HP 8656A fan questions

[bdunham7]

I recently purchased a cheap, tired and beat HP8656A for a project, which it was.  Sold as 'parts or repair only', it not only needed repairs but was missing a lot of internal hardware and some parts.  I have it all working now (all with parts purchased at 4 different hardware vendors--grommets, ball bearings, screws, locknuts and washers) but I'm mystified by the fan.  The old one was louder than my kitchen blender, so I replaced it with a slightly less deafening one that is functionally identical.  The problems are that there is only one way to put it in and the way it is (and was) makes the fan blow IN not OUT, which seems odd.  As I was sort of looking it over to see if that seemed reasonable, I realize that while there are internal holes, there is no way for air to escape or enter the case, so the fan can't move any air in either direction.  The fan housing seems to only have the function of resonating and amplifying the fan noise.  I do have a bit of air coming out the holes where the feet are missing, but those would normally be sealed up.  So how does this work and should it blow in or out?

Also, what gives with parts availability on these?  Is nobody parting out these no matter how broken they are?  Or are they getting ground up for the gold?  I'd like to find a handle strap, side covers, feet and maybe some key switches, but those are all either unobtanium or priced like gold ($50 for each plastic switch....). 

[MadTux]

Tell me what you need, I might have a broken one somewhere around.
The last time I shipped some small stuff from Germany, shipping prices to US were OK too.

In my opinion, HP8656/7 is pretty much low end HP crap. At least nowadays, when you can get instruments such as 8662A/8663A or 8644/8645 for rather little money. Specs are bad as well, inside looks ugly and boards are bad to reach/measure on.

[pigrew]

My 8656B blows OUT, though I'm pretty sure I replaced the fan so perhaps I screwed up the direction? It has a NMB 3110SB-05W-B50 with the label such that it is visible through the back grill.

You're right that there are almost no vents. There are a few small holes in the chassis behind the fan, and some small gaps around the back panel, and a few 3mm holes too. It seems like bad thermal design. I think the fan is mostly just to cool the transformer and transistors (LDOs?) in mounted in the back area. The current design must be really reducing the airflow from the fan.

My unit was also bad when received. It turned out that it had a bad diode in the phase detector of the PLL and one of the huge electrolytic capacitors on the power supply board had leaked. In retrospect, I should have gotten a 8657A/B model since they go >= 1 GHz.

I ordered feet direct from Malaysia on eBay. They are about two dollars each, search for 5041-8801 or 5041-9167.

In terms of repair, I was pretty happy with the ease. The boards fold apart, leaving everything accessible when the unit is balanced on the side. Also, nearly everything is discrete so it can be fixed (other than the N-fractional PLL ASIC).

[bdunham7]

Tell me what you need, I might have a broken one somewhere around.
The last time I shipped some small stuff from Germany, shipping prices to US were OK too.

In my opinion, HP8656/7 is pretty much low end HP crap. At least nowadays, when you can get instruments such as 8662A/8663A or 8644/8645 for rather little money. Specs are bad as well, inside looks ugly and boards are bad to reach/measure on.

Thanks, I'll work up a short list.  I've put on some rubber feet I had and fixed most everything else, so the main things would be the side handle strap, end covers and maybe side covers if you have some with fewer broken tabs than mine.  Also rear bumpers if they are mint, mine are OK but a bit cracked. As far as other stuff I might want to stock, I suppose keyswitches, attenuator contacts and cam levers would be nice, but that would require disassembly work on your part and I don't need them for mine at the moment. So only if you are willing.

As far as how good it is, I paid $75 for mine locally and I think it is worth every penny.      But I've heard the same complaints from others and I can see why--it is a huge clunker and I can imagine it not being reliable in an industrial production environment.  The repairs weren't terribly awful, but you need a HUGE clean bench (clean being the obstacle for me) to work on it when it is all unfolded.  And reassembling those attenuators is not for the fat-fingered or the nervous!  Still, it took very little adjustment after repairs to get fairly stable, level outputs to the best of my ability to measure.   

[bdunham7]

My 8656B blows OUT, though I'm pretty sure I replaced the fan so perhaps I screwed up the direction? It has a NMB 3110SB-05W-B50 with the label such that it is visible through the back grill.

You're right that there are almost no vents. There are a few small holes in the chassis behind the fan, and some small gaps around the back panel, and a few 3mm holes too. It seems like bad thermal design. I think the fan is mostly just to cool the transformer and transistors (LDOs?) in mounted in the back area. The current design must be really reducing the airflow from the fan.

