs. sales volume.
They just force many to look around for the same information published in other locations on the 'net. Too bad patents were not easier to read!
I'm sure that app notes are a good place to look too. That is how I keep up with the devices and techniques I use.
Hi Ken,
Of course, just like everything else.
I think you'll find that careful matching and circuit design will minimize that effect. So will careful design in the voltage and error amplifier sections. Don't forget that many higher power amplifiers are designed for what holds together and mass production does not allow for hand matching. That is one reason why the newer On-Semi parts have me so happy. They are all close right out of the tube.
Ken, I have an idea for the Mission amp that cooked itself. I might be able to rebuild it slightly differently in a way that excessive heat is minimized. That output stack must go for one.
-Chris
anatech said:
Hi Chris,
You're right on all counts. Matching output transistors, both among the same sex and from sex to sex, is usually a good thing for THD. My comments were solely directed at the degree of need for transistor matching for purposes of bias stability in light of a given choice of base stopper resistor and emitter resistor.
You are also correct about emitter resistors being more effective at promoting uniform current sharing at high current swings. Current sharing under quiescent conditions does, however, become important in amplifiers with high rail voltages and high quiescent bias current.
Cheers,
Bob
jam said:
Hi Jam,
This is a good question and does point to an area where vertical MOSFETs are at a bit of a disadvantage, at least in my experience. While it is quite possible to design a BJT amplifier with emitter resistors of reasonable value that will ensure adequate current sharing among paralleled BJTs that are not specifically matched, this is not as much so with vertical MOSFETs.
Reasonable size source resistors, like 0.5 ohm or less, do not adequately promote sharing among paralleled vertical MOSFETs because the possible gate voltage offsets can be quite a bit larger than the delta Vbes among BJTs.
For this reason, the use of reasonably matched vertical MOSFETs (at least for Vgs at the planned quiescent current) is important. The devices should be checked before use. The good news is that modern vertical MOSFETs are made with quite good processing, so that most of the devices in a tube will be like peas in a pod. Nevertheless, this must be confirmed by measurement.
Cheers,
Bob
Hi Bob,
Yes, everything you've mentioned is what I am seeing in practice. Especially the point about matching sex to sex beta as much as possible. I think this is more important than may be realized by many. Take that a step further by matching the driver compliments and any other same sex pairs in the design. It seems that this reduces the amount of imbalance in the differential pair, thus reducing odd order THD. You can measure a small drop in THD after dong this, but the real difference is in the listening. I was both curious and surprised for many years as I did the work this way.
I should mention that I was trained in amplifier service by some good technicians at Superscope/Marantz here in Canada. That and application notes by Motorola. As a young technician, I was very lucky to have had the guidance I did.
There is one thing that there is debate over that you may be able to weigh in on Bob. I was trained to match beta in both small signal and power transistors (BJT of course). In tested incoming parts for grading, I have seen that when beta matches, so does Vbe. Of course, Vbe is always within a few mV anyway, so I don't see this as a valid measure to grade parts on. AndrewT and I disagree on this basic point. The current sharing is better when matched for beta in my experience, and so is the balance of a differential pair. I use a steady state current measurement method, not the "transistor test" function on a multimeter. I have also made jigs to encompass higher voltage and current levels, as well as a jig to balance transistors in long tail pairs. The circuit is simple a long tail pair running open, no feedback of any kind. The parts are in thermal contact and pairs matched in this way do react to the matching of emitter degeneration resistors. That seems to be an extremely close match, and it works well in practice.
Is it possible that matching Vbe also matches beta to some degree? I would think that beta changes are greater than Vbe, and so parts are easier to differentiate by using beta as the tested parameter. Different viewpoints on this are welcome.
Hi Andrew,
Have you ever tried to measure beta to compare how close Vbe measures? Just curious. I have, as mentioned above. I would say that matches in beta area subset of matches in Vbe.
