Harrison High-Stability Lab PSU with Oven Stabilized Voltage Sense - HP 6102A

[quantumvolt]

The Hewlett-Packard Harrison 6102A is in my opinion a state of the art scientific laboratory power supply for medium voltage electronics and physics. The instrument was released in 1966 and I challenge you all to show me a general purpose lab bench power supply unit of higher overall quality and versatility.


I will use this tread to present documentation to support my view on the instrument's very high professional standards. In this first post I will point to only the main feature of this 47 years old precision instrument design - it's stability:

0-40 Volt at 0-500 milliAmpere - 0.001% line and load regulation (fixed error within 100 microVolt)

Long time stability 0.012% / Month (fixed error within 120 microVolt)

That means that this power power supply can be used as a 10 Volt Reference Voltage accurate within 1.32 mV a month after calibration. 

NB! My writing and/or math could be in error. The values are in the manual linked below. I do not use +- values for errors because the pdf-file of the Users Manual from Hewlett-Packard dated 1966 does not specify one- or two-sided errors. This will have to be checked later.



For those interested - and for my own convenience - here are links relevant to this instrument:


1. Manual 6102A
http://cp.literature.agilent.com/litweb/pdf/06102-90001.pdf

2. Specification sheet 6101A-6116A
http://www.teknetelectronics.com/DataSheet/HP_AGILENT/HP_6106a.pdf

2. Catalogue 1966-67
http://www.hparchive.com/Catalogs/HP-Catalog-1966-1967.pdf

3. Manual 6112A (in principle the same instrument with built-in mechanical digital programmer)
http://cp.literature.agilent.com/litweb/pdf/06112-90001.pdf




So - I hear you say - looks like any other old PSU. Yep it does ... 






But look at the back terminal strip     (Who can beat that one ...)






And on the inside is a temperature regulated oven keeping your Voltage-Sense-and-Compare goods at 65 degrees Celsius  .

Enjoy ...

[quantumvolt]

Voltage and Current - Series and Parallel

(If you want to refer to the manual - please open link "1. Manual 6102A" in Post #1 over and keep it for reference in a separate window.)

Two units can be connected in series to obtain a voltage of 80 Volts. Two units can be connected in parallel to get current 1 Ampere. More than two units can be connected together.

As described on pages 15 and 16 in the manual, one unit can figure as Master controlling the other (Slave). This way two units connected in series will work as a Dual Voltage Tracking Supply.

As mentioned in paragraph 3-34 on page 15 both Remote Programming and Remote Sense can be used simultaneously with Master/Slave Mode.

The unit also allows Auto Tracking Ratio. This means that two units can supply for instance V and V/2 where voltage is regulated at the Master unit, and the half voltage at the Slave tracks the Master. See pg. 16.

With certain limitations, the unit can be series/parallel connected to other supplies of different type, voltages and currents (even MC and car batteries). This way one can obtain voltages higher than 40 Volts at current of several Ampere. This however should be done with caution.

[robrenz]

Very nice but you put the challenge out there.  The 6114A & 6115A have much better specs than this and have digitaly setable voltages.  Power Designs Precision supplies are also in this class but are not as good on paper (some have an ovenized reference). This thread shows that they can be much better than thier specs.

But I agree with your general premise, You can't buy buy new supplies today with these specs (regulation, ripple, drift, temp coeficient) without spending a small fortune. 

EDIT: added link to 6114A - 6115A manual

[quantumvolt]

Very nice but you put the challenge out there.  The 6114A & 6115A have much better specs than this and have digitaly setable voltages.  Power Designs Precision supplies are also in this class but are not as good on paper (some have an ovenized reference). This thread shows that they can be much better than thier specs.

But I agree with your general premise, You can't buy buy new supplies today with these specs (regulation, ripple, drift, temp coeficient) without spending a small fortune.