My unit was also bad when received. It turned out that it had a bad diode in the phase detector of the PLL and one of the huge electrolytic capacitors on the power supply board had leaked. In retrospect, I should have gotten a 8657A/B model since they go >= 1 GHz.

I ordered feet direct from Malaysia on eBay. They are about two dollars each, search for 5041-8801 or 5041-9167.

In terms of repair, I was pretty happy with the ease. The boards fold apart, leaving everything accessible when the unit is balanced on the side. Also, nearly everything is discrete so it can be fixed (other than the N-fractional PLL ASIC).

Thanks for checking!  The 'B' must be different, mine has a 120VAC 92mm fan, so different than yours.  I'm tempted to cobble in a lower voltage fan for less noise since it doesn't cool much anyway.  I'm also tempted to cobble in the existing fan to blow out and see if it is quieter.  At the level it is at now, it should be mining bitcoins!

The feet you mentioned don't appear to work on mine--I have two holes in the front and three in the back--and the part numbers listed are different.  Am I missing something?

[pigrew]

Thanks for checking!  The 'B' must be different, mine has a 120VAC 92mm fan, so different than yours.  I'm tempted to cobble in a lower voltage fan for less noise since it doesn't cool much anyway.  I'm also tempted to cobble in the existing fan to blow out and see if it is quieter.  At the level it is at now, it should be mining bitcoins!

The feet you mentioned don't appear to work on mine--I have two holes in the front and three in the back--and the part numbers listed are different.  Am I missing something?

I had completely forgotten the two pegs thing. I cut off the front peg from the foot and it fits nicely. I have not tested it with a bail, though. See the attached image.

Yeah, they completely changed the fan. I don't know if reducing the fan speed is good or not (I wouldn't), but I'm guessing it'll be fine since it would have been designed to operate in a rack of hot equipment. Perhaps you can add a series resistor to lower the fan speed if it internally is a DC fan (i.e. has an internal rectifier). My understanding is that AC motors do not like undervoltages, and it may be difficult to synthesize a lower frequency AC source......

-Nathan

[duak]

About slowing down AC fans -  Most if not all small ones are impedance protected so if they stall or the bearings seize, they won't catch fire.  I've operated a number of them on reduced voltage either from taps on the power transformer or with a series resistor.  Just make sure the resistor doesn't get too hot. 

I have an hp 6002A power supply with a howler of a fan.  Changing it out makes a heart transplant look easy,  The fan AC cable end mates with 0.025" pins so I soldered a couple of pins to the power transformer secondary connections on the PCB and just plugged the cable in.  I think the voltage is something like 70 VAC and fan is much quieter.

[pigrew]

I learn something new every day. I found a long thread about slowing shaded pole induction motors. It looks like they normally operate with so much slippage that they can reasonably be slowed with lower voltage (as long as they have enough torque to initially start). It also suggests using a series film capacitor (of sufficient voltage rating) to reduce the voltage (probably easier than a resistor and it wouldn't heat up nearly as much).

[HalFoster]

I have had very good luck changing out the AC fans for a comparable DC version - same performance for much less noise.  A couple of things to watch for, though:  Never power it from the regulated side of a rail since, even if it doesn't over stress the rail, it will inject a *lot* of noise.  Usually once the large filter caps have been replaced with something a fifth of their original size there is usually more than enough room left to add a small 12 VA transformer and rectifier/filter.  A 18V transformer with a bridge and cap output generates right at 24V for a 24V fan. If you want to slow it down you can always add a LM317 adjustable, but usually it's better (easier & cheaper) just to add a series 5W or so resistor in series to drop the voltage (and fan speed) to where you want. I would personally stay away from any of the DC/DC converter bricks even if they are a really small and inexpensive power source - introducing a switching power supply into a piece of precision test gear usually doesn't end well.

Anytime you replace a fan the biggest "gotcha" by far, though, is to only look at the CFM rating; the CFM really doesn't mean spit unless it can deliver it at the same static pressure the particular system's thermal design requires.  This is critically important with HP's RF gear in particular - it tends to have the fans blowing into the case and creating a large positive pressure to force air through the (small) cooling channels in the RF blocks.  More than one piece of that gear will overheat *fast* if it is operated with the cover off - RTFM to know for sure.  BTW, the HP 3335A is a perfect example of this; those tiny screens on each end of the RF cards are the air inlet/outlets.  It takes a whole honkin' lot of air, under pressure, to keep it all cool and working - and the power supply output transistors tend to cook even with the cover on even at the best of times.

