Bob Cordell's Power amplifier book

Member
Joined 2007
Paid Member
Hi all ... & thanks for your feedbacks. As it is I actually also bought them from Polida 2008, however, only after REALLY insisting that he positively confirmed that they would be genuine Sanyos. But, given that the casing should be TO-126 (& no heatsink) they cannot be.

To this end: Do any of you then know where to get genuine Sanyo 2SC3503s - and not too many as for a start I just need them for tests? I already have Fairchilds.

Cheers & thanks again for your feedback - very helpful!

Jesper
 
Any New News on the BC-1 Amp?


By the way, the "search" on DIYAUDIO will not find anything when you put in "BC-1" !

The BC-1 has been on hold for awhile, as the DH-220C has taken up more time than expected. We had been working on them in parallel, but more recently have decided to focus on getting out the DH-220C, as working on both at the same time was a bit of a distraction.

The DH-220C is very close to release; just doing some clean-up, design centering and documentation at this point. The boards are on their second rev and looking very good. I do not expect another rev for the boards unless it is something very trivial in terms of cleanup. We have spent quite a bit of time paying attention to detail. If others are going to build it, we want things to go smoothly and for them to be successful.

The BC-1 is very solid, but is still on the first version of the boards. There are some physical design issues we want to work out to make it more straightforward to build and integrate into available heatsinks and enclosures. One thing that I believe would be nice is to be able to package it in one of the enclosures available on DIYaudio, maybe one that comes with heatsinks. A big advantage we had with the DH-220C was that all of that basic infrastructure was already there.

Cheers,
Bob
 
Mark Johnson- the Sanyo picture of the TO-126 is the top side.
The metal plate is on the underside.
All TO-126 devices I have seen have the metal plate. - except fully isolated devices, which are relatively recent.
I haven't checked the JEDEC spec yet, but I would be surprised if it does not include the metal plate.
 
Member
Joined 2011
Paid Member
They (BC-1 and DH220-C) are in Bob's textbook (2nd edition), and his Burning Amp 2016 video presentation (screengrab below).


_
 

Attachments

  • cordell.jpg
    cordell.jpg
    58.1 KB · Views: 265
Hi
I haven't followed the discussion in this thread for a while, could some tell me what the "BC-1" and "DH-220 C" is?

A simple question to Bob:
Why on earth do you call an OPS that uses a multi sinh linearizing technique "DoubleCross Tm" I cant see anything new.

S

I don't think it is properly classified as "multi-sinh". I call it Doublecross because it goes through two crossover points that are each displaced from zero. A conventional class AB design has a single crossover point that occurs in the vicinity of zero net current output. These differences are shown in the wingspread diagrams in my book. It is also useful to compare the Double Cross (DX) with Self's Crossover Displacement (XD) approach. Both approaches push the crossover point away from the zero net current output condition. XD forces it away from zero by pulling current out of the output stage in a manner not unlike that used with op amps where a resistor to negative supply or a current source pulls current out of the output stage to force it to operate in class A for small signals.

Cheers,
Bob
 
Ok
I have the 2nd edition, but can't find anything in the index.

S

The BC-1 is described in Chapter 4, which is entirely dedicated to the BC-1 as an amplifier that can be built, based on a particular variation of one of the example amplifier topologies of chapter 3. Chapter 4 includes design theory, implementation, testing, measurement and troubleshooting of the BC-1. It is a straightforward implementation of the Thompson topology with two output pairs in an output triple (3 EF) Locanthi arrangement. It also uses a 2 EF VAS. There is no magic in the design, and it achieves very good performance largely due to the use of the output triple, fast output transistors, high driver bias current and some attention to detail. Protection schemes are also included and described.

The DH-220C was described in part in two chapters. Its IPS-VAS was described in the IPS-VAS chapter. The remainder of it was described in the MOSFET power amplifier chapter as an example of a lateral MOSFET power amp. Its original design was done in 2016 and was presented at Burning Amp.

Cheers,
Bob
 
I don't think it is properly classified as "multi-sinh". I call it Doublecross because it goes through two crossover points that are each displaced from zero. A conventional class AB design has a single crossover point that occurs in the vicinity of zero net current output. These differences are shown in the wingspread diagrams in my book. It is also useful to compare the Double Cross (DX) with Self's Crossover Displacement (XD) approach. Both approaches push the crossover point away from the zero net current output condition. XD forces it away from zero by pulling current out of the output stage in a manner not unlike that used with op amps where a resistor to negative supply or a current source pulls current out of the output stage to force it to operate in class A for small signals.