Thank you very much for your comment. By taking me up on this, you give me a chance to follow up and learn. I have done a little homework:


According to http://www.testequipmentdepot.com/usedequipment/hewlettpackard/powersupplies/6114aspecs.htm the 6114A has 

Drift    CV, 0.0015% + 15µV per 8 hours, 0.0075% + 30 µV per 90 days.

I agree that this is marginally better. But with a

Temperature Coefficient CV, 0.0001% + 15 µV/°C; CC, 0.02% + 50 µA/°C

it is probable that 6114A has one or more linear IC's. The 6102A on the other hand contains only discrete transistors,and can be improved to ANY practically obtainable precision for small money by refined matching/design modifications helped by the oven's fixed temperature . This is not possible for the 6114A unless you build discrete units to replace the op-amps.


In http://www.testequipmentdepot.com/usedequipment/hewlettpackard/powersupplies/6114a.htm they write "A four-digit pushbutton switch increases or decreases the output voltage in unit steps, and switches go directly from "9" to "0" without backing down."  This precision programming is nothing but resistors added to the programming lead. I can outperform this any day by using external resistors measured to better than 0.01% and calibrate the PSU's voltage with a 5 1/2 digit DMM. HP also produced external programmer that can be used by the 6102A (see the "Datasheet-link" in the first post in the tread).


I found a link to the Manual for the 6114A http://www.to-way.com/teqman/HP/6114A/5950-5976.pdf and it shows that there are integrated circuits. While it is a better instrument spec-wise, I put more weight on the discrete nature of all components of the 6102A (except one single double transistor that can be made by hand-matching) and the controlled environment of the oven.


Another fine PSU



---

I have also studied the PD 2020B. The manual is here http://www.ko4bb.com/Manuals/09%29_Misc_Test_Equipment/Power_Designs/Power_Designs_2020B_5020_C500_Power_Supply_Service_Manual.pdf. If you look at page 3 of the pdf-file you will see that they only mention a 1 week timespan for drift. Although this unit is very "sexy" and also has an oven, I prefer Harrison for it's scientific simple design and documentation. If you have the time, please compare the manuals for the 6102A and 2020B, and you will see what I mean.

You write "This thread shows that they can be much better than thier specs.". This is generally true for all instruments - the specs are supposed to catch some 90-99% (or whatever) of the produced units. But imo depending on a single specific unit with serial no. ABC xyz because it turned out to be a "good one", is against all that metrology is about.


---

Conclusion: I thank you for your post and do agree that in general I must admit that both of the instruments that you mention and a bunch more in most cases produces as good or better results as the 6102A. But I still chose the latter because of all-discrete design, open modular building block circuit (see the schematic in the "Manual" link in my post #1), the oven-feature and the documentation (relative to the PD unit).

[robrenz]

Thank you for having a gentlemanly discussion.  To compare apples to apples you need to be comparing the 6102A to the PD 2005A.  It is all discrete components and has a ovenized reference. You will find that the 2005A has significantly better specs on paper.  It also is a calibrated output with 0.1% +1mV maximum error.  One caveat is the 2005a is only 20V compared to the 40V of hte 6102A

Edit: I have 2 PD 2005A units and they perform much better than spec.  2005A manual link here

[quantumvolt]

I have done a mix on the oven featured Power Designs unit. It is not the 2020B that has an oven - it is the 2005A. The manual for the latter is http://www.ko4bb.com/Manuals/09%29_Misc_Test_Equipment/Power_Designs/Power_Designs_2005A_Power_Supply_Service_Manual.pdf.

The 2020B gets it's good specs from analog integrated circuits just like the HP 6114A.

Still,  both the PD manuals I have given links to mention - as far as I can see - nowhere anything about stability for more than 1 week. The HP instruments give specs for 1 month.

[quantumvolt]

Thank you. With 2 minutes difference in posting time, it seems we detected my error on the same time.

As you can see, I do in many ways agree with you that all the instruments you mention probably performs better than the 6102A.