Hal



 

[duak]

The electrical noise from DC fans can be a real pain.  I was responsible for the design of the power distribution in an assembly that had a number of DC fans.  I knew about the current transients but there wasn't any space for an isolated supply.  A simple RC filter of a few ohms in series with a few hundred uF in parallel was all that was needed.  Make sure the fan power leads are twisted to minimize loop area as the radiated current transients are sharp enough to induce noise in sensitive circuits.

[HalFoster]

duak, for new gear - especially after an SMPS -  you are absolutely correct, and for most newer test gear it will work and it's my impression that is what you were referring to, but it could also be taken by others to mean it's OK to do in *all* circumstances.  For the older gear I was referring to, even if it doesn't cause a immediate release of Magic Smoke, throwing an additional load like a DC fan on the rail(s) of equipment like we are talking about is definitely not trivial and can cause subtle issues later on.  On the digital side, the uP will often either lock up or behave erratically if the digital rail has noise or the power-on profile is different.  Hanging it on one of the analog rails (also known as "inviting guaranteeing trouble) in not something to be done lightly, and preferably not at all.  Some of them can source enough current, but a surprising number are regulated by a zener shunt and are limited to low currents.  I'm sure you are aware of the issues, but for many of those reading this who may just be getting into test gear and electronics it is a very risky thing to do.

Hal

[bdunham7]

Thanks for the input on my 'project'.

Duak -- both the old and new fans do say 'impedance protected', I suppose I can just put in a 5uF motor capacitor from a ceiling fan or something.  It's what I did with my attic exhaust fan to change its speed when it was resonating with something.

HalFoster -- thanks for the input.  The fan does seem to be operating at a very high static pressure,  I just can't see how the airflow works on this beast.  I know they need to seal it up to prevent RF leakage, but I just don't see where the air can get out.  I think I can slow the fan down a bit if I want to go to the trouble, since I'm not running it continuously or in a hot environment.  And if it did burn up, then I'd get to find out exactly how much gold there is in there.

Pigrew -- thanks for the photo.  I think I'm going to keep my rubber feet (see photo) as I hardly have a collectible, mint unit in any case.

[HalFoster]

bdunham7, Sorry I was talking more in general than specifics.  In the case of your 8656A you should be fine slowing the existing fan down some; it's in HP's higher-end, bleeding edge (at the time) stuff that there is not much margin to play with.  If I remember right, the 8656A bottoms out at around -120 or -130 dBm which, while not as low as the bleeding-edge gear, is still pretty much equivalent to "nothing" in my mind.  The problem, though, is noise of the electrical type - to go that low and still have purity of your signals requires shielding - lots of heavy, expensive and hard to cool blocks of aluminum enclosing the circuit blocks/cards.

Hal

[tverbeure]

Anytime you replace a fan the biggest "gotcha" by far, though, is to only look at the CFM rating; the CFM really doesn't mean spit unless it can deliver it at the same static pressure the particular system's thermal design requires.  This is critically important with HP's RF gear in particular - it tends to have the fans blowing into the case and creating a large positive pressure to force air through the (small) cooling channels in the RF blocks.  More than one piece of that gear will overheat *fast* if it is operated with the cover off - RTFM to know for sure.  BTW, the HP 3335A is a perfect example of this; those tiny screens on each end of the RF cards are the air inlet/outlets.  It takes a whole honkin' lot of air, under pressure, to keep it all cool and working - and the power supply output transistors tend to cook even with the cover on even at the best of times.

if I want to replace the HP 8656A with an equivalent one that is hopefully quieter (I'm counting on 40 years of technological improvements here), what do I need to look at in addition to the specs in the service manual: 115V, 50/60Hz, 36-CFM?

I also want to experiment with adding a series resistor to slow the fan down. What kind of order of magnitudes are we looking at? ChatGPT hallucinated a 120 Ohm, 7 Watts (that's with a 2x safety margin) resistor to reduce the voltage from 120V to 100V for a 20W motor. It provides the mathematical derivation, and the math looks reasonable, but I have no intuition about the real power consumption of a fan like this.

As an alternative to adding a resistor, I'm wondering if it would make sense to connect the fan to different connections of the transformer instead.

Attached is the schematic for the case where the 120V selection is active. You can see now the fan is fed straight from the 120V input (after going through the line filter.)

Instead of connecting the fan straight to the red connection, I could connected it to connection 3 of the transformer. This is normally used when the 100V selection is active. So I'd be using the transform as a voltage divider to reduce the fan voltage to 100V as well. Does that sound reasonable?

The 8656A is already doing that in 240V mode (also attached). In this case, two windings of the transform are in series, and the purple wire is connected to the middle of those two windings.