-Chris
Yes, everything you've mentioned is what I am seeing in practice. Especially the point about matching sex to sex beta as much as possible. I think this is more important than may be realized by many. Take that a step further by matching the driver compliments and any other same sex pairs in the design. It seems that this reduces the amount of imbalance in the differential pair, thus reducing odd order THD. You can measure a small drop in THD after dong this, but the real difference is in the listening. I was both curious and surprised for many years as I did the work this way.
I should mention that I was trained in amplifier service by some good technicians at Superscope/Marantz here in Canada. That and application notes by Motorola. As a young technician, I was very lucky to have had the guidance I did.
There is one thing that there is debate over that you may be able to weigh in on Bob. I was trained to match beta in both small signal and power transistors (BJT of course). In tested incoming parts for grading, I have seen that when beta matches, so does Vbe. Of course, Vbe is always within a few mV anyway, so I don't see this as a valid measure to grade parts on. AndrewT and I disagree on this basic point. The current sharing is better when matched for beta in my experience, and so is the balance of a differential pair. I use a steady state current measurement method, not the "transistor test" function on a multimeter. I have also made jigs to encompass higher voltage and current levels, as well as a jig to balance transistors in long tail pairs. The circuit is simple a long tail pair running open, no feedback of any kind. The parts are in thermal contact and pairs matched in this way do react to the matching of emitter degeneration resistors. That seems to be an extremely close match, and it works well in practice.
Is it possible that matching Vbe also matches beta to some degree? I would think that beta changes are greater than Vbe, and so parts are easier to differentiate by using beta as the tested parameter. Different viewpoints on this are welcome.
Hi Andrew,
Have you ever tried to measure beta to compare how close Vbe measures? Just curious. I have, as mentioned above. I would say that matches in beta area subset of matches in Vbe.
-Chris
Re: Vgs & Vbe matching of paralleled pairs/sets
I think Bob was saying that it is more necessary in VMOSFETS because their Vgs varies more from unit to unit than the Vbe of BJTs. So the remaining mismatch between BJTs can be taken care of by relatively low Re values. For VMOSFETS - if not matched - you would need unduly large Re values to take care of the mismatch if not matched.
jd
AndrewT said:
I think Bob was saying that it is more necessary in VMOSFETS because their Vgs varies more from unit to unit than the Vbe of BJTs. So the remaining mismatch between BJTs can be taken care of by relatively low Re values. For VMOSFETS - if not matched - you would need unduly large Re values to take care of the mismatch if not matched.
jd
Hi Jan,
That's exactly how I think things work.
Hi John,
Firstly, how are your eyes doing? I'm wishing you a complete recovery.
You have a view that's interesting, I'm not saying whether it's right or wrong, just interesting. Why is this comment interesting to me?
Well, as the designer who created the complimentary differential pair, or long tailed pair as it was probably known as back then, the arrangement is extremely picky about transistor matching. From long experience with this arrangement, and probably why I don't really care for it as much as a single differential pair, it's been my experience that most commercial amplifiers (not yours, I haven't seen one that I know of) will sound a great deal better once a careful matching is done involving both pairs. Most amplifiers I see have mis-matched transistors in this area, and that seems to cause a very "brittle" sound quality. They have that sound that limits your listening time. Please understand that I am referring to your design that is not executed properly. Not your own work.
So, once all four transistors are matched using beta measurements, the sound quality is greatly improved, having lost the annoying quality that the amplifier had in stock trim. This is, of course, a very time consuming procedure that involves many signal transistors. I have seen the Vbe vary by a couple mV or so, but this does not seem to affect the match. As a side benefit, the DC offset of a matched set shows itself through very low DC offsets, less than 5 mV unless something else is going on. In fact, a DC offset of a couple mV is common. To my way of thinking, this would seem to show that matching by beta is correct.
I know this conflicts with your view on this topic, so I'm trying to reconcile the apparent different view points.
What I am thinking is that a couple mV mismatch in Vbe may create a couple mV DC offset at the output. If the design has gain at DC with no servo, then the offset is amplified by the same factor as the intended program material. In the past, I did match on Vbe. I never did get good results this way, not terrible but not good either. Are transistors with matching beta a subset of transistors with similar Vbe? That might explain it.