But I still regard the latter as being in a class of its own because of it's circuit design, openness, simplicity and documentation. I need some time to prepare my post so I will have to come back later.

Thanks a lot anyway - I enjoy this even if I have to admit my errors ...

[robrenz]

Looks like we were posting the same thing at the same time. 

I am not trying to win here, they are both very capable supplies and far exceed the capabilities of anything costing less that several thousand dollars today.

[blackdog]

Hi quantumvolt, robrenz

May i also play ;-)
This is my Harrison 6111a, after cleaning about one year ago, and its in top condition 

The drift over 9hours is about 30uV.
I made some pictures so you can see the drift and the oven switching.

The HP/Harrison 6111a


Switching of the oven afther 2 hours power on and 16 minutes collecting data


This is 9 hour of collecting data.


The output voltage was 11.00000V
And yes this powersupply is Kicking But

And i am designing a powersupply that is better than HP/Harrison /PD
Why, because i like designing precision stuf.

The structure is the same as the HP / Harrison power supply's.
I tuned it with modern components to kill any hum and noise.
It is clean @ switch on or off.
Noise is < 5uV @ 22Khz bandwidth.
Stable @ dynamic load, also in currend mode.
Drift/stability depend on the LT1021 reference en the voltage set potmeter.

I'am optimizing the design at the moment, also i need to design the comparator circuit to
switch the voltage tabs on the transformer to keep the dissipation low.

Take a peak at the schematic
www.bramcam.nl/NA/NA-SCH-PSU-12b.png

Kind regarts,
Blackdog

[robrenz]

Welcome blackdog

My first PD was a 2010 here are the drift tests fresh from ebay just cleaned the switches.
My 8846A shorted input DC 152 hr stability min-max span  is 1.08 uV with a SD of 74 nV
100mV setting.   stat mode on the 8846A 6.5 digit 100nplc digital filter 
1 hour AVG 99.99839 mV    1.75 uV min-max span   SD of .392 uV
3 hour AVG 99.99873 mV    2.93 uV min-max span   SD of .667 uV
6 hour AVG 99.99935 mV    3.33 uV min-max span   SD of .855 uV

Your min-max span for 9 hours was 92.14 µV
How do you plan to measure that 5 µV ripple?  Here is my attempt at measuring the ripple of a PD2005.

Schematic looks nice and simple but I am not qualified to comment electronicaly.

Edit: May I say you have very good taste in bench multimeters

[quantumvolt]

You are very welcome. I am grateful for others joining - this can be fun.

This weekend coming I won't get much done as it my turn to have the children. But I am preparing a kind of comparison of Harrison and PD - the main point being that in PD the oven is powered by the mains, switched on/off by a thermo-sensitive mains switch device, and the heated board houses both voltage comparison circuit and reference. The Harrison unit has a more sophisticated oven low voltage control system, and the heated difference amplifier is separated from the voltage ref. In principle the Harrison can run on external ref. I have great joy from reading the manuals, and especially the HP manual is a very beautiful piece of documentation.

But when I get the cash and chance, I will also buy a PD 2005. Having you coming here with custom stuff makes it even better.

But my children first, these DC babies must wait until next week

[blackdog]

Hi robrenz,

I measure the noise with AP, Portable One Plus, (i hope you like it)


De DC shift of 92uV is because of the oven switching and temperatuur shift in my lab ~10C delta.

Here in the Netherlands its inpossible to get a PD powersupply.
Last month I bought a KEPCO power supply which is also rare in the Netherlands.

The two Tektronix multimeters are here 24/7/365 switcht on.
After dc nulling they are within 1 digid on the 12V range are equal

On the lower ranges about 3 digits, mutch beter than the specs, maybe I let them calibrate this year.
But it's so expensive...

Sorry for bad english, I'm a dyslexic monkey 

Kind regarts,
Blackdog

[quantumvolt]

De DC shift of 92uV is because of the oven switching and temperatuur shift in my lab ~10C delta.