Cheers,
Bob

Hi Bob, as someone who worked closely with Barrie Gilbert at ADI, the DoubleCross^TM is a multi-sinh technique. DoubleCross is a cool name, but it is still multi-sinh.

In input stages, where this is most commonly used, a mult-sinh differential pair input stage has two or more points where a diff-pair tips completely to one side or the other. You could call these points the 'Cross-over' points of the diff-pairs. This is implemented with an array of diff pairs which have built-in off-sets in opposite senses.
The DoubleCross output stage closely resembles the n=2 Kermit (Barrie came up with the name), where n=2 is the number of diff-pairs or cross-over points.

Regarding Doug Self's output stage, that is a totally different story. I actually prefer the multi-sinh approach.

Best, Sandro
 
The index is lacking references to BC-1 & DH-220C, next rev :)
Chapter 4, first paragraph makes mention of BC-1.
DH-220C: IPS-VAS is in section 9.11,fig 9.20 (p214), servo fig 10.5, (p223), OPS section 14.10,fig 14.17 (p346)

I figured out that the old JEDEC TO-126 was replaced by TO-225 which has two variants "AA" & "AB" the "AB" is the thicker package which matches the Fairchild drawing. So much for standards and applying them.
I have to use a steel chisel, to spread the Aavid 577500B00000G heatsink, plus some thermal compound in order to make it go on the Fairchild parts as the HS was designed for the TO-126 & TO-225 "AA" package = PITA, so a chisel gets called out in the tools list for a build. Lots of fun.
 
Last edited:
Hi Bob, as someone who worked closely with Barrie Gilbert at ADI, the DoubleCross^TM is a multi-sinh technique. DoubleCross is a cool name, but it is still multi-sinh.

In input stages, where this is most commonly used, a mult-sinh differential pair input stage has two or more points where a diff-pair tips completely to one side or the other. You could call these points the 'Cross-over' points of the diff-pairs. This is implemented with an array of diff pairs which have built-in off-sets in opposite senses.
The DoubleCross output stage closely resembles the n=2 Kermit (Barrie came up with the name), where n=2 is the number of diff-pairs or cross-over points.

Regarding Doug Self's output stage, that is a totally different story. I actually prefer the multi-sinh approach.

Best, Sandro


Hi Sandro,

As you probably know, sadly, Barrie Gilbert passed away last month. He was a great guy and a true innovator. I last exchanged email with him over a year ago.

An input stage and a class AB-like output stage are two rather different things. However, I don't know enough about the multi-sinh technique to say yes or no based on similarity. Multi-sinh sounds more like a mathematically-defined technique than the DoubleCross, which just tries to implement two crossovers where each attempts to somewhat obey the Oliver condition.

However, I do think I see what you are talking about in using multiple transistors with slightly different bias to linearize the diff pair transfer function. But I think that Pat Quinn's Cascomp is a better approach to achieving this. He came up with this at Tek under Barrie's mentoring. Before he went to Tek, Pat and I both worked at Bell Labs in Holmdel and were classmates at Stanford for our Master's degrees.

Do you have an example of a power amplifier that uses a multi-sinh output stage? Apart from semantics and issues of relative novelty, the most important thing about the DoubleCross is that it works.

A key feature of the DoubleCross is that you can run a standard class-AB stage at higher bias current, yielding a larger class A region. But this pushes it into gm-doubling, which is usually undesirable. With gm enhanced (usually less than fully doubled) in the crossover region, the other pair of output transistors in the DoubleCross begin to contribute make-up transconductance as one of the original pair begins to turn off. As a result, the wingspread exhibits two crossover points displaced from the center. Between those crossover points, things are fairly linear, as that is effectively a widened class-A region.

Cheers,
Bob
 
Hi Bob,
I did know about Barrie's passing. Big loss to the industry and Analog Devices. We designed the AD8338 together, and I learnt a great deal from him. As you know he was the master of non-linear circuits. A bit of an odd man, but well, all analog engineers are, me included ... ask my wife.