I would expect models that you design do leave the factory with properly matched transistors. Here is where Japanese signal transistors really excel. The consistency between individual parts in each batch, and better control between batches allowing similar performance between them.
-Chris
That's it for me I think tonight. My UPS has saved this post at least three times now and I can't upload yet.
Another try ...
Edit: (I'm pushing my luck with this storm).
I normally do match my parts at the expected current levels for both signal transistors and power transistors. A jig for each polarity and type, and I'm thinking of a better one for power devices. It is also possible to apply the same supply voltages within reason. I should have grabbed a dead Carver M 1.5 (or similar) for it's supply voltages when I had the chance. That would have given me high voltages that could be easily varied with high current capability. I can match 4 power transistors at a time right now. I control the temperature by varying the current and voltage during the warm up stage using a thermocouple temperature meter. I can also get an idea of thermal performance that way as I also have a dual unit. It's a pain in the drain. The next one will have controlled heaters for that spaced along the heat sink.
That's exactly how I think things work.
Hi John,
Firstly, how are your eyes doing? I'm wishing you a complete recovery.
You have a view that's interesting, I'm not saying whether it's right or wrong, just interesting. Why is this comment interesting to me?
Well, as the designer who created the complimentary differential pair, or long tailed pair as it was probably known as back then, the arrangement is extremely picky about transistor matching. From long experience with this arrangement, and probably why I don't really care for it as much as a single differential pair, it's been my experience that most commercial amplifiers (not yours, I haven't seen one that I know of) will sound a great deal better once a careful matching is done involving both pairs. Most amplifiers I see have mis-matched transistors in this area, and that seems to cause a very "brittle" sound quality. They have that sound that limits your listening time. Please understand that I am referring to your design that is not executed properly. Not your own work.
So, once all four transistors are matched using beta measurements, the sound quality is greatly improved, having lost the annoying quality that the amplifier had in stock trim. This is, of course, a very time consuming procedure that involves many signal transistors. I have seen the Vbe vary by a couple mV or so, but this does not seem to affect the match. As a side benefit, the DC offset of a matched set shows itself through very low DC offsets, less than 5 mV unless something else is going on. In fact, a DC offset of a couple mV is common. To my way of thinking, this would seem to show that matching by beta is correct.
I know this conflicts with your view on this topic, so I'm trying to reconcile the apparent different view points.
What I am thinking is that a couple mV mismatch in Vbe may create a couple mV DC offset at the output. If the design has gain at DC with no servo, then the offset is amplified by the same factor as the intended program material. In the past, I did match on Vbe. I never did get good results this way, not terrible but not good either. Are transistors with matching beta a subset of transistors with similar Vbe? That might explain it.
I would expect models that you design do leave the factory with properly matched transistors. Here is where Japanese signal transistors really excel. The consistency between individual parts in each batch, and better control between batches allowing similar performance between them.
-Chris
That's it for me I think tonight. My UPS has saved this post at least three times now and I can't upload yet.
Another try ...
Edit: (I'm pushing my luck with this storm).
I normally do match my parts at the expected current levels for both signal transistors and power transistors. A jig for each polarity and type, and I'm thinking of a better one for power devices. It is also possible to apply the same supply voltages within reason. I should have grabbed a dead Carver M 1.5 (or similar) for it's supply voltages when I had the chance. That would have given me high voltages that could be easily varied with high current capability. I can match 4 power transistors at a time right now. I control the temperature by varying the current and voltage during the warm up stage using a thermocouple temperature meter. I can also get an idea of thermal performance that way as I also have a dual unit. It's a pain in the drain. The next one will have controlled heaters for that spaced along the heat sink.
anatech said:
I once heard Doug Self talk about that. One point he made was that if you use a servo to correct DC offset from a non-matched diff pair, you also base the pair off of the center point where the transfer function is as linear as it gets. Any off-centering (which the servo would do) would place the pair in a less linear region. Intuitively I feel that the effect should not be very great, and there is still the global nfb to straighten things out, but the effect is there and can be measured on the ol function.
jd
- Status
- Not open for further replies.