Here in the Netherlands its inpossible to get a PD powersupply.

If I understand it correctly, the PD was outputting 100 milliVolt, but the Harrison 11 Volt? That is a factor of 110. If so the 90 or so vs. 3 or so microVolt min-max logs are in favor of the Harrison? Or am I wrong? Are we talking about DC drift or higher frequency noise? If it is drift then there is a 110:1 ratio in output and a 30:1 ratio in drift. Please comment.

---

One thing that I find strange about the PD is that the mains transformer is for 110 V AC only. In my view - no serious lab instrument maker would just I-don't-care-about-Europe-and-the-rest-of-the-world. So in a sense the PD is more like a "sexy" American gadget to me. If and when I buy one, I will have to get a quality step-down transformer from 220-110 V AC of at least 100-200 VA. Will be pricey.

[robrenz]

If I understand it correctly, the PD was outputting 100 milliVolt, but the Harrison 11 Volt? That is a factor of 110. If so the 90 or so vs. 3 or so microVolt min-max logs are in favor of the Harrison? Or am I wrong? Are we talking about DC drift or higher frequency noise? If it is drift then there is a 110:1 ratio in output and a 30:1 ratio in drift. Please comment.

Nobody said we could not cheat with our specs   I am running a dc drift test right now on a PD2005 at 10V I will post results.

Edit: @ blackdog, that analyzer is quite nice, I am jealous now

[quantumvolt]

No problem with a little favorable data presentation Look forward to see your results for the PD.

It is 6.33 AM Saturday here, and I am leaving to no-internet-land for the weekend. So maybe no posting for me. I have got email that my 6102A (USD 35 + 60 shipping) is now in the hand of USPS on it's way to Thailand.

My whole instrument collection consists of two pieces: Agilent 34401A and HP Harrison 6201A (and some cheap DMM's, voltage ref's and precision resistors).

I have now started researching my Poor-Man_High-Precision_Arduino_DataLogger:

High input impedance low drift instrumentation differential amplifier op amp for voltage sense. The minus input is clamped to a fixed high precision low drift voltage reference rounded-off value.

If I want to measure drift on 10 Volt, I clamp the amplifier minus-input at 9.9xxx Volt, and adjust the amplification factor for range width: for instance 9.95 to (9.95+0.1= 10.05) for FS 100 milliVolt. Then I amplify this 10x (or 50x) in a second stage to minimize FS error on the Arduino analog input. I have got an SD card reader, so I can store several hours/days on card, and then import the data series to the laptop.

Blackdog: Your design is beautiful and truly modular simplistic. More so than Harrison, and even more than the PD. But as robrenz says - the apples are different because you you non-discrete component.s So your modern era DIY will be the benchmark.

Time to go. Will be back ...

[blackdog]

Hi,

I measured the output impedance of the powersupply i am designing today.
See the schematic in my other post.
This kind of picture, measurements, you dont see from HP/Kepco/PD 
The opamp is not NE5534a but NE5532a, little brainfart...



The next step is to make the resonant piek smaler.
This is how i wil try to do it, i wil design a output circuit like this...


If you like to see more, take a look at this Dutch website, i place a lot of info about building/designing of powersupply's
A lot of pictures en schematics
http://www.circuitsonline.net/forum/view/110029/7

Kind regarts,
Blackdog

[quantumvolt]

@ Blackdog: I looked at your website in Dutch - translated. I like your DIY equipment - using glue and soldering the components together w/o any PCB or contact row if that is what seems to give the best result. Your work reminds me of vintage electronics before the PCB came. Good!

Beautiful work - never mind the people who criticize the boxes because the knobs and LEDs are not aligned. Let them go buy a stereo rack to look at ...

@ robrenz: Read through most of the PD-thread. Very impressive instrument collection you have, and I also note you show serious measurement techniques knowledge. Impressive!