The cascomp is a great technique (I did not know Pat Quinn invented the idea, kudos to him) and does a great job at linearization. The only drawbacks are:
- Headroom: The cascode eats some headroom, but there are folded cascode implementations that get around this.
- Noise: By construction, you need to degenerate the input stage with this technique.
In audio, neither one is a big issue, but in op-amp design it can be. If say you are gunning for a very low noise amp, cascomp is a bad choice. A doublet input stage (The basic multi-tanh doublet. | Download Scientific Diagram) will do better in noise, but trades off linearity.

Regarding amplifiers that use the multi-sinh technique in their output stage, the Yamaha M2 (1979) uses the technique. You can find the schematic within the service manual. Are they the first to use the technique?, I have no idea. Also, I have not seen it used in op-amp output stage design. For example, for my ADA4932, I just used a tweaked version for Oliver's relation. All its predecessor's (AD8138, AD8132, etc) are the same. Even the AD797 is the same, Scott uses a different technique to linearize the OPS.

Finally, this idea of double-cross is more pervasive than you may think. Wing-spread diagrams are used in multiple of circuits, not just output stages, to show linearity. Here are a couple of examples:
- Complementary transmission gates (MOSFET switches) - You can plot the resistance of the switch vs. Input voltage. You get two "cross-overs" due 3 regions of operation, one when the P-MOS is working, one when the n-mos is working, one when the two are working. There are some very interesting technique to linearize these, but don't apply to amps.
- Doublet, triplet, kermit input stages: The paper from Barrie Gilbert in the reference above talks about this a lot, I just scratched the surface. I also got a patent on a version of these: US8692615B2 - Enhanced transconductance circuit
- Google Patents


Bottom line, semantics don't bug me (if anything DoubleCross is a neat name, I give you that), I am not arguing about your technique being novel or not (I like it either way), and I do agree it has a lot of merits, hence why I like it a lot better than Doug's. All I am saying is that I disagree when you say this is not a multi-sinh/tanh based circuit when it is, and why I am reacting on this statement. :happy1:.

Best, Sandro
 
Last edited:
Sandro,

Your work at Analog Devices is truly impressive!

Have you seen the spreadsheet for Double Cross? It was mentioned in Post 9464 here Bob Cordell's Power amplifier book
It uses the W-function to plot the currents and Gm's for the DoubleCross circuit in Bob's book Fig.13.8(a). It is an exact general solution for a common emitter DoubleCross using 4 transistors. I don't have a solution for the Common Collector follower version yet. BTW most of the spreadsheet is taken up getting the W-function for each transistor because the W-function is not a standard function in my spreadsheet.

Also of interest to you and visitors here might be a DoubleCross simulation with emitter resistors bypassed by a low cost, compact 10F/3V supercapacitor, see Post 45 here Emitter resistor in HexFet OPS
 
Thanks Ian!
I downloaded the spreadsheet about a month ago (PAK307 I believe), just haven't had to time to play with it... but it is in my to do list :)

I'll take a look at the "Emitter resistor in HexFet OPS" post. Thanks for the heads up and for the great contributions.
Edit: I just read the post... super interesting. Need to ponder on your findings.

BTW, if you have time, take a look at my hobby video series.
SW-VFA-01: Audio power amplifier video series

Best, Sandro
 
Last edited:
Hi Bob,

Regarding amplifiers that use the multi-sinh technique in their output stage, the Yamaha M2 (1979) uses the technique. You can find the schematic within the service manual. Are they the first to use the technique?, I have no idea. Also, I have not seen it used in op-amp output stage design. For example, for my ADA4932, I just used a tweaked version for Oliver's relation. All its predecessor's (AD8138, AD8132, etc) are the same. Even the AD797 is the same, Scott uses a different technique to linearize the OPS.

Best, Sandro

Hi Sandro,

Thanks for bringing the Yamaha M2 schematic to my attention. It does indeed resemble the DoubleCross, with 3 output pairs that appear to turn on in succession as a result of the offsets in the base bias voltages. Interestingly, they use fairly high-value 0.47-ohm emitter resistors. It is a little unclear how heavily they bias the inner pair to form a large class A region, and to what degree the inner pair creates gm doubling that must be corrected by the other pairs as they turn on in succession. It would be nice to see a wingspread of this output stage with it biased as intended.

Cheers,
Bob