@ robrenz & Blackdog. I just got a small job that demands full attention for a few weeks. And as I have seen your skills, knowledge and equipment, I don't think it is so    important that I finish my 'comparison'.

Let's just say that:

1. Line stability can be handled by right regulation.

2. Load stability can be obtained by proper dimensioning.

3. Temperature is best handled by 'freezing' the temperature at a level higher than any ambient operating temperature and adjusting the reference / comparison circuit / etc. to a small tempco for this fixed temperature.

The makers of Blackdog, PD and Harrison PSUs as well as we understand this. I

'll be around - but my family can't eat  ppm's, so I'll better get this job done

[don.r]

Beautiful work - never mind the people who criticize the boxes because the knobs and LEDs are not aligned. Let them go buy a stereo rack to look at ...

Yeah, needs MOAR dancing LED spectrum analyzer..  O0

3. Temperature is best handled by 'freezing' the temperature at a level higher than any ambient operating temperature and adjusting the reference / comparison circuit / etc. to a small tempco for this fixed temperature.

The makers of Blackdog, PD and Harrison PSUs as well as we understand this.

The oven in a PD 2005 is a thing of beauty. Might make a good starting point for a temp. stabilized voltage reference.

[robrenz]

Results from my first PD supply a 2010. as recieved from fleabay other than cleaning and calibration.
Voltage set to 10.0000 V  actual voltage was 10.00078 so 78ppm off.  My 8846A is only a 23ppm meter but cal at 10V was spot on. (I know it does not mean its spot on now, dont start a rant on that)  I tested my 8846A shorted input drift over 152 hours at 1µV total span with a SD of 73nV.

8 hour PD2010 @10V stability was 23.23µVpp   2.3ppm over 8 hours is not bad

[blackdog]

Hi robrenz,

It's looking good!   
Nice stability, if your room temp was relativ stable al the drift is your PD power supply.

You need to test the Fluke on the range you use, thats the 10V range and look at its drift, and that is difficult, do you have a Josephson Array laying around? 
Based on my experience with two Tektronix DMM4050 multimeters I can say that on the 10V range, there is olmost no drift at room temp.

Kind regarts,
Blackdog.

[robrenz]

The total temperature span over that test time was 1.7 deg.C  A interesting thing is that the 2010 does not have a ovenized reference.  I am letting one of the PD2005 units stabilize for a test now.

I didn't feel like setting up my josephson array for checking the 8846A drift at 10V, too much trouble

[quantumvolt]

Nice work all.

I haven't done much since my 6102A is still over the Pacific somewhere  . Also I have to work long days, so I can only do a little bit at night.

Having followed eBay prices in general (all brands) and also Agilent Malaysia's 2nd-hand website, I have decided to go basic. Simple is beautiful - they say ... After all I am a mathematician  / physicist - not an engineer.

So no big buy's or fancy stuff. My oscilloscope-money just went to buy 2 tablets with learning programs and games for the children. My young naive wife said it was a good investment because of the learning potential, but I haven't seen anything but games in the devices history-stack 

This will be my first DIY hunt for precision:




It will be a voltage ref. The set up is a 5 dollar DC unregulated 12 Volt wall supply powering a 7805 regulator stepped up with a diode to some 5.6 Volt which finally feeds a 3 to 8 dollar monolithic reference (in this case REF195E 5 Volt - two refs in series on separate supplies to get 10 Volt - but I am testing several chips - very promising is 4 x 2.5 Volt refs).

The enclosure is a water bottle oven powered by 12 Volt unregulated into 9 pcs. 1/4 Watt resistors with a total resistance of ca. 250 Ohm. Total power a bit more then 1/2 Watt.


Ref.:




Heater:




NB /// Ignoring the 34401A basic (best) spec of 19 ppm. It is the only precision instrument I have//

I have followed it for days. With ambient  temp changes from around 20-35 deg Celsius (in Asia the walls are cement-painted cardboard thrown sand on - and the windows/doors are open all the time / no air-con ), it drifts from 4.99966 to 4.99971 Volt.  This is 50 microVolt or 10 ppm for 5V. I do not have a thermometer (true!), so I don't yet know the oven's average temp and drift. Very important is that the resistors are just "feelable" warm - maybe 45+ deg Celsius. I cannot measure noise, but the DMM gives 0.4 milliVolt AC. NO measures taken whatsoever about battery power, shielding and cables.

I will increase the temperature a bit - some of the refs have a sweet point with flat tempco-curve. I will replace the bottle with a multilayer box. I will build a very small inner feedback high time-constant heater with 1 transistor-junction as temp sense and the other as heater (0.1 Watt) to be put in the smallest enclosure opposite the ref board. The resistor heater will be inside the outer layer (even at low temp the 2 outer layers should be fire proof).

I will achieve at least a few ppm's. Change of LiPo every week (2 batteries in diode-parallel). May be less than 1 ppm. Philosophically and system-wise this is IMO far better than the LTZ1000. The latter is as it is because of the industrial temp range and intended use in all kinds of devices. But in the end the goal is to eliminate temp changes in the ref - not to make some highly sophisticated combined heater-ref-sensor chip that is aggressively controlled with high gradient feedback.

I have found a nice guy in the US that will measure these devices for  USD 5 a piece when I send them in with the USD 25 ref I bought from him (you all know who he is). I will then - if needed - get 10 (!) independent refs measured at NIST-traceable 3458A spec for USD 50 and 20 or something for shipping Thailand-USA-Thailand. By series-parallel I can start my own cal-lab

And when I get some cash (really broke now ...) I will order a (USD 10-20) LTC2400 24-Bit ADC and hook it up to Arduino as a data logger.

This should be good enough to start to study and modify the Harrison 

The IMO ultimate voltage reference enclosure 





(I will build it from corro/styro as rectangular boxes with 1 side open: the smallest rectangular box slides in to the 2nd smallest open-end first (hence closing the smallest), the 2nd smallest slides in to the 3rd smallest open-end first and so on. Cables and electronics for the layers follow the boxes' flat sides by folding (as in ironing and folding?. The two outer layers from copper-clad raw PCB).

Will be back later...

[blackdog]

Hi, :-)

I am designing a PSU Short Circuit Tester so i can measure the behavior of a Power Supply after a short circuit.
I did some initial testing, maybe its interesting...

Yes its in Dutch, but again, a lot of pictures en you can use a translator 

http://www.circuitsonline.net/forum/view/113455/1

Kind regarts,
Blackdog

[quantumvolt]

A very nice project. No problem to read after translation. Real science - half physics and half electronics. Nice you found use for the small buck/boost converter. I have 2 from eBay. Will look at them later.

I have put away my HP Harrison for a while. When I got it from the USA, it turned out that the seller had used a small flat-rate box with almost no protection, and the PSU must have fallen at least a meter or two - the fins/ribs on the cooling unit were broken as were the knobs on the front. So I got angry and put it away for now. It is 40 years old, so it can wait a little bit more ...

I am now working with 10 V refs. I have around 5ppm/24 hour with references costing just a few dollars. It is nice working with microVolts and kiloAmperes  (your PSU killer elcap discharge)

My next project will also be kitchen bench science. I am studying litterature for expensive many thousand dollars LCR bridges. I have started slowly to make my own using the Agilent 34401A as voltage sampler / display and Arduino/Processing/C++ for control and processing.

So then you know - I won't start to repair the PSU now. Next year ...

[calin]

I know I am reviving an old thread .. but .. i can't stop it !!! .. sorry.


I just saved from the dump an "AS-IS HP 6112A", for 25$ plus 14$ shipping. Was on fleabay and stumbled over it ... says it does not power up so I hope is fixable.


Wish me luck    .. and yes Rob is all your fault, if wife kills me for spending again money on "my crap" I have someone to blame


EDIT: Added the before